Hunter-Gatherer Economics and the Formation of a Housepit Floor Lithic Assemblage
Bibiiotheque nationale du Canada
Acquisitions and Direction des acquisitions et Bibliographic Services Branch des services brbliographiques 395 b1elling!on Stre? 395, rue Well~ngton Ottawa, Ontario Otiawa (Ontario) KIA ON4 KIA ON4
NOTICE AVIS
The quality of this microform is La qualite de cette microforme heavily dependent upon the d6pend grandement de la qualit6 quality of the original thesis de la thke sournise au submitted for microfilming. microfilmage. Nous avons tout Every effort has been made to fait pour assurer une qualite ensure the highest quality of supbrieure de reproduction. reproduction possible.
If pages are missing, contact the S'il manque des pages, veuillen university which granted the communiquer avec 11universit6 degree. qui a confere le grade.
Some pages may have indistinct La quaiit6 d'impression de print especially if the original certaines pages peut laisser a pages were typed with a poor desirer, surtout si les pages typewriter ribbon or if the originales ont &e university sent us an inferior dsctylographiees a I'aide d'un photocopy. ruban use ou si I'universitb nous a fait parvenir une photocopie de qualit6 inferieure.
Reproduction in full or in part of La reproduction, meme partieile, this microform is governed by de cette microforme est soumise the Canadian Copyright Act, a la Loi canadienne sur le droit R.S.C. 1970, c. C-30, and d'auteur, SRC 1970, c. C-38,et subsequent amendments. ses amendements subsequents. HUNTER-GATHERER ECONOMICS AND THE FORMATION OF
A HOUSEPIT FLOOR LITHIC ASSEMBLAGE
William C. Prentiss
B.A., University of South Florida 1982
M.A., University of South Florida 1986
THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
In the Department of ARCHAEOLOGY
@ William C. Prentiss 1993 SIMON FRASER UNIVERSITY
March, 1993
All rights reserved. This work may not be reproduced in whole or in part, by photocopy or other means, without permission of author. National Library Bibliotheque nationale of Canada du Canada Acquisitions and Direction des acquisitions et Bibliographic Services Branch des services bib1iograph;ques 395 Wellington Street 395, rbe Welmglsn Ottawa, Ontano Ottawa (Ontarlo) KIA ON4 KIA ON4
The author has granted an L'auteur a accorde une licence irrevocable rton-exciusive licence irrevocable et nen exclusive allowing the National Library of perrnettant A la Biblioth&que Canada to reproduce, loan, nationale du Canada de distribute sr sell copies of reproduire, priiter, distribuer ou - his/her thesis by any means and vendre des copies de sa these in any form or format, making de quelque maniere et sous this thesis available to interested quelque forme que ce soit pour persons. mettre des exemplaires de cette thhe a la disposition des personnes interessees.
The author retains ownership of L'auteur conserve la propriete du the copyright in his/her thesis. droit d'auteur qui prothge sa Neither the thesis nor substantial these. Ni la these ni des exttaits extracts from it may be printed or substantiels de celle-ci ne otherwise reproduced without doivent 6tre imprimes ou his/her permission. autwment rsproduits sans son -- autsaisation.
ISBN 0-315-91101-8 APPROVAL
NAME: William Clark Prentiss
DEGREE: Doctor of Philosophy (Archaeologyi
TITLE OF THESIS: Hunter-Gatherer Economics and the Formation of a Housepit Floor Lithic Assemblage
EXAMINING COMMITTEE:
Chairman: David Burley
~kianHayden sfhior Supervisor Professor
---- -I-v-t------L ~ardd'~ibb1e External Examiner Professor Department of Anthropology University of Pennsylvania PARTIAL COPYRPGHT LICENSE
1 hereby grant to Simon Fraser University the right to tend my thesis, project or extended essay (the title of which is show^^ bc!ow) to users of the Simon Fraser University Library, and to make partial or single copies only for such users or in response to ri rcqucst f~om the library of any other university, or other education:iI institution, on its own behalf or for one of its users. I further agree that permission for multiple copying of this work for schol;wly purposes may be granted by me or the Dean of Graduate Studies. it is understood that copying or publication of this work for financial gititi shall not be allowed without my written permission.
Title of Thesis:
HUNTER-GATHERER ECONOMICS AND TIfE FORMATION OF A HOUSEQIT FLOOR L.IrF'f-ILC ASSEMBLAGE
William Clark Prentiss November 19, 1992 ABSTRACT
This thesis is designed to present and test the
effectiveness of a method for recognizing the culling sf flakes for tool use. A modified version of Sullivan and Rozen's debitage typology and three flake utility indices are used to anticipate the character of the archaeological record under different formational conditions. The primary eozmation processes considered focus on the effects of hunter-gatherer economic behavior on the archaeological record, especially as reflected in the decisions on how to produce, use and reuse stone tools, Throughout the thesis, it is argued that the concept of risk management best explains the variety of economically oriented strategies and decisions made by individuals and groups attempting to avoid the possibility of reduced access to critical subsistence resources. The method is presented in three phases of research. First, lithic technological organization is demonstrated to be a critical component of larger systems cf risk management in hunter-gatherer societies. The ethnographic people of the Middle Fraser Canyon on south-centmak British Columbia provide an example hsw the globally described risk management tactics can be applied to a single case study. Second, experimental lithic research is presented, designed to develop and test
~ethedsfor recognizing prehistoric fithic reductfen techniques, flake culling and flake tool reuse/recysling, More specifically, a ~eliabflityand validity analysis is conducted of the modified version of Sullivan and Rozen's iii debitage typology and the three lithie utility indices and mathematically derived debitage expeetations are developed for
recognizing patterning in fhe archaeological record, Third, the applicability of the methods for recognizing pxehistoric
economic behavior is tested through an analysis of a Late Prehistoric housepit floor lfthie assemblage from the #eatley
Creek site in south-central British Columbia. A variety of reduction and flake culling strategies are recognized in different parts of the floor. Results of the analysis support the argument that winter
hausepit occupation resulted in a high degree of raw material
conservation through the reduction of small semi-prepared spheroid cores, recycling exhausted cores and intensive reuse of previously discarded flake tools. Pt is argued that lithic raw material conservation allowed the requisite tools and clothing to be produced during winter to reduce the risk of
spring shortages in subsistence resources. In general, it is concluded that the methods tested and applied to these data are useful for future archaeological research. Recommendations are made for further experimentation and refinement. I have received a tremendous amount of support as a
yraduate student at Simon Fraser University. My senior supervisor, Dr. Brian Hayden, provided financial support through employment in the field and laboratory on the Keatley
Creek Project and recommendations for fellowship support from S.F.U. Dr. Hayden helped provide me with the opportunity to teach Archaeological Theory and Lithic Teehnolcgy to a large proportion of the unsuspecting undergraduate population at S.F.U. Dr, Hayden provided considerable information on lithic technology and Plateau archaeology and ethnography, and above all, encouraged me ro txy out new analytical techniques with act -:ials from his excavations at Keatley Creek. For these reasons, I am most grateful.
I thank Dr. Jack Nance for his support of my research. f have been most fortunate to have his input particularly in the area of reliability and validity assessment. During his tenure as Chair of the Archaeology Department at S .F.U., Dr. Nance has also helped provide me with research funding (S.F.U. f~llowships)and teaching positions. I was most fortunate to have Dr. Knut Fladmark
( internal-external examiner) and Dr. Harold Dibble f external examiner) as members of my examining committee. I thank them both for their time spent reading my thesis and their challenging questions asked during my defense. I thank Dr. David Burley for serving as chair of my V examining committee and for his support throughout my tenure as an S.F.U. graduate student.
Many fellow students and colleagues have provided input during the time spent researching this thesis. I offer tha~ks to all. Many of the ideas developed in this thesis originated
from a conversation between Gene Romanski and myself held on a cold night in 1985 at the now defunct Washakie Hotel Bar in Worland, Wyoming. Gene's input was critical in my early thinking regarding the use of "distinctive assemblage1' approaches. Ian Kuijt worked with me on several research projects dealing with Sullivan and Rozen's typology and was always willing to play devil's advocate to help refine new ideas. Likewise, Dr. David Pokotylo offered useful advice on lithic technology and information regarding ethcographic stone tool use in the Northwest Territories of Canada. Dan Amick,
Ray Mauldin and Dr. Lewis Binford, offered commentary and encouragement to an early pilot study preceding the research developed here. Eric Ingbar and Dr. Mary Lou Larson offere6 productive critical commentary on some components on this research. Dr. Jonathan Driver provided very productive early encouragement (from a zooarchaeological perspective) which aided tremendously in this research. Jim Welch provided space and support in the completion of this thesis. Numerous graduate and undergraduate students provided me with information and/or listened and commented on my ideas. I thank, in particular, Diana Alexander, Dave Crellin, Rob
Gargett, Martin Handly, W. Karl Hutchings, Dr. Chris Knusel, v i Karla Kusmer, Malcolm James, Dana Lepofsky, Dr. Yvonne Marshall, Donna Morrison, Laurie Milne, Kike Richards, Mike Rousseau, Rick Schulting, Greg Sullivan, and Jason Turner. I extend special thanks to Jim Spafford, whose research into housepit floor lithic artifacts paralleled my own in some ways. Jim drew all of the figures in this thesis (other than lithic illustrations) and was always willing to listen and offer advice, critical cotnmentary or encouragement to my prognostications regarding lithic analysis, Plateau archaeology, archaeological theory and many other topics. I extend special thanks to Dr. John P. Cook for his encouragement and offer of financial aid to conduct research
in the Atigun Pass area of the Bureau of Land Management's Arctic ~istrict.
The Archaeology Department staff at S.F.U. have been an favaluable source of aid in the completion of this thesis. In particular, I thank Ingrid Nystrom, Linda Bannister, Andrew Barton and Linda Przybyla. I owe a special debt of thanks to Ingrid, who has been supportive since the day I arrived at S.F.U. My sincere thanks go to my parents, Bill and Sally Prentiss, my brother and sister, Charley and Liz Prentiss, and to my wife's parents, Esther and Jim Backhouse. I thank Dr. Jim Backhouse for his interest and insightful comments on the logical structure of my arguments. Josie Backhouse has provided emotional and intellectual support to me throughout my tenure as a graduate student at S.F.U. Thank you Josie. To my daughter Thecla, you have added a new dimension to my life and whether you know it or not, 1 thank you for your good humor and love that allowed me to complete this thesis.
This Ph,D.'s for you! +rl ..-4 .Ti X
rn al a4 a E. W 0 JJrn "Ti d Flake Volume Index ...... 108
Acute Angle Edge Length ...... 165 High Angle Edge Length ...... I10 Validity Study Data ...... 111 Sullivan and Rozen Typology ...... 111 Modified Sullivan and Rozen Typology ..129 Flake Volume Index ...... 133 Acute Angle Edge ~ength...... 13 4 High Angle Edge Lenyth ...... 135 Reliability Analysis ...... 136 Assessing Reliability ...... 337 Assessment of Reliability ...... 151 Sullivan and Rozen Typology ...... 1.51 Modified Sullivan and R~zerTypology . . 1.53 Flake Volume Index ...... 162 Acute Angle Edge Length ...... 167 High Angle Edge Length ...... i71 Reliability Retests ...... 175 Flake Volume Index ...... 176 Acute Angle Edge Length ...... 176 High Angle Edge Length ...... 181 Validity Analysis ...... 188 Assessing Validity ...... 188 Assessment of Validity ...... 195 Sullivan and Rozen Typology ...... 195 Modified Sullivan and Rozen Typology ..203 Flake Volume Index...... 215 Acute Angle Edge Length ...... 222 X High Angle Edge Length ...... 229 summary ...... 235
Chapter 4: Experimental Utility Index Sequences ...... 238 Lithic Reduction Experiments ...... 239 Construction of Utility Index Sequences ...... 245 Utility Index Sequences ...... 307 Stage 2 Biface ...... 308 Stage 3 Biface ...... 411 Soft Hammer Flake Retouch ...... 413 Hard Hammer Flake Retouch ...... 416 Stage 3 Biface Pressure Flaking ...... 419 Flake Edge Pressure Flaking ...... 421 Prepared Block Core Reduction (Medium Size Flake Production Goal) ...... 422 Unprepared Block Core Reduction (Medium Size Flake Production Goal) ...... 424 Prepared Block Core Reduction (Large Size Flake Production Goal) ...... 427 Unprepared Block Core Reduction (Large Size Flake Production Goal] ...... 429 Bipolar Core Reduction (Medium Size Flake Production Goal) ...... 431 Bipolar Reduction on Bifacial Core (Medium Size Flake Production Goal) ...... 433 Summary ...... 435 Chapter 5: Analysis of the Debitage and Flake Tools from a Housepit Floor ...... 441 m~LII~ Keatley Creek Site ...... 442 Analytical Methods ...... 449 Analysis ...... 461 Debitage Analysis ...... 462 Xl
List of Tables
Table 1. Total raw and rescaled Sullivan and Rozen typology reliability data...... e.lOO Table 2. Total raw and rescaled Modified Sullivan and Rozen Typology reliability data...... 101 Table 3. FVI reliability data...... 103 Table 4. AAEL reliability data...... 104 Table 5. HAEL reliability data...... 105 Table 6. Reliability summary statistics...... e..106 Table 7. Sullivan and Rozen typology validity analysis raw flake count data..,...... ll2 Table 8. Sullivan and Rozen typology validity analysis rescaled data...... ,...... e..l13
Table 9. Modified Sullivan and Rozen typology validity analysis raw flake count data...... 114 Table 10. Modified Sullivan and Rozen typology validity analysis rescaled data...... A6 Table 11. FVE validity analysis raw index data for available flake types ...... 119 Table 12. AAEL validity analysis raw index data for available flake types...... 123 Table 13. HAEL validity analysis raw index data for available flake types ...... ,...... A26 Table 14. Sullivan and Rozen Typology reliability analysis correlation matrix...... ,P54 Table 15. Sullivan and Rozen typology reliability analysis.initia1 statistics...... 155 Table 16. Sullivan and Rozen typology reliability analysis rotated loadings...... 156 Table 17. Theta coefficient data...... 157 Table 18. Modified Sullivan and Rozen typology reliability analysis correlation matrix...... l58 Table 19. Modified Sullivan and Rozen typology reliability analysis initial statistics...... l60 Table 20. Modified Sullivan and Rozen typology xiii reliability analysis rotated loadings .....=..161 Table 21. FVI reliability analysis correlation matrix.,l64 Table 22. FVI reliability analysis initial statistics..l65 Table 23. FVI reliability analysis rotated loadings ....166 Table 24. AAEL reliability analysis correlation matrix.168 Table 25. AAEL reliability analysis initial statistics.169 Table 26. NLreliability analysis rotated loadings ...170 Table 27, HAEL reliability analysis correlation matxix.172 Table 28. HAEL reliability analysis initial statistics.173
Table 29. MJBL reliability analysis rotated loadings ...174 Table 38. FVZ reliability analysis c~rrectedcorrelation matrix...... ,...... ,..177 Table 31. FVI reliability retest initial statistics....l97 Table 32. FVf reliability retest rotated loadings ...... 178 Table 33. Reliability retest theta coeficient data.....179
Table 34. AAEL reliability analysis corrected correlation mtrL~.~...... ,.,....,.....~..l80 Table 35. AAEL reliability retest initial statistics...l82 Table 36. AAEL reliability retest rotated loadings .....183 Table 37. HAEL reliability analysis cozrected correlation matrix...... e...... e...... *.. 184
Table 38. HAEL reliability retest initial statistics...l85 Table 39. HAEL reliability retest rotated loadings .....186 Table 40. Sullivan and Rozen Typology validity analysis correlation matrix...... c187
Table 41- Sullfvan and Rszen typolsgy validfty analysis initial statistics...... 197 Table 42, Sullivan and Rozen typology validity analysis rotated loadings ...... e*...198 Table 43. Sullivan and Rozen typology validity analysis factor scores...,....,...... 199 Table 44. Modified Sullivan and Rozen typology xiv validity analysis correlation matrix...... 200 Table 45. Modified Sullivan and Rozen typology validity analysis initial statisti~s,..~..=.~205 Table 46. Modified Sullivan and Rozen typology validity analysis rotated loadings ...... 206 Table 47. Modified Sullivan and Rozen typology validity analysis factor scores...... 207 Table 48, FVI validity analysis corrected correlation matrix ...... 210
Table 45. FVI validity analysis initial statistics.....217 Table 50. FVI validity analysis rotated loadings...... 218 Table 51 PVI validity analysis factor score data matrix .....O...... 219 Table 52. AAEL validity analysis corrected correlation - - matrix ...... 221 Table 53. AAEL valid-ity analysis initial statistics....224 Table 54. AAEL validity analysis rotated loadings ..-...225 Table 55. AAEL validity analysis factor score data matrix...... U...... e.....226 Table 56. HAEL validity analysis corrected correlation matrix...... 228 Table 57. HAEL validity analysis initial statistics....230 Table 58. HAEL validity analysis rotated loadings...... 231 Table 59. HAEL validity analysis factor score data IM~L~X...... 232 Table 60. Stage 2 biface utility index data sequence...246 Table 61. Stage 3 biface utility index data sequence ...251 Table 62. Soft hammer flake retouch utility index data sequence ...... 257 Table 63. Hard hammer flake modification utility index data sequence ...... *.e...e...... 262 Table 64, Biface pressure flaking utility index data sequence ...... -267 Table 65. Flake edge pressure flaking utility index data sequence ...... -270 xv Table 66. Prepared block core (medium flake) utility index data sequence..,,,,.. ...*IIII..Z?3
Table 67. Unprepared core reduction (medium flake) utility index data sequenceo...*...... 278 Table 68. Prepared con reduction (large flake) utility index data aequence.,...... 283
Table 69. Unpfe ared core reduction (large flake) utili% y index data sequence ...... 38 Table 70. Bipolar core reduction (medium flake) utility index data sequence ...... 293 Table 71. Bipolar neduction (bifacial core, medium flake utility index data sequence ...... Y..Z98
Table 72- Construction of basic split flake indices from trampled unprepared core reduction (large flake) (Table 69)...... C....303 Table 73. Raw debitage data matrix from the floor of housepit 7...... ,...... 463 Table 74. Rescaled debitaqe data matrix from the floor of housepit 7..,.....,...... 469
Table 75. Housepit 7 debitage analysis coxrelation mtrix...... ,...477
Table 76. Housepit 7 debltaqe analysis initial statistics...... 479 Table 77. Housepit 7 debitage analysis rotated loadings ...... e...... 480 able 78. Housepit 7 debitage analysis factor scores...481 Table 79. Housepit 7 debitaqe analysis factor 3 experimental sequence A...... o...... 502
Table 80. Housepit 7 debitage analysis factor 3 experimental sequence B...... 506
Table 81, General su-ny of debitage analpla conclusians on a case by case basis...... 519 Table 82. Raw HSRT tool aata r#atrix...... 535 Table 83. Rescaled HSRT tool data mtrix...... 537 Table 84. Housepit 7 flake tool analysis correlation mtrix ...... :...... 540 XVl Table 85. Housepit ? flake to~lanalysis rotated loadings ...... 0...... 5 42
Table 86. Housepit 7 flake tool analysis initial statistics...... *543 Table 87. Housepit 7 flake tool analysis facton scores.544 Table 88. Housepi? 7 tool analysis factor 1 experimental sequence A...... ,...... *.. 549 Table 89. Housepit 7 tool analysis factoz 1 experimental sequence B...... o...... e...... A2 Table 90. Housepit 7 tool analysis factor 2 experimental Sequence ...... ea...e. 556 Table 91. Housepit 7 tool analysis factor 3 experimental sequence A...... e.....a...... a.560 Table 92. Housegit 7 tool analysis factor 3 experimental sequence %...... 562
xvi i List of Figures Figure 1 . Middle Fraser Canyon study area ...... 26
Figure 2 . Complete flake ...... 68 Figure 3 . Proximal fragment ...... 69 Figure 4 . Medial/Distal fragment ...... 70 Figure 5 . Nonorientable fragment ...... 71 Figure 6 . Split flake ...... 72 Figure 7 . Hard hammer modified flake ...... 86 Figure 8 . Soft hammer modified flake ...... 87 Figure 9 . Pressure modified flake ...... 88 Figure 10 . Biface types ...... 89 Figure 11 . Unprepared core types ...... 90 Figure 12 . Prepared core types ...... 91 Figure 13 . Sullivan and Rozen Typology validity analysis factor score plot ...... 204 Figure 14 . Modified Sullivan and Rozen Typology validity analysis factor score'plot: factors 1 and 2.208 Figure 15 . Modified Sullivan and Rozen Typology validity analysis factor score plot: factors 3 and 4.209 Figure 16 . Flake Volume Index validity analysis factor score plot ...... 220 Figure 17 . Acute Angle Edge Length validity analysis factor score plot ...... 227 Figure 18 . Obtuse Angle Edge Length validity analysis factor score plot ...... 233 Figure 19 . Comparison of untrampled and trampled MSRT stage 2 biface reduction data ...... 309 Figure 20 . Comparison of rescaled FVI. AXEL and HAEL data from stage 2 biface reduction ...... 310 Figure 21 . Comparison of rescaled FVIxAAEL and FVIxHAEL data from stage 2 biface reduction ...... 311 Figure 22 . Comparison of FVI. AAEL and HAEL residual distributions for stage 2 biface reduction.,312 xvi i i Figure 23. Comparison of FVIxAAEL and FVIxHAEL residual distributions from stage 2 biface reduction.313
Figure 24. Comparison of trampled FVI, and HAEL residual distributions from stage 2 biface reduction ...... 314 Figure 25. Comparison of trampled FYIxAAEL and FVIxHAEL residua1 distributions from stage 2 biface seduction ...... 315 Figure ?6. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from stage 2 biface reduction ...... 316 Figure 27. Comparison of trampled rescaaled FVIxAAEL and FVIxHAFL distributions from stage 2 biface reduction...... 317 Figure 28. Comparison of untrampled and trampled MSRT stage 3 biface reduction data...... 318 Figure 29. Comparison of rescaled FVI, AAEL and HAEL data from stage 3 biface reduction...... 319 Figure 30. Comparison of rescaled FVIxAAEL and FVIxHAEL data from stage 3 biface xeduction...... 320
Figure 31. Comparison of FVI, AAEL and HAEL residual distributions for stage 3 biface reduction..321 Figure 32. Comparison of FVIxAAEL and FVIxHAEL residual distributions from stage 3 biface reduction.322 Figure 33. Comparison of trampled FVI, AAEL and HAEL residual distributions from stage 3 biface reduction...... 323 Figure 34. Comparison of trampled FVIxAAEL and FVIxHAEL residual distributions from stage 3 biface reduction ...... 324 Figure 35. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from stage 3 biface reduction...... 325 Figure 36. Comparison of trampled rescaled FVIxAAEL and FVPxHAEE distributions from stage 3 biface reduction ...... 326 Figure 37. Comparison of untrampled and trampled MSRT soft hammer flake reduction data...... 327 Figure 38. Comparison of rescaled FVI, AAEL and HAEL data from soft hammer flake reduction...... 328 xix Figure 39. Comparison of rescaled FVIxAAEL and FVIxHAEL data from soft hammer flake reduction...... 329
Figure 40. Comparison of FVI, AAEL and HAEL residual distributions from soft hammer flake reduction ...... 330 Figure 41. Comparison of FVIxAAEL and FVIxHAEL residual distributions from soft hammer flake reduction...... reduction...... ,,...... -....331
Figure 42. Comparison of trampled FVI, AAEL and HAEL residual distributions from soft hammer flake reduction...... 332 Figure 43. Comparison of trampled FVIxMEL and FVIxHaEL residual distributions from soft hammer flake reduction ...... 333 Figure 44. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from soft hammer flake reduction ...... 334 Figure 45. Comparison of trampled rescaaled FVIxAAEL and FVIxHAEL distributions from soft hammer flake reduction...... 335 Figure 46. Comparison of untrampled and trampled MSRT hard hammer flake reduction data...... 336 Figure 47. Comparison of resealed FVI, AAEL and HAEL data from hard hammex flake reduction...... 337
Figure 4b. Comparison of rescaled FVIxAAEL and FVIxHaEL data from hard hammer flake reduction...... 338 Figure 49. Comparison of FVI, AAEL and HAEL residual distributions for hard hammer flake reduction...... 339 Figure 50. Comparison of FVIxAAEL and FVIxHAEL residual distributions for hard hammer flake reduction...... 340 Figure 51. Comparison of trampled FVI, AAEL and HAEL residual distributions from hard hammer flake reduction ...... =*s=*w...... '..341 Figure 52. Comparison of trampled FVIxAAEL and FVIxWAEL residual distributions from hard hammer flake reduction ...... 342 Figure 53. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from hard hammer flake reduction ...... 343 Figure 54. Comparison of trampled rescaaled FVIxAAEL and FVIxHAEL distributions from hard hammer flake reduction ...... 344 Figure 55. Untrampled MSHT biface pressure flaking data ...... 345
Figure 56. Comparison of rescaled FVI, AAEL and HAEL data from biface pressure flaking ...... 346 Figure 57. Comparison of rescaled FVIxAAEL and FVIxHAEL data from stage 2 biface reduction ...... 347 Figure 58. Comparison of FVI, MEL and HAEL residual distributions for biface pressure flaking ...348 Figure 59. Comparison of FVIxAAEL and FVIxHAEL residual distributions from biface pressure flaking..349
Figure 60. Untrampled MSRT flake edge pressure flaking data ...... e,...... -...350 Figure 61. Comparison of rescaled FVI, AAEL and HAEL data from flake edge pressure flaking ...... 351 Figure 62. Comparison of rescaled FVIxAAEL and FVIxHmL data from flake edge pressure flaking ...... 352
Figure 63. Comparison of FVI, AAEL and HAEL residual distributions for flake edge pressure flaking ...... 353 Figure 64. Comparison of FVIxAAEL and FVIxHAEL residual distributions from flake edge pressure flaking ...... 354 Figure 65. Comparison of untrampled and trampled MSRT medium flake prepared block core reduction..355 Figure 66. Comparison of rescaled FVI, AAEL and HAEL data from medium flake prepared block core reduction ...... 356 Figure 67. Comparison of rescaled FVIxAAEL and FVIxHAEL data from medium flake prepared block core reduction ...... 357 Figure 68. Comparison of FVI, AAEL and HAEL residual distributions for medium flake prepared core reduction ...... 358 Figure 69. Comparison of FVIxAAEL and FVIxHAEL residual distributions from medium flake prepared core reduction ...... 359 Figure 70. Comparison of trampled FVI, AAEL and HAEL XXI residual distributions from medium flake prepared core reduction ...,.,,..,a.s*msa.e.-360
Figure 71. comparizon of trampled FVIXAAEL and FVIXAAEL residual distributions from medium flake prepared core reduction ...... ,.....361 Figure 72. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from medium flake prepared core reduction...... 362
Figure 73. Comparison of trampled rescaled FVIxkAEL and FVIxHAEL distributions from medium flake prepared core reduction ...... 363 Figure 74. Comparison of untrampled and trampled MSRT medium flake unprepared core reduction data.364
Figure 75. Comparison of rescaled FVI, AAEL and HAEL data from medium flake unprepared core reduction...... 365
Figure 76. Comparison of rescaled FVIxkAEL and FVIxHAEL data from medium flake unprepared core reduction...... 366
Figure 77. Comparison of FVT, AAEL and HAEL residual distributions for medium flake unprepared core reduction...... 367
Figure 78. Comparison of FVIxAAEL and FVIxHAEL residual distributions from medium flake unprepared core reduction ...... 368
Figure 79. Comparison of trampled FVI, AAEL and HAEL residual distributions from medium flake unprepared core reduction ...... 369
Figure 80. Comparison of trampled FVIxAAEL and FVIxHAEL residual distributions from medium flake unprepared core reduction...... 370
Figure 81. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from medium flake unprepared core reduction...... 371
Figure 82. Comparison of trampled rescaled FVIxAAEL and FVIxHAEL distributi~nsfrom medium flake unprepared core reduction...... 372 Figure 83. Comparison of untrampled and trampled MSRT large flake, prepared core reduction...... 373
Figure 84. Comparison of rescaled FVI, AAEL and HAEL data from large flake prepared core reduction...... 374 xxi i Figure 85. Comparison of rescaled FVIxAAEL and FVIxHAEL data from large flake prepared core reduction ...... -...,...... e.375 Figure 86. C~mparisonof FVI, AAEL and HAEL residual distributions for large flake prepared core reduction ...... ,...... 376
Figure 87. Comparison of FVIxmL and FVIxHAEL residual distributions from large flake prepared core reduction Figure 88. Comparison of trampled FVI, AAEL and HAEL residual distributions from large flake prepared core reduction ...... 3 78
Figure 89. Comparison of trampled FVIxAAEL and FVIxHAEL residual distributions from large flake prepared core reduction ...... 379 Figure 90. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from large flake prepared core reduction ...... l...... ,...... 380 Figure 91. Comparison of trampled rescaled FVIxAAEL and FVIxHAEL distributions from large flake prepared core reduction ...... 381 Figure 92. Comparison of untrampled and trampled MSRT large flake unprepared core reduction data..382 Figure 93. Comparison of rescaled FVI, AAEL and HAEL data from large flake unprepared core reduction ...... ,...... 383 Figure 94. Comparison of rescaled FVIxAAEL and FVIxHAEL data from large flake unprepared core reduction ...... -*...... 384 Figure 95. Comparison of FVI, AAEL and HAEL residual distributions for large flake unprepared core reduction ...... 385
Figure 96. Comparison of FVIxAAEL and FVIxHAEL residual distributions from large flake unprepared core reduction ...... 386
Figure 97. Comparison of trampled FVI, AAEL and HAEL residual distributions from large flake unprepared core reduction ...... 387
Figure 98. Comparison of trampled FVIxA&*L' and FVI xHAEL residual distributions from large flake unprepared core reduction ...... 388 xxi ii Figure 99. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from large flake unprepared core reduction ...... 389 Figure 100. Comparison of trampled rescaled FVIxAAEL and FVIxHAEL distributions from large flake unprepared core reduction ...... 390 Figure 1 I. Comparison of untrampled and trampled MSRT bipolar core reduction data ...... 391
Figure 102. Comparison of rescaled FVI, AAEL and HAEL data from bipolar core reduction ...... 392 Figure 103. Comparison of rescaled FVIxAAEL and FVIxHAEL data from bipolar core reduction ...... 393 Figure 104. Comparison of FVI, AAEL and HA!ZL residual distributions for bipolar core reduction ....394 Figure 105. Comparison of FVIxkAZL and FVIxHAEL residual distributions from bipolar core reduction ...395 Figure 106. Comparison of trampled FVI, AAEL and HAEL residual distributions from bipolar core reduction ...... 396 Figure 107. Comparison of trampled FVIxAAEL and FVIxHAEL residual distributions from bipolar core reduction...... a...... 397 Figure 108. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from bipolar core reduction ...... 398 Figure 109. Comparison of trampled rescaled FVIxAAEL and FVIxHAEL distributions from bipolar core reduction ...... 399 Figure 110. Comparison of untrampled and trampled MSRT bipolar core reduction data (bifacial core1.400 Figure 111. Comparison of rescaled FVI, AAEL and HAEL data from bipolar core reduction (bifacial core) ...... 401 Figure 112. Comparison of rescaled FVIxAAEL and FVIxHAEL data from bipolar core reduction (bifacial core) ...... 4 02 Figure 113. Comparison of FVI, AAEL and HAEL residual distributions for bipolar core reduction (bifacial core) ...... 403 Figure 114. Comparison of FVIxAAI2L and FVIxHAEL residual distributions from bipolar core reduction xxi li (bifacial core) ...... 404
Figure 115. Comparison of trampled FVI, AAEL and HAEL residual distributions from bipolar core reduction (bifacial core) ...... 405
Figure 116. Comparison of trampled FVIxAAEL and FVIxHAEL residual distributions from bipolar core reduction (bifacial core)...... 406
Figure 117. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from bipolar core reduction (bifacial core)...... 407 Figure 118. Comparison of trampled rescaled FVIxAAEL and FVIxHAEL distributions from bipolar core reduction (bifacial core) ...-...... 408 Figure 119. Keatley Creek Archaeological Site location in the Middle Fraser Canyon of south-central British Columbia...... 443 Figure 120. Housepit 7 floor map, Keatley Creek Archaeological Site ...... 445 Figure 121. Housepit 7 floor artifact density distribution (adapted from SpaEford 1991) ...... 451 Figure 122. Division of analytical units on the floor of housepit 7...... 452 Figure 123. Relationship between converted MSRT debitage distribution and FVI cull model for early stage biface production (from Table 60) ..... 457 Figure 124. Relationship between converted MSRT debitage distribution and AAEL cull model for early stage biface production (from Table 60) .....458 Figure 125. Relationship between converted MSRT debitage distribution and HAEL cull model for early- stage biface production (from Table 60)..... 459 Figure 126. Housepit 7 floor debitage analysis factor score plot: factors 1 and 2...... 484 Figure 127, Housepit 7 floor debitage analysis factor score plot: factors 3 and 4...... 485 Figure 128. Housepit 7 floor debitage analysis factoi score plot: factors 5 and 6...... 486 Figure 129. Comparison of case 6 with the stage 2 biface trampled AAEL residual distribution...... 490 Figure 130. Comparison of case 20 with the medium flake, XXV prepared core reduction, trampled, FVIxAAEL residual distribution ..,...... ,.....492
Figure 131. Comparison of case 52 with the trampled, hard hammer flake retouch data set ...... 495 Figure 132. Comparison of case 21 with the mixed prepared and bipolar core reduction FVIxAAEL residual distribution ...... 499 Figure 133. Comparison of cave 61 with mixed prepared core reduction (FVIxAAEL residual) and bipolar core reduction (HAEL residual) distribution ...... 504 Figure 134. Comparison of case 97 with the mixed trampled biface and bipolar core reduction AAEL residual distribution...... 508 Figure 135. Comparison of case 14 with the medium flake, prepared core reduction, trampled FVIxAAEL residual distribution ...... 511 Figure 136. Comparison of case 86 and the medium flake, prepared core reducti.on data set ...... 515 Figure 137. Distribution of analytical units interpreted to be associated with tool maintenance/resharper,ing ...... 523 Figure 138. Distribution of analytical units interpreted to be associated with'biface reduction...... 524 Figure 139. Distribution of analytical units interpreted to be associated with prepared core reduction...... 525 Figure 140. Distribution of analytical units interpreted to be associated with bipolar core reduction ...... 526 Figure 141. Distribution of analytical units not interpreted to be associated with trampling ...... 527 Figure 142. Distribution of analytical units interpreted to be associated with acute edge angle flake culling ...... 528 Figure 143. Distribution of analytical units interpreted to be associated with high edge angle flake culling ...... 530 Figure 144. Distribution of analytical units interpreted not to be associated with any form of flake culling ...... 531 Figure 145. Comparison of flake tool analysis case 11 with xxvi factor 1 experimental trampled flake tool discard distribution (Part A) ...... 549 Figure 146. Comparison of case 3 and the factor 1 experimental trampled curation distribution (Part B) ...... 554 Figure 149. Comparison of case 13 with the factor 2 modelling sequence trampled final discard distribution ...... 558 Figure 148. Comparison of case 7 with the mixed biface and prepared core redustion flake, trampled cull distribution ...... 563 Figure 149. Comparison of case 8 with the factor 3 experimental trampled curation distribution.564
Figure 150. Distribution on the housepit 7 floor of areas with flake tool assemblages interpreted to be the result of prepared core reduction flakes only ...... 567 Figure 151, Distribution on the housepit 7 floor of areas with flake tool assemblages interpreted to be the result of prepared and bipolar core reduction flakes ...... 568 Figure 152. Distribution on the housepit 7 floor of areas with f'lake tool assemblages interpreted to be the result of prepared core and biface reduction flakes...... 570 Figure 153. Distribution on the housepit 7 floor of areas with flake tool assemblages interpreted to be the result of some degree of larger flake tool curation...... 571 Figure 154. Example of flake tool assemblage formation based on the reduction and breakage of a single tool ...... 587 Figure 155. Debitage and tool assemblage distribution, analytical unit 96 (sector 13)...... 604 Figure 156. Debitage and tool assemblage distribution, analytical unit 97 (sector 13) ...... 604
xxvi i CHAPTER 1
The intent of this thesis is to present and test the
effectiveness of a new method for recognizing the culling of flakes for tool use. The method requires the use of a modified version of Sullivan and Rozen's (1985; Sullivan 1987) debitage typology and three utility indices to predict or
anticipate the character of the archaeological record under different formational conditions. The primary archaeological formation processes in which I am interested derive from the economic behavior of hunter-gatherers, especially as reflected in their decisions on how to produce, use and reuse tools. I argue that the concept of risk management best explains the variety of economically oriented strategies and decisions made by individuals and groups attempting to avoid the possibility of reduced access to critical subslvtence resources. Llthic technology is an important part of this process as it provides tools for directly procuring and processing resources as well as tools for making the often more complex organic tools required to insure survival. Archaeologists interested in explaining the role of lithic technology in hunter-gatherer economic systems must be able to recognize evidence for the use of different technclogical strategies in the archaeological record. These strategies are manifested through the procurement of lithic raw materials, and the I production, use and reuse/recycling of stone tools. This dissertation is concerned with two poorly understood
components of this process: 1) the recognition and explanation
of lithic flake removal or culling and 2) the recognition and explanation of flake tool production, reuse and recycling.
STUDYING RISK MANAGEMENT WITH LITHICS
To understand risk management strategies in prehistoric contexts, the archaeologist is faced with two fundamental problems: (1) How is raw material acquired, used and conserved
(if at all); and (2) how is the time and energy invested in stone tool production and use translated into more effective means for ensuring access to critical resources? In other words, how do lithic raw materials move through an archaeological context and in what forms are tools produced and used in the acquisition and preparation of important nonlithic resources. Ultimately, these kinds of information are required for us to understand the role sf lithic technological organization within the larger system of risk management in any prehistoric context. Solving these problems requires a thorough understanding of the flow of lithic raw materials through archaeological contexts. This includes lithic procurement, core reduction/tool blank production and tool production, use, discard, recycling and curat ion, scavenging, and flake/core/tool removal. Archaeologists have become quite adept in recent years at recognizing many of these factors in the archaeological record. Research on lithic procurement strategies is well developed (Ericson and Purdy 1984; Francis
1983; Luedtke 1976; Vehik 19851. A variety of techniques are now available for understanding lithic reduction strategies including analyses of technological aspects of tool forms
(Callahan 1979; Cotterell and Kamrninga 1987; Frison and Bradley 1981; Dibble 1987; Isaac 1977; Johnson 1981; Knudson 1983; Muto 5971) and technological analyses of debitage (Ahler
1989; Fish 1981; Hayden and Hutchings 1989; Ingbar et a1 1989;
Johnson 1981; Magne 1985; Magne and Pokotylo 1981; Mauldin and
Amick 1989; Odell 1989; Prentiss and Homanski 1989; Sullivan 1987; Sullivan and Rozen 1985). Tool use has been approached from a number of perspectives. The most prominent Is prsl2ably still the analysis of use-wear to determine tool function
(Hayden 1979; Keeley 1980; Knudson 1983; Odell 1981; Tringham et al. 1974; Vaughaun 1985; Yerkes 19871, though the analysis of residues is gaining some prominance (Loy 1983). Tool use and recycling is being approached from a tool use-history perspective aEten combining technological, use-wear and refitting analyses (Audouze 1987; Bamforth 1986; Dibble 1987,
1991; Hayden 1989; Keeley 1982; Knudson 1983; Shott 1989;
Wilrnsen and Roberts 1984; Yerkes 1987). Conttributins to the study of curation have been largely theoretical in nature
(Bamforth 1986; Binford 1973, 1977; Shott 19891, Lithis toel scavenging remains a somewhat neglected area of archaeological research despite the early pioneering efforts of Schiffer
[1976:167,170-171). Fortunately, some studies of lithic tool scavenging are appearing, however, in studies of large scale surface distributions (c.f. Bamforth 1990; Camilli 1988). The
removal of lithic items from camps, workshops, quarries, etc. is a problem area which has often been noted to be important and worthy of continuing investigation (c.f. Ferring 1976;
Frison and Bradley 1981; Ford 1987; Henry 1989; Kuhn 1991;
Teltser 1991). Refitting studies have dramatically improved conclusions on the removal of lithic items from sites (c.f.
Bamforth 1990; Leach 1984; Singer 1984). It should be clear now that a wide variety of analytical approaches are available to the archaeologist interested in studying the movement and trans•’ormation of lithic raw materials through archaeological contexts. Two problem areas remain, however. First, it has not been possible to study an unmodified debitage assemblage to identify which flakes were chosen or culled for use as tools, without recourse to comparlsons with associated tool assemblages. As the recognition of flake use/flake tool production is important to the understanding of overall lithic economies, this has created a situation where archaeologists have attempted to draw such conclusions from comparisons between debitage and flake tool assemblages, knowing that many tool types have left the site area and many others are transformed to a state entirely different from their original flake form. The potential for error in recognizing flake culling strategies grows higher as the degree of flake and tool curation, recycling and scavenging increases. Thus, the linkage between debitage assemblage production and flake tool production has not been adequately considered. Methods for recognizing not only general flake culling, but different culling strategies, without recourse to comparisons to flake tool asemblages would be very useful. It is well known, ethnographically, that depending on organizational context, flakes of a variety of sizes and shapes are chosen for use as tools and that this has important implications for our understanding of lithic technological organization (c.f. Allchin 1957; Binford 1986; Fortuine 1985; Gallagher 1977; Gould 1968, 1971, 1977; Gould et al. 1971; Hayden 1977, 1979; Miller 1979; Strathern 1969; Thompson 1954; Webley 1990; White and Thomas 1972). The second major problem area is that of the flake tool uue/diucard/reuue cycle. While many analyses have coped well with the problems of lithic tool transformation, these studies have tended to center on the more curated components of lithic tool assemblages (Dibble 1987; Hayden 1989; Keeley 1982; Shott 1989). Though use-wear studies often show multiple useages of the same flake tool (c.f. Yerkes 1987), the systems of flake tool use/reuse remain poorly understood. On some living surfaces, depending on occupational intensity and sedimentation rates, discarded flake tools remain as critical resources for continued reuse. Thus, depending on morphology and discard context some classes of flake tools could be expected to be reused on multiple occasions, This could vary with other classes of flake tools expected to be more heavily curated. Study of the archaeolcgical result of this process would be knowledge of the occupational history of a given land surface (i.e. Camilli 1983) and a better understanding of the
total use of lithic raw materials from an economic perspective.
THESIS OUTLINE
In this thesis, I am concerned with the development of techniques for allowing the recognition and assessment of prehistoric lithic reduction, flake culling and flake tool reuse/recycling in archaeological contexts. I acomplish this through three distinct phases of research. First, I demonstrate that lithic technological organization is a critical component of larger systems of risk management in hunter-gatherer societies. I discuss, first, the concept of risk and variation in risk management strategies in hunter-gatherer societies. Next, I discuvv the ethnographic people of the Middle Fraser Canyon of south-central British Columbia, to show how the globally described risk mangement tactics can apply to a single case study. This ethnographic information is used later to compare the results of the archaeological studies from this area in order to note any similarities or differences between the Late Prehistoric and Contact period use of the area, as reflected primarily through lithic reduction, flake culling and flake tool use/reuse strategies. Second, experimental lithic research is presented,
6 designed to develop and test methods for recognizing prehistoric flake culling and flake tool reuse/recycling. I accomplish this, First, with an analysis of the reliability
and validity of Sullivan and Rozenls (1985; Sullivan 1987) debitage typology. I next develop and test the reliability
and validity of three flake utilicy indices. Reliability and validity analyses are important to the development of any technique, such as these, designed to measure variability in a quantitative manner, Reliability assessment tells the researcher if the technique or instrument provides consistent results if reapplied to the same phenomena. Validity assessment indicates whether or not the instrument functions or provides xesults which match theoretical expectations (Carmines and Zeller 1979; Nance 1987). I conclude that when used in conjunction with a modified form of the Sullivan and
Rozen typology (Modified Sullivan and Rozen Typology [MSRT'), mean utility index scores can be reliably and validly used to predict the approximate utility of given flake classes (defined by size and breakage in the MSRT). Rescaled utility index profiles are then used to predict possible culled flake assemblages, depending on technology employed. Further transformation of these data sets allows the prediction of residual flake assemblages, or those which have had flakes removed from them. As trampling is expected to have severe effects on these data sets, cull and residual distributions are als~developed for trampled contexts. This research provides the basic pattern recognition criteria for understanding archaeological assemblages. Third, the applicability of the methods is tested with ?n
analysis of a Late Prehistoric housepit floor assemblage from the Keatley Creek site in the Middle Fraser Canyon of south-central British Columbia. Spatial groupings among artifacts are defined following Spafford (1992). Principal components analysis is used to facilitate pattern recognition from the complex data matrices of debitage and flake tools quantified in MSRT categories. The results of this study indicate that archaeological patterning is complex but understandable in reference to the methods developed in the experimental phase of this study. A series of technological strategies is recognized with prepared core reduction occurring ubiquitously across the floor and tool production activities relatively tightly associated with hearth features. The effects of flake culling are found throughout the floor, focussing on the removal of larger acute edge angled flakes throughout the floor and to a lesser degree high edge angled flakes primarilly in the southwest and northeast corners. Trampling is identified throughout except in portions of the floor around the edge, in some hearths and in areas of dense posthole patterning. Flake tools exhibit two primary sequences of reuse/recycling. On the east and west edges of the floor, flake tools exhibit patterns of more intensive curation (long term intensive use of tools before discard) while those in the middle portions of the floor are used more expediently (short term use of tools followed by discard), but are recycled more intensively (resulting from repeated short term use episodes). In other words, these latter tools are consistently culled from discard contexts. Flake tools appear to have been discarded in roughly the same areas from which the original flake blanks were produced and culled. I suggest that the methods developed are useful for further experimental and archaeological study. Two lines of research beyond this thesis are recommended. Further experimental research is required for better understanding the use of lithic utility indices. This could include actualistic culling experiments, such as blind tests for recognizing different archaeological patterns, as well as flake tool use/reuse experiments. Future archaeological research could begin to focus on new methodological approaches such as combining more traditional "distinctive artifact" (Sullivan and Rozen 1985) approaches with utility index distributions for more detailed perspectives on archaeological assemblages. As well, refitting studies could be combined with utility index focussed studies to form more complete perspectives on archaeological lithic assemblage formation. CHAPTER 2
RISK MANAGEMENT, LITHICS AND WINTER HOUSEPIT OCCUPATION IN THE CENTRAL FRASER CANYON
In this chapter, I will discuss the role of lithic technology in the economic system of the aboriginal inhabitants of the Middle Fraser Canyon of south-central British Columbia during the ethnographic period. I will demonstrate that lithic r3w material acquisition, tool production, use and discard were only part of a larger economic system organized to effectively manage risk. This is accomplished by first, examining the overall economic system of the people of the Middle Fraser Canyon in reference to risk management; and second, integrating lithic technological organization into this system and examining it from the perspective of risk management. ~inal.ly,I deal specifically with the role sf flake culling and flake tool use/reuse in this system. The chapter begins with a theoretical consideration of risk management, focussing first, on a justification for the choice of risk management over other optimization principles for explaining the adaptations of hunter-gatherers. Second, I outline the role of risk management in structuriric~foraging, group mobility and aggregations, food and equipment storage, technology and intra- and inter-group interactions. I do not attempt to be globally comprehensive in my discussion of variation in risk management strategies. Rather, ny g3al is
10 to demonstrate the importance of risk in structuring economic decisions made by hunters and gatherers.
Next, I introduce the environment and salient economic strategies of the historic inhabitants of the Middle Fraser Canyon, as known from the ethnographic record. I review the role of risk management in mobility and vinter aggregation, storage, technology and intra- and inter-group relations. Finally, I provide a more comprehensive discussion on the role of risk *.aa~agementin late fall and winter lithic technological organization. More specifically, I review the role of flake tool production and use from stored raw material sources. Conclusions are drawn on the role of lithic technological organization in the overall organization of Middle Fraser Canyon economies.
Binford (1981) has argued that the quest for understanding the general principles of cultural organization is the domain of general theory. This chapter is oriented towards providing a general theoretical understanding of the adaptations of the protohistoric inhabitants of the Middle Fraser Canyon. It also illustrates the theoretical framework for which I will view prehistoric behavior.
RISK MANAGEMENT
This section is designed to introduce the concept of risk. To accomplish this I first define the term and argue for the relevance of its use as a logical extension of sinpler energy maximization p~inciples. Having theoretically demonstrated the importance of the risk management concept, I then review four primary ways in which hunter-gatherers manage risk. Hunter-gatherers, making economic decisions, must cope with two critical problems: "uncertainty due to imperfect information and risk due to the consequences of unavoidable variation" (Smith 1983:638). While the concept of uncertainty refexs directly to problems to be overcome in the course of foraging, risk deals with the effects of foraging success on human groups. More formally, risk can be defined as "the magnitude sf the stochastic variance associated with particular choices*' (Winterhalder 1986:384). In more general terms, this refers to the probability of a loss or "unfortunate outcome" (Keene 1981:179). In subsistence terms, this can mean a failure to meet dietary needs (Terrence
Clearly, risk is something to be avoided or at least to be minimized, in any type of economy. This can be accomplished by foragers in two possible ways: either through risk averse or prone behavior (Carsco 1981). Risk averse foragers are concerned with reducing dietary variation while risk prone foragers are more concerned with increasing variation (Smith 1953:639!. These options appear to be operationslized by foragers in contexts defined by what has been called the "extreme variance rule" (Stephens and Charnov
1982) whereby foragers make decisions on the degree of foraging variance depending on whether or not they anticipate returns above or below a critical threshold (e.g.
Winterhalder's 119861 starvation threshold). Foraging variance is maximized where returns are predicted to drop expected to be above the threshold. In other words, where abundance is predicted, less attempt is made to produce high variation. Where shortage is predicted, effort is made to add variation. This indicates that as greater shortage is predicted, foragers are willing to take greater risks to acquire enough food sources to meet minimum requirements (Smith 1983:639). Notably, it has been suggested that the optima derived from risk sensitive foraging strategies (as above) are often little different from that of simpler energy efficiency-maximization models (Smith 1983; Stephens and
Charnov i.982). This leads to the questimi: why bother with complex foraging models concerned with risk in the explanation of hunter-gatherer decision making, when simpler models will suffice?
Winterhalder 11986:390) has argued for a distinction between the behavior of individual hunters or gatherers involved in making prey choices and the overall behavior of groups seeking to minimize risks of dietary security. He may increase their efficiency of resource acquisition as well as their degree of security by foraging in a more risk-prone
13 manner and offsetting that risk by sharing. (Winterhalder
1986:390). The primary implication of Winterhalder's (1986) study is that simple energy efficiency models do work to explain the resource procurement behavior of individual foragers. Explaining the more complex behavior of consumer groups requires more attention, however, with special regard to risk management strategies.
Wiessner (1982) has outlined four fundamental stratecies for reducing risk: prevention of loss; transfer of risk or loss; storage; and pooling of risk. It has been argued that
technology (Terrence 1988) and mobility (Kelly 1983) are also useful means for reducing risk. I review these tactics under Wiessnerfs prevention of loss category. I now review the role of each of these risk management strategies in hunter-gatherer economic and socio-political organization.
PREVENTION OF LOSS
Wi2ssner (1982:172) defines prevention of loss as "reduction of hazard or minimization of actual loss." She suggests that prevention of loss can be accomplished through rituals (e.g. religious rites designed to insure success in hunting, gathering or fishing), resource control through burning, allocation of land rights and territorial defense. The primary function here is to guarantee as much access as possible critical resources. Outside of ritual behavior and burning, prevention of loss is largely the result of
14 inter-group interactions of land access. Cashdan (1983:48-49) has asserted that access to land can be managed by either actual perimeter defense or what she and ~tkers(Peterson
1975) have termed social boundary defense. Perimeter defense occurs in contexts of more intensive competition over resources and is indicated by boundary marking and control of Lerritorial space. This form of defense typically occurs where resources are more dense and predictable, as on the
Northwest Coast (e.q. Ames 1985; Arima and Dewhirst 1990:392; Hayden et al. 1985). Social boundary defense occurs where resources are sparse and/or unpredictable and involves the control of access to the social group living in an area
(Cashdan 1983:49; Peterson 1975). Thus, in environments where survival depends largely upon sharing within a group, an effective way of controlling access to limited resources is by limiting those who are permitted access to the social group. Ritual control of resources and burning to provide predictable access to resources, as means for preventing loss, are examples of high degrees of planning depth (Binford 1987). Planning depth refers to the degree of advance planning required to avoid risk in providing adequate short and long term access to resources. This continual process can be expected to heavily affect ritual behavior and landscape modification (such as burning). It can also be expected to have major effects on technologies and mobility strategies
Kelly (1983) has argued that hunter-gathexer mobility strategies are structured largely to provide information to consumer groups regarding availability and access conditions to resources. Groups can then adzpt their food getting
options to these parameters. Kelly (1983), following Binford
(1980), defines two separate types of mobility, residential and logistical. Residential mobility refers to the movements of entire consumer groups, while logistical mobility refers to specialized task group movement on a landscape. Logistical mobility is the primary means for gathering information, as well as for procuring food in many contexts. Less residentially mobile groups with correspondingly higher rates of logistical mobility position consumer groups in critical places to provide access to a variety of resources, depending on information collected by logistical task groups. Groups employing these mobility strategies, termed collectors
(Binford 1980), are often found in areas of clumped resources which may be accessible for only short time periods during the year. More highly mobile residential groups can continually take in information by short term logistical moves and increased rates of residential mobility (see also Binford 1982). These groups tend to be found in localities of seasonally ubiquitous but spatially unpredictable resources. I In either case, however, risk is minimized by reducing the uncertainty of resource availability and access by continual information gathering through mobility strategies. Optimal foraging research has supported some of the conclusions drawn by Binford and Kelly on mobility strategies. 16 Using Horn's (1968) foraging model, Heffley (1981) demonstrated that three northern Athabaskan groups modified
their settlement patterns (mobility strategies) to solve resource availability and accessibility problems. Mobile, aggregated and unpredictable resources were best procured by aggregations of foragers, while more evenly distributed unpredictable resources required less group aggregation and higher degrees of mobility. Aggregated, predictable resources such as some types of fish produced high degrees of storage and thus also resulted in high rates of human aggregation. Heffley (1981:146) notes that much of the variation in mobility and human aggragation resulted from human information sharing, a conclusion paralleling those drawn by Binford (1978, 1980, 1982) and Kelly (3-9831 regarding the significance of mobility in information gathering and sharing. Recent research by Binford (1991) and Whitelaw (1991) has demonstrated, however, that hunter-gatherer decision making may be more variable than has been indicated by Heffley or Kelly. Future researchers will need to consider the roles of kinship, labor organization and uncertainty decisions as partial conditioners of residential and logistical mobility strategies. Torrence (1983) has added technology as a means of preventing loss. In essence, she says that technologiea are used to maximize the chances of success in resource procurement and processing where success is defined as the avoidance of potential risk (Torxence 1989:59). She suggests that the nature of a technology should be responsive to first, the timing of risk and second, the severity of risk involved. Technological design parameters will center on quality, depending on prey mobility and annual availability and quantity of vtechnological in~estrnent,~'dependent on "the severity of the consequences of failing to procure the resource in question" (Torrence 1989:60). The timing of risk can be further divided into spatial and temporal considerations. Each can have different effects on the character of the technological response. For example, where resources are immobile and spatially dispersed (e.g. plants), foragers tend to use tools of limited complexity, termed instruments (Torrence 1983, 1989 1. As critical resources become more mobile and less easily predicted, foragers adopt more weapons and facilities (such as traps) of more complex design (Torrence 198'9: 60). As seasonality has a greater effect on resource access, foragers shift from the use of weapons and instruments in situations of reduced seasonality to tended (fish traps, caribou or bison pounds) and untended facilities (small game traps, deer fences) in contexts of high seasonality (Torrence 1989:61). coupled with risks associated with resource access is the severity of risk. Torrence argues that the severity of risk affects the complexity and the specialization of specific technologies. Thus, where risk is severe, as in most high latitude contexts, technologies will be complex and tailored to specific resources. Where short-term risk is limited, the technological focus will be reduced and other social mechanisms may be more prominant (Torrence 1989:62),
The effects of risk character and severity on technologies are well illustrated by considering the ways in which different technologies can be organized. ~leed(1986) has offered the concepts of reliable and maintainable technologies, suggesting that in conditions of temporal stress on resource access, reliable technologies will be used, while in situations of spatial unpredictability, more maintainable technologies will be used. His definitions for the two center on two components, the role of manufacture and maintenance and the role of effectiveness in application. Reliable technologies are produced by specialists during times between periods of use. They are designed so as not to be used to
full capacity and are not repaired during use. Maintainable technologies may be repaired during use, contain modular replacement parts and are repaired by non-specialists. Torrence (1989:63) correctly notes that Bleed's examples of reliable technologies contain maintainable parts as well as modular replacement parts, considered by Bleed to be characteristic of maintainable technologies. She suggests that Bleed's categories should be viewed more as variables conditioning the make-up of technologies as opposed to types
of technologies. AS reliability refers prlmarlly to the ability of a tool to get a job done, it is linked to risk severity. Maintainability is linked to risk character as it refers primarily to the degree to which that technology can be modified to procure different resources during a single period of use. Thus, hunter-gatherers may incorporate aspects of both reliability and maintainability into the production of tools used to solve problems and reduce risk. Tool manufacture and maintenance scheduling and raw material acquisition can also be expected to be organized to solve problems associated with risk. The importance of scheduling, in reference to solving technological problems has been considered by numerous researchers (Barnforth 1986; Binford 1977, 1979, 1982; Bleed 1986; Hayden 1989; Shott 1989; Toxrence 1983, 1989). In reference to risk, the salient point appears to be that scheduling of manufacture and maintenance depends largely on the timing of required reliability in technological options. Thus, although there is an important component of maintainability in scheduling decisions, reliability will always be a major factor. Actual scheduling decisions may also be made more complex by problems of raw material access (Bamforth 1986; Hayden 1989 1 and processing requirements (Hayden 1989). It has been amply demonstrated that access to raw material for tool production may require a high degree of planning to avoid rlsk. scheduling of tool production and maintenance cannot operate independently of raw material access. Thus hunter-gatherers employ different strategies of xaw material procurement, ranging from those embedded into other pursuits (Binford 19791, to direct raw material procurement trips designed to provide stockpiles of raw material for use at leisure (e.g. Parry and Kelly 1987). Use of raw materials can range from expedient to highly curated (Binford 1979) depending on short and long term anticipated needs. Lithic tool recycling can also be expected to occur in contexts where access to raw materials is limited either through occupational intensity and scheduling constraints (Binford 1979; Hayden 1988; Parry and Kelly 1987), territorial unfamiliarity (Goodyear 19791, seasonality (Binford 1979; 1983) or excess tool breakage due to brittle raw materials (Hayden 1992, personal communication). Flake culling may be extremely focussed or limited to a few specific flakes in contexts of anticipated high raw material accessibility, or it can be broader including culls of primary reduction flakes and a range of other secondary by-products of the reduction process where specialized needs are present or raw material shortages are ant iclpated.
RISK TRANSFER
Wiessnerts (1982) second strategy for reducing risk involves transferring risk from one group to another. Thls can occur either through status competition and exchange or through taking of resources from others by force. Potlatching on the Northwest Coast of North America may be an examplz of status competition and exchange of goods as a means of transferring risk. Piddocke (1965) has argued that potlatching, in times of short term food deficiency, served to alleviate the risk of starvation by transferring wealth back and forth between groups. The transfer of wealth through
potlatching allowed groups facing food shortage to purchase
needed supplies and where anticipated shortages were only
short term, it maintained stability between the potlatching groups. This strategy does not appear to have been a useful
one for long term risk management however, where, among some groups, resource deficiencies were long term. When this
occurred, these potlatch groups could not achieve renewed
status through wealth accumulation and thus declined in
prestige and eventually would be forced out of a potlatching system (Piddocke 1965:261). Cannon (1992) has demonstrated the relationship between
periods of resource shortage and the use of conflict to gain
access to salmon resources on the Canadian Plateau. In
Wiessner's (1982) terms, this has 'the effect of transferring risk of shortage of salmon from one group to the next by one
group actually taking the salmon of the other group. This
strategy of reducing short term risk through active conflict appears to have been common among hunter-gatherers and
horticulturalists throughout much of North America (c.f.
Barnforth 1988; Mi1ne.r et al. 1991).
STORAGE
Wiessner's (1982) third means of risk reduction is through storage of resources to cover anticipated losses of
22 resources. Storage among hunter-gatherers can be accomplished in three general fashions. First, portable stores can be
created by drying foods such as meat for transport by mobile
groups. This strategy was commonly used on the North American Plains (Lowie 1954; Soffev l989). Second, caches of foods can be created as insurance (Binford 1978, 1983) against potential shortages. This strategy was common among northern hunter-gatherers who often rely on frozen meat caches in times of shortage (Binford 1978). Third, caches of foods are created and used as primary subsistence material (Binford 1983; Soffer 1989). This is again typical of many northern peoples relying on intensive fall game and plant harvests for winter survival (Arnes 1985; Binford 1983; Frison 1978; Hayden et al. 1985; Smith 1978). The role of storage is intimately tied to other strategies for managing risk including foraging strategies (Smith 19831, mohility~strategies (Blnford 1380; Kelly 19831, technology (Hayden 1981; Torrence 1989) and intergroup interactions (Soffer 1989).
RISK POOLING
Wiessner's (1982) fourth and final means of reducing risk is that of risk pooling or sharing (see also Winterhalder 1987). Sharing can take a multitude of forms, ranging from generalized reciprocity (Sahlins 19721, to more complex trade partnerships (Binford 1983:214-220) and central authority redistribution (Earle 1989; Feebles and Kus 1977). The result is a pooling of risk whereby no individual or group is entirely respansible for mitigating adverse effects of high
risk situations. Thus, as Winterhalder (1986) points out, foragers can partake in risk-prcne behaviors to obtain specialized resources where sharing can be counted on to minimize the effects of risk. By extension, individuals and groups within food producing (horticultural) or intensive harvesting societies (Northwest Coast/Plateau) can specialize on certain resources, knowing that exchange and reciprocity will minimize the adverse effects a•’ their special9zatlon, by adding additional resources to their economies. These strategies for reducing risk rarely occur in isolation. Gxoups of people also respond differently depending on whether risk is short or long term. Wiessner
(1982:173) notes that risk reducing strategies can be tied together in general packages depending upon resource type and access. She notes that individual risk pooling is typically associated with Binford's (1980j foraging strategy, while storage, risk transfer and centralized pooling tend to occur more of ten societies using the collector's approach. also clear Lhat risk management is not a normative concept. Groups continually adapt to new situations given fluctuations in the natural and social environments (Winterhalder 19801.. Thus a range of options must always be left open to guarantee survival. I now consider the role of risk management in the economic and socio-political organization of the aboriginal inhabitants of the Middle Fraser Canyon, on the western edge of the Canadian Plateau.
RISK MANAGEMENT IN THE MIDDLE FRASER CANYQN
In this section I consider the role of risk management in Middle Fraser Canyon economies and socio-political organization. To accomplish this, I first provide a brief overview of relevant environmental information from this area. I focus on the availability and accessibility of several critical plant and animal resources including deer, salmon, trout and some species of berries and roots. I then review the importance of each of Wiessner's (1982) risk avoidance strategies in central Fra3er Canyon economy and socio-political organization. More specifically I discuss the role of territoriality, foraging, mobility, technology, potlatching, warfare, storage and -trade as rlsk avoldance or minimization tactics. I conclude with a discussion of the role of lithic technology in this overall system. Hayden (1992) has defined a Middle Fraser Canyon study area as extending from Texas Creek, at the southern end, north to Kelly creek (Figure 1). The western boundary 1s the coastal mountains rising steeply from the Fraser Canyon, while the eastern boundary lies across the peaks and eastern slopes of the Clear and Marble Mountain Ranges. This area lies on the western border of the british Columbia Plateau (Holland
1964) and in the northwest portion of what Ray (1939) has referred to as the Plateau culture area. Even though
25 Figure 1. Middle Praser Canyon study area. 26 geographically and geologically distinct from much of the B.C. and southern Plateau areas, the cultures of the Middle Fraser Canyon area unquestionably belong to that of the Plateau (Hayden i992; Ray 1939:2). The Middle Fraser Canyon study area ranges in elevation from 198 meters (650 feet) in the southern portion of the river valley to 2332 meters (7650 feet) at the top of the highest alpine area (Hayden 1992). The canyon is the result of continuous fluvial erosion over millions of years. Benchlands along the river consist of unconsolidated Quaternary aged sediments resulting from Late Pleistocene glacial activity, modified through Holocene mass wasteage and Eluvial processes (Ryder 1978:56). The study area has been divided into seven environmental zones: Alpine Tundra, Montane Parkland, Montane Forest, Intermediate Grasslands, Intermediate Lakes, River Terraces and River Valleys; based on temperature and precipitation patterns (Alexander 1992a; Mitchell and Green 1981). I now summarize each of these zones, relying primarily on Alexander f lCJgZa). The Alpine zone occurs generally at elevations above 1980 meters (6500 feet); a zone characterized by long severe winters and short growing seasons. Vegetation in this zone is dominated by low shrubs, sedges and grasses, though a few subalpine tree species can be found in low numbers. Soil formation is generally poor. Mammal species used aboriginally Include grizzley and black bear, wolf, coyote, wolverine, weasel, yellow-bellied marmot, deer, elk, bighorn sheep and possibly mountain goats. These are found only in the summer
due to adverse weather conditions in all other times. The Montane Parkland zone lies at the transition between the Alpine and Montane Forest zones at elevations ranging from
1525 to 2135 meters (5000 to 7000 feet). Climate of this zone is only slightly less severe as that of the Alpine zone. The effects of high winds are reduced due to the presence of trees. In addition to shrubs, sedges and grasses, a variety of tree species are found including subalpine fir, lodgepole pine, Engelmann spruce and whitebark pine. Mammals found in this zone important to aboriginal peoples include all those listed under the Alpine zone as well as hare, porcupine, red squirrel, flying squirrel and a variety of small predators. This zone was especially important as summer range for deer.
A variety of plant species used as aboriginal food sources ~ipe found in this zone (Alexander l992a).
Montane Forests range in elevation from about 1980 tc 610 meters (6500 to 2000 feet). Thus much of the Middle Fraser Canyon study area defined by Hayden (1992a) is covered by this zone. Climate is somewhat warmer and drier than in the Montane Parkland. Two major vegetative associations are found here including the Engellmann Spruce-Subalpine Fir and the
Interior Douglas Fir (Mitchell and Green 1351!, As expected, this zone contains a wide variety of trees and shrubs, many of which were used by aboriginal people (Turner 1979, 1992).
Montane Forests contain the same range of mammals as that of the higher elevation zones with the exception of mountain goats. Most larger game seasonally congregate around the margins of this zone. I The Intermediate Grasslands are found within the Interior I Douglas Fir vegetation zone at elevations of 915 to 1370 I meters (3000 to 4500 feet) (Alexander 1992a). These areas are characterized by open grasslands with deciduous trees growing only in riparian contexts and along meadow edges. Winters are described as "cool and snowyM while summers are "warm and I dry". Mammals found here are little different from that of I the Montane Forests with the exception of a major reduction in the more alpine inclined species such as bighorn sheep or yellow bellied marmots. Several species of grouse have been common in this area. A variety of edible berry and I root-bearing plants can be found in the adjacent meadows and I in riparian contexts. I The Intermediate Lakes area is found within the Douglas Fir vegetation zone and in some open areas along its margins. It consists of lands adjacent to lakes and their adjacent inlet and outlet streams. Water level may fluctuate widely based upon a number of factors (Alexander 1992). Climate is I similar to that of the Intermediate Grdsslands. Animal life is similar to the adjacent grasslands with the addition of many species oriented partially or entirely to aquatic living. Most critical to aboriginal inhabitants were the high numbers of trout present in the lakes and streams. It has been noted that the Indians probably even stocked the lakes with trout 29 I (Teit 1900:348). A variety of wetland plant food species were also present in this area (Alexander 1992a). The River Terraces zone contains all of the terraces of the Fraser River valley, ranging in elevation from 300 to 600 meters (1000-2000 feet) (Alexander 1992a). This area is extremely arid with cold winters and very hot summers. Thus, along with the River Valley zone, it contains the greatest seasonal variation in temperature. Precipitation is minimal and occurs primarily as snow during the winter. Ponderosa Pine is the dominant tree in all but riperian areas, where a wider variety of types are found. A variety of other shrubs and grasses are common including Bluebunch wheatgrass, antelopebush and sagebrush (see Alexander 1992a; Beil et a;. 1976:54), Mammals are most common on the margins of this zone, but in general are not common. Plant foods are not common (Alexander l992a . he River valley z~:~~-leis defined by Alexander (1992a those lands immediately adjacent to the Fraser River, less than 60 meters (200 feet) above the river. With the exception of the area around the town of Lillooet, the River Valley zone is extremely steep sided. Where present, the vegetation is little different from that of the River Terraces zone. A variety of mammals use the river valley as a travel corridor and bears tend to be found here when salmon are spawning. Mussels are present as are a variety of fish. Most noteworthy are the massive numbers of salmon moving through the Fraser River at different times between spring and fall.
30 To understand the role of risk management in cultural systems one must first have some understanding of the natural fluctuations in availability and accessibility of critical natural resources (Winterhalder 1980, 1981). Ideally, one would want to rank the relative contribution ~ndlabor investment for obtaining all resources used. Since these data are not currently available, I focus on only a few key
resources to illustrate the role of risk management. Although a wide variety of food resources were used by the aboriginal
peoples of the Middle Fraser Canyon, several stand out as far more critical to survival than others. These include deer,
salmon and a variety of berries and roots (Hayden 1992b; Kew 1992; Romanoff 1990, 1992a,b; Turner 1978, 1992). I now review, briefly, the annual accessibility and availability of these resources. By availability, I refer to overall population size (or total biomass.of that resource). By accessibility, I refer to temporal and spatial constraints on the availability of that resource (Wiant and Hassan 1985). Deer were critical resources for year-round subsistence is well a3 for items such as clothing. There are no quantified estimates for deer population fluctuations in the Middle Fraser canyon area (Hayden 1992b). It is known however, that deer populations in nearby areas do fluctuate dramatically through 18 year cycles, Alexander f1992a) has provided estimates for deer population fluctuations in the nearby Hat Creek Valley, ranging from about 125 to 600 animals. Given a predictable 18 year population fluctuation 31 a a, U a3 6) U h l-i d rCI U 4 +,m m a2-( w 0 cn U rrj %
Q) Q) U C C, ah a
0 t-i t-i 0 w U IT5 -Im a .rl R w &) Q) UC) 4J Ur amm C UUU 4 C, C, C, FCC m C a aJ r; 4 U U d E 0 'r3 Thus, even though in absolute numbers, salmon were essentially inexhaustible using pre-European contact technology (Hayden 1981), yearly fluctuations in numbers of salmon did require long term planning for any intensive reliance on this resource. Salmon accessibility is conditioned by a number of factors. First, major salmon runs occur only during short periods of the year (Kew 1992). Second, fishing access fox sockeyes and springs is limited to relatively few places in the zteep Fraser canyon where individuals could construct fishing platforms (Kennedy and Bouchard 1992; Romanof f l992a). Third, as a fall and winter survival resource, salmon had to be processed (drying, oil extraction) for long term use (Romanoff 199Za). Thus, personnel and time were required for intensive processing activities. Fourth, adequate technology for capture, butchering, drying and storage had to be in place to facilitate salmon use (Hayden 1992b). Thus, accessibility depended not only on opportunity to capture fish, but also on the time and ability to properly process it for long term use. Plant foods were used in two fashions, as immediate consumption staples and as stored long-term survival foods. "Root" plants appear to have been used primarily as immediate consumption staples (Hayden 1992b; Turner 1992). Bezries were used both as immediate consumption foods and as dried storage food to accompany dried salmon (Romanoff 1992b; Turner 19921. Availability of plant foods was conditioned by seasonality as well as the degree of exploitation in prevlous years. Turner (1992) notes that many root crops were best used during periods before or after flowering, Thus, many root crops were procured during late spring and early summer, depending on elevation. Spring beauty for "mountain potatoesw) could be
harvested throughout the summer, though they were hard to find when not in flower. Berries are best used when ripe, but not over-ripe. Thus, depending on berry type and elevation, berry harvesting could be done continuously throughout the summer f Turner 1992). Fluctuation in plant resources on an annual basis may have been conditioned by climatic variability as well rates of harvesting, especially of root crops such as spring beauty corms. Turner (1992) comments that there is little evidence for instability in root yielding plant resources in the recent past. Hayden f1992b) notes that berry species may be highly fluctuatlonal in their yearly production. Accessibility to critical plant resources appears to have been conditioned primarily by time constraints. Often other critical resources (salmon and deer) required attention at the same time as plant foods. Thus, scheduling appeafs to have been the primary constraint on accessibility. The Middle Frager Canyon was occugfed ethnographically by the Upper Lillooet and Canyon Division Shuswap (Teit 1900, 1906, 19091, all of whom reliec? on a strategy of intensive storage, logistically organized resource collecting and a biseasonal pattern of *dinter sedentism and spring, summer and fall residential mobility (Alexander l992b). It is also known
34 that these groups had a fair degree of socio-political complexity with ownership and control of certain critical resources (Romanoff 1985; 1992a; Teit 1906:254), slavery iTeit 1906:264), household crests (Teit 1906:2561 and extensive trade and warfare controlled by community leaders (Cannon n.d.; Hayden 1992 a,b; Teit 1909:576). Wiessner's (19823 risk avoidance strategies were used independently and in different combinations to cope with risks associated with resource availability and accessibility. I now review the role of each strategy in the Middle Fraser Canyon from an ethnographic perspective. Territoriality, sharing, potlatching, trade and warfare are linked together in the Middle Fraser Canyon as critical means for reducing risk of food shortage. The integration of these factors illustrates the critical role of different risk reduction strategies used together in this region. Territoriality was practiced through physical defense of tribal owned territorial boundaries (Cannon 1992; Romanoff 1992a; Teit 1906, 1909). Thus, any tresspasser could be killed if caught in rival territory without some special arrangement. Some places, however, such as the Six-Mile Rapids, which produced far more salmon than any single tribe could use, often served as free-access areas for more than one tribe (Romanoff 1992a). The concept of territory is not truly understood without considering its role at the intra-group level. Among the Canyon Division Shuswap personal property did not exist, but fishing-places and trapping grounds belonged to crest groups within the nobility class (Teit 1909:582). Hunting territory, berry-gathering places, root-digging grounds and camp-sites in the mountains belonged to the nobility of the band. Commoners and strangers had to pay rents to the crest groups of the nobility for the use of these lands, In the form of skinu, dried fish and oil. More distant lands were tribally owned with free access for everyone from that group. The Upper Lillooet did not have quite the same level of stratified ownership of resources as displayed by the Canyon Shuswap. Teit (1906:256) notes that all hunting and root digging areas, as well as trails, were considered to be tribal property. Important fishing sites were clan owned. Romanoff (1992a) indicates that some places were tribal property and thus available for all (sockeye salmon fishing spots), others were indeed privately owned (spring salmon fishing locations). Although it has been documented that others could use a critical fishing spot after the principal owner finished with it (Romanoff 1992a1, access to enough food for winter survival could still be in question, if the next person or persons to use that fishing spot did not have enough labor available to process the fish. -Since lower class people did not have access to larger labor pools, this was often the case. This situation was exacerbated by the fact that hunting was principally an activity of the noble class. Thus, even though commoners had access to land, they rarely received the training or had the time to engage in hunting to a significant degree (Hayden 1992b; Romanoff 1992b).
The result of variation in resource access was inequality in food availability for the different classes of people in Lillooet and Shuswap groups. Without adequate access to the spring salmon fishing spots or larger game, lower status groups had more problems surviving without some alternate means for obtaining food. This axternate means of obtaining
food could have been through sharing, trade, patlatching and warfare.
During tines of shortage, families who could not acquire enough food had to request food from those who had surplus. This had two results. First, those reduced to begging for food lost status (Romanoff 1992a). Second, those who repeatedly helped others gained prestige. Romanoff (1992a) argues that gift-giving of rations of fish was an acceptable practice until it became too unbalanced thus giving rise to such degerogatory terms as Mmoocher,fi There was a tight link between predictable access to food staples and individual and group status. Potlatching was not as developed as on the coast and the goods distributed, were primarily of the subsistence variety (Romanoff 1992b). Teit
(1906:258) notes ho.wever, that some Lillooet potlatches were indeed elaborate affairs with chiefs of different clans attempting to outdue one another with lavish gifts. It appears likely that the potlatch may have accomplished two things in the Middle Fraser Canyon. First, wealth accumulation and alternately, subsistence resource acquisition 37 through potlatching may have served as a risk transfer mechanism in much the same way as postulated among the
Southern Kwakiutl (Piddocke 19651. More typically, however, potlatching in this area may have served more crucially as a risk pooling mechanism for redistributing deer meat from higher rankfng families to commoner families regardless of short-term salmon availability (Romanoff 1992b). It also encouraged continuous training far deer hunters regardless of the availability of deer in a given year.
To anticipate shortages of certain foods (most typically salmon), many groups actively engaged in trade. People of the Middle Fraser Canyon found themselves in the ideal position as "middlemen" between the people further inland and those at the lower end of the Fraser River system (Cannon 1992). Thus, they obtained a wider variety of' trade goods and they were far less likely to be raided than other groups (Cannon 1992). In general, trade allowed group3 go obtain food not available is adequate quantities in their own territories and it reduced the likelihood of warfare. Typically, people traded with
,immediately adjacent groups and raided those whose teritories did not abutt their own.
One final, crucial inter-group means for reducing the risk of food shortage was to raid for food, thus transferring risk tot other groups by stealing their food. Cannon (1992) has documented a tight correlation between limitations in salmon supply and the frequency of warfare. The seriousness of this warfare is attested to by Teit (1900:246,267, 1909:542) and it is clear that in many contexts, offensive raiding was seen as "an advantageous complement to tradew (Cannon 1992324). Thus far I have considered the more complex inter- and intra-group mechanisms for minimizing risk (territorial boundary protection), transferring risk (warfare, potlatching) and pooling risk (food sharing, patlatching, trade). I now turn to the technological and mobility strategies more closely associated with solving problems of resource availability and accessibility.
Alexander (1992bf and Teit (1906:224) have described the seasonal round of the aboriginal inhabitants of the Middle Fraser Canyon. Briefly, the Upper Lillooet and Canyon Division Shuswap year begarA in November, when people moved back into their winter villages. As houses were prepared, people hunted and ,fished in the immediate vicinity. They generally spent December, January, February and occasionally parts of March in semi-subterranean structures (pithouses) living from stored foods (primarilly dried salmon and berries .[Romanoff 1992a, 1992bl). stored food could be supplemented, when weather permitted oktdoor activities, by some local deer hunting and ice fishing around the Intermediate Lakes. As is typical for most northern latitude hunter-gatherers, late winter 'was a critical time when stored foods were becoming exhausted and game and plant foods were as yet difficult to obtain. After collecting plant resources at lower elevations,
39 families spread out in the spring, a few individuals fished for early spring salmon, while most moved to trout fishing
stations in the Intermediate lakes or to the Montane Parklands for root collecting and deer hunting. Late spring (June) saw intense berry collecting and by mid-July people were back in the river valley focussed entirely on salmon fishing and berry collecting. Following this, families moved back to the Montane Parkland to continue hunting and plant collecting. Intensive hunting and sporadic fishing and plant collecting in many environmental zones continued until November, when families again returned to their winter residences (Alexander
Kelly (1983) has drawn a distinction between the seasonal round concept and the mobility strategy. He notes that while a seasonal round describes the geographic movements of a group of people, the mobility strategy actually refers to the decisim making procegs behind residential group and task group movement. To understand risk management one must therefore consider the role of the mobility strategy. The Lillooet and Shuswap of the Middle Fraser Canyon followed a classic biseasonal pattern (see Gilman 1987) of group mobility. During the period of December through February people were residentially sedentary using only short logistical trips to gather supplemental resources where possible. By Iiarch and April, residential mobility increased dramatically as did logistical mobility. Residential moves were often made to the Intermediate Lakes and occasionally to
4 0 the Intermediate Grasslands during April or May (Alexander
1992b). March and April logistical mobility ranged widely throughout the Terraces, Intermediate Lakes and Grasslands and
River Valley depending on access to plant resources, game and early salmon runs. May saw both residential and logistical use of the Montane Parkland zone. Some families moved up into the
Montane parkland for extended hunting and root collecting trips, while most concentrated residentially on trout in the
Intermediate Lakes sending logistical parties to the Montane parklands for deer and roots. By June, intensive berry collecting and drying had begun. Some families remained residentially stable in the Intermediate Lakes, sendins out logistical parties to the Terraces, Grasslands and Parkland.
Others were slightly more mobile eutablluking residential camps back on the Terrace. During this time, residential groups continued to focuss less on the Montane Parkland, while logistical use of this zone probably increased (see Alexander
1992b:36-37) as deer grew fatter and roots and lichens more accessible. Mid-summer saw a sudden residential shift to the River Valley for intensive salmon harvesting and processing.
Logistical mobility at this time was limited only to short berry gathering trips in the local vicinity of the salmon fishing spots (Alexander 1992b).
By mid-August, the most intense salmon fishing had ended and residential bases were shifted to the Montane Parkland for a period of intensive deer hunting, trapping and plant collecting. Multi-family residential camps were established
in this area and logistical mobility focussed on hunting trips within the same environmental zone and into the Alpine zone (Alexander 1992bi. Some female logistical groups spent brief periods in the Intermediate lakes collecting cattails (Alexander 1992bf. This pattern continued probably until some time during September when deer began to move to lower
elevations. The period of mid-September to early November was
probably one of relatively high residential mobility by small
single to multi-family ucits as hunters attempted to follow the deer in their annual movements. Logistical activity focussed on continual hunting and trapping in the Montane
Forest and Parkland for deer as well as other large and smaller game (Alexander 1992b). Some logistical fishing activity may also have centered around late salmon runs (Kew 19921 and gathering of late appearing plant resources (Turner 19921. It should be emphasized that many small game and plant resources were gathered casually during the course of other more critical activities (Alexander 1992b). At the advent of cold weather, families moved back into their winter residences . Logistical and residential group movement and resource collecting 93s centered on obtaining foods for immediate consumption as well as for obtaining foods for storage and raw materials for equipment. Alexander's ethnoarchaeological
42 research indicates that residential mobility was focussed on placing families in locations central to obtaining resources
for long term use. Immediate subsistence resources could be
extracted from that primarily designated for storage or they could come from other sources. While early spring residential group m~bilitywas designed primarily to gain access to resources for immediate consumption, late spring, sumrner and gall mobility was heavily oriented towards acquiring massive quantities of storable resources such as deer, berries and salmon. Immediate subsistence depended upon use of the above resources as well
as other locally available foods such as roots. It seems likely that root crops were a critical dietary resource for supporting family groups engaged in gathering foods intended for storage. Logistical mobility solved problems occurring when critical resources were available simultaneously in different parts of the landscape. In these situations, some family members could remain in the vicinity of residential base camps collecting less mobile resources such as roots and small game, while others established short term camps for deer hunting or even salmon fishing. It is clear from the ethnographic research conducted in this region, that survival would not have been possible without a highly flexible warm season strategy of mixed residential and logistical mobility. Thus, the risk of food shortage could be greatly minimized through continu~uslong and short range planninq.
43 Wiessner (1982) and Torrence (1989) have argued that the
avoidance of zisk Is best accomplished, respectively, through
storage and technological design parameters and organization.
I now review the roles of storage and technology as risk avoidance strategies in the Middle Fraser Canyon. Storage was a critical survival strategy in the Middle Fraser Canyon (Teit 1906:223). Not only were foods stored, but caching strategies were used also for tool storage. A variety of food caching techniques were used and it useful to consider the role of insurance versus storage caching strategies for organizing food storage. Teit (1906:223) describes both storage and insurance caches. Storage caches are designed for continual winter use, while wsurpluswcaches are designed for longer term storage. This distinction between continuous use and long term storage is reminiscent of Inuit insurance and storage caches (Binford 1978; Gubser 1965). Teit (1906:223) notes that the surplus caches contained food not required during winter. It is likely however, that these served as means for insuring late winter or early spring survival. Teit comments that they were not used until spring.
Elevated caches have been described for the Montane Parkland zone (Teit 1906:215; 1909:495) and in areas adjacent to winter villages (Hill-Tout 1978:110). Many of these, especially those in the Montane parkland may have served as insurance caches of meat for logistically organized hunting and root gathering groups. Others, especially those
44 associated with winter villages may have preserved foods such as dried deer meat for regular winter use, especially if deer meat was saved for potlatches (Romanoff 1992b). Elevated caches often contained gear stored in anticipation of use in the mountains (Alexander 1992b; Teit 1900:198) or stored in villages for protection from dogs (Teit 1900:199, 1909:495f. Storage, as a critical component of a society employing logistically organized resource collecting, allows anticipation of temporary shortages. Storage caching requires some degree of sedentism, while insurance caching can permit some degree of mobility as well as serving as backup during sedentary periods. Middle Fraser Canyon peoples avoided risks of food shortage by storing critical foods such as salmon, deer and berries for winter consumption. They apparently also, anticipated temporary shortages on food collecting trips by caching dried foods at critical places on the landscape, such as in the Montane parkland zone. Finally, they attempted to ensure winter survival by caching extra foods for late winter and spring use. Storage however, could not be seen as an effective risk avoidance tactic without technological strategies equally designed to avoid risk. I now consider the critical role of technology. I review the role of technology in relation to Torrence's (1989) expectations of the relations between technological variation and risk character and severity. I consider the role of lithic technology in a subsequent discussion.
Torrence (1989:60-61) has argued that certain aspects of tool assemblage structure will be heavily conditioned by the character of risk. Risk character has two components: temporal variation and spatial variation. Where spatial variation is the only problem to be solved, people tend to rely more on instruments and simple weapons. As temporal variation becomes more evident, weapons and facilities are relied upon to a greater degree. In the Middle Fraser Canyon, temporal variations have a far greater effect on technological organization than do spatial variations. In other words, resources are generally not spread out requiring high search times and reduced processing time. Critical resources become available in large quantities for short time periods, often in conflict with the availability of other resources. For example, the most critical resource, salmon, occurs in vast quantities during a very few weeks during the summer and fall. Often the procurement of salmon conflicts with access to berries, deer, roots and small game. Likewise, another important resource, trout, became accessible at a time when root gathering and deer hunting are also important. People must optimize their technological capability to procure these resources in large quantities during short periods of availability. Technological strategies should also be preferred that enable people to cope with the problem of simultaneously available resources. Salmon fishing was accomplished using both tended facilities and weapons, including nets, traps and spears (Teit
46 1906:227-228, 1909:526-530). Trout fishing was accomplished with traps, spears and hook and line (Teit 1906:228,
1909:526-529). Big game hunting (especially deer) was accomplished using both game drives and individual ambush hunting, aided by spears and bows and arrows (Teit 1906:226, 1909:521-522). Deer and bears were also trapped using simple snares (Teit 1906:226) as well as deer fences (Alexander n.d.b; Teit 1909:521-522) . Smaller game were obtained using traps such as dead-falls (Teit 1906:227). They may also have been opportunistically hunted. Plant foods were gathered using baskets and digging sticks.
As predicted by Torrence (1989:60-61), high time stress on food access resulted in a series of technologically sophisticated strategies for obtaining critical resources. Fishing and big game hunting (including deer) were carried out with both tended and untended facilities and weapons. Small game was procured using weapons and untended facilities. Plant foods were procured using instruments. The severity of risk has high implications for the technological options employed in the Middle Fraser Canyon. Where risk is severe, as is the case for resources such as salmon, procurement tools are clearly specialized, designed with multiple components and probably produced by specialists (e-g. nets, traps). As use of a resource such as salmon also depends heavily on processing time, processing equipment may also have been optimized with many of the design components of the procurement tools (i.e. high reliability and specialization). It is unknown as to what tools were used in the interior -
Many of these same design parameters can be seen in hunting equipment as well. Deer and bear snares required a high degree of specialization in their design. As well, their construction had to be able to withstand a high degree of applied force from a large animal attempting to escape. Thus, they had to be overdesiqned to some degree. Some weapons illustrate a fair degree of reliability in their design as in the case of the Lillooet beaver spear (Teit 19063226). Maintainability is seen in many aspects of Middle Fraser Canyon mobile technology (e.y. those tools which are carried by foragers). ~ostbows, arrows, spears, digging sticks, baskets and bags (Teit 1906:205-223), could be adjusted for use on a range of-different resources. They could also be easily repaired during use. In this way, a wider variety of resources could be gathered, none of which had the attached risk severity that a resource such as salmon had. Combined as a single unit however, the variety of smaller game, fish and plant foods obtaine3 using a primarily maintainable technology, may have been nearly as critical as that of salmon or deer. Thus, the combination of maintainability and reliability in Middle Fraser River technologies ensured that the rizk of reduced ~ccezzto foods through technological inadequacy, could be avoided. LITHIC TECHNOLOGY
In this section I review the ethnographic evidence Eor the use of lithic technology in the Middle Fraser Canyon. Next, I demonstrate that lithic technology was an important part of the complex system of risk management in this region. This is done by considering the role of lithic implements as primarily tools for making other tools, as occurs during wood working or hide working. I note that embedded lithie raw material procurement and winter conservation of raw materials was critical for complete preparation for warm season activities and that this process contains implications for the recognition of economic systems through debitage and flake tool analysis.
Teit (1900, 1906, 1909) and Moxice (1893) have provided ethnographic descriptions of lithic technology on the Interior Plateau of B.C. Some of this information has been summarized by Magne (1985). 1 provide a brief review of some of the major tool types produced and reduction strategies followed. Both groundstone and chipped-stone tools were produced. Groundstone tools included hammers, mauls, axes, adzes, polnts and knives (Teit 1900:183, 1906:203, 1903:473). Teit has described bipolar core reduction (1900:182), hard hammer percussion flaking (1900:182), soft hammer percussion flaking (1909:4731 and pressure flaking (1900:182). Chipped stone tools produced include pressure flaked arrow points (Teit 1900:182), stone knives (1900:183), scrapers (1900:184-1851,
49 1906:203, 1909:473), arrow smoothers f1909:519) and chisels (1900:183). Raw materials used centered on 'glassy basaltw (Teit 1306:203) as well as a range of other types including chert, chalcedony, quartz, jasper, serpentenite, nephrite, "greenstonettand obsidian (Teit 1306:203, 1909:473). Morice (1893:65) describes some quarry sources as privately owned among the Thompson, while Teit makes no reference to raw material source ownership among any group. With the exception of stone arrow and spear points and some knives, chipped stone tools were primarilly produced as tools for making other tools. Teit's ethnographic descriptions indicate a primary focus on wood-working and hide-working using such tools as 'chisels, carving-knives, scrapers and arrow smoothers for wood working and knives and scrapers for working hides. It can be assumed therefore, that much lithic reduction behavior was oriented towards production of suitable tools fox these purposes. Spring, summer and fall were periods of extreme activity among people of the Middle Fraser Canyon. Gaining access to critical foods and other resources required mobility as well as properly prepared gear and clothing. Thus, it is likely that winter served as "down time" (Binford 1979; Bleed 1986) for production and maintenance of gear and clothing so that it might 'be ready for immediate use in the spring. Indeed, Teit describes the winter as time for manufacturing weapons
Parry and Kelly (1987) have argued that sedentary groups
50 in North America typically stockpiled raw materials and then
used that raw material at their leisure. The result was a
relatively high degree of expediency in tool production and
little effort expended on raw material conservation. Several
factors can be expected to have produced a somewhat similar strategy in the Middle Fraser Canyon. First, a high degree of food and equipment storage was rel.ied upon, resulting in cold season sedentism. Second, as warm season mobility was critical, winter was used for intensive to31 and clothing manufacturing and maintenance activities. Given this situation, it is possible that raw materials may have been stockpiled during the fall in a similar fashion to that described by Parry and Kelly (19871 for Late Prehistoric groups in the Southeastern, Southwestern and Plains areas of
North America and in parts of Mesoamerica. This would have been advantageous to temporarily sedentary people as it would have provided a long term source of raw materials for the production of stone tools to be used primarily in working organic materials.
Parry and Kelly (1987) argue that with raw material stockpiling, there is an ever increasing focus on expediency in tool manufacture and use. They cite the repeated presence of amorphous unprepared and bipolar cores coupled with drastic reductions in hiface production as indicators of increased expediency in lithic reduction. In some wzys this argument is deceivingly simple. Bipolar core reduction can indicate expediency, but it can also indicate the need to recycle and 51 extend raw material use-life (Goodyear 1989). Thus bipolar core reduction may even indicate some degree of curation given raw material shortages. This could be expected to occur among sedentaxy peoples where access to raw materials is temporarilly reduced due to adverse weather conditions. Some of the patterning cited by Parry and Kelly from more northern contexts (Ahler and VanNest 1985; Binford and Quimby 1963) may be at least partially due to this. If access to lithic raw materials is substantially reduced or temporarily eliminated, one would expect to see an increasing focus on raw material conservation. This might be indicated by higher degrees of edge preparation in block and hifacial core reduction, coupled with salvaging of flakes from exhausted cores and tools using bipolar techniques. In this scenario, flake tools might continue to be used, but in a more curated fashion. This could be indicated by more intensive resharpening of some tools and reuse for new purposes of other previously discarded (and often trampled 1 tools. Archaeological assemblages resulting from this process would contain a range of both heavily retouched and broken but minimally retouched flake tools. On initial inspection these assemblages could appear to represent largely expedient tool use, while the actual formation proceses may have been far more complex with some tools undergoing curated use and many others used expediently on multiple occasions (or serial expedient use). Teit's descriptions of a range of different types of specialized flake tools indicates that this could be likely. From this ethnographic perspective, flake culling can be predicted to have operated in three fashions. First, reoccupation of old house floors may have resulted in scavenging of flakes produced during earlier occupations. Second, lithic reduction likely focussed on production of primary flakes for either curated or expedient use. Thus, culling focussed initially on those flakes, which were probably larger with either high or acute edge angles, depending on needs. Flakes culled for hafting and exceptionally long use probably had edge angles that facilitated further reduction and shaping, Third, specialized tool needs and late winter shortages in raw material probably encouraged people to intensively use secondary byproducts of the reduction process. These are broken flakes resulting from accidental breakage of either primary or plat for^^ preparation flakes. In general, lithic debitage assemblages from winter housepits were probably culled fairly intensively. the removal of items such as flakes or animal parts from a larger population of items. There is no intent implied for prehistoric culling to have been oriented towards any form of population management is in herd culling by livestock managers .
To summarize, I argue that the primary role of chipped-stone technology in the Middle Fraser Canyon was Eor the production and maintenance of other organically based tools including arrows, spears, traps, nets, digging sticks,
baskets and hide bags and clothing. A substantial amount of
manufacture and maintenance of these items was conducted during the period of winter sedentism. This required enough lithic raw materials to be available for continuous use over a
period of at least three months. Thls was accomplivhed by first stockpiling raw materials in the form of flake cores in winter housepits during the fall and second, by producing specialized flake tcols for curated and serial expedient use during this period. Late winter shortages in raw materials were dealt with using bipolar reduction techniques to salvage I additional flakes from exhausted cores and worn out bifacial and flake tools.
SUMMARY
In this chapter I have sought to construct an argument for the relevance of risk management as a general theoretical construct used to explain some aspects of the interrelationships between intra- and inter-group I interactions, mobility, technology and storage and problems in the seasonal availability and accessibility of critical I subsistence resources among the ethnographic peoples of the Middle *Fraser Canyon. The primary purpose of this exercise I has been to demonstrate that the organization of fall and I winter lithic raw material acquisition and tool production and I I use was part of the overall economic system of these people. I This modelling will provide a theoretical backdrop to be compared to conclusions drawn from the archaeological record regarding li.thic raw material use, subsistence, mobility strategies, etc. The archaeological record may provide indications of both similarities and differences to this system. From this, we can expect to learn about the shifts in adaptations which occurred between the late Prehistoric and Contact Periods and about limitations of the models used. To summarize, I have drawn the following conclusions regarding risk management, cultural systems and lithic technology:
1. Risk Manaqement. Risk of food shortage is a problem dealt vith in many components of cultural systems. Territorial boundaries are controlled to limit other group's access to resources. Mobility and technology help to avoid risk by placing people in resource procurement locales with the appropriate technologies to acquire and process adequate amounts of that resource. Storage aids in avoiding risk by anticipating periods of resource shortage. Warfare and potlatching work to transfer risk to external groups or between internal groups, respectively. Trade and sharing pool risk such that any group can accomodate for anticipated or sudden unexpected shortages.
2, Manaqinq Risk in the Middle Fraser Canyon. A variety of specialized and often interrelated tactics were used by inhabitants of the Middle Fraser Canyon to minimize and avoid risk, Territorial boundaries were maintained at community and
household levels. A biseasonal strategy of winter sedentism
and warm season logistically organized collecting was employed to pzovide access to a wide variety of resources, often available for short time periods and in restricted geographic localities. Technologies allowed intensive salmon and trout fishing, a wide range of plant collecting and large game (especially deer) hunting. Storage was used to provide winter sustenance as well as some components of subsistence during other times of the year. Potlatching provided deer meat to many members of the communities as well as promoted a form of risk transfer not unlike that of some coastal groups. Warfare allowed some groups to take advantage of others, by stealing critical rescurces. Trade and food sharing promoted intra-group interaction and risk pooling. It also reduced the chances of warfare between contiguous groups.
3. Risk and Lithic Technoloqy. Aside from the production of projectile points, butchering implements, and some specialized groundstone items, lithic technology served primarily to aid in the production of other tools made from organic components (hide, wood, etc.). Winter lithic reduction activities must have been conditioned by the availability of lithic raw rmteriais collected and stored for winter use. Cores were used to produce flake tools and a relatively high degree of curation or recycling of many of these flake tools seems
56 likely. Tool and core use-lives were possibly extended in late winter by bipolar reduction activities. Winter production of tools and gear was important as preparation for spring food gathering activities, as food was typically short at this time and success in foraging, critical. In the next chapter I develop and test methods for interpreting the formation of debitage and flake tool assemblages independently of the conclusions drawn here. I then apply the methods to archaeological contexts in ozder to evaluate the usefulness of the methods and to assess the archaeological record for the possibility that some similar economic strategies, as those described here, could have been operative in Late Prehistoxic winter pithouse villages. CHAPTER 3 RELIABILITY AND VALIDITY ANALYSES
The goal of this chapter is to develop and test instruments to be used in measuring variation in flake culling and flake tool reuse/recycling. The instruments to be used include a modified version of Sullivan and Rozenfs debitage typology (Sullivan 1987), hereafter referred to as the Modified Sullivan and Rozen Typology (MSRT) and three flake utility indices, designed to measure variation in flake size
(Flake Volunae Index t FVf I), length of edge with acute edge angle (Acute Angle Edge Length [EhAELl) and length of edge with high edge angle (High Angle Edge Length [HAELI). An experimental study is conducted to determine first, if the instruments provide consistent results (reliability), and second, if they measure what they were intended to measure (validity). These analyses are important because without them it is impossible to know whether observed variability in the experimental outcomes and among archaeological data sets is the result of real variability in the phenomenon of interest (different reduction techniques) or just the result of other variables that are not of interest in the research (e.g., skill level, knapper errors, errors derived from the measuring instrument, analyst error). I begin the chapter with a theoretical justification for the use of Sullivan and Rozen's typology, the MSRT, and the utility indices. I argue that, currently, there is no other
58 way to identify flake culling from archaeological debitage assemblages alone, without the use of the MSRT and the utility indices. Further, based on ethnographic descriptions of stone tool production and use, I argue that variation in flake culling is best predicted and understood using utility indices that focus on size and edge attributes rather than formal shape characteristics. Finally, I suggest that archaeological flake culling could be recognized through the use of pattern recognition criteria derived from experimental utility index data applied to flake types from the MSRT. Ideally, using these experimental pattern recognition criteria, archaeologists should be able to interpret MSRT data as the cumulative result of technological behavior, culling and scavenging and taphonomic processes. I next analyze the Sullivan and Rozen typology (SRT), the MSRT, and the util-ity iildices for reliability and validity.
Because the SRT is not expected to achieve a high level of validity, its reliability and validity are studied in order to help justify the use of the MSRT. The experiments provide model assemblages and indices, produced under controlled conditions, to which archaeological assemblages may be compared. In order to fulfill their function the model data must possess two properties.
~irst,the experiments must produce debitage assemblage profiles and utility indices that vary in distinctive ways depending on the relative importance of specific reduction techniques. The amount of reduction technique-linked
59 variation that is subject of the research, must be sufficient to permit accurate (valid) inferences about reduction technique from the observation of assemblage profiles and utility indices. The purpose of validity testing is to determine whether or not the experimental assemblages are sufficiently distinct to permit recognition of reduction technique, culling, etc., from the study of assemblage profiles and utility indices. Second, the experimental data must not contain excess random (error) variability because this would mask variability among experiments that is the result of variation in reduction technique. The purpose of the reliability studies is to assess the amount of random variability present in the experimental data. Statistical techniques used in the reliability and validity analysis -are derived primarily from sources in education and psychology (Carmines and Zeller 1979; Green and Carmines 1980; Nance 1987). Instruments are first tested for reliability using principal components analysis and the coefficient theta (Green and Carmines 1979). When unacceptably low reliability is identified, the correction for attenuation (Nance 19871 is applied and the instrument is retested. Validity testing is accomplished using principal components analysis. This chapter is critical to this thesis because it provides the justification for use of the basic analytical techniques. The MSRT (and the utility indices) must not contain excessive random error in their applications as this would skew any archaeological interpretations. Likewise, they
must produce data distributions from debitage assemblages which are predictably distinct depending on reduction technology. It is important to be able to link distinctive utility index profiles to distinctive reduction strategies as
origins of either culled flakes (as in flake tool assemblages) or unmodified flake assemblages which have had flakes removed. Effective utility indices will permit the study of multiple formation processes on debitage and flake tool assemblages in a way analagous to Binford's (1978) use of faunal utility indices. Utility indices will also permit the archaeologist to better assess economic decision making on the part of the prehistoric stone-worker. It is important to identify reduction technique, but even more important to explain what that reduction technique was intended to accomplish and why it was conducted in the first place. In this way, we move closer to an increased anthropological understanding of archaeological lithic assemblages.
INSTRUMENT DESIGN
~ostarchaeologists involved in lithic arialysis acknowledge the importance of recognizing flake culling lithic tool assemblages (c.f. Ahler 1989:213-216; Audouze 1987:193; Bamforth 1990:94-95; Barton 1990:28; Dibble .
1991:248; Ferring 1976:226; Frison and Bradley 1980:18; Fladmark 1984:148; Henry 1989:1+9; Isaac 1977:174; Knudson
1983:ZO-21; Kuhn 1991:88; Lewenstein 1987:194; Lothrop
1989:116-117; Odell 1981:332, 1989:225; McDonald 1985:62-65; Moss 1986:118; Patterson 1990:557; Potts 1991:f60; Schiffer 1976:107; Singer 1984:43-44; Teltser 1991:370; Yerkes
1987:129-30). Identification of flake culling behavior has generally been attempted by means of comparisons of individual tool morphology and debitage assemblage characteristics (c.f.
Ferring 1976; Frison and Bradley 1980; Knudson 1983; Kuhn 1991; Lothrop 1989). While this technique is useful and has improved our understanding of the technological origins of tool assemblages (especially flake tools), there is some room for error. Flake culling can be missed when culled flakes have been removed altogether. Likewise, flake culling may be misidentified if tools have been heavily ,ecycled or processed into new forms so that original flake morphology is not recognizable (i.e. bifaces). Removal of entire classes of flakes from production locales is a difficult process to recognize. With the use of a refitting analysis, Singer (1984:43-44) has noted that while core reduction debris is present at several sites in southeastern California, many flakes, cores and most tools were apparently removed by the prehistoric knappers. The implication is that without the refitting analysis it would have been difficult to identify the effects of culling on
62 these lithic assemblages. Singer's (1984) analysis still does not permit the prediction of the morphology of the removed flakes. In a similar analysis, Bamforth (1990) has identified the removal of flakes, bifaces and cores from llthic procurement localities in southern California. Bamforth attempts to draw inferences about the form of removed flakes by assessing refitted core and flake morphologies. ~houghhis conclusions regarding the removal of *'cores and useable flakes" ( Bamforth
1990:95) appear to be well founded, he does not provide any conclusions on the predicted morphology of the "useable flakes." Thus, it is difficult to evaluate their intended roles in cultural systems. Bunn et al, (19801 have identified flake removal in Lower Paleolithic assemblages from East Africa through refitting analyses. Though some large pieces are recognized as missing they do not speculate further on morphology. Clearly, methods are required for predicting missing flake morphology in archaeological contexts. Some archaeologists have described flake culling and removal to different use-locales in ethnographic contexts. For example, Binford observed, among the Alyawara of Central Australia, that "The blanks that the men wanted to save for particular manufacturing needs were generally placed near the right knee of the worker and picked up wnen the worker abandoned the core ...During lulls Setween different episodes of workers reducing the core, other men and boys would recover from the arc of debris, flakes that were taken to other family camps at Bendaijerum 63 Binford and OtConnell note (also for the Alyawara), ".. . informants made it clear that the manufacture of blades. ..was normally done at the quarry. The resulting blanks would then be introduced to the residential site as manufactured items (1984:415)". These comments indicate that the identificatian of culling may reflect on the role of planning depth in technological organization. Clearly some flakes were culled by the original knappers in anticipation of specialized use. Other flakes were culled by observers with other intentions, which could include more expedient type uses. It is likely
that variability in the morphology of culled flakes may reflect different types of culls, Clearly, the identification of culling is potentially important to archaeologists, as are the types of culls and conditions associated with culling. Flake tool recycling and more intensive forms of tool production such as biface manufacture may transform culled flake assemblages into new tool assemblages from which it is difficult to reconstruct the original flake morphology. Dibble (1987:37, 1991; Roland and Dibble 1990:485) has commented extensively on the formation of flake tool assemblages through resharpening and recycling. It is clear from his discussibns that flake tools can be intensively used and reused to the point that the tool eventually does not bear any resemblance to its initial form. Part of the use/reuse cycle may also involve zzworkfng or reuse of broken portions of flake tools (cf. Audouxe 1987:194). Thus, the initial form of a flake tool may not necessarilly be a good predictor of 64 its final form (c.f. Hayden 1989; Shott 1989). The inherent problems in recognizing culling strategies from the analysis of flake tools again argues for methods of identifying and explaining culling which do not rely 3n debitage/•’lake tool assemblage eompar isons. This is not to say, however, that studying the formation of flake tool assemblages is unimportant. In most archaeological contexts, lithic reduction and flake culling has certainly been accompanied by both flake tool production, use and discard (and possibly recyclicgl and flake/flake tool exportation
(Johnson and Morrow 1987; Knudson 1983; Lothrop 1989; Schiffer
1976; Wilmsen and Roberts 1984). To understand the full range of raw material movement, archaeologlsts must be able to identify both the complete range of reduction and recycling strategies on-site as well as which materials have been moved elsewhere. Faunal analysts have relied on utility lndices to aid in the prediction and explanation of animal part culling from kill/butchery contexts, Binford (1978) devised a series of caribou utility indices for identifying the weconomic anatomy" of the animals. Some of the indices developed include measures of meat utility, marrow utility, bone grease utility and drying utility. Probably the most useful index has been the modified general utility index [MGUI) (Binford 1978:74! which attempted to measure not only the utility of individual parts, but also attempted to account for the realities of butchering whereby some parts of lesser utility are often culled simply because they are attached (Vidersn) to parts sf maximum utility.
Speth (1983) used the modified general utility index to identify culling of high meat and fat content parts from bison semains at the Late Prehistoric Garnsey bison kill site. Todd (1987) has accomplished the same thing using the MGUI at the Horner bison kill site (Paleoindian). The advantage which faunal analysts have over lithics researchers in this case is that animals have an anatomical inventory which is always the same. Thus, all things being equal, where parts are missing, they can be assumed to have been removed by some human or nonhuman agency. Utility indices help to identify patterning in thete removal processes which-can be linked to human economic behavior.
Up until this point, lithics researchers have been stuck without any analogous structure-for developing utility indices. It has been assumed that lithic assemblages simply do not come in pre-packaged distributions resembling that of animal skeletons (c.f. Binford 1978). Earlier attempts to characterize debitage assemblages resulting from specific reduction techniques were based primarily on linkages between flake size, flake attributes and human behavior patterns (c.f. Magne and Pokotylo 1981). Other studies focussed on size iiistributitns across entire assemblages ic.f. Ahler 1984; Stahle and Dunn 1982). Patterning in flake attributes is useful for detecting variability in technological behavior, 1 but it does not provide a typological framework analogous to
66 an animal skeleton for applying flake utility indices. Designation of bfface thinning flakes, bipolar flakes, etc. is
not entirely useful either as it excludes large portions of debitage assemblages which cannot be identified in any way other than their size or degree of breakage. Size distributions are potentially useful. Flake size Is certainly
a potentially good criterion for choosing flakes fcr tools (c.f. White and Thomas 1972). It appears to be a relatively good indicator of reduction behavior (Ahler 1989). As this approach can produce a size-hased flake typological breakdown, it could be used as a framework for applying utility indices. Unfortunately, a wide range of processes may be involved in producing even a single size class. For example, small sized flakes (1-4 square centimeters) can be the result of edge preparation, percussion resharpening, pressure flaking, or the shatter of piatforms and bulbs of force, etc. It is hard to unambiquousfy link assemblages measured by size distribution alone with the multiple processes which produce the archaeological record. Sullivan and Rozen (1985; Sullivan 1987) have argued that distinctive reduction techniques should produce equally distinctive debicage assemblages. They use a simple typology which includes complete flakes, proximal fragments (broken
flakes), medial/distal fragments ( flake fragments 1, nonorientable fragments (debris) and split flakes to characterize debitage assemblages as the result of tool or core reduction [Figures 2-61. The assumption here is, like Figure 2. Complete flake. Figure 3. Proximal fragment. Figure 4. Hedial/distal fragment. Figure 5. N~norientable fragment. Figure 6.. Split flake- that of the flake size distributions, that the same reduction techniques will always produce the same distribution of these flake types in the debitage assemblage (for a given type of raw material). Different reduction techniques should result in predictably different assemblage characteristics. One can assume from this, that if different flake type profiles exist, one could also predict different potential utility index profllev (measured across Sullivan and Rozen's typology). Though simplistic, the typology appears to hold the promise of ultimately belng useful as a basic structure for use in conjunction with flake utility indices for predicting the culling of flakes in archaeological contexts. It has been argued that Sullivan and Rozen's typology is. too simple tzo accomplish what it is intended to zccomplish (Amick and Mauldfn 1989; Ensor and Roemer 1989; Prentiss and Romanski 1989). Indeed, Prentiss et al. (1988) employed a version of Sullivan and Rozen's typology coupled with a dichotomous size typology which divided flakes into small and large groups based upon a median size criterion of four square cm. The result was an enhanced view of archaeological contexts allowing the identification of size-sorting through sheetwash erosion, technological variability and one possible instance of flake culling. Experimentation to date (Baurnler and Downum 1989; Prentiss and Romanski 1989; Prentiss et al.
1988) on the typology has indicated that though potentially useful, too many factors may produce ambiguity in data sets structured only by the five basic flake types.
73 Anticipating ambiguity problems in the basie Sullivan and
Rozen typology (SRT), I propose a more complex version which adds four size variables. Expanding from Prentiss et al. (1988), the new sizes include small (less than 4 square cm,), medium (4-16 square em.), large (16-64 square cm.) and extra-large (greater than 64 square cm). This creates a more zompfex typolcgy of 20 flake types, thus combining the flake wcompleteness" (Ingbar et al. 1989) approach of Sullivan and
Roaen (1985) with a flake size distribution approach similar in some ways to those of Ahler (1989) and Stahle and Dunn (1982). This approach differs from that of the latter researchers in that I am interested in the distribution of the larger flakes more so than the small flakes (focussed on by Ahler and Stahle and Dunn). By considering flake breakage in conjunction with a broad range of flake sizes I expect to be able to recognize a broader range of technological and taphonornic processes than was possible by Sullivan and Razen or Ahler and Stahle and Dunn. Edge preparation and retouch, pressure flaking, and breakage from larger categories are expected to be represented in the small category. Medium and large categories should contain both larger flakes purposefully produced, as well as breakage from the largest categories. The extra-large category is expected to contain only flakes purposefully produced (though occasionally broken during production). By examining the distribution of SRT flake types across the four size categories, much ambiguity in interpretations should be removed. Constracting flake utility indices requires, first, some understanding of the criteria used by people involved in the production and culling of flakes. For this information, I turn to the ethnograghic record, focussing on flake selection criteria. Thgugh ethnographic descriptions of lithic flake production, selection and use are rare, some excellent accounts are available from Australia (Allchin 1957; Binford 1986; Blnf~rdand o'Con~-11 1984; ~ould1968, 1971, 1977, 1980; Gould et al. 1971; Hayden 1977, 1979; O'Connell 1977; Thomson 1964), New Guinea (Strathern 1969; White 1968; White and Thomas 1972; White et al. 1977), Ethiopia (Gallagher 1377), Brazil (Miller 1979), the Northwest Territories of Canada (Pokotylo and Hanks 1989, Pokotylo 1992 pers. cornm.), Alaska (Fcrtuine 1985) and Guatemala (Hayden 1987; Hayden and Nelson 1981). I rely on these accounts to illustrate the range of criteria used in choosing flakes for use as tools, Flake size appears to have been an important factor in many contexts. Gallagher (1977:4101 has described hafted hide scraper blanks, used for removing layers of dermis, from Ethiopia as being about 6-7 cm. in length and 4-5 cm. in width. Slightly smaller flakes were used in Australia as hafted adzes for .use in woodworking (c.f . Gould et a1 . 1971:154). Pokotylo and Hanks (1989:56) have described somewhat larger flakes (15 by I1 cm.3 used as hafted hide scrapers (used for softening rather than removal of layers of dermis) in the Northwest Terrf tories. Small flakes were commonly hafted for cutting/shredding (Strathern 1969:317; White and Thonas 1972s272) or for drilling (White and Thomas
1972:279) in New Guinea. They were hafted for a variety of
cutting purposes including surgery in many Alaskan societies (Fortuine 1985:38-39). Hafting is associated with some degree
of curation ard asscciated long use-life (Barnforth 1986; Binford 19*i7, 1979; Torrence 1983; Shott 1989). It iu tempting to link litkic flake size with curation decisions. However, it has been demonstrated that artifact size is not directly related to length of use-life. Rather, artifact size often zesponds to other variables including actual functional needs and anticipated use-intensity (Hayden 1989; Shott 1989). Length of use-life tends to be more associated with manufacturing cost (Shott 1989:22-23) and other organizational variables associated with risk management (Torrence 1989 ) . Choice of large versus small flakes thus may depend more on intended use. Both large and small flakes have been documented as having been used as hand-held tools. Where tools were required for intensive cutting or scraping operations, larger flakes were often chosen to provide both a substantial working edge as well as a gripping area (Gould et al. 1971; Hayden 1979; Miller 1979; Strathern 1969; Thomson 1964). Hand-held small flakes tend to be used in contexts of more delicate or specialized tasks including wood cutting
(White and Thomas P972:278), surgery and bloodletting !Deal and Hayden 1987; Fortuine 19853 and bone and antler wozking
(Deal and Hayden 1987 ) . Size, however, is not always the primary criterion for 76 choosing a flake for use as a tool. Numerous accounts' demonstrate that the primary criterion for flake selection was length of avcilable edge with acute or high edge angle.
I Miller (1979:402-403) notes that Xeta informants chose flake tools dlmost entirely on the basis of edge angle criteria.
Acute edge angle flakes were designated as '?Good for cutting meat, hides, vegetablesw (Miller 1979:403), while flakes with 65 to 85 degree edge angles were chosen as wood-working tools.
White and Thomas (1972) and Strathern (1969) describe Highland New Guinea flake choice as contingent largely on the presence of acute versus high edge angles. Host critical here is the fact that overall flake or core shape was unimportant, as long as a substantial acute or high angle edge was available. Acute edge angles were generally preferred for cutting while high edge angles were preferred for scraping or planing.
White ek al. (19771 comment that where hafting decisions were required, flake tools with intermediate edge angles were preferred, thus adding some degree of functional flexibility to a tool produced with a higher degree of effort expenditure. These conclusions regarding the importance of edges with acute versus high angles are mirrored in the Australian literature. Gould scates that
,"The aborigines classify their hafted stone tools into two categories based en the cross-section of the working edge: flakes with thick, steep working edges suitable for scraping wood Ipurpunpa) and flakes with thin sharp edgeo suited for slicing and cuttlng tasks (tjimari)" f1968:111).
Adzes for woodworking typically, also, had steep edge angles 77 (Allchin 1957; Gould 1977; Gould et al. 1971; Hayden 1979). Like ?he New Guinea examples, edge morphology was often more
important than overall flake shape (unless hafting was intended) (Gould ek al. 1971). If flake utility is largely related to size and length of available edge with acute and high edge angles, then valid flake utility indices must take chis into account. I now focus attention on the construction of three utility indices. I propose, first, an index of overall flake size called the Flake Volume Index (FVI). The index is constructed simply by multiplying maximum flake length by width by thickness. The second, which measures length of available edge with acute edge angle, is called the Acute Angle Edge Length index
(AAEL). The third, which measures the length of available edge with high edge angle is known as the High Angle Edge
Length index (HAEL). These indices have the advantage of usefulness on their own as well as In conjunctlati with one another. Clearly, flake culli~gdecisions are often based on size and edge morphology criteria. Thus, to be realistic, joint indices, combining the FVI and the AAEL or HAEL, should be constructed. The adequacy of the joint indices, howevez, depends on the basic reliability and validity of each basic index. The indices are to be used in conjunction with the
Modified Sullivan and Rozen Typology (MSRTI (assuming that it is more reliable and valid for recognizing technological variability than the SRT). For example, in the reduction of a 78 core, a number of flakes are assigned to each flake type in
the MSRT. Each individual flake receives also an FVI, WL
and HAEL score. The mean FVI, AAEL or HAEL score provides the estimate of the utility of that flake class. I expect that different reduction techniques will provide not only different MSRT data sets, but will produce distinctly different utility index profiles across the MSRT flake types. Demonstrating that the MSRT and the flake utility indfces provide consistent results when reapplied to the same phenomena and predictably variable results when applied to different phenomena requires the design of experiments to assess their respective reliability and validity. I now turn to this problem.
DESIGN OF EXPERIMENTS
The field of experimental design has been developed to aid researchers in the construction of controlled experiments. When coaclusions based on experiments are presented, critics can either argue against the interpretation of the results of a given experiment or they can critique the design or logical structure of an experiment (Fisher 1960:2). To avoid a critique sf the design of an experiment, researchers must follow a series of procedures beginning with the identification of the research prcblem (Montgomery 19751. It is here that the researcher must determine the intended contribution of the experiment and the scale at which the experiment is to be conducted. Next, the researcher must choose an instrument of measurement, or device or procedure for "assigning numbers to variables that represent attributes or properties of subjects or treatments" (Specton 1981:12). Instru~entsof measurement in archaeology are used to measure variation in archaeological materials produced by human behavior or other processes responsible for the variability in the archaeological record.
As these are devices or procedures for gathering quantitative information, they are sgbject to random and systematic ercor
(Montgomery 1976:2-3; Nance 19871. Random error results •’ram such things as limits in precision of the instrument and errors in the use of the instrument. Systematic errors are repeated and directional. They occur as the result of idiosyncratic tendencies of the operator and bias in the design of the instrument (hick et al. l989:3 1. Before applying a measuring instrument-to the archaeological record, it is important to determine the degree of random and systematic error inherent in its measurements, This is accomplished through the experimental analysis of reliability and validity,
The concepts of reliability and validity have become well established in the literature of the social sciences, especially psychology, sociology and education as the means for identifying random and systematic error (Carmines and Zeller 1979; Cohen 1960, Guilford and Fruchtex 19?3; Robinson 1957; Roulon 1939; Schuessler 1971; Stanley 19711. Despite some early statements on the importance of reliability (Cowgill 19701, these concepts have only recently begun to receive any serious treatment in archaeology (Amick et al.
1989; Nance 1987; Nance and Ball 1986, 1989). Reliability and validity are concepts used to describe the fundamental nature of measurements or procedures used to make measurements (Nance
and Ball 1986:461). Reliability describes the degree of consistency in a series of measurements. Reliability assessment provides a statistical assessment of the replicability of a measurement. If the same phenomenon can be measured repeatedly with llttle random error, it is considered to be reliable. High degrees of unreliability occur where repeated measu,res of the same phenomenon produce highly variable results. Validity describes the degree of accuracy in a series of meas:-trements. The assessment of validity provides an indication of the degree of concordance between the actual behavior of the measuring instrument and the theoretical
expectations regarding that behavior. A valid instrument results in a high degree of concordance between theoretical expectations and the actual behavior of the instrument. Once independent experiments have been conducted to demonstrate instrument reliability and validity, these instruments can be used in other experiments resulting in conclusions and recommendations for better understanding some phenomenon (Montgomery 1976). Data collection and analysis in any experiment, whether to test the reliability and validity of an instrument or to 81 use a tested instrument to better understand some phenoinenon, requires the construction of an experimental design. The experimental design defines acceptable results, sample size, the order in which data are obtained and the randomization process (Montgomery 1976:4). The randomization process is especially critical as it allows the researcher to "average out" (Montgomery 1996:3) the adverse effects of confounding (biassing) variables. Confounding variables inject unwanted systematic error (bias) into experiments. Unless the effects of these variables are purposely controlled or randomized, these effects will reduce the validity f an experiment (Spector 1981). In the following section, I detail the design for experimental data collection to be used in testing the reliability and validity of the SRT, MSRT, FVI, AAEL and HAEL My intent here is to provide a d.etailed discussion of terminology, sampling issues, potential confounding variables, randomizing and actual data collection techniques. I consider formal definitions and statistical issues related to reliability and validity analysis in a later section.
DATA COLLECTION
The data collection phase of this study involved a series of controlled reduction events. The goal was to produce two groups of assemblages: one for the reliability study and one for the validity study. Potential effects of confounding variables were randomized: sequences of reduction, blanks
selected for reduction events, and time of day when reduction
was accomplished. I now review the process of experimental
data collection starting first with a series of basic
definitions of terms. I then review the primary variables to be controlled to provide useful data for later reliability and validity assessments, the procedures used to avoid the effects
of confounding variables (randomization) and the actual procedures implemented during the data collection process.
The term assemblage refers to a group of flakes produced during a single reduction event. Reduction event refers to the uninterrupted percussion or pressure flaking of a core/blank to produce a group of 'flakes for later analysis
(the degree of reduction is discussed below).
A core/blank is a nucleus of raw material destined to be reduced during a reduction event? Core/blanks are divided into flakes, bifaces and prepared and unprepared cores.
Flakes are "any piece of stone removed from a larger mass by the application of force" (Crabtree 1982:36), Force, in this case, is intentional on the part of the knapper. Flakes used as blanks in the validity experiment have edge angles measuring between 10 and 40 degrees.
A biface is an "artifact bearing flake scars on both facesw (Crabtree 1982:16). All bifaces used in this study start at Callahan's (1979:16) stage 2, where roughly centered edge angles of between 55 and 75 degrees are present and the width/thickness ratio is 2.0 or more.
83 A core is "a mass of material ...p reformed by the worker
to allow the removal of a. ..flakew (Crabtree 1982:30). I
separate prepared from unprepared cores on the hasis of Crabtree's (1982:49) definition of platform preparation ar "the grinding, polishing, faceting, beveling of that part of the platforrn to receive applied force." Unprepared cores do not have any platform preparation. In other words, unprepared core reduction flakes are removed direcly from platforms without benefit of platform shaping through small flake removals or abrading. In this experiment, unprepared cores have dorsal/platform edge angles of 60 to 80 degrees while prepared cores have dorsal/platform edge angles of 80 to 90 degrees. The experiments rely on one type of raw material, Glass Butte obsidian. As the intent of the experiments is to examine the performance of the measu~inginstruments, it is important to hold raw material constant. Examining the effects of raw material variability is a problem for future studies. Three variables are of critical importance in structuring these experiments: hammer type, flake size goal and working edge morphology (flakes, bifaces and prepared and unprepared cores). Previous experimental research (Prentlss et al. 1988; Prentiss and Romanski 1989) has demonstrated that these are the most important variables conditioning variation in Sullivan and Rozen's debitage typology. I have already provided ethnographic evidence that flake size and edge morphology are the most critical variables conditioning the choice of flakes for tools. Stone workers often use specific techniques to produce flakes of a given size and shape for later use as tools. With this in mind, an important goal of this study is ko test the validity of the assumption that a combination of specific hammer type, reduction goal (flake size goal) and core/blank morphology will produce a predictable distribution of flakes,
measured in terms of volume (FVI), completeness (Wozen and
Sulllvan 1989:170) and edge morphology (AAEL and HAEL). The variables hammer type, flake size goal and core/blank edge morphology are combined to produce a series of experimental debitage assemblages (see Figures 6-11 for basic core types). I make a distinction here between debitage
assemblage and analytical assemblage. A debitage assemblage is the result of a single reduct-ion event, while an analytical assemblage is produced from the combination of three debitage assemblages. The use of three debitage assemblages to make up each analytical assemblage derives.from the need to reduce analytical bias and sampling error. The choice of three debitage assemblages to make up each analytical assemblage was done to introduce randomly chosen variability and thus eliminate sources of systematic error (see below). Thus, for the reliability study, where consistency in measurement is assessed, a total of 30 biface seductisn assemblages are collapsed into 10 analytical assemblages. For the validlty study, a more complex combination of hard and soft hammer, Figure 7. Hard hammer modified flake. zigure 8. Soft hammer modified flake. Figure 9. Pressure modified flake. Figure 10. Biface types (A=soft hammer reduced; B=hard hammer reduced; C=Pressure flaked). Figure fl: Unprepared core types (b=harci hammer reduced; B=s~ft hammer reduced; C=pressure flaked). Figure 12. Prepared core types (A=hard hammer reduced; B=pressure flaked; C=soft hammer reduced). pressure flake, extra-large, large and medium flake size goal and flake, biface and prepared and unprepared care assemblayev
(Figures 7-12) are considered and the resulting 60 debitags assemblages collapsed into 20 analytical assemblages as follows:
Flake Size Unprep . Prep. Goal Flake Biface Core Core ____------Extra- Large 1. H.h. 2. H.h.
Large
Med ium 9. H.h. 12. H.h. 15. H.h. 18. H.h. 10. S.h. 13. S.h. 16. S.h. 19. S.h. 11. Pr. 14. Pr. 17. Pr. 20. Pr.
Where : H.h. = Hard Hammer S.h. = Soft Hammer Pr. = Pressure 1-20. = Analytical Assemblage Numbers Flake size goals refer to the intention of the knapper to produce flakes in a given size range. Intended size ranges followed in this study are defined by the general size classes of the MSRT. Completed debitage assemblages include the flakes successfully produced in these size ranges as well as all others produced during a reduction event.
The fundamental issue to be considered in developing a design fcr experimental data collection is the problem of confounding variables, or those uncontrolled factors that might influence the outcome of the study in undesirable ways.
Variability in any of these factors will affect the results a•’ the analyses and my result in unwarranted conclusions. For example, excess random variation will produce unacceptably low reliability and this will mask variability of interest. Systematic variation due to extraneous variables can likewise severely affect the validity of experimental results. More specifically, in this study, bias could come from uncontrolled variation in knapper behavior, core/blank morphology and hammer or pressure-•’laker morphology. Bias in knapper behavior could come from differences in lighting, knapper ability, mood, strength, and health. Bins in core/blank morphology could come from variation in size, shape and edge morphology. Finally, bias in hammer and pressure flaker morphology could come from size/weight, shape, hardness and condition. To cope with the problem of potential confounding variables, a randomization process was embedded into the experimental design. To combat potential variation in knapper behavior, reduction time was randomized such that each reduction event was conducted at a randomly chosen time during the day. All reduction was accomplished over a period of two weeks by a single knapper (myself). Potential influence of core/blank morphology was controlled by using similar sized core/blanks for the production of each analytical assemblage (Appendix A). Next, l each analytical assemblage was produced from the reduction of three core/blanks. To accomplish this, the total number of core blanks for each reduction class (total of 9 in the
9 3 validity analysis and 1 in the reliability analysis) were randomly assigned to a reduction event such that nc judyement was involved in the choice of a core/blank for p-articular reduction events (Appendix B). Different combinations of three hammers or pressure-flakers were used in :he production of each analytical assemblage to reduce potential error from uncontrolled variation in hamrner/pressuxe-flaken morphology. Hard hammers included three large hammer-stones for extra-large flake production (one granite, one slightly eroded quartzite and one high quality quartzite). Hard hammers •’or large and medium flake production were slightly smaller and included one slightly eroded andesite, one granite and one quartzite. Soft hammers included one caribou antler shaft, slightly damaged on the ends; one moose antler tine with very little modification from use, but ground intg a rounded shape at its working end; and one hardwood billet with little end modification. Pressure-flakers included one hafted small deer antler tine, one unhafted elk antler tine and one hafted copper wire. Each hammer or pressure flaker was randomly assigned to an appropriate corehlank for each reduction event
(Appendix B). In this way, bias could not enter into the experimental process through preferential association of specific hammers and pressure-flakers with particular
The goal of each reduction event was to produce approximately 40 flakes following seiving through a 1/4 inch
94 mesh screen (thus, each core was not reduced to exhaustion). Using estimates on the relationship between individual blows
to a nucleus and the number of flakes actually removed (from
Amick and Mauldin I19881, Callahan [19793 and Magne [19851), predictions were made regarding the number of actual blows to a core/blank it would take to produce the requisite number of 40 flakes. These are as follows:
All pressure flaking = 25 pressure events Flakes = 25 blows Bifaces = 10 blows Prepared and Unprepared Cores = 7 blows
In the case of bifaces and prepared cores, edge shaping and platform preparation flake removals were included in the debitage assemblages to be analyzed but were not counted as actual blows. To summarize, fur the reliability experiment, 30 biface production debitage assemblages were collapsed into 10 analytical assemblages by choosing groups of three reduction assemblages for each analytical assemblage. Reduction was accomplished using soft hammers aimed at the production of medium sized flakes. Platform abrasion was accomplished using a small eroded quartzite hard-hammer. Blanks were relatively small stage 2 (Callahan 1979) bifaces (Appendix A). The goal here was to introduce as much consistency as possible across the ten analytical assemblages and to randomize the potential confounding variables, such that the tests for reliability focus only on the instruments free of bias that would be
9 5 introduced by confounding variables. In the validity experiment, 60 debitage assemblages were collapsed to form 20 analytical assemblages. Reduction was accomplished using a wide range of soft and hard hammers.
Core/blank size varied with flake size goal. Larye cores were uu& to produce extra-large flake assemblages. Medium sized cores were used for large flake assemblages and small cores were used to produce medium flake asssemblages (Appendix A). Likewise, larger bifaces were used to produce large flake assemblages, while smaller bifaces were used to produce medium flake assemblages. Since flake reduction centered on medium (actual results were often small, however) flake assemblages, flake blank size remained constant (Appendix A). The goal of this program of lithic reduction was to introduce a high degree of variability in order to test the ability of the instruments to identify that variability. Randomizing reduction sequences and blanks and hammers was carried out to reduce the possibility of systematic error entering into the data set producing problems with the validity solutions. Following the production of these assemblages, flakes were sorted into each of the Sullivan and Rozen flake types (Sullivan 1987). Within each of these groups, flakes were also visually subdivided into the MSRT size classes. Maximum length, width and thickness measurements were made with calipers. Maximum thickness was calculated at a location of maximum thickness off the striking platform. Maximum length extended from the striking platform to the distal tip. Total flake length was preferred over measurements of point of fracture initiation to distal tip (Arnick and Mauldin 1989), as
I am interested in potential flake utility, as opposed to purely technological problems. Maximum width was measured anywhere it occurred on the lateral margins (transverse to the axis of force) of a flake (including the platform). Edge length measurements depended first on the characterization of edge angle. I used a hand-held Ward's Contact Goniometer to estimate edge angle. I use the term edge angle for convenience, but more precisely refer to spine plane angle or what Knudson (in Hayden 1379) refers to as production angle.
Following Odell (1979), I calculated all edge angles at a point 2 mm. from the actual edge of the flake. Edge lengths for edge angles less than 45 degrees were combined on each flake to form AAEL scoreu. Edge lengths for edges of greater than 45 degrees were combined to form HAEL scores. Actual edge lengths were measured with the same calipers on a point to point basis such that small undulations in edge shape were collapsed into single measurements. This has minimized time expenditure and although some random error has been undoubtedly introduced, I consider it to be minimal.
DATA
In this section, I review the initial results of the experiments. This will provide an initial assessment of these data for their applicability to the statistical analysis of
97 reliability and validity. It will also provide an initial assessment of their respective distributions and their
potential effects on the reliability and validity analysis. This section serves as an introduction to the data sets used to measure reliability and validity. It does not constitute the actual analysis of reliability and validity.
I RELIABILITY STUDY DATA
All data compiled for the reliability assessment of the
Sullivan and Rozen typology, the Modified Sullivan and Rozen Typology, the Flake Volume Index, the Acute Angle Edge Length Index and the Kigh Angle Edge Length Index are presented in Tables 1-5. For each instrument I review descr:.ptive statistics (from raw data anly) as well as actual data distributions. I focus on the identification of variability
which might adversely affect the rellahllity a~~?s~ii~ct-~t.80th raw and rescaled data are reviewed (SRT and MSRT). Raw data
were rescaled to facilitate multivariate analysis. This was accomplished by converting the category with the highest number of flakes to a score of 100 and scaling all others in relation this (Binford 1981; Draper 1985). This is done to avoid the effects of differences in numbers of flakes in the multivariate analysis. It also makes visual comparisons between data sets easier. I consider this point further in the actual analyses of reliability and validity. Sullivan and Rozen Typology
Data values for the flakes falling within the SRT types reveal a distribution containing numerous medial/distal and proximal fragments and very few complete or split flakes
(Tables 1 and 6). No nonorientable fragments were produced. This is consistent with Sullivan and Rozen's (1985) prediction that a high rate of flake breakage can be expected with toc>l production, thereby reducing the complete flake count proportionately. The variance/mean ratio was calculated to indicate the degree of clustering versus uniform dispersal across the ten analytical assemblages, The results (Table 6) indicate relative uniformity for all categories, with the exception of the medial/distal fragments, with a score of 4.65 indicating a high degree of clustering. This category contains the highest number of flakes. Clustering is expected here as breakage will vary to some degree depending on flake size and thickness. Some variation across analytical assemblages may therefore be expected given random variation in flake removal success. The medial/distal category is not considered in the multivariate analysis, since its converted value is always 100.0, which has a variance of 0. As its raw data variance is the highest, it would have been instructive to consider it in the multivariate analysis. ~ortunately,the rescaling process scales all other data categories in relation to variability in Table 1. Sullivan and Rozen Typology reliability analytical assemblage data from biface reduction iCF=Complete Flake, PP=Proximal Fragment, MDF=Medial/Distal Fragment, NF=Nonorientable Fragment, SF= Split Flake, 1-lO=flake counts from cases, Resc.=rescaled flake counts).
1 Resc. 2 Resc. 3 Resc. 4 Resc. 5 Resc.
CF 4 10.3 7 10.8 1 1.3 3 4-5 3 3.5 PF 28 71.8 23 35.4 19 24.4 28 42.4 29 34.1 MDF 39 100.0 65 100.0 78 100.0 66 100.0 85 100.0 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 2 5.1 5 7.7 3 3.8 3 4.5 3 3.5 Total 73 100 101 100 120
6 Resc. 7 Resc. 8 Resc. 9 Resc. 10 Resc.
CF 3 3.4 1 1.3 2 2.1 1 1.2 1 1.1 PF 21 24.4 20 25.3 22 22.7 29 34.5 20 21.1 MDF 86 100.0 79 100.0 97 100.0 84 100.0 95 100.0 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 3 3.5 2 2.5 6 6.2 2 2.4 3 3.2
113 102 127 116 119 Table 2. Modified Sullivan and Rozen Typology reliability analytical assemblage data from biface reduction (CF=Complete Flake, PF=Proxhmal Fragment, MDF=Medial/Distal Fragment, NF=Nonorientable Fragment, SF=Split Flake, 1-lO=cases [flake countsl, Resc.=rescaled flake counts).
1 Ress. 2 Resc. 3 Resc. 4 Resc. 5 Resc. Large CF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 PF 1 3.1 2 3.7 0 1.0 1 2.0 0 1.0 MDF 1 3.1 0 1.0 0 1.0 1 2.0 1 1.4 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 Med i urn CF 2 6.3 3 5.6 0 1.0 1 2.0 0 1.0 PF 10 31.3 10 18.6 5 8.2 6 12.5 14 13.7 MDF 6 18.8 11 20.3 17 27.9 17 35.4 13 18.3 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 0 1.0 1 1.9 0 0.0 1 2.0 2 2.8 Small CF 2 6.3 4 7.4 1 1.7 2 4.2 3 4.2 PF 17 53.1 11 20.4 14 23.0 21 21.1 15 21.1 MDF 32 100.0 54 100.0 61 100.0 48 100.0 71 100.0 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 6.3 4 7.4 3 5.0 2 1.4 1 1.4 ...... 2 Total 73 100 101 100 120 Table 2. Contnd.
6 Resc. 7 Resc- 8 Rex. 9 ReSG. 10 Resc.
Large CF 0 0.0 0 0.0 0 0.0 0 0.0 0 8.0 PF 2 2.8 1 1.7 1 1.2 1 1.5 1 1.2 MDF 0 1.0 0 1.0 0 1.0 2 2.9 1 1.2 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 0 0.0 0 0.0 0 0.0 0 0.0 0 8.0 Med i urn CF 1 1.4 0 1.0 0 1.0 0 1.0 0 1.0 PF 5 7.2 5 8.3 3 3.5 7 10.3 6 7.4 MDF 17 24,6 19 31.7 12 14.1 14 20.6 13 16.0 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 1 1.4 1 1.7 3 3.5 1 1.5 0 1.0 Small CF 2 2.8 1 1.7 2 2.3 1 1.5 9 1.2 PF 14 20.3 14 23.3 18 21.2 21 30.9 13 16.0 HDF 69 100.0 60 100.0 85 100.0 68 100.0 81 100.0 NF 0 0.0 0 0.0 0 0.0 0 0.0 0 0.0 SF 2 2.8 1.7 ...... 1 3 3.5 1 1.5 3 3.7 Total 113 102 127 116 119
Table 4. Acute angle edge length reliability data (Data from analytical assemblages [l-POI; CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
Cases 1 2 3 4 5 6 7 8 9 10 Large PF 82.7 99.1 0.0 90.5 0.0 78.8 97.8 112.3 120.7 87.3 MDF 77.1 0.0 0.0 99.7 78.5 0.0 0.0 0.0 91.7 0.0 Med i um CF 64.5 60.1 0.0 72.8 0.0 76.4 0-0 0.0 0.0 0.0 PF 34.2 45.4 50.3 39.6 47.4 25.0 37.1 34.4 32.7 40.1 MDF 47.9 34.1 29.9 36.7 47.5 42.4 29.9 32.4 36.6 27.3 SF 0.0 56.8 0.0 15.3 31.0 21.6 0.0 21.6 14.8 0.0 Small CF 24.4 25.1 33.5 22.2 29.9 36.8 32.4 20.8 11.7 30.6 PF 16.2 14.2 16.3 15.0 12.0 9.0 13.8 13.9 14.3 11.2 MDlF 14.6 20.1 13.1 16.8 15.2 13.0 10.6 11.1 12.0 11.7 SF 18.7 16.6 5.1 20.9 22.5 6.9 17.3 19.0 0.0 12.6 Table 5. High angle edge length reliability data (Data from analytical assemblages [ 1-101; CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake). Cases 1. 2 3 4 5 6 7 8 9 10 ...... Large PF 19.6 99.1 0.0 18.6 0,O 27.3 59.1 34.3 24.5 35.1 MDF 46.4 0.0 0.0 34.5 53.8 0.0 0.0 0.0 39.1 49.5 Med ium CF 17.0 3.6 0.0 11;2 0.0 0.0 0,O 0.0 0.0 0.0 PI? 37.7 23.7 46.2 35.2 29.7 28.6 36.0 31.5 33.9 25.9 MDF 32.0 48.4 46.7 38.8 35.8 32.9 49.2 42.0 47.5 40.2 SF 0.0 25.6 0.0 46.5 37.9 28.4 55.0 36.6 40.7 0.0 Small CF 0.0 5.3 5.8 0.0 0.0 6.1 0.0 11.7 7.3 0.0 PF 14.3 21.5 18.4 14.4 18.6 19.4 17.8 18.0 19.7 18.7 MDF 20.9 23.5 23.8 20.1 23.6 22.3 26.3 26.4 28.4 25.2 SF 20.6 22.6 22.2 12.4 28.5 40.4 27,O 14.5 42.5 16.0 Table 6. Reliability summary statistics (CF-Complete Flake, PF=Proximal Fragment, MDF=MediaP/Distal Fragment, SF=Split Flake; CV=Coef f icient of variation; means derived from flake analytical assemblage flake counts [SRT and MSRTI and analytical assemblage utility index data [FVI, AAEL, HAELI). Modified Sullivan an3 Rozen Typology Flake Volume Index Variance/ Mean Variance Mean Ratio Mean Variance CV
Large PF 1.0 .44 .44 MDF .6 .49 .81 Med i urn CF .7 1.12 1.60 PF 7.1 10.77 1.52 26.8 24.65 18.5 MDF 13.9 14.54 1.05 20.9 32.90 27.4 SF 1.0 .89 .89 Small CF 1.9 3.60 1.89 2.9 2.91 58.8 PF 15.8 16.99 1.08 4.0 1.35 29.0 MDF 62.9 292.27 4.65 2.4 .67 34.1 SF 2.2 1.73 .79 3.0 2.70 54.8 ......
Acute Angle Edge Length High Angle Edge Length Mean Variance CV Mean Variance CV Medi urn CF PF 38.6 58.10 19.7 32.2 39.36 19.5 MDF 36.5 53.35 20.0 41.4 42.22 15.7 SF Small CF 24.7 115.08 43.0 3.6 17.56 116.4 PF 12.6 14.70 30.4 17.1 15.15 22.8 MDF 13.8 8.62 21.0 24.1 6.67 10.7 SF 14.0 56.91 53.9 24.7 104.62 41.4
Sullivan and Rozen Typology Variance/ Mean Variance Mean Ratio ...... CF 2.6 3.60 1.38 FF 23.9 16.39 .71 MDF 77.4 232.27 3.78 SF 3.2 1.73 -54 the highest scoring category (medial/distal fragments). For example, in two assemblages, one with 90 medial/distal
fragments and 16 proximal fragments and the other with 40
rnedial/distal fragments and 16 proximal fragments, the
rescaled values of the proximal fragments would be 17.78 and
40.0 respectively, reflecting the variability in medialldistal
fragment scores. Thus, the variability lost by excluding the
medial/distal fragments is reflected in the rescaled complete
flake, proximal fragment and split flake data. I consider the
rescaled data therefore to be an accurate reflection of the
actual performance of the typology. It is important to study
the reliability of the SRT using the converted data as it is
in this form that the validity analysis is also conducted.
Were, sample size differences are somewhat more extreme.
Modified Sullivan and Rozen Typology
BiEace reduction accomplished for the reliability analysis produced a distribution of size classes including small, medium and large flakes (Table 2). No extra-large
flakes were produced. Mean values indicate similar proportions cf high medial/distal fragments, reduced proximal
fragments and low numbers of complete and split flakes in both the med'ium and small size classes (Table 6). Very few large proximal or medial/d istal fragments were produced. No large complete or split flakes were produced. Complete flakes are more consistently produced in the small class than in the medium class. This may be the result of edge shaping and platform preparation resulting more often in small thicker flakes which are less susceptible to breakage during production. Large flakes and most medium flakes are typically broken to some degre?. These flakes are primarilly biface thinning flakes which are thin and more susceptible to breakage (Sullivan and Rozen 1985).
variance/rnean ratio scores (Table 6) indicate relatively little clustering in all flake categories except the small medial/distal category, where clustering is substantial. As was the case with the SRT distribution, this appears to be due to the large sample size and the fragmentary nature of the flakes in this category.
As in the case of the SRT reliability analysis, the small medial/distal category was dropped from the multivariate analysis of MSRT reliability. Follcwing the arguments already presented, I anticipate no problems.
Flake Volume Index
Mean FVI scores were calculated from only those MSRT flake categories where all analytical assemblages contained at least one flake (Tables 3 and 6). Using this criterion, six categories were chcsen for this analysis (Table 3) os well as in all other utility index reliability studies. This is justifiakle since it includes the primary contributors to the makeup of the assemblages.
108 Medium flake types contain by far the highest FVI scores,
while small categories score low. Both small and medillin
proximal fragments score respectively higher than small and medium medial/distal fragments. This is to be expected as proximal fragments are typically somewhat larger than
medial/distal fragments, which come in a wider range of sizes. The low score in the small split flake category may be
indicative of the fact that smaller flakes are less likely to
be broken during flake removal during biface reduction.
Variances (Table 6) indicate some degree of variation in
the medium categories. Coefficients of variation (Table 6)
indicate that t:,is index tends to be relatively more variable,
however, in the small size class. This may be due to the fact
that small flakes are byproducts of a flaking process focussed
more towards the production of larger flakes. In general,
variability appears to be low.
Acute Angle Edge Length
Mean AAEL scores for the six MSRT flake categories
indicate high edge lengths with low edge angles in the medium
categories as well as in the small complete flake category
(Tables 4 and 6). Small proximal and medial/distal fragments and split flakes have relatively low AAEL scores. Larger
flakes and complete flakes have comparatively longer edges
with low edge angles in biface reduction debitage assemblages,
than small f raqmentary flakes. Q) a, 0-J - aJ 0 ul I: 4 m Y. L: c a, 4J -4 a u 4 Y A 2L a,CA .r) TCi CaJclrr , Q) w 3 cdw .SG~ @mE 10 .-I CU 4 t(] &aa'l dD4J ul Q, x a H a ~c wrn.-~aa,tn rc w AzaJWa, a c 0 MAU5 O.CclCIJZC, *dC truaca.ncu .d F m C' u a) 03 rtr U rnwa LJL-r'drCal Q)d)cm 4JaI-i 4 E: Q,CUC-lLlW4-'0" ~rdrn~aal2 a .-ICnoY-Law U C x 24 .-ISU3 u-4 rd u L, moklrl a, c' a o .r( +r-tu a c aa~r-c 3U-U W .Q ak3.d fl LC@ 12 C3.ALlOM L2G d 3d tnD3 oo+'mtnuxx 5 wommc .d r( al (URIaC .c( 02 -4 .cI C, clhaIUB E WrtlaUUl mn!cu~la&~~t3 cd lu Mrl)go .d b4LlUdsd,f2rn coOwuf5E k 'J-J 0 rn 0 XI P 0 rn C OUf-i..r( aJu?! CW-4 tn*>.-ir:mmu XLls=m@am,n m -4 4, Q, m rsj darnur cUa,uI=r,maxrlInr 03 C, IOM 4rlC,C.~OdUl 4.1 a) m ~.~rlomom.~aa?a~a cu 0 TJWWUICUU~~U\~5rh.1 .GEE!\ a, rtl 4 G~UIEPC)~ C, .ddY htJl 5 hodm.4 rdxm~~a)aa,cuu~r(cmato LI O.P(LCX Q, m.4 o map -a,ro 03krdC, @US@ mr6Q,d WC,~ ~GW--Q,O E a,rc~rldd m z m k Q, CDL,ww XCCI ~uEaax3 fa orcwmrcl C: 11;1 .d dc,OOa,Uri rla,E~U !sa.d~.-rL.(AC-( w 6, a, -4 rc cv ~LJP am2BU mx~ .rl 2 4 a- .c(+)Wrnoc,nfo 9 iLna cc.4 tr, r-r Q) h c=am --EcUUQ)cUdC.Qt(] XU>c.cd,G(=(I)$ +).~.c~a0 0 a tsr Clr U.d(O o~C)Tdv).d.cIG a a .r( .ri ~3 L,-taw u urdu ECCCdUlCf5UUl a W '(3U.G --i m C~~dCQ,@OOwCl+JC atn~ a~rtlxfiu rma m c?mr=m ca,m@ r: m a, (13 c W m 0 tr r-( rl OrlaJUdL, 4 v, r( .dac w cr ..4ww.c, =CO,rn-r*L,OJ.d9 atnu a,b)urn rl a, a, u m Usaariurla, QL.(~)-c3.GC, .r( Mad rrjl$)-c mo oatrc rlaouawmo EcUVlUa.dO QIGUXEO-~UU --IIlrrjlUlNCU V]~(~UIP)V]UPIW~ that complete and split flakes, which have substantial lengths
of unbroken edges, often with acute edge angles, vary the
greatest when examined with the HAEL index.
VALIDITY STUDY DATA
I review now the raw (FVI, AAEL, HAEL) and rescaled (SRT and MSRT) data produced for the validity analysis (Tables 7-13). I focus primarily on the recognition and explanation of variability in these results in anticipation of more detailed consideration in the statistical analyses of validity. This discusion does not constitute the actual analysis of validity.
Sullivan and Rozen Typology
11iitial Inspection of these data (Tables 7 and 8) reveals a striking similarity across most of the experimental data cases. With few exceptions medial/distal fragments score highest, followed by proximal fragments. Proximal fragments are most common in three of four pressure flake assemblages. In only one case (#17, unprepared core) do proximal fragments eclipse medial/distal fragments as the most common elements in the assemblage. Modzrate numbers of complete iiakev occur with some consistency. Hard hammer reduction of prepared platforms appears to be related to slight increases in complete flakes. Comparatively higher numbers of complete
111 Table 7, Sullivan and Rozen Typology validity analysis analytical assemblage data (Data represent flake counts from analytical assemblages [ 1-20 I ; CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, NF=Nonorientable Fragment, SF= Split Flake 1.
Cases HH HH HH SH HH SH HH SH HH SH PC UPC BF BF UPC UPC PC PC FL FL 1 2 3 4 5 6 7 9 10 ------8 CF 8 18 19 5 11 4 22 8 43 20 PF 15 28 22 22 14 14 34 21 3 12 MDF 134 94 114 114 74 93 127 106 39 38 NF 19 14 12 5 8 9 13 15 0 1 SF 13 10 13 7 9 21 7 14 14 4
Cases PR HH SH PR HH SH PR HH SH PR FL BF BF BF UPC UPC UPC PC PC PC 11 12 13 14 15 16 17 18 19 20 CF 10 20 6 4 10 9 9 13 17 2.2 PF 23 12 11 10 14 fO 36 1'9 18 22 MDF 29 38 45 34 79 78 32 55 71 30 NF 0 1 0 0 13 17 1 12 0 1 SF 19 4 6 I0 1Q 11 9 4 7 20 Table 8. Rescaled Sullivan and Rozen Typology validity analysis data (Data from analytical assemblages [I-20%; CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, NF=Nonorientable Fragment, SF= Split Flake, HH=Hard Hammer, SH= Soft Hammer, PR=Pressure, PC= Prepared Core, UPC=Unprepared Core, BF=Biface, FL=Flake Tool 1. Cases HH HH HH SH HH SW HH SH HH PC UPC BF BF UPC UPC PC PC FL _____--_------1 2 3 4 5 6 7 8 9 CF 5.9 19.3 16.7 4.4 14.9 4.3 17.3 7.5 180.0 PF 11.2 29.8 19.3 19.3 18.9 15.1 26.8 19.8 7.0 MDF 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 90.7 NF 14.2 14.9 10-5 4.4 10.8 9.7 10.2 14-2 1.0 SF ...... 9.7 10.6 11.4 6.1 12-2 22.6 5.5 13.2 32.6 Cases SH PR HH SH FR HH SH PR HH FL FL BF BF BF UPC UPC UPC PC ...... 10 11 12 13 14 15 16 17 18 CF 21.7 34.5 52.6 13.3 11.8 12.7 11.5 25.0 34.5 PF 39.1 79.3 31.6 24.4 29.4 17.7 12.8100.0 30.9 MDF 100.0 100.0 100.0 100.0 100.0 100.0 100.0 88.9 100.0 NF 1.0 1.0 2.6 1.0 1.0 16.5 21.8 2.8 21.8 SF ------10.9 65.5 10.5 13.3 29.4 12.7 14.1 25.0 7.3
CF 23.9 40.0 PF 25.4 73.3 MDF 100.0 100.0 NF 1.0 3.3 SF ------9.9 66.7 Table 9, Modified Sullivan and Roaen Typology validity analysis data (Data represent flake counts from analytical assemfllages [ 1-20 I ; CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Dfstal Fragment, NF=Nonorientable Fragment, SF= Split Flake, HH=Hard Hammer, SH= Soft Hammer, PR=Pressure, PC=Prepared Core, UPC=Unprepared Core, BF=Biface, FL=Flake Tool). Cases HH HH HH SH HH SH HH SH HH SH PC UPC BF BF UPC UFC PC PC FL FL 1 2 3 4 5 6 7 8 9 10 Extra Large CF 4 PF 2 MDF 0 NF 0 SF 2 Large CF 2 PF 7 MBF 10 N? 0 SF 7 Med i urn CF 2 PF 3 HDF 31 NF 0 SF 2 Small CF 0 PF 3 MDF 93 NF 19 SF 4 Table 9. Continued.
cases PR HH SH PR HH SH PR HH SH PR FL BF BF BF UPC UPC UPC PC PC PC 11------12 13 14 15 16 17 18 19 20 Extra Large CF 0 0 0 0 0 0 0 0 0 0 PF 0 0 0 0 0 0 0 0 0 0 MDF 0 0 0 0 0 0 0 0 0 0 NF 0 0 0 0 0 0 0 0 0 0 SF 0 0 0 0 0 0 0 0 0 0 Large CF 0 1 0 0 3 3 0 4 1 0 PF 0 0 0 0 4 0 0 0 0 0 MDF 0 0 0 0 2 1 0 1 1 0 NF 0 0 0 0 0 0 0 0 0 0 SF 0 0 0 0 1 2 0 0 0 0 Pled i urn CF 0 7 3 0 4 3 0 7 6 0 PF 0 2 1 0 4 3 0 10 3 0 MDF 0 5 6 0 14 10 0 6 24 0 NF 0 0 0 0 1 0 0 0 0 0 SF o o a o 3 2 o 1 1 o Small CF 10 12 3 4 3 3 9 8 10 12 PF 23 10 10 10 6 7 36 6 15 22 MDF 19 33 39 34 63 67 32 49 47 30 NF 0 1 0 0 12 17 1 12 0 1 SF 19 10 20 ------4 6 6 9 9 3 6 Table 10. Rescaled Sullivan and Rozen Typology validity analysis data (Data represent rescaled flake counts from analytical assemblages 11-201; CF=Complete Flake, PF=Proximal Fragment, MDF=Medfal/Distaf Fragment, NFlNonorientable Fragment, SF= Split Flake, HH=Hard Hammer, SH=Soft Hammer, PR=Pressure, PC=Prepared Core, UPC=Unprepared Core, BF=Biface, FL=Flake Tool).
Cases HH HH HH SH HH SH HH SH HH PC UPC BF BF UPC UPC PC PC FL 1 2 3 4 5 6 7 8 9 Extra Large CF 4.3 15.9 1.0 1.0 1.8 1.0 1.0 1.3 PF 2.2 1.0 1.0 1.0 1.0 1.0 1.0 1.0 SF 2.2 1.0 1.0 1.0 1.0 1.0 1.0 1.0 Large CF 2.2 2.9 1.1 1.2 12.5 6.3 10.7 5.2 PF 7.5 4.3 2.2 1.2 17.9 12.7 2.9 2.6 MDF 10.8 11.6 2.2 9.6 10.7 12.7 5.8 9.1 SF 7.5 4.3 2.2 1.0 7.1 3.2 1.0 2.6 Med i urn CF 2.2 2.9 3.3 1.0 1.0 1.0 1.0 1.3 PF 3.2 5.8 11.0 6.0 1.8 4.8 2.9 9.1 MDF 33.3 24.6 23.1 27.7 21.4 34.9 17.5 28.6 NF 1.0 1.4 1.1 1.0 1.8 1.6 1.0 1.3 SF 2.2 2.9 3.3 3.6 3.6 4.8 6.8 2.6 Small CF 1.0 4.3 15.4 4.8 3.6 1.0 9.7 3.9 100.0 PF 3.2 30.4 11.0 19.3 5.4 4.8 27.2 15.6 7.5 MDF 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 97.5 NF 20.4 18.8 12.1 6.0 12.5 12.7 12.6 18.2 1.0 SF ------4.3 8.7 3.8 4.8 5.4 25.4 6.8 13.0 3.2.5 Table 10. Continued
Cases SH PR HH SH PR HH SH PR HH FL FL BF BF BF UPC UPC UPC PC 10 11 12 13 14 15 16 17 18 Extra Large CF 1.0 1.0 1.0 1.0 1.0 1.8 1.0 1.0 1.0 PF 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 SF 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 Large CF 1.0 1.0 3.0 1.0 1.0 4.8 4.5 1.0 8.2 PF 1.0 1.0 1.0 1.0 1.0 6.3 1.0 1.0 1.0 MDF 1.0 1.0 1.0 1.0 1.0 3.2 1.5 1.0 2.4 SF 1.0 1.0 1.0 1.0 1.0 16 3.0 1.0 1.0 Med i urn CF 2.3 1.0 21.2 7.7 1.0 6.3 4.5 1.0 14.3 PF 4.5 1.0 6.1 2.6 1.0 6.3 4.5 1.0 20.4 MBF 4.5 1.0 15.2 15.4 1.0 22.2 14.9 1.0 12.2 NF 1.0 1.0 1.0 1.0 1.0 1.6 1.0 1.0 1.0 SF 1.0 1.0 1.0 1.0 1-0 4.8 3.0 1.0 2.0 Small CF 22.4 34.5 36.4 7.7 11.8 4.8 4.5 25.0 16.3 PF 36.4 79.3 30.3 25.6 29.4 9.5 10.4100.0 12.2 MDF 100.0 100.0 100.0 100.0 180.0 100.0 100.0 88.9 100.0 NF 1.0 1.0 3.0 1.0 1.0 19.0 25.4 2.8 24.5 SF ------11.4 65.5 12.1 15.4 29.4 9.5 13.4 25.0 6.1 Table 10. Continued
Extra Large CF PF SF Large CF PF MDF SF Med i um CF PF MDF NF SF Sma 11 CF PF MDF NF SF Table 11. Flake Volume Index validity analysis data (Data represent mean FVI scores derived from analytical assemblages [I-201; CF=Complete flake, PF=Proximal Fragment, MDF=Medial/Diutal Fragment, NF=Monor ientable Fragment, SF= Split Flake, HH=Hard Hammer, SH=Sof t Hammer, PR=Pressure, PC=Pregared Core, UPC=Unprepared Core, BF=Biface, FL=Flake Tool).
Cases HH HH HH SH HH SH HH PC UPC BF BF UPC UPC PC P 2 3 4 5 6 7 Extra Large CF 4529.5 3434.2 1456.7 2343.7 PF 5013.4 SF 2160.7 Large CF 609.3 746.8 140.0 99.2 970.1 724.5 701.7 PF 801.6 598.1 129.5 110.8 456.6 568.4 369.7 MDF 288.1 614.8 102.3 139.2 390.2 159.9 302.8 SF 571.6 1062.9 113.4 332.3 41.9 Med i urn CF 34.6 97.9 40.1 59.1 PF 123.0 98.8 48.0 22.8 80.2 35.2 181.4 MDF 70.5 35.2 33.5 31.9 45.6 26.5 43.5 NF 126..2 47.3 5.3 5.1 SF 48.4 10.1 37.6 23.3 42.5 34.8 52.6 Sma 1l CF 3.2 2.7 1.4 2.3 5.5 PF 1.9 4.2 4.9 4.3 3.5 1.7 5.5 MDF 4.3 3.4 4.0 3.3 5.3 4.8 5.3 NF 5.3 9.8 3.6 3.5 9.4 6.5 9.2 SF 6.4 2.1 7.1 2.5 3.7 5,1 5.7 Table 11. Continued
cases HH SH PR HH SH PR PC FL FL FL BF BF BF 8 9 10 11 12 13 14 Extra Large CF 2062.6 PF SF Large CF 304.5 FF 67.1 MDF 126.0 SF 217.2 Med i urn CF 16.5 PF 34.4 MDF 19.7 NF 16.9 SF 36.2 1.0 Sma 11 CF 2.3 1.7 .9 .9 5.0 1.6 1.1 PF 6.4 1.7 1.6 1.0 3.4 4.2 1.1 MDF 4.2 1.5 4.3 -5 2.7 2.2 -8 NF 5.5 3.4 Table 11. Continued Cases HH SH PC HH UPC UPC UP@ PC 15 16 17 18 1 Extra Large CF PF SF Large CF 593.9 146.4 261.4 50.7 PF 228.7 MDF 306.8 27.6 45.9 42.6 SF 369.1 236.1 Med iurn CF 67.3 14.1 47.0 23.7 PF 56.7 19.3 20.8 24.5 MDF 40.6 17.3 20.0 18.6 NF 68.8 SF 88.8 16.4 5.3 8.5 Small CF 3.8 1.4 1.8 3.4 2.4 2.6 PF 3.5 2.2 1.9 3.7 5.0 2.2 MDF 3.1 2.9 1.6 2.4 3.2 1.2 PIF 3.1 4.0 .8 3.2 .2 3.3 1.6 1.6 SF ...... 3.5 1.1 5.4 Table 11. Continued
caueu HH SH PR HH SH PR UPC UPC UPC PC PC PC 15 16 17 18 19 20 Extra Large CF PF SF Large CF 593.9 146.4 261.4 50. 7 PF 228.7 MDF 306.8 27.6 45.9 42.6 SF 369.1 236.1 Med i urn CF 67.3 14.1 47.0 23.7 PF 56.7 19.3 20.8 24.5 MDF 40.6 17.3 20.0 18.6 NF 68.8 SF 88.8 16.4 5.3 8.5 Small CF 3.8 1.4 1.8 3.4 2.4 2.6 PF 3.5 2.2 1.9 3.7 5.0 2.2 MDF 3.1 2.9 1.6 2.4 3.2 1.2 NF 3.1 4.0 .8 3.2 .2 SF 3.5 3.3 1.1 5.4 1.6 1.6 Table 12. Acute Angle Edge Length validity analysis data (Data represen? mean AAEL scores derived from analytical sssembhges [I-20;; CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, NF=Nonorientable Fragment, SF=Split Flake, HH=Hard Hammer, SH=Sof t Hammer, PR=Pressure, PC=Prepared Coze, UPC=Unprepared Core, BF=Bi•’ace, FL=Flake Tool).
Cases HH HH HH SH HH SH HH PC UPC BF BF UPC UPC PC 1 2 3 4 5 6 7
Extra Large CF 12.0 20.4 0.0 80.6 PF 25.3 SF 0.0 Large CF 26.1 37.7 95.5 123.0 33.6 41.5 38.3 PF 18.9 14.5 51.3 55.6 38.0 36.9 25.1 MDF 15.9 7.6 72.0 63.0 25.8 50.0 28.1 SF 24.7 0.0 56.2 15.3 89.6 Med f urn CF 20.7 0.0 38.6 33.1 PF 31.5 29.4 36.9 26.1 40.5 5.9 17.5 MDF 12.1 22.6 28.0 36.2 14.6 21.6 21.9 NF 0.0 16.2 0.0 0.0 SF 27.5 21.4 55.3 10.6 12.0 2.7 21.8 Small CF 8.9 22.6 18.2 20.8 18.8 PF 10.7 8.4 20.8 10.3 8.4 9.5 10.2 MDF 9.4 7.2 11.8 12.1 9.2 13.7 11.9 NF .7 .8 11.6 5.6 0.0 1.7 3.2 SF 4.7 15.9 8.8 10.6 7.0 5.7 9.3 Table 12- Continued
cases SH HH SH PR HH SH PR PC FL FL FL BF BF BF 8 9 10 11 12 13 14 Extra Large CF PF SF Large CF PF MDF SF Medium CF PF MDF NF SF Small CF PF MDF NF SF Table 12. Continued
Cases HH SH PR HH SH PR UPC WPC UPC PC PC ec ...... 15 16 17 18 19 20 Extxa Large CF PF SF Large CF 36.3 93.6 80.5 56.0 PF 42.2 MDF 74.6 61.1 64.3 125.9 SF 0.0 76.0 Medium CF 38.3 63.7 66.5 70.7 PF 20.1 16.2 23.7 73.1 MDF 20.8 39.9 43.0 51.0 NF 0.0 SF 9.5 28.9 0.0 59.4 Small CF 15.1 40.5 17.7 21.6 30.5 24.7 PF 6.4 13.6 13.9 12.8 18.0 16.2 MDF 8.9 14.1 9 .5 14.7 17.2 9.4 NF 5.2 6.8 0.0 7.5 0.0 SF ...... 11.4 10.9 6.0 7.1 12.2 5.9 High Angle Edge Length validity analysis data (Data represent mean HAEL scores derived from analytical assemblages [I-201; CF=Complete Flake, FF=Pr~ximalFragment, MDF=Medial/Distal Fragment, NF=Nonorientable Fragment, SF=Split Flake, HH=Hard Hammex, SH=Soft Hammer, PR= Pressure, PC=Prepared Core, UPC=Unprepared Coxe, BF=Biface, FL=Flake Tool). Cases HH HH HH SH MH SH HH PC UPC BF BF UPC UPC PC ------1 2 3 4 5 6 7 Extra Large CF 220.7 235.6 230.2 110.5 PF 269.8 SF 233.3 Large CF 100.0 131.8 7.3 0.0 117.9 122.1 145.7 PF 124.4 136.9 50.1 70.8 86.6 99.9 91.6 MDF 143.3 170.2 44.8 70.1 112.4 102.0 115.8 SF 155.0 191.8 67.7 112.3 4.3 Medium CF 63.4 85.4 38.4 50.2 PF 49.1 59.5 30.7 42.0 55.9 40.4 77.0 MDF 77.1 60.8 46.8 42.9 57.7 47.4 58-4 NF 112 :2 75.8 60.0 62.4 SF 74.6 41.5 27.5 61.7 51.4 74.8 50.6 Small CF 22.0 6.2 3.7 6.5 12.5 PF 21.7 19.8 14.6 18.6 23.9 12.0 21.3 MDF 30.1 28.6 23.1 25.0 29.1 28.0 28.3 NF 40.9 35.9 23.6 28.3 40.7 35.8 22.9 SF 24.3 10.5 27.3 22.5 22.1 27.4 31.6 Table 13. Continued Cases SH HH SH PR HH SM PR PC FL FL FL BF BF BF 8 9 10 11 12 13 f 4 Extra Large CF PF SF Large CF PF MDP SF Pled 1urn CF PF MDF NF SF Small CF 9.4 4.1 5.6 4.1 3.3 5.8 9.2 PF 17.5 11.4 13.6 15.3 17.0 15.0 10.3 HDF 28.3 16.0 20.1 16 6 19.6 19.1 16.1 NF 22.7 24.6 SF 31.6 11.3 19.4 13.0 8.3 25.9 19.0 Table 13. Continued caYes HH SH PC WH SH PR UPC UPC UPC PC PC PC ------15 16 17 18 19 20 Extra Large CF PF SF Large CF 130.8 35.8 50.6 14.4 PF 81.8 MDF 95.7 67.6 73.0 22.4 SF 124.4 74.1 Med i urn CF 37.8 3.8 12.9 9.9 PF 62.1 41.5 32.2 16.9 MDF 57.9 51.2 34.3 27.5 NF 17.2 SF 59.7 44.3 50. 0 0.0 Small CF 17.6 7.4 4.8 5.5 2.3 5.7 PF 18.7 16.0 13-0 14.3 15.0 12.0 MDF 23.2 25.1 18.7 20.9 21.1 18.0 NF 29.1 29.8 22.8 26.3 20.0 SF 18.3 22.3 13.4 34.8 16.7 17.9 flakes in association with hard hammer reduction and prepared platforms are found in both core reduction and biface reduction. This is especially notable in the comparison between hard and soft hammer reduction of thin flake edges
(assemblages #9 and #I0 respectively). In the case of hard hammer flake edge reduction, complete flakes are the most frequent flake type produced. Production of nonorientable f rayments 13 relatively cons istent throughout the experimental data matrix, with few exceptions. No nonorientable fragments were produced during the reduction of a flake edge
(assemblages 9 through 11) and few to none were produced during medium flake oriented bi face reduct ion (assemblages 12 to 14). Split flakes are also quite consistent in their occurrence throughout the matrix. High numbers of split flakes, however, are narrowly associated with all pressure flaking assemblages (asuemblaqes 11, 14, 17, 20) and the hard hammer flake edge reduction assemblage (#9). Split flakes appear to be the result of a process similar to that described by Frison and Bradley (1980) as a radial fracture. Essentially, fractures initiate in several directions from a single impact by a hard indentor when the core is supported firmly by the knapper .
Modified Sullivan and Rozen Typology
I provide a description of the MSRT data matrices (Tables
9 and 10) by size class. All size classes are represented and unprepared, large flake, core reduction assemblages contain more proximal and medial/distal fragments than prepared, large flake, core assemblages. The same difference is found in the hard hammer, medium flake production core assemblages where unprepared cores produce few complete flakes and numerous proximal and medial/distal fragments compared with the numerous complete flakes and fewer proximal fragments found in prepared core assemblages. The difference between medium flake, prepared and unprepared cores is reduced in the soft hammer categories with very few flakes of any kind present. All assemblages, excepting the pressure flake assemblages contribute to the medium flake category. The primary source of variability across the matrix appears to be the result of flake size goal differences. All extra-large and large flake goal assemblages produce high numbers of medium sized flakes, each distributing in a similar fashion, with very high numbers of medial/distal fragments, reduced proximal fragments and split flakes and very few complete flakes or nonorientable fragments. The majority of these flakes are the result of breakage in the larger categories. They are generally not produced during platform preparation. Neither do they successfulPy produce large or extra-large flakes. Pt is not surprising that they follow this esnsbstent pattern of breakage. Patterning in the medium flake goal assemblages is variable depending on the presence of platform preparation and hammer type. Platform p~eparation,again, has the effect of producing assemblages with high numbers of complete flakes and reduced split flakes (c.f. medium flake goal biface reduction
[assemblages 12 and 131 and prepared core reduction assemblags
[assemblages 18 and 191). Proximal and medial/distal fragments seem to be variable and conditioned to a greater degree by hammer type. Hard hammer, again appears to have the effect of producing more proximal fragments. For example, hard hammer, prepared core (#18) and biface (#I21 assemblages contain numerous complete flakes. Each also contains more proximal fragments than in the case of soft hammer reduction
(assemblages 19 and 13, respectively). This is a clear indication of goal oriented behavior which does not show up in the medium size class of the large and extra-large flake goal assemblages. The small flake category is affected similarly to those described above. Variability is found in the large and extra-large flake production assemblages. Here, the presence of platform preparation is clearly noted. All assemblages with platform preparation contain high numbers of proximal fragments and increased numbers of complete flakes. These are the result of small blows of force aimed at removing small relatively thick flakes to shape platforms. A fair number of these flakes retain platforms and my be identified as proximal or complete. This differs from situations where no preparation has occurred, resulting in very few complete or proximal fragments. The same pattern is found in medium flake goal, core reduction assemblages. A major source of variation in the matrix are those assemblages resulting from pressure flaking and hard hammer reduction of a flake edge. In all but the latter case, no flakes larger than the small size category were produced. In the latter case, only a few were produced and the distribution of these is very distinctive. These assemblages are the result of a single process at a small size scale. Thus, only
one "ti*rM of flake breakage is produced. his is in contrast to situations such as prepared core reduction, where platform preparation and large flake removal promotes a three tiered situation of primary flakes or those which were intended to be
removed, residual f ragments of those primary flakes from poorly designed or executed flake-removals, and residue from from platform preparation. The single tiered assemblages are least likely to be masked by eliminating the size criterion as is done by the SRT.
Flake Volume Index
Only thirteen (assemblages 2, 3, 4, 5, 7, 10, 12, 13, 15,
16, 18 and 191 of the possible 20 assemblages produced enough data for the FVI validity analysis, due to missing values in many of the MSRT flake categories (Table 11). This matrix retains. enough reduction variability that validity analysis is still feasible. There are two major types of patterns in the raw data matrix. The first, is a consistent and not surprising reduction in index values from the extra-large to small flake categories. Typically, proximal fragments are larger than medial/distal fragments. Within the small flake category, nonorientable fragments and split flake scores are often large since these flake types are often relatively thick. This compensates for their small size, resulting in high FVI scores. Small complete flakes seem to fluctuate to some degree, which may be due to actual variability is size as well as thickness.
The second pattern of variation in the matrix appears to between assemblages. Core reduction assemblages typically produce more massive flakes than biface reduction assemblages. Likewise, hammer type appears to have some effect, as hard hammer produced core reduction assemblages seem to contain larger flakes than soft hammer assemblages.
The same thirteen cases were considered in the AAEL validity analysis as in the FVI analysis (Table 12). As was the case with the FVI, the AAEL is distributed along a flake size related vector. Commonly, larger flakes have large AAEL =cores. This is not always the case, however. Many medium complete and proximal fragments exceed their counterparts in the large size category. This is not the case in the i comparison between the small and medium flake categories. I Small split flakes are also quite comparable in AAEL scores to medial/distal and proximal fragments. Small nonorientable fragments, typically, have very low scores. Variation also exists across the different reduction assemblages. Hard and soft hammer biface reduction (assemblages 3, 4, 12 and 13) and prepared soft hammer core reduction (assemblages 8 and 19) appear to produce greater AAEL scores than all other reduction strategies. This type of variation is not standardized across flake size boundaries and substantial variability exists between size categories. For example, large flake oriented prepared core reduction from soft hammers produces high AAEL scores in the rnedi-um proximal and medial/distal categories (assemblage 8) in comparison to the hard hammer version of this reduction type (assemblage 71, yet these are comparable in small flake scores. On the other hand, hammer type does not appear to greatly affect large flake oriented biface assemblages (3 and 4), while it does appear to have some effect on medium flake oriented biface assemblages (12 and 13) .
High Angle Edge Length
The same cases and variables were used in the analysis of the HAEL as in the FVI and AAEL (Table 13). In many ways, the structure of the HAEL data distribution is similar to the AAEl distribution, although its content is exactly its opposite. Whereas the AAEL measured edge lengths for acute adges, the OAEL is concerned with edge lengths for obtuse edge angles. This includes snapped edges. In a similar fashion to the
AAEL, HAEL scores descend with reduction in overall flake
size. Large flake categories have larger flake scares than medium or small flake categories. Within the medium and small categories, nonopientable fragments contain the highest scores, due to heavy breakage. Otherwise scores within each size category are quite consistent. There appeaars to be clear variation between reduction strategies. Biface and soft hammer core reduction assemblages have consistently lower scores than those of hard hammer core reduction assemblages. Platform preparation appears to also have some effect. For example, prepared, medium flake, core
reduction assemblages (18 and 19) have scores which are far lower than those of their unprepared counterparts (assemblages
15 and 16).
RELIABILITY ANALYSIS
In this section, I present first, a theoretical and mathematical definition of reliability. This includes a discussion of the statistical techniques to be used in this analysis, I next present the reliability analyses of the SRT,
MSRT and the FVI, AAEL and HAEL indices. Corrections for attenuation are produced for each index and reliability retests are undertaken. Recommendations for applicability to
- the validity analysis are provided. ASSESSING RELIABILITY
To be reliable, measurements must be consistent. Thus, in a series of measurements, one would expect some degree of similarity between the first measurement and the second. The third and all other measurements should also be similar to the first two. A perfectly reliable measuring instrument will produce exactly the =am results with repeated appll@atic>tion the same subject (Nance and Ball 1986:461). An acceptably reliable instrument will produce nearly the same results with repeated applications to the same subject. Thus, if reliable, parallel measurements of the same phenomenon will provide the same or nearly the same results. Reliability, consistency and replicability are synonymous. I follow Nance and Ball
(1986:461) in defining a "score" as "the value of a datum arising from an observation on some variable of interestf1and a '*testw as "the experimental procedure employed in obtaining scores," No standard exists for reliability testing of lithic typlogies or technological indices. A number of lithics researchers have attempted to evaluate consistency of observations focussing primarily on edge angle measurements (Amick 1986; Burgess and Kvamme 1978; Dibble and Bernard 1980; Fish 1978) and use-wear analysis (Mcquire et al. 1982). None have applied any of the method and theory available on reliability measurement. Thus, with the notable exceptions of
Cowgill (1970) and Nance (19871, there is little literature in
137 lithic archaeology for dealing with problems of reliability assessment. In this study, I rely heavily on Nance (l987),
Nance and Ball (1986) and on literature originating in psychology and education (Armor 1974; Carmines and Zeller 1979; Cronbach 1951; Green and Carmines 1980; Schuessler
1971). Choosing an approach for assessing reliability depends on the measuring instrument to be assessed. In this case 1 am assessing the reliability of two flake typologies (SRT and
MSRT) designed to measure variation in flake size and shape that arises as a function of technological variation. The indices rely on a flake typology to measure variation in debitage across entire assemblages. Understanding archaeological variability, in both cases, is a multivariate problem, since the typologies and the utility indices depend upon variation in percentages of, or scores on, multiple flake types. Interpretation of the flake typologies and the indices depend on the probability, P, that a specific predictable distribution of flake type percentages or index scores will be produced from a given reduction episode (e.9. biface production, prepared core reduction). Successful production of that specific distribution depends on a high degree of control. over the lithic reduction process and raw material type (Cottereli and Kamminga 1987; Prentiss and Romanski 1983; Speth 1972). The probability that a specific recognizable SRT, MSRT, or utility index distribution will occur single reduction episode, thus, depends on the ratio of systematic control of reduction strategy and raw material (XI) to the total available reduction strategies and raw material variation (m). The probability of characteristic distribution production for each assemblage is P=xl/m. By characteristic, I am referring to production of a distinctive distribution resulting from a specific process, unlike those of a different process. when extended across a series of n debitage assemblages, the estimate P, for the probability of a dlstlnctive debltage dlstributlon, is derived as a composite those n debitage assemblages. This will not be an exact estimate of the true nature of the distinctive distribution due to the presence of random var iahi 1 i ty among sxpar lrnantal outctmea . The term "precis ion" is used to denote the amount of.sampling error present in this type of estimate. Nance and Ball (1986:462) argue that precision is generally stated as an absolute value called the standard error of the estimate. Precision is a statement of variability arising from the "repeated sampling of a population at a given sample size" (Nance and Ball 1986:462). Thus, estimates with small standard errors denote greater precision than those with more substantial error estimates. Since I am attempting to measure variation in a multivariate population, it is critical that I be able to segregate variation due to measurement error from true variability due to reduction strategy. Variability could be 139 expected to be introduced by several means (as discussed earlier) :
a. Variation in knapper behavior could produce variable results. Error could occur through idiosyncratic variation in knapper behavior caused by extraneous factors as lighting, temperature and knapper ability, mood and physical condition. b. Another factor producing variation could be variability in blank size, shape and edge configuration. Variability in experimental outcomes could be introduced by variation in these factors producing variation in flake type, size and shape distributions. c. Variation in hammer size, shape, hardness, weight and condition could have an effect on the shapes of distributions. d. Application of the typology or index could produce error, either through faults in these instruments or faults in application by the researcher. e. The final source of variability is variability caused by purposive shifts in reduction strategy and raw material use. This is the actual variability to be measured by the proposed instruments. It is potentially masked by error arising from factors a.-d. The potential effects of variation in factors a.-d. are randomized in the design of experiments. Potential error due to operator variability is partially controlled since all data have been gathered by one person (Prentiss) in good health, working with adequate lighting. The goal of this analysis is to determine if faults in the instruments will prevent the recognition of true variability (factor e.). The precision of the overall estimate derives from a combination of factors a,, b., c., d. and e. since it involves the total error in that estimate (Nance and ball 1986:463). The goal of reliability assessment is to separate out the actual (true) variability from error variability. The goal shifts from one aimed at discovering error as associated with' the concept of precision to one oriented towards finding the contribution of the actual phenomenon of interest. Nance and Ball argue:
"A reliability coefficient, then, is an estimate of true inter-subject variability, exclusive of that part of the varia~ilityin the observations that is due to (random) measurement...error . More precisely, a reliability coefficient measures the proportion of variance in observations that is due to the true differences arnonq the observational units themselves" (1986:463),
The observational units studied in this case are experimentally-produced debitage assemblages- To provide a more complete understanding of reliability assessment and to facilitate an explanation of the statistics to be used in this study, I shift now to a discussion of the formal definition of reliability focussing on classicaS test theory. Assuming that in all measurements, some degree of
I random error is present, any observation can be expressed as I
where y is an observed score on a given variable, t is a non-random "truew score and e is random error. True scores are unknowable quantities representing the actual values one
is attempting to measure (Green and Carmines 1980; Nance
1987). Random error produces deviations, either positive or
negative from the true value.
since reliability relates to 'a series of measurements,
then each of the values yl, tl and el will be variable. Thus,
a series of repeated variable measurements are best expressed
in terms of variances or VAR. If so, equation (1) may be
restated as
If errors are independent of true scores, then the covariance
term in (2) will equal zero so that
VAR(~)= VARlt) + VAR(e). (3)
In other words, the observed total variance equals the sum of
the true score and error variances. The reliability (ryy) of (y) as a measure of (t) is expressed as the ratio of true score variance to observed
variance :
Following Nance (1987:2511, this may be rewritten as an expression of error variance:
Equation (5) indicates that the reliability is 1.0 minus 'the ' proportion of the variance in the observed scores due to the presence of random error. Thus, reliability of clove to 1.0 indicates the ~resenceof minimal random error. Since true variance is a product of observed variance and reliability, it is possible to estimate the unobserved true variance:
Since this study relies on multiple attributes (or flake types) to provide information on a single phenomenon such as a lithic reduction strategy and, as suck, is multivariate in nature, it is important to restate the above equations in multivariate terms. In the multivariate case, multiple true scores are present indicating the behavior of different phenomena. Given a randomized series of observations:
Where :
Y = Matrix of observations
T = Matrix of true scores
E = matrix of errors of measurement
Assuming that ZCOV(T,E) equals zero, (i.e., that errors are
independent of true scores) equation (7) may be xestated in
variance terms where the variance of T is denoted as Ct, the
variance of Y as CO and a diagonal matrix of error variance a3 E2:
In this case, the observed and true score covariances are identical, but the observed variances are inflated by random .error (Greene and Carmines 1980:163). Thus, any observed
correlations will be attenuated (Cowgill 1970). Multivariate reliability can be expressed in terms similar to the above by considering a composite m as a linear combination of yl or observed values combined with a nonnull vector of weights (w) such that
m = Yw. The variance of m is
Where :
w = any nonnull p x 1 vector of weights
wtCo = the matrix of observations
wtCo = the matrix of true scores
This is the same as equation (3) except that here multivariate scores are calculated using matrices of values. Reliability is calculated in a similar fashion:
In terms of error variance, this translates to
This is the multivariate version of equation (5) which displays reliability as 1.0 minus the ratio of error variance to observed score variance. Here again, a high degree of reliability is indicated by a score of close to 1.0, f ndicating minimal random error (Greene and Carmines 1980 1 .
a series of loadings for each item (variable) informing the
researcher about the strength with which that item is
correlated with that component. The summary statistic for
each extracted component is termed an eigenvalue and is the
sum of the squared loadings for all variables on that
component. It is no surprise then that the components gradually decrease in importance, and in corresponding
eigenvalue, from the first to the last (Carmines and Zeller
1979:60).
The results of the principal components analysis can be
interpreted to assess reliability assuming that each experiment is attempting to measure a single phenomenon. With
this as a hypothesis, several aspects of the initial unrotated
factor matrix could be assessed. According to Carmines and
Zeller (1979:60), the first extracted factor should capture a
large portion of the total variance (>40%). Additional
factors should capture gradually decreasing amounts of that
variance. The majority of the items involved should load on
the first factor at a level of at least .3. Finally, all or
most of the items should have higher loadings on the first
component than on the other components.
The data produced during the principal components analysis can be ~sedto produce a statistic for a concise
summary statement of the reliability of the original set of
observations. The summary statistic is known as the coefficient theta (Armor 1974). Coefficient theta is derived
from a different statistic known as Cronbach's alpha (Cronbach 1951). Before one can understand theta, Cronbach's alpha must I be considered . i Cronbach's alpha belongs to a group of approaches to reliability assessment known as "internal consistencyff
(Carmines and Zeller 1979). Probably the most simple way of
measuring reliability iu using palred data sets derived from either a systematic or random split of a single data set or
from replicated data collection. Correlations between the two
halves provides a direct indication of the proportion of the
total observed variance due to true scores versus that due to
random error (Carmines and Zeller 1979; Nance 1987; Nance and
Ball 1986). The results, however, may depend on the technique
used to derive the samples used in the analysis. Though the
results would be similar, it is also useful to have a unique
indicator of reliability (Nance 1987:263). Internal
consistency methods allow this type of assessment to occur.
Internal consistency methods depend on high
intercorrelation indicates low random error content and thus
high reliability. When applied to the correlation matrix,
Cronbach's alpha is produced as follows:
where n equals the number of items and p equals the average
interitem correlation. Application of this formula to a
matrix of correlations among observations assumes that as correlations increase, indicating reduced random error, and n also increases, reliability will increase. Coefficient theta is a version of Cronbach's alpha, specially constructed to assess reliability from principal components analysis. It depends on a weighting vector (w) chosen within the principal components analysis and thus, it
is described as a "maximized alpha coefficient" (Greene and Carmines (1980:164). Since it depends on variances rather than correlations, the form of the coefficient is closer to that of the variance/covariance version of Cronbach's alpha (Greene and Carmines 1980:equation 18). The formula for coefficient theta is as follows:
where
t = theta
p = number of items
x = largest extracted eigenvalue
Theta reliability accomplishes the same things as alpha, except that it maximizes its output relying on X as a maximized score. Like alpha, theta reliability is produced by high correlations in the original correlation matrix and a related high initial eigenvalue (X) on the first extracted factor. Also like alpha, it produces an internal consistency
149 statistic indicating the amount of variance captured by the true scores, unattenuated by random error, in the original ~et of observations. Reliability, as measured by theta, is positively related to the size of n and the reduction in random error. Interpretation is the same as alpha, where scores close to 1.0 indicate reliability or little random error (Carmines and Zeller 1979).
Cowgill (1370) has forcefully argued that where observations of a variable are unreliable, the correlation of that variable with any other variable will be attenuated or reduced due to random error present in that original observation. However, Nance (1987:268) has demonstrated that it is possible to use a correction for attenuation on observations on variables where excess random error is present. He argues that when a correlation is attempted between two variables and one contains random error, that cnrrelatioli will aotorimtieally be reduced. However, if the reliabilities for these variables are known, then the correlation can be corrected. The correction for attenuation is:
where .
r(XY1 = observed correlation between X and Y
rtXX') = reliability of X r(YYt) = reliability of Y
~(xYT)= correlation between X and Y corrected for attenuation
This equation will have the effect of increasing the correlation in an amount that is related to the amount of random error present in measurements on Y and X.
Azsesument of Reliability
I now proceed with the analysis of reliability for the SRT and MSRT and the FVI, AAEL and OAEL indices. For each instrument, I present the principal components analyses, followed by the calculation of coefficient theta. Where unacceptable degrees of random error are identifled, corrections for attenuation are calculated and the instrument(s) are retested.
Sullivan and Rozen Typology
Principal components analysis was conducted using the SYSTAT statistical package (PC version ESystat 1nc.l). All raw data were rescaled following Binford (1981) (Table 1). The aed-ialidistai fragment category was dropped from the analysis since it is made up completely of scores of 100.00, which have a variance of 0 and are not amenable to factor analysis. No further transformations such as conversion to natural logs (Draper 1985) or chi-square scores (Binford 1989) were made of these data in an attempt to stay clo~eas to the original raw data. A correlation matrix was produced from the converted data matrix (Table 14). Factors were extracted from the correlation matrix, an eigenvalue criterion of 1.0 was set and a significant loadings criterion was set at .3 (Carmines and
Zeller 1979). These parameters are used in all subsequent principal components analyses unless otherwise noted. Initial statistics and loadings vectors are presented in Tables 15 and
16.
Carmines and Zeller's (1979:60) criteria for the principal components analysis assessment of reliability expects a high proportion of the variance to be accounted for on the first factor (>4O%). It also expects most of the variables to load on the first factor. If this is the case, then principal components analysis of the SET Indicates hiqh reliability since only one factor was extracted containing very high loadings on all variables. Coefficient theta was calculated using the eigenvalue score of 2.159, producing a reliability coefficient (Table 17) slightly above .8. Narice and Ball f1986:468) argue that reliability coefficients less than about .8 are undesirable as they may contain unacceptable amounts'of random error. Given this result, I argue that the
SRT is indeed reliable. In other iwrds, with an acceptably small amount of random error, it consistently provides the same scores given consistency in lithic reduction behavior. 152 Modified Sullivan and Rozen Typology
A correlation matrix (Table 18) was produced from the
rescaled converted data (Table 2). All zeros were converted to ones to facilitate a more efficient principal components analysis. Small medial/distal fragments were eliminated through rescaling to a consistent score of 100.00 (Table 21, which has a variance of 0. The correlation matrix was subjected to a principal components analysis from which three factors were extracted (Tables 19 and 20). Factor 1 captured approximately 50% of the total variance while seven of the nine variables loaded heavily on it. Although this meets the- minimum Carmines and Zeller (1979) criterion, some variability among the variables is present. The greatest vaviation is exhibited by the medium medial/distal fragments and small split flakes which load more heavily on factors two and three. Medial/distal fragments are the least predictable in terms a•’ expected breakage rates as they may result from platform shattering and flake snapping during reduction and upon ground impact. Complete flakes have not been broken and typically are produced during biface reduction during platform preparation. Proximal fragments result from flake snapping during reduction and upon ground impact. Nonorientable fragments occur where hard percussors and high force are used, which produces shat~eredplatforms and bulbs of force. They occur very rarely in biface production using soft hammers. 153 Table 14. Sullivan and Rozen Typology reliability analysis correlation matrix (CF=complete flake, PF= proximil Fragment, SF=Split flake). Tabf e 15. Sullivan and Rozen typology reliability analysis initial statistics.
Variance Percentage Explained by of Total Variance Factors Eigenvalues components- Explained 1 2.159 2.159 71.975 2 0.754 3 0.087 ...... Table 16. Sullivan and Rozen Typology reliability analysis loadings. Table 17. Theta coefficient data.
Assemblage Type Coefficient theta Sullivan and Rozen Typology .81 Modified Sullivan and Rozen Typology .88 Flake Volume Index '74 Acute Angle Edge Length .58 High Angle Edge Length .74 Table 18. Modified Sullivan and Rozen Typology reliability analysis correlation matrix (CF=Complete Flake, PF=Proxlmal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake). Large Large Medium Med i urn Medium PF MDF CF PF MDF ------_ Large PI? MDF Medium CF PF MDF SF s~ia
Med i urn Small Small Small SF CF PF SF
Med i urn SF 1.000 Small CF 0.079 1.000 PF -0.226 0.427 1.000 SF -0.269 9,688 0.302 1.000 Thus, complete flakes and proximal and nonorientable fragments are affected by fewer processes than medial/diutal fragments and are somewhat more reliable indicators of reduction variation. Split flakes occur only where a specific combination sf edge configuration, impact angle and percussor morphology is achieved (Cotterell and Kamminga 1987; Prentiss and Romanski
1989). Split flakes appear to be the result of a process similar to that described by Frison and Bradley (1980:89) as a radial break or percussor impact to a core surface at or near xight angles producing fractures extending from a central impact point in several directions. The actual fracture process propogates through the raw material at a right angle' to the primary flake removal fracture (flake ventral surface) with probably a compression-controlled propagation, ending with an axial termination (Cotterell and Kamminga
I387:G91-693), Is other words, they are fairly specialized fractures which may be expected to occur at varying rates of consistency depending on context. The results of this study indicate that small split flakes may be more useful indicators of biface reduction than medium split flakes with small split flakes loading heavily on factor one and medium split flakes loading more strongly on factors two and three. The positive loadiny on factor three is interpretable in relation to the strong negative loading of medium medial/distal fragments on factor three. Examining the raw data (Table 21, medium split flakes vary somewhat inversely with medium medial/distal Table 19. Modified Sullivan and Rozen Typology reliability analysis initial statistics. Variance Percentage Explained by of Total Variance Factors Eigenvalues Components Explained ------1 4.526 4.526 SO. 284 2 1.641 1.641 18.229 3 1.249 1.249 13.872 Table 20, Modified Sullivan and Rczen Typology reliability analysis loadings (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
Large PF 0.804 MDF 0.540 Pled i urn CF 0.969 PF 0.871 MDF -0.035 SF -0.290 Small CF 0.877 PF 0.696 SF 0.741 fragments. This relationship may not be more than an artifact of random variation in medium split flakes related to similar random variation in medium medial/distal fragments. The production of small split flakes appears to be more consistent and may be the result of a more regular combination of edge configuration, percussor and knapper behavioral variables. Medium split flake prodnction seems to be controlled in a less
consistent manner such that fewer of these flake types are produced. These differences may be the result of differences between small flake production aimed at shaping edges where percussors impact platforms at angles more likely to produce split flakes (as occurs in core reduction) and medium flake production aimed at producing thinning flakes, where percussor
impact is more oblique and less likely to produce ~pllt flakes.
The coefficient theta score (Table 17) 13 high (.88)
indicating minimal random error in the solution as a whole. The problems of variability in medial/distal and split flake counts appear to be minimal. Although there is some random error present, it does not appear to be of the magnitude which would prevent extremely reliable recognition of biface reduction. I conclude that the MSRT is reliable.
Flake Volume Index
A correlation matrix was produced from the raw FVI data set (Table 21). The correlation matrix was subjected to a 162 principal components analysis from whicn two varimax rotated factors were extracted (Table 22 and 23). Factor one captured
approximately 44% of the variance and high positive loadings on four of the six variables. One additional variable produced a high negative loadlng. Factor two contains a high negative loading on two variables. Following Carmines and
Zeller (1979), the solution appears to achieve the minimal criteria for reliability. The small split flake category appears to be the primary source of variability within the matrix. Small split flake FVI scores vary in proportion to those of small proximal fragments, while they vary inversely to all other flake type distributions. The overall size of these flakes is conditioned heavily by the size of the platform present and the location of the breakage. These factors appear to be more critical conditioners of error variation in small flake size thaii md3uri~flake size as small proximal and split flakes my consist of not only small broken edge shaping and platform preparation flakes, but also of extremely fragmentary thinning flakes with comparatively greater thicknesses. Small complete flakes and medial/distal fragments would not contain this type of variability. Small complete flakes are the result of edge shaping and are typically small and contain low error variation. Small medial/distal fragments may contain fragmentary edge shaping and platform preparation flakes as well as fragments of larger thinning flakes. Their high numbers, however, hide this variation when comparing mean Table 21. Flake Volume Index reliability analysis correlation matrix (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
Med i urn Med i um Small Small Small PF MDF CF PF MDF I ------Med i urn PF 1.000 MDF 0.742 1.000 Small CF -0 .211 -0.286 1.000 PF 0,061 0.462 -0.047 1.000 MDF 0,572 0.660 -0.218 0.237 1.000 SF -0.353 -0.020 -0.057 0.198 -0.303 ------Table 22. Flake Volume Index reliability analysis initial statistics. Variance Percentage Explained by of Total Variance Factors Eigenvalues Components Explained Table 23. Flake Volume Index reliability analysis loadings (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
Medium PF 0.841 0.276 MDF 0.910 -0.240 Small CF -0.385 0.191 PF 0.390 -0.700 MDF 0.833 0.115 SF -0.290 -0.816 scores. Medium proximal and medialidistal fragments are very consistent in size distributions as they are the result of one form of flake removal, biface thinning, The coefficient theta score (Table 17) is slightly below
-8, indicating a slightly problematic amount of random exsor. This is partially a function of the low number of flake categories analyzed as theta will increase with additional var iableu . Nonetheleu=, the aritc)urit of random error p~esent may attenuate further analyses and a correction for attenuation iu recomrnended fn further studies.
Acute Angle Edge Length
A correlation natrix was produced from the raw data matrix (Table 24). The correlation matrix was then subjected to a principal components analysis producing three varimax rotated factors (Tables 25 and 26). Factor one captured only 32% of the total variance and four of the six variables with significant loadings. This does not meet Carmines and
Zeilerfs (1979) minimal criteria for reliability. Each variable contributing strongly to factor one, also contributes significantly in the negative or positive dimensions on factors two and three. Fracture location and length appears to be quite consistent in the small categories as indicated by high loadings of small rnedial/distal fragments on factor one. Some variability in fracture size in the medium size group is indicated by the low loading of meriiurr~ Table 24. Acute Angle Edge Length reliability analysis correlation rcatrix (CF=Cornpiete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake). Medium Med i urn Small Small Small PF MDF CF PF ...... MDF Medium P F MDF Small CF PF MDF SF
Small SF ------Small SF 1,000 ------Table 25. Acute Angle Edge Length reliability analysis initial statistics.
Variance Percentage Explained by of Total Variance Factors Eigenvalues Components Explained ------1 1.911 1.911 31.849 2 1.435 1.436 23.937 ------Table 25. Acute Ancjle Edge Length reliability analysis loadings (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
1 2 3 ------Med i urn PF 0.689 -0.256 0.579 PlDF 0.208 0.727 -0.490 Small CF -0.248 0.425 0.797 PF 0.642 -0.589 -0.218 MDF 0 .761 0.309 -0.017 SF 0.583 0.469 0.093 ------medial/distal fragments on factor 1. The low loading of small complete flakes on factor one ma-J indicate that platform size is partizlly linked to variability in AAEL scores. This contention is suppo~tedalso on factor three, where small complete flakes and medium proximal fragments load strongly in the positive dimension.
The coefficient theta score is correspondingly low (Table
17). Random error appears to be high in the use of the AAEL index and thus it is not reliable. Further application will require a correction for attenuation.
High Angle Edge Length
Raw data were used to produce a correlation matrix (Table 27). The correlation matrix was subjected to a principal components analysis and two varimax rotated factors were extracted (Tables 28 and 29). Factor 1 capture? 448 ~f the variance and significant loadings on five of the six variables. Factor two also has a high loading on medium proximal fragments. Regardless, the solution meets most of the relevant criteria for principal components reliability.
Medium proximal fragment HAEL scores are primarily the result of length and location of breakage, In some cases, breakage is very minimal, perhaps only at the termination of the flake, producing a reduced HAEL score. In other cases, breakage is a lateral snap, which results in a greatly increased score. This would be expected to more greatly Table 27. High Angle Edge Length reliability analysis correlation matrix (CF=Cmtpiete Flake, PF=Pxoximl Fragment, MDF=Medfal/Distal Fragment, SF=Split Flake), Medium Med i um Small Small Smaf P PF MDF CF PF MDF
Med i urn PF 1.000 MDF 0.283 f .800 Sm11 CF 0.068 0.271 1.000 PF -0.333 0.490 0.433 1.000 MDF 0.037 0.641 0 - 464 0.558 1.000 SF -0.030 0.031 0.184 0.483 0.348
Small SF Small SF 1.000 Table 28. High Angle Edge Length reliability analysis initial statistics. Table 23. High Angle Edge Length reliability analysis loadings fCF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
1 2 ------Medium PF -0.011 0.875 MDF 0.707 0.498 Small CF 0.645 0.079 PI? 0.837 -0.373 MDF 0.860 0.139 SF 0.520 -0.387 ------affect medium sized flakes than small flakes since the r'3ngc of variation would be greater.
The coefficient theta score is just below .8 (-74) (Tablr
17) indicating the presence of a small amaunt of random esxor.
Again, the reader is cautioned that theta reliability rises with increased numbers of variables such that this reliability score may be unrealistically low. However, variability in medium proxiri~alfragment and mall split flake scores LY indicative of some degree of random error which could have attenuatit>rl effects (Cc~wgill1970). A correction for attenuation is recommended for further testing the index.
RELIABILITY RETESTS
The correction for attenuation formula prevented by
Carmines and Zeller (1979:48-49) and Nance 11987:268) was used to correct the correlation matrices associated with the zeliability analyses of the FVI, AAEL and HAEL indices. I chose not to use Armor's (1974) scaling approach to coefficient theta calculation since it assumes the presence off more than one underlying factor and this analysis attempts to isolate only one factor, soft hammer biface reduction. Correspondingly, Armor's approach was not applied to correcting the original matrices for attenuation, as the effectiveness of the correction would have been substantially reduced by considering individual behavior of the var iables rather than the group as a vhole. Following production of the 175 corrected carrelation xtatrices, all were retested with new principal cnmponents analyses. This was done in order to gauge the effectiveness of the correlations before applying them to the validity analysis.
The corrected correlation matrix (Table 30) was subjected to a principal compan~ntsanalysis which produced two varimax rotated factors (Tables 31 and 32). Factor one captured 52% of the variance and retained four of six variables with significantly high positive loadings and one with a hiqh negative loading. Factor two contained high loadings on medium proximal fragments (positive) and small split flakes (negative). This result is similar to that produced in the original analysiz. The co-loading of medium proximal frayment~on factors one and two and the small split flake negative loading on factor two indicates that while random error remains, its effects have been minimized.
The new coefficient theta score (.82) (Table 33) indicates that an acceptable amount of random error has been compensated for. I conclude that the FVI is reliable with the correction for attenuation.
Acate Angle Edge Length
The corrected -4AEL correlation matrix (Table 34) was
176 Table 30. Flake Volume Index reliability analysis corrected correlation matrix (CF=Complete Flake, PF=Pra,~imalFragment, MDF=Medial/Distal Fragment, SF=SpLit Flake).
Med ium Med ium Small Small Small PF HDF CF PF MDF ------.----- Medi urn PF 1.000 MDF 1.000 f .000 Small CF -0.290 -0.390 1.000 PF 0.080 0.620 -0.060 1.000 MDF 0.770 0.890 -0.290 0.320 1.000 SF -0.480 -0.030 -0.080 0.270 0.410
Small SF ------Small SF 1.000 ------
Table 31. Flake Volume Index rellablllty retest initial statistics.
Variance Percentage Explained by cf Total Variance Fact~is Eigenvafues Components Explained ------______------1 3.110 3.110 51.834 2 1.515 1.515 25.248 ------______-______------Table 32. Flake Volume Index reliability retest loadings (CF=Complete Flake, PF=Proximal Fragment, MDF=MedLal/Distal Fragment, SF=Split Flake).
Medium PF 0,889 0.527 MDF 1.029 0.039 Small CF -0.452 0.014 PI? 0.491 -0.519 MDF 0.902 -0.234 SF 0.038 -0.955 ------Table 33. Reliability retest theta coefficient data.
Assemblage TYPE Coefficient theta ___-__-______----______------______---- Flake Volume Index .82 Acute Angle Edge Length .74 High Angle Edge Length -82 Table 34. Acute Angle Edge Length reliability retest correlation matrix (CF=Complete Flake, PF=Proxirnal Fragment, MDF=Medial/Bistal Fragment, SF=Split Flake).
------Small SF 1.000 subjected to a principal components analysis producing three
varimax rotated factors (Tables 35 and 361. Factor one accounted for 43% of the total variance and captured
significant loadings on four of the six variabies. Four
strong loadings sgain occur on factor cwo. High positive loadings on factor three again include medium proximal
fragments and small complete flakes. From a reliability standpoint, the result is improved and it does appear to nwek Carmines and Zeller's minimal criteria for principal components reliabi 1 i ty.
The new coefficient theta score is slightly below "8
(.74) (Table 33) indicating that the effects of random error have been greatly reduced compared to the original analysis. I recommend that validity testing using the correction for attenuation with the caution that the variability in medium proximal and medial/distal fragment and small complete flake behavior be assessed closely before drawing interpretations, Some minimal attenuation effects may be present.
High Angle Edge Length
The new corrected HAEL correlation matrix (Table 37) was subjected to a principal components analysis producing two varimax rotated factors (Tables 38 and 39). Factor one explained 53% of the total variance and produced significant loadings on five of six variables. Factor two contains significant loadings on two variables (medium proximal and 181 Table 35- Acute Angle Edge Length reliability retest initial statistics. Yar iance Percentage Explained by of Total Variance Factors Eigenvalues Components Explained
-1 2.582 2.582 43.030 2 1.757 1.757 29.283 Table 30. Acute Angle Edge Lenqtn reliability retest loadings (CF=Complete Flake, PF=Froximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
------__-______II______II~~~~~~~~~~~ Med i urn PF 0.801 -0.283 0.622 MDF 0.242 0.803 -0.528 Small CF -0.288 0.471 0.856 PF 0.746 -0.652 -0.233 MDF 0.885 0.341. -0.019 SF 0.678 0.519 0.098 ...... Table 37. High Angle Edge Length reliability retest corxelation matrix (CF=Complete Flake, PF=Proxlmal Fragment, MDF=Medial/DistaP Fragment, SF=Split Flake). Med i urn Medium 5m11 Smlf Small PP MDF CF PF MDF Hed i urn PF 1.000 MDF 0.382 4.000 Small CF 0.091 -0.366 1.000 PF -0.450 0.662 0.585 1.000 MDF 0.050 0.866 0.627 0.754 1.Q80 SF -0.040 8.042 0.249 8.652 0.470
Table 39. High Angle Edge Length reliability retest loadings fCF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
------_------_____ Medium PF -9.013 0.912 MDF 8.780 0.520 Small CF 0.711 8.082 PF 0.923 -0.388 MDF 0.950 0.145 SF 0.573 -0.404 Table 40. Sullivan and Rozen Typology validity analysis correlation matrix (CF=eomplete flake, PF= proximal Fxagment, SF=Split flake), rnedial/distal fragments 1, one of which (medium proxima1 fragments) is significantly independent of the loadings on I factor one. These results provide a clear indication of compensation for the estimated reliability of the HWEL. Medium proximal fragments continue to produce some error I variability, but the effects of their overall contribution is seduced. The coefficient theta score (Table 33) is sufficiently
high (.82i to corroborate the above conclusions. I conclude
that the HAEL index is reliable with the correction for attenuation.
VALIDITY AHALYSIS
I now present a definition of validity, followed by a
discussion of my proposed validity analysis of the SRT and
MSRT and the FVI, AAEL and HAEL indices. Subsequently, the actual validity analyses are discussed.
ASSESSING VALIDITY
The validity of a measure depends on its ability to
provide the type of result for which it is intended. Thus,
validity depends on the relationship between the measuring
instrument and the phenomenon measured (Nance and Ball 1986:465). While reliability measures consistency or
replicability in the measuring instrument, validity provides an assessment of the usefuLness of a measuring instrument far
actually measuring variability. Using Nance and Ball's (13861
test pit example, one could demonstrate high reliability in test-pit sampling by showing that one consistently find
nothing using this technique. However, it would also be invalid since this technique is intended to allow archaeologists to find buried archaeological sites,
Validity deals with assessing the presence of systematic
erroz (bias) in observations or measurements. Systematic errors provide predictable deviations fr~mintended results.
In the above example, systematic error would be the cause for not finding any sites. This differs from random error which reduces the consistency in results. Translating this into terns applicable to this research,
I am concerned with the ability of the typologies and indices to measure variability among debitage assemblages chat parallels variation in lithic reduction strategies. Thus, the validity of the SRT and the MSRT will depend on their ability to segregate variability in lithic reduction strategy, while the utility index profiles will also vary to some degree by reduction strategy.
A number of types of validity exist (Carmines and Zellex
1379; Nance 1987). Content validity is a subjective appraisal
of a body of information to determine if it is actually the information desired. For example, one might question the content validity of minimum numbers of individuals (MNI) as a measure of dietary importance of different animals and
189 conclude that it was invalid since it generally provides a low estimate on the number of animals present and does not take
into account the use of different parts as opposed to whole animals INance 1987).
Criterion-related validity involves testing the
relationship between the measuring instrument and some
external criterion. For example, an archaeologist might want
to demonstrate that an artifact type is a valid chronological
marker by examinimg its distribution in stratigraphic
contexts. If it is determined that the artifact class was
modified by plowzone disturbance and does not reflect the
chronology at that site, the artifact class will not have achieved criterion related validity (Nance 1987:285).
Construct validity is probably the most useful to many scientific disciplines as it involves the assessment of a measuring instrument to determine if its results reflect a given theoretical construct. It works by forming theoretical expecL~tionsabout a phenomenon of interest. These expectations state how measurements should behave cjiven specific circumstances. After data collection, if these expectations are met the instrument is considered to be valid.
If not, then either the data were collected improperly, the theory is wrong or the measurement does not reflect the phenomenon of interest (Nance 1987:287). This type of validity assessment is critical to a developing discipline such as archaeology, as it results in theory building (Nance
1987:288). In experimental contexts it helps produce the type
130 of theoxy known as middle range theoxy (Binfoxd 1981). Some ethnoafchaeological researchers have produced studies focussing heavily on the validity of certain
theoretical constructs. Binford (1978) for example, in his Nunarniut research, tested numerous propositions linking theories about economic behavior to archaeological patterning. By operationalizing the theoretical constructs with
quntitatlve models based on faunal utility indices, he was able to directly assess their validlty with Nunarniut faunal
da3 tar Validity studies remain relatively rare in Lithic
technology research. Most reseaxch of the last two decades is best described as exploratory (Burton 1980; Dibble and
Whittaker 1981; Hayden and Hutchings 1989; Magne and Pokotylo
1981; Patterson 1982; Patterson and Sollberger 1978; Newcomer
and Seiveking l.980; Shelley 1990 to name a few). These
studies are important as we are only now determining which attributes are useful for measuring specific kinds and sources of variation in Pithic reduction.
Some studles have tested specific hypotheses about the relationships between theoretical expectations and instrument behavior. Stahle and Dunn (1982) tested the relationship between theoretical expectations regarding flake size distributions and biface reduction stages, concluding that their instrument was valid for measuring variation in
reduction stage. Magne (19851, while in many ways exploratory, still tested several hypotheses about the 19 1 characteristics of debitage assemblages and the teshnicjues used to produce them. Conclusions were drawn regarding the
validity of some of these flake characteristics for measuring reduction stages and an instrument was constructed. Interestingly, Magne's instrument has been recently tested by
Ingbar et al. (1989) and found to lack validity. Debate regarding Sullivan and Rozenls (1985) debitage typology has centered on its construct validity or its ability to match flake type distributions predictably and uniquely to reduction strategies (Amick and Mauldin 1989; Baumler and
Downum 1389; Ensor and Roemer 1989; Ingbar et al. 1989; Prentiss and Romanski 1989; Rozen and Sullivan 1989a; 1989b3.
Although to date, no formal test of the validity of the typology exists in the literature, some preliminary experimental research has been accomplished (Baumler and
Downum 1989; Prentiss and Romanski 1989; Prentiss et dl, 1988). This experimental research has suggested that a host of reduction, raw material and taphonomic variables condition flake type output using the typology.
Testing the SRT (as well as the MSRT and utility indices) for construct validity involves first, defining, theoretically, the expected behavior of the instrument(s1.
Sullivan and Rozen (1985) argue that the typology is useful for segregating core from tool reduction. In other words, in their research, it is used to segregate out lithic reduction aimed at producing large flakes for later use as tools from lithic reduction aimed more towards production of formal tools, such as bifaces from flakes. At the most minimal level then, to achieve construct validity, the SRT must be able to separate "t oolqlproduct ion from "coretBreduction in a controlled experimental setting. The MSRT has been proposed as a more complex alternative to the basic SRT in anticipation the the SRT might not be valid for secognizing an extended range of behavioral and taphonomic variables Fn debitage assemblage formation. It has also been proposed for use in conjunction with the utility indicev which depend on variation in flake size au well as shape. Thus, it serves as a "skeletonw framework for application of the utility indices for debitage assemblage prediction in much the same way that caribou skeletal anatomy served as -3 framework for the application of BlnfordF3 (1978) utility indices. Construct validity will be achieved if the MSRT can also be demonstrated to segregate "tool" production fr~m"corew reduction. The utility indices measure different aspects of variation in basic flake size and edge configuration^ The FVI estimates overall volume of each flake. When it is used to characterize an entire assemblage, mean values for each of the
MSRT flake type categories are used. Likewise, the kAEL and the HPBL rely on mean edge length values for each MSRT flake type torcharacterize an entire assemblage. Construct validity will be achieved if the mean index values can, in a controlled experimental context, segregate large flake from small flake production (FVI) and low edge ar.gle from high edge angle assernbPages (AAEL and MAEL). This is important if the utility indices are to be used for predicting technologically unique cull assemblages or those flake assemblages, culled for tool-use according to their size or edge morphology, deriving from a single reduction strategy. The results are expected to be heavily conditioned by variation ln the use of oft verzuu hard hammers, Soft hammer use has been demonstrated to produce consistently thinner flakes than that of hard hammer reduction (Hayden and Mutchings 1989). By implication, mean edge angles across all flake types produced should vary with flake thickness. Thin flakes should have consistently lower edge angles than those of thicker flakes. Results may also be a•’f ected by core size and the use . edge preparation. Since each of the instruments measared depends on multiple variables to characterize different phenomena (reduction strategies), construct validity assessment must depend 01-1 a multivariate technique for identifying this type of variability. Like reliability assessment, factor analysis may also be used to assess construct validity (Carmines and
Zeller 1973). To accomplish this, empirical data are collected from the observation of phenomena thought to condition variability in typology and index output. Using empirical data, patterns of intercorrelations are analyzed using principal components analysis (Nance b989:288). Construct validity is assessed from patterning in the intercorrelations corresponding to the variability in the phenomena measured. This study depends on an examination of individual factor loadings and plots of factor scares. Factor
loadings are used to indicate the significant contributiun of
variables to the different factors. F~C~CEscores arc used to indicate the contribution of each individual case to the
overall solution (Kim and Mueller 1978; Rurnrnel 1970). By comparing factor scores to factor loadings, variability can be
assessed on a case by case basis as well as in terms of the contribution of the individual variables. Thus, cases which
contribute to a given factor in very different ways can easily
be separated depending on the behavior of variables identified as significant on the factor loading matrix. This is an ldeal situation for measuring validity of this type since the analysis seeks to identify the presence of different sources of variability. If the variability is present, but the instrument cannot identify it, the instrument will not have accomplished its desired goals and will suffer from a low degree of validity.
ASSESSMENT OF VALIDITY
I now discuss the principal components analyses of validity, highlighting the role of factor scores as indicators of variability. Constnuct validity depends on the correspondences between the theoretical predictions and observed results.
Sullivan and Rozen Typology I The converted data {Table 8) were used tc2 produce a correlation matrix (Table 40). This matrix was subjected to a
priqcipal components analysis which resulted in the extraction
of two varimax rotated factors (Tables 41 and 42). Factor one
contains high positive loadinys on proximal fragments and
split flakes and a high negative loading on nonorientable
fragments. Factor two contains a high positive loading on complete flakes and a high negative loading on rnedial/distal
fragments. This mat~ixis best interpreted in relation to the factor score matrix (Table $3) and plot (Figure 13).
since factor one measures Y iynif icant variability oti
p~oximaland nonor ientable fragments and split flakes, it segregates only three pressure flake assemblages from the
entire group of assemblages. These include the flake (If),
unprepared core (17) and prepared core (20) assemblages,
containing high numbers sf proximal fragments and split flakes
and few nonorientable fragments.
Factor two, with its identification of variability in
complete flakes and rnedial/distal fragments, separates two assemblages as substantially different and one as somewhat
different f rotn %he larger group of assemblages. Assemblage 3
(hard ha~nmer, flake edge) is most noticably distinct, with
high complete flakes and reduced medial/diutal fragments.
Assemblage 17 (pressure, unprepared core) is also clearly
distinct with a reduced number of rnedial/distal fragments.
Assemblage 12 (hard hammer, medium flake goal, biface Table 41. Sullivan and Rozen Typology validity analysis initial statistics. Variance Percentage Expla i ned by of Total Variance Factors Eigenvalues Rotated Components Explained
1. 2,434 1.882 37.639 2 1.078 1.638 32.595 Table 42. Sullivan and Rozen Typology validity analysis rotated loadings !CF=Complete %lzk@, PZ=Proximai Fragment, MDF=Meciai/Distzi Fragment, NF=Nonorieniabie fragment, SF=Split Flake 1.
CF PF MDF NF SF Table 43. Sullivan and Rozen Typology validity analysis factor scores (ELFzExtra-large flake; LF= Large flake; MF=Hedium flake; Hh= Ward hammer; Sk=Soft hammer; Pr= Pressure; UPC=Unprepared core; PC= Prepared core, B=Biface; F=flake).
Case Asse~blage Factor Factor Type 1 Number------2 Case 1 ELF,Hh,UPC -0.789 -0.560 Case 2 ELF,Hh,PC -0.469 -0.349 Case 3 LF,Hh,B -0,428 -0.140 Case 4 LF,Sh,B -0.352 -0.472 Case 5 LF,Hh,UPC -9.486 -0.330 Case 6 LF,Sh,UPC -0.173 -0.639 Case 7 LF,Hh,PC -0.496 -0 267 Case 8 LF,Sh,PC -0.515 -0.599 Case 9 MP,Hh,F -0.66Q 3.721 Case 10 MF,SH,F 0.225 -0.116 Case 11 MP,Pr,F 2.539 -0.397 Case 12 MF,Hh,B -8.198 0.747 Case 13 MF,Sh,B 0.027 -0.257 Case 14 MF,Pr,B 0.593 -0.435 Case 15 MF,Hk,UPC -0.690 -6.478 Case 16 MF,Sh,UPC -0,941 -0.573 Case 17 MF,Pr,UPC 1.481 1.386 Case 18 MF,Hh,PC -0.882 -0.033 Case 19 MF,Sh,PC -0.112 0.044 Case 20 MF,Pr,PC 2.325 -0.253 Table 44. ~odiffedSullivan and Rozen Typology validity analysis correlation matsix (CF=Carnplete Flake, PF=Pnoximal Fragment, ~DF=Medial/B%stabFragment, SF=Split Flake).
Extra- Extra- Extra- Large Large Large Large Large ------CF PF SF CF PF Extra- Large CF 1.000 PF 0.163 f .BOO SF 0. f 63 1.000 1.800 Large CF -0,032 -0.094 -0.094 I. 000 PF 0.136 6.211 0.211 0.636 1.000 MBF 0.4'73 0.334 0.337 8.446 0.703. SF 0.417 0.627 0.627 0.390 0.764 Med i urn CF -8,l.l.l -0.106 -0.106 -0.811 -0. 281 PF 0.015 -0.093 -0.093 Q. 252 -0.125 MDF 0,185 0.269 0.269 0.240 0.372 NF 0,245 -0 .f 30 -0.130 0.566 0.857 SF 0.046 -0.053 -0.653 0,646 0.460 Sm13. CF -0,196 -0.180 -0,188 -0.359 -0.373 PF -0.041 -0.220 -0.220 -Q. 409 -8.415 MDP 0.082 0.864 0.064 6.213 0.152 NF 0 309 0.284 8.284 0.534 0.323 SF -0,183 -0 .I89 -0. f 89 -0.393 -0.240
Large MDF 1.000 SF O.?OS Med f urn CF -0,380 PF 0.054 MDF 0.598 NF 0.617 SF 0.528 Sarall CF -0.518 PF -9.442 EIDF 0,225 NF 9.4'92 SF -0.389 Table 84. Contnd.
Med i urn Hed i urn Small Small Small NF SF CF PF MDF Med P urn NF 1,000 SF 0.484 1,000 Small CF -0.371 -9.325 1 ,000 PP -0.373 -0.506 0.237 I. 088 PaDF 0.157 8.228 -0.255 -0.586 1,800 NF 0.369 0.484 -0.517 -0.530 0.241. BF -0.220 -0.419 0.493 0.652 -0 118 reduction) is only slightly different from the larger group
with a slightly increased number of complete flakes.
Thus, the SuIlivan and Rozen typology is incapable of segregating "tool" production from "coreu reduction. In this
analysis, both core and tool reduction debitage assemblages are grouped togethex. Only two very specific and specialized reduction strategies are separated as different from the total group. Pressure flaking is distinctive in its production of
high numbers of split flakes and proximal fragments. Hard
harmer reduction of a thin flake edge is aP~odistinctive in its production of high numbers of complete and split flakes. In all other cases, the variability produced among the five
flake types is not enough to warrant the identification of
$8 distinctive assemblages" (Sullivan and Rozen 1985:755). I conclude that the SRT does not achieve construct validity. In theory, the assessment of variation in flake breakage rates, as an indicator of lithic reduction variation, is gcod. Different hammer types, edge configurations and force application techniques should produce very different rates and types of flake breakage. However, the actual breakage patterns are highly variable and are not characterized well by five flake types. The key problem appears to be that variability is graded by flake sizes. For example, in the reduction of prepared cores, two distinct data sets are produced. One is associate3 with platform preparation and the other is the result of larger flake removal facilitated by that platform prepration. Thus, prepared platform core reduction assemblages contain two dimensions of variability only measurable by a typology designed to capture that variation with the inclusion of a size variable. It is not surprising that the analysis reported here identified pressure flake assemblages as unique. Pressure flake assemblages were the only ones falling within one size class (as identified by the MSRT), The SRT is too simple for application to archaeological problems.
Modified Sullivan and Rozen Typology
The converted data matrix (Table 10) was used to produce a correlation matrix (Table 441. .The correlation matrix was subjected to a principal components analysis which extracted four varimax rotated factors (Tables 45 and 46). Factor one contains high positive loadings on medium complete flakes, medium proximal and medial/distal fragments and small medial/distal and nonorientable fragments. It contains high negative loadings on small proximal fragments and small split f lakss. Factor two contains high positive loadings on extra-large proximal and medial/distal fragments and large split flakes. Factor three contains high positive loadings on large complete and split flakes, large proximal and medialidistal fragments, medium nonorientable Eragments and medium split flakes. Factor four contains high positive loadings on extra-large complete flakes, large rnedial/distal fragments and small nonorientable fragments. It contains a
203 3 I
li
2 om
Factor 1 12l :40
lo. 3. 0 - 19 6, a 4 14 8 2 a3 0. 77 15, 9 1 e 7 a 1 a -L 168 I I 1 1 I 0 1 2 3 4 Factor 2
Figure 13. Sullivan and Rozen Typology validity analysis factom score plot (Factor scores are tabulated In Table 43). Variance Percentage Explained by of Total Variance actors- Eigenval~es Rotated Coinponents Explained
1 6.182 3.141 18.477 2 2.531 2.563 16.926 --? 2.241 4.204 24.729 4 1. I55 1.952 11.494
--____--__-__------___------_-_-_I______I___II_II rn,is~le ' 45. Modified Sullivan and Rozen Typology validity analysis rotated loadings (CF=Complete Flake, PF=Proximal Fragment, MDF=MedialiDistal Fragment, SF=Split Flake).
Extra- Large Cf PF SF Large CF PF MDF SF MeSlum CF PF MDF NF SF Small CF PF MDr" NF SF Table 47. ~odifiedsullivan and Rozen Typology validity analysis factor scores (ELF=Ext+a-large flake; CF=Large flake; ~f=Medlumflake; Wh=Mard hammer; Sh= Soft hammer; Pr=Pressure; UPC= Unprepared core; PC=Prepared core; B=Bifaee; F=Flake).
Case Assemblage Factor Factor Factor Factor Number Type 1 2 3 4 Case I ELP,Nh,UPC 0.313 4.118 -0.259 0.473 Case 2 ELF,Hh,PC -0.287 -0.341 0.004 3.263 Case 3 LF,Hh,B 0.779 -0.238 -8.272 0.117 Case 4 LP,Sh,B 0.300 -0.173 -0.077 0.393 Case 5 LF,Hh,UPC -0.448 0.214 2,876 -0.390 Case 6 LF,Sh,UPC -0.124 -0.078 1.985 -0,111 Case 7 LF,Hh,PC 0,267 -0,575 0.990 -0.318 Case 8 LF,Sh,PC 0.501 -6.275 0.376 0.676 Case 9 MF,Hh,F -0.432 0,128 -0.328 -1.915
Case 10 MF,Sh,F -0.267 -0.143 -0.6% -0 8, 275 Case 11 MF,Pr,F -1.617 -0.148 -0-506 -0.540 Case 12 MF,Hh,B 0.999 -0.157 -0.962 -0.982 Case 13 MF,Sh,B 0.070 -0.050 -0.655 -0.386 Case 14 MF,Pr,B -0.728 -0.048 -0.434 -0.435 Case 15 MF,Hh,UPC 0.609 -0.513 0.943 -0.124 Case 16 MP,Sh,UPC 0.594 -0,116 -0.093 0. I68 Case 17 MF,Pr,UPC -2.071 -0.579 -0.992 1.100 Case 18 MF,Hh,PC 2.193 -0.810 -0.799 0,294 Case 19 MF,Sh,PC 0.951 -0.093 -0.666 -0.403 Case 20 MF,Pr,PC -1.551 -0.116 -0.485 -0.621 ------Factor 2
Figure Modified Sullivan and Rozen Typology validity analysis factor scone plot: factors 1 and 2 (Factor scores are tabulated in Table 47).
Table 48. Flake Volume Index validity analysis corrected correlation matrix (CF=CornpPete Flake, PF=Proximl Fragment, HDP=Medial/DfstaP Fragment, SF=Spiit Flake).
Medium Med f urn Small Small Small PF MDF CF PF MDP
Med i urn PF 1,800 MDF 1.000 1.000 Smll CF 0.868 0.549 l. 000 PF 0.523 0.401 0.447 1.000 MDF 0.859 0.782 0.162 0.360 1,600 SF 0.397 0.381 0.557 0,050 0.408 high negative P~addngon small complete flakes. These patterns are hnterpreted with the aid of factor scones (Table
47; Figures 14 and 159.
In a continuum of vazhabl%iky, factor one segregates pressure flaking assem&lagea (assemblages 11, 14, 19 and 209
fnsa Rand hammer biface reduction (assemblages 3 and 12% and hard and soft hamex medium flake goal, core reduction
tassemblages 15, 16, $8, 19). The bf dace and medium f fake, core reduction assemblages produce flake distributions centering heavily on medium flake categor%es as well as nebativehy high numbers of small medfal/d4stal and nonorimtable fragments. They contain relatively few split flakes and variable numbers od proximal fragments depending on the presence of platform prepsratican, En contrast, pressure flake assemb'bages contain no medium size flakes and numeroepa small split flakes and pzoxfma% fragments.
Factsw two is easy to interpret as it clearly serves to defineate the unprepared extra-large flake, core reduction assemblage (I) f rsm all other assemblage types;. This assemblage contains comparatively high numbers of extra-large proximal and med$a%/distal fragments, as well as Parge split flakes. Factor three separates large 89ake core reductfan iassemblagaa 5, 6, 7, 8) from all ather rcduieti~nstrategfea. ft ioads heavily on ail contributing Parge flake types a% well as medium nonorientable fragments and split flakes. Lange flake core reduction is segregated from laxqe flake, bifac~ production (assemblages 3 and 4) and extra-large flake, esze
reduction Cassewblages 1 and 2) in that each sf these contains relatively low numbers sf zedbum nsnamientablle Eaagmmts and split flakes while these flake types are increased in the
Barge $lake csxe assemblages. Due to a high nuWer sf medium split flakes and large complete flakes and proximal and aaedieP/distal fragments, the medium flake, hard hammer core seduction asserablage Q15P falls also within this group.
Factor four clearly identifies two variant forms af
~eduction: prepared, extra-large flake, core reduction (2) and
hard hammer, flake edge reduction (91. Heavy loadings on extra-large complete flakes, large mdial/distal fragments and sm3.1 wonorientable fragments are consistent with gxepared extaa-large flake, core reduction. In %Re hard kamer, flake edge reauction assemblage, small complete flakes are dominant and small split flakes aze numerous. This varies with the larger form sf core reduction fa which small complete and split flakes are Pew.
Fsnr general implications sf tkfa analysis my be discussed, These concern the segnegathsa sf *to~l~production from @conetqreduction, the use sf had versus soft hamen, flake size goals and the role of platform prepa~atfon. The presence ~f each is reflected in the solution, from al,l other form rf reduction vithin the analysis, Hame% type does not appear to have significantly affected the identification of core reduction, as factox one identifie8 all medfumrrt flake cores, factor three isolated aPP large Elake cores and factors two and four identiffed the extra-laxge flake cores- Tool production asse;;a$hges are designated in several ways. Fimst, gnessure flake asae identidied on factor one* Hard hamex Elake edge reduction and $&%acereduction are isolated on factor four. Hard hamsf
$iface reduction also occurs on factor one, as in some ways It resembles that of medium flake oriented core reduction. Soft hammer $iface and flake edge reduction are wat identified as unique, akthough they are consistently separated from a31P form of core reduction and pressure flaking. In general, the HSRT segregated Wtoolffproduction from @*corewreduetian assemblages.
There are strong indicators that the use of hard kaslraaaer in bifaee reduction and flake edge reduction will affect the overall MSRF distribution. For example, on factox ens, haxd hammer bifaees group closely with medium flake production cores. Ow factor four, some hard hamaen %sol assemblages axe sepazated front the other dorms sf tool psaduction. Hamex type does not appear to significantly affect the identification sf core reduction. Hard and soft hamen core
~eductionassemblaqes are grouped together on factors one and three.
Flake size goal is extremely important. Faets~sone, three and two and four monitor different flake size goal categories sf core zeduction. Pressure assemblages are akaa segregated on factox one, paskially due to BPEEarensas in flake size goal. Flatform preparation 1% clearly an important conditioner of core reduction assemblage variability. Factors two and four separate extra-large flake, core reduction assemblages entirely on the basis of platform preparation. Factor three provides an indication of separation between prepared and unprepared large flake goal core reduction assemblages. Likewise, some minimal differences are evident on factor one between prepared and unprepared hard hammer, medium flake, core reduction, The distiaction comes from the fact that prepared cores produce more complete flakes in the desired size range than do unprepared cores. Unprepared cores produce numerous proximal and medial/distal fragments in the desired ' size range as a result sf reduced control in flake removal from a lack prepared platforms . Platform preparation produces small complete and proximal fragments in prepared core reduction assemblages which are largely absent in unprepared core reduction assemblages. In this analysis, the MSRT has been able to separate all forms cf core reduction from other forms sf lithic reduction. It has also proven to be useful in identifying variation in nammer type, flake size production goal and platform preparation. With these successes in mind, I conclude that the MSRT has achieved construct validity. Some caution should be observed however, in the identification sf soft hammer biface reducticn, as it forms something of a genez-ic assemblage, resembling a number of others in the presence ~f high numbers of rnedial/distal and proximal fragments, often
across several size classes.
Flake Volume Index
The reliability analysis demonstrated the need for the
use of a correction for attenuation due to the presence of
random error. The correction was constructed for the six
retained flake types (medium proximal and medial/d istal
fragments and small complete and split flakes and proximal,
medial/distai and nonirentable fragments. These flake types
are also used in the analysis of the -EL and HAEL indices. I
assume that patterning in these flake data will reflect
variation in all flake categories.
A correlation matrix (corrected for attenuation) (Table
48) was produced. This matrix was subjected to a principal components analysis from which two varirnax rotated factors
were extracted (Tables 49-51; Figure 161. Factor one contains
high positive loadings on medium proximal and medial/diutal
fragments and small mediah/distal fragments. Factor two
contains high positive loadings Dn small complete and split
flakes and proximal fragments,
Factor one separates assemblages producing extremely
large flakes from those producing moderate to small flakes.
Hard hammer, large flake production (assemblages 5 and 7) assemblages and the extra-large flake, core reduction assemblage (asrernblage 2) are separated from the other 215 assemblages. Interestingly, each of these contains numerous
small medial/distal fragments with relatively high FVI scores.
This may be the result of the production of numerous extremely fragmentary but thick flakes, Soft hammer reduction acd biface reduction are grouped together with one. Soft hammer, medium flake, biface reduction (13) is isolated at the
opposite end of the factor score distribution as it contains by far the smallest flakes.
Factor two separates hard hammer, prepared core
(assemblages 7 and 18) and biface (assemblages 3 and 12) assemblages from soft hammer assemblages (assemblages 4, 10, 16 and 19) and hard hammer unprepared core assemblages
(assemblages 5 and 15). Bifaces and prepared cores both result in distributions of small platform preparation flakes,
Notably, many of these retain platforms or fragments of platforms and are classified as complete, proximal or split.
Factor tt~ocontains high loadings on ~rrtallctampletx and split flakes and proximal fragments, indicating that these flakes have greater size in hard hammer, prepared core and biface assemblages than in soft or hard hammer unprepared core reduction assemblages. Two factors contribute to this. First, few small complete and split flakes and proximal fragments are produced during unprepared core reduction.
Those which are produced are often smaller since they axe secondary detachments and not the result sf any systematic behavior. Many are produced during any form of lithic reduction where a substantial amount of platform preparation
216 Table 49. Flake volume Index validity analysis initial statistics. Variance Percentage Explained by of Total Variance Factors Eigenvalues Rotated Components Explained
1. 3.831 2.720 45.329 2 1.133 2.245 37,412 Table SO. Flake Volume Index validity analysis rotated loadings (CF=Complete Flake, PF=Proximal Fragment, HDF=Mediai/Distal Fragment, SF=Split Flakej.
Med i urn PF 0.923 0.448 MDF 0.921 0.291 Small CF 0.378 0.730 PF 0.172 0.830 MDF 0.907 0.100 SF 0.161 0.853 ------Table 51. Flake Volume Index Factor score data matrix (ELF=Extra-large flake; iF=Largs Clake; MF=Medium Flake; Hh=Hard hammer; Sh=Soft hammer; UPC=Unprepard core; PC=Prepared core; B=Biface; F=Flake).
Casz Assemblage Factor Factor Number Type 1 2 ------______2 ELF,Hh,PC .842 -. 340 3 LF,Hh,B -.042 1.072 4 LF,Sh,B -. 001 -. 583 5 LF,Hh,UPC 1.657 -. 857 7 LF,Hh,PC 1.791 1.061 8 LF,Sh,PC -. 637 1.276 10 MF,Sh,F -058 -1.869 12 MF,Hh,B .815 .599 13 MF,Sh,B -1.19 3 -. 135 15 MF,Hh,UPC .230 -. 033 15 MF,Sh,UPC -.456 -.925 18 MF,Hh,PC -1.028 .814 19 MF,Sh,PC -. 451 -. 213 ------0.0 - Factor 1
Factor 2
Figure 16. Flake Volume Index validity analysis factor score plot (Factor scores are tabulated in Table 51).
220 Table 52. Acute Angle Edge Length validity analysis corrected correlation matrix (CF=Complete Flake, PF=Proximal Fragment, MDF=Mediaf/Distal Fragment, SF=Split Flake).
Medium Ned i urn Small Small Sma l 1 PF MDF CF PF MBF
Med i urn PF 1.000 MDF 0.645 1.000 Small CF 0.059 0.804 1.000 PP 0.902 1.000 0. 800 1.000 MPF 0.717 I. 000 I. 000 I. 000 I. 000 SF 0.125 -0.085 -0.064 0.089 0.306
Small SF
Small SF takes place. The second important factor is the use of hard hammers to shape platfonms. Soft hammer flakes are typically
thinner than hard hammer flakes (Hayden and Hutchings 1989). If flake thickness can be expected to reduce breakage
(Sullivan and Rozen 1385) then hard hammer platform preparation assemblages should contain significantly higher numbers of complete flakes as well as eornparatively larger complete and split flakes and proximal fragments.
Demonstration of FVI validity depends on the ability of the index to separate large flake production strategies from smaller %Bake ~rientedstrategies. Pn this analysis, the FVP separated core reduction strategies aimed at large flake production from other reduction strategies. It also highlighted the importance of platform preparation and hammer type as a conditioner of flake size variability. Thus, I conclude that the FYI achieves construct validity.
Acute Angle Edge Length
The raw data matrix was used to produce a correlation matrix fcorrected for attenuaticsn) (Table 52). This matrix was subjected to a principal components analysis which resulted in the extraction of two varimax rotated factors
(Tables 53 and 54). Factor one contains high positive loadings on medium proximal and medbal/distal fragments and small complete flakes and proxfmal and medial/distal fragments. Factor two contains high positive loadings on medium proximal f~aqmentsand small split flakes, These factors are interpreted in reference to factor scores (Table
55; Figure f 7 ) . Factor one segregates bi%acereduction and soft hamer core reduction from hard hammer core reduction. Medium flake, soft hammer, prepared and unprepared core (assemblages 16 and 19) and biface reducti~n(assemblage 13) assemblages are isolated at the positjve end of the factox one factor score distribution. While large and extra-large flake, hard hamer, core reduction assemblages fall at the opposite end. This cleanly indicates that thick flakes produced during hard hammer reduction also have higher edge angles. Given this conclusion, it is curious that the large flake bdface
(assemblages 3 and 4) and large flake, soft hammer, prepared csre reduction (assemblage 8) assemblages were not distributed higher In this distribution, This is due to the fact that, in
Barge flake production contexts, the medium flakes zep~eaent residual breakage from the larger category. As such, edge angles are not consistently low enough to produce high WL scores. However, the large flake types contain extremely high AAEL scores indicating that these assemblages still fit the expected pattern for soft hamer csre reduction or biface production. Unfortunately, the large flake category was not included in the validity analysis due to its removal from the reliability analysis and its inconsistent appearance throughout the validity raw data matrix. Factor two Ls difficult to interpret and captures only Table 53. Acute angle Edge Length validity analysis initial statistics. Variance Percentage Explained by of Total Variance Factors Eigenvalues Rotated Components Explained Table 54. Asute Angle Edge Length vaiidity analysis rotated loadings (CF=Complete Flake, FF=Froximai Fragment, ME~F=Hedial/Disi&LFraqmenc, SF=Split Flake).
1 A!. Table 55. Acute Angle Edge Length Factor score data matrix (ELF=Extza-large flake; LF=&arge flake; MF=Mediurn Flake; Hh=Hard hammer; Sh=Soft hammer; UPC=Unprepard core; PC=Prepaxed core; B=Biface; F=Flake).
Case Assemblage Factor Factor Number TYP~ 2 2 ELF, Hh, PC &F,Hh,B EF, Sh,B kF,Hh,UPC LF,HhpPC LF,Sh,PC MF,Sh,F IQIF, Hh, B MF,Sh,B MF, Hh,UPC MF,Sh,UPC MF,Hh,PC MF, Sh,PC Factor
Factor 2
Figure 17. Acute Angle Edge Length validity analysis factor scoxe plot (Factor scores are tabulated in Table 55) . Table 56. High Angle Edge Length validity analysis corrected carrelation matr fx (CF=Complete Flake, PF=Proximal Fragment, MDF=Hedial/Distal Fragment, SF=Splft Flake).
Medi urn Medium Small Small Small PF MBF CF PF MDF
Hed f urn PF a. ooo MDF 1.008 1 .000 Small CF 0.950 9.972 1.000 PF 1.000 1,000 0.589 1.000 MDF 1. 000 f .800 0,914 1.008 1.900 SF -0.131 -0.157 -0.328 -0.299 0.089 about 21% of the total variance (as opposed to Eactox one with
698 of the variance). It appears to be primarily concerned with variation in aLscores on small split flakes. Both medium flake oriented biface assemblages iassembLages 12 and
13) and the extra-large flake prepared core assemblage
(assemblage 2) group together with high positive factor scores. Given the low robusticity of this factor and the potential for random error remaining ln the AAEL, this association is considered to be fortuitous and not worthy af further discussion, Factor one is extremely robust and indicates a strong pattern of separation between soft hamer core reduction/biface production assemblages and all hard hamrnes core reduction assemblages. To achieve construct validity, the A&EL needed to demonstrate exactly this type of separation. T conclude that the AAEL is valid for further use.
High Angle Edge Length
The raw data matmix was used to produce a correlation matrix (corrected for attenuation) (Table 56). This matrix was subjected to a principal components analysis producing two varimax3rotated factors (Tables 57 and 58). Factor one is exceptionally robust, capturing 78% of the variance and producing high positive loadings on all variables except for the small split flakes. This solution is interpreted in Table 57, High Angle Edge Length validity analysis initial statistics.
Variance Fsezeewtage Explained by sf Total Variance Factors Ehgenvalues Rotated Components Explained
E 4.744 4.674 77.900 2 1.086 1.856 19.269 Table 58. High Angle Edge Length validity analysis rotated loadings !CF=Cornplete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake). Table 59. I igh Angle Edge Lengt h Factor score data matrix iELF=Extra-large flake; LF=Large flake; MF=Medium Flake; Hh=Hard hammer; Sh=Soft hammer; UPC=Unprepard core; PC=Prepared core; B=Biface; F=Flake).
ELF,Hh,PC LF,Hh,B LF,Sh,B LF, Hh, UPC LF,Hh,PC LF,Sh,P@ MF,Sh,F MF,Xh,B MF,Sh,B MF, Hh, UPC MF,Sh,UPC MF,Hh,PC MF,Sh,PC 0.0 Factor 1
Factor 2
Figure 1%. High Angle Edge Length validity analysis factor score plot (Factor scores are tabulated in Table 59). reference to the factor scores (Table 59); Figure 18). The factor score distribution for factor one contains a
continuum from hard hammer core reduction to soft hammer biface and core reduction. There is a clear distinction between those assemblages produced with hard hammers, favoring production of thick, high edge angle flakes and those produced
with soft hammers resulting in thin low edge angle flakes. Hard hammer biface assemblages fall roughly in the center of the continuum as do soft hammer unprepared core reduction and haxd hammer, prepared core reduction [medium flakes). Factor two identifies variation in small split flakes. The extra-larye flake, prepared core (assemblage 2) and medium flake, hard hammer, biface reduction (assemblage 12) assemblages group closely together with high positive factor scores. Since the factor two loading on small split flakes was negative, the high positive factor score indicates a low split flake score on both reduction types. Since this factor is not very robust and many of the high split flake scores are found in other prepared core and biface assemblages, the association is considered to be fortuitous and not worthy of further discussion.
To achieve construct validity, the HAEL needed to separate assemblages according to variation in their contribution of high edge angled flakes. It has succeeded in separating hard harmer core reduction assemblages from soft ha~nmer core reduction and biface production assemblages, thereby highlighting the importance of hammer type, core size 234 and edge preparation in the production of thick, high edge angle flakes. I conclude that the HAEL has achieved construct validity.
SUMMARY
This study provides a classic case of a measuring instrument achieving reliability but not validity. The Sullivan an3 Rozen typology consistently identified biface production assemblages with minimal random error. When it was required to distinguish between tool and core reduction, it was only able to separate some pressure flake assemblages and one hard hammer retouch assemblage from a much larger group which included forms of biface and core reduction. Thus, it was identif ied as invalid for its intended application. The fundamental problem appears to be its simplicity. It cannot cope with complex reduction strategies which produce multiple size tiers within a single assemblage. The lumping of different sized groups of flakes together serves to mask important variation, rather than to highlight it. The problem is largely eliminated with the use of the Modified Sullivan and Rozen typology or MSRT. The MSRT is both reliable and valid. It identified biface reduction, with less random error than that of the SRT. It was able to segregate core reduction assembiages according to platform preparation, flake size goal and to some extent, hammer type, It was also able to identify hard hammer biface reduction and 235 all pressure flaking assemblages. The success of the MSRT is due to its ability to recognize flake size variability.
Because of this, complex forms of Lithic reduction, such as prepared core reduction, can be easily recognized by largely independent patterning in the small, medium, large and occasic)nally extra-large flake size categories. More simple forms of lithic reduction such as pressure flaking form different patterns at single size levels, With the conclusion of MSRT reliability and validity, it is applicable for use with the utility indices (FVI, AAEL and
HAEL). The utility indices were designed to provide quantitative profiles of some basic morphological characteristics of specific flake types. The MSRT provides the flake type structure facilitating the use of the utility indices. The analysis of the utility index reliability and validity was predicated on the assumption that if the MSRT was reliable and valid, such tl-~akcertain technoloqical strategies could provide predictable assemblages of flakes, then, if the utility indices were reliable and valid, they would provide predictable morphological profiles regarding flake size and edge morphology. Having achieved this situation, archaeologists would then be able to produce a series of quantitative expectations linking lithic reduction behavior and patterning in flake removal, discard and storage strategies depending on organizational contexts. In essence, archaeologists could stand on more firm ground in assessing prehistoric dynamics from the perspective of lithic debitage patterning. The reliability analyses demonstrated the presence of
some random error in each of the utility indices. I computed corrections for attenuation and retested the utility indices. This substantially reduced the amount of random ernor present in each, allowing me to cautiously draw reliability conclusions on each. All utility index validity analyses employed correction for attenuation, thereby reducing the effects of random erxor presumably to acceptable limits. The results were robust and easy to interpret and each index achieved construct validity. These conclusions indicate that it is possible to produce a series of utility index based debitage assemblage models linking reduction strategy to flake utility profile. In other words, using a combination of lithie reduction data and utility indices, it is possible to predict MSRT profiles for assemblages which have already had flakes removed or culled for specific uses. Following from this, it should be possible to predict what groups of removed or culled flakes wi 11 look bike given specific selection criteria. Taphonomic controls can be added regarding such pzocesses as trampling (Prentiss and Romanski 1989) to provide comprehensive predictions on the nature of variability in debitage assemblage formation. I turn to this problem in the next chapter. CHAPTER 4
DEBITAGE UTILITY INDEX SEQUENCES
In this chapter, I will discuss the construction of the vitreous traehydaeite (commonly referred to as vitreous basalt [from Cache Creek, B.C.1) debitage utility indices and corresponding debitage assemblage composition sequences. Debltage assemblage composition sequences are mathematically constructed expectations of what different groupings of flakes might be like under different conditions. As the utility indices operate as a "reference dimensionM (Binford 1978:%9) for evaluating some aspects of decision making regarding the usefulness of certain classes of flakes, the assemblage composition sequences serve to place the utility indices in more of a real world context. This is accomplished by considering the effects of behavioral variability in lithie reduction, flake culling and trampling. The basic indices and corresponding utility index sequences serve as a link Between complex behavior and potential patterning in the archaeological record. First, I review the program of vitreous trachydacite experimentation. This discussion includes a consideration of sampling, confounding variables, and lithic reduction techniques. Second, I discuss the construction of the indices and distributions themselves. Here, I review and justify the techniques used to mathematicafly transform the raw MSRT and utility index data into different assemblage composition
238 distmibutions. I also discuss the limitations of these data and their potential applicability. Finally, 1 review the
results of each of the 12 sequences of index construction. The general purpose of this is to gain some understanding of the processes which condition the actual transformations in
the data. This chapter is designed, literally, as a bridge, linking the organization of lithic reduction, trampling and flake culling with potential patterning in the archaeological record. This chapter is, like the previous one, oriented towards producing middle range thesxy (Binford 1981). A body of middle range theory is necessary Eor independent interpretation of any aspect of the archaeological record and it allows us to avoid the use ~f "folk knowledgew in our interpretations of lithic assemblages (Ambck et al. 1988:26). Rather, we develop an "observational languageti(hick et dl.
1488; Binford 1981) for classifying and interpreting the archaeological record.
EITHIC REDUCTION EXPERIMENTS
Since breakage and other debitage characteristics may vary with different types of raw material (quartz, quartzite, cherts, ,igneous flows, etc) , lithic reduction experiments with vitreous trachydacite were carfied out to produce basic data for the assessment of MSRT distributions and the construction of debitage utility indices and sequences. These experiments were designed to provide a range of data on reduction techniques centering primarily on core reduction (prepared and unprepared block core and biiasial esnei and resharpening (pressure, soft hammer and hard hammer) techniques. The experimental design was also conditioned by the need to produce litkis reduction data which could be easily Pinked to specific reduction techniques while av~idingthe pxesence of excess amounts of random or systematic error fn the data. Lithic reduction experiments were designed to examine the appearance of debitage assemblages resulting from a wide range of core reduction strategies (all core and hamerstone size data in Appendix Cf (platform prepared on all prepared cores and bifaces by grinding and minor shaping; all preparation debris included in analysis):
I. Two stage 2 bifaces (Callahan 1979) reduced with small eroded granite hard hammers to produce medium (4 to 16 square cm.) flakes through 10 successful flake removals each. 2. Two stage 3 bifaces (Callahan 1979) reduced with a moose antler billet to produce medium sfzed flakes through 15 successful flake removals each. 3. One small prepared block core (Callahan 1979) reduced witn a medium sized slightly eroded quartzite hammerstone to produce*rnedium sized flakes through 15 successful flake reinovais.
4. One small unprepared spheroid core Ccallahan 1979) reduced with a medium sized slightly eroded quartzite hammerstone to produce medium sized flakes through 15 sueeessful flake removals.
5. One larger prepared block core reduced with a large sized slightly eroded quartzite hamerstone to produce large (16 to
64 square cm.! sized flakes through 15 successful flake removals.
6. One larger unpnepared spheroid core reduced with a medium sized slightby eroded quartzite hammerstone to produce large sized flakes through 15 suceessfu$ flake removals, 7. Two bipolar cores reduced with a large slightly eroded quartzite hammerstone to produce medium sized flakes through 10 successful flake removals each.
8. Two bipolar cores reduced from resharpened stage 3 biface blanks with a medium slightly eroded quartzi te hamme~stoneto produce medium sizea flakes (where possible) through 10 successful flake removals each.
A series of additional experiments were conducted focussing on bifaee and flake resharpening and shaping techniques (a11 platforms prepared by grinding and minor shaping; preparation debris included in analysis):
1. Two stage three bifaces, pressure flaked, using a hafted deer antler tine, for 35 suecessfui flake removals each (no goal in flake removal sizz other than as large as possible, which never exceeded 4 square crn.).
2. Two large flakes, pressure flaked, using a hafted deer antler tine, for 35 successful flake removals each (no goal in
flake removal size other than as Large as possible, which
never exceeded 4 square cm,).
3. Two large flakes reduced along the margins using a moose antler billet, for 15 successful flake removals each (no flake size goal).
4. Two large flakes reduced along the margins using small eroded granite hammerstones for 15 successful flake removals each (no flake size goal).
SuccessfuP flake emo ovals occur when a percussor or pressure flake^ impacts a nucleus of raw material producing a flake in the desired size range (measured on the negative scar left on the cone). I emphasize here that the ideal size range applies only to the total flake removal as evidenced on the negative scar. The actual flake itself may or my not be intact. I continue to employ the concept of flake size goal as presented in chapter 3. Flake size goal refers to the decision by the knapper to attempt to produce a flake of a given size. The actual technique used to produce that flake may also affect other aspects sf that flake8s overall morphology.
The vitreous trachydacite experiments have been designed to monitor purposefully created variation. I am interested in variability in reduction technique associated with a certain degree of consistency in hammer or pressure flaker type. Where possible, data from one nucleus, only, was used in each experiment. Where more than one nucleus was required, combined data from two cores or flakes were used and an attempt was made to ensure that the size and shape sf each of the tvs was as isometric as possible- For each reduction technique and core size class, an attempt was made to use the same hammerstones, billets or pressure f lakexs throughout the experiments. All core reduction was accomplished using slightly eroded hamnekst~nesof different sizes, depending on core and flake goal size. AIL s~fthammer reduction was accomplished using a single moose antler billet. 411 pressure dlaking was done with a single thin deer antler tine.
As the goal of the vitreous traehydacite experiments was to examine technological variability, there was still the need to randomize some aspects of the experimentation. The sequence in which the reduction experiments were carried out could be expected to produce systematic error if not randomized- For example, this could result from unconscious biases on the part sf the knapper moving systematically through reduction continua (e.g. stage 2 and stage 3 bdface followed by pressure flaked biface). To avoid this source of error, the 20 nodules to be reduced ox modified were randomized such that the actual sequence af reduction through these 20 refated in no way to any intuitively normal sequence of lithic reduction (core reduction, too% production, re~harpenfng)~Thus, each type of reduction was carried out as independently as possible from other similarly related forms of reduction. Another source of erron was the time of day that Lithic reduction was conducted. Rather than have a data set be partially conditioned by moods of the knapper, lighting conditions, etc. from subjectively assigned periods of daytime basalt knapping, I decided to randomize the time of day when lithkc reduction experiments were conducted. Thus, the reduction of each nodule was conducted at same randomly assigned time sf day during the period of one week, (November of 1990, at the Simon Fraser Universfty Archaeology DepartmenC flaking pit), This was accomplished by first designing the randomized reduction sequence and then assigning each reduction experiment to a randomly chosen one houx time slot between the hours of 9:QB A.M, and 4:QO P.M. during the period of a five day week. During the experimentation, all flakes were saved on a cotton sheet. At the completion of each experiment, all debitage were seived through a 114 inch (6.3 square m.) screen. Basis MSRT and utility index (FMI, AAEL and HAEL) data were gathered using the techniques described in chapter three. W trampling experiment was conducted to introduce a new source of variability in these data, Teitt% (1900, 1986, 1909) descriptions on the use of winter residential structures are indicative of situations where many people are regularly traversing across surfaces which aka sexve as wo~kareas. Trampling of lithic materials is an inevitable result af this process. The appearance of archaeological remains in the floor deposits sf winter housepits is expected to be highly
244 influenced by trampling. To understand lithic reduction and flake culling and scavenging behavior, one must be able ta
recognize it even where trampling has made modificatLons to
that original data set. Trampling data in and of itself may also serve as a useful indicator sf variability in floor srganlzatlon (Gffford-Gonzalez fit al. 1985; Neilson 1991; Prentiss and Romanski 1989). Following the complete cslaection of the raw MSRT and
utility index data, each experimental assemblage was subjected
to a trampling experiment, This was accomplished following
techniques described in Gifford-Gonaalez et al. (1985) and
Prentiss and Romanski (1389). A layer of mixed coarse and
fine sand, approximately 5 cm. deep was first spread over a
cloth sheet in a one by two meter rectangle. Each experimental flake assemblage (already analyzed for untrampled
MSRT values) wzs spread in a roughly circular pattern in the center of each area. One person wearing soft sofed tennis shoes walked over the scatter for 15 minutes. Each trampled assemblage was then screened through a 1/4 inch mesh. MSRT data were gathered on each trampled assemblage.
CONSTRUCTION OF UTILITY INDEX SEQUENCES
Twelve utility index sequences were undertaken using raw and rescaled MSRT and utility index data (Tables 60-72). The raw data rescaling is the same as that described in chapter
three (see also Binford 1978). This rescaling process serves Table 60. Stage 2 Biface utility indices (CF=CampSete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Sphit Flake; Resc.=rescaled data).
FVX xAAEL U'tiIity 'Indices Index ------.------* ------FVL
a Raw FVI AAEL HAEL x Data Resc. FVI AAEL WAEL Rese. Resc. Resc. AAEL Resc. ( 1. 1 (2) (31 (4) (5 (6) (7) (8 1 (91% (10)
N P Large cn CF PF Med i urn CF PF MDlF SF Small CF 12.0 60.0 7.0 21.8 13.1 6.0 28.4 23.1 15.3 1.7 PF 13.0 65.0 6.1 15.7 18.4 5.2 20.5 32.4 9.6 1.0 MDF 20.0 180.0 4.3 13.2 22.1 3.7 17.'2 30.9 5.7 .6 NF 5.0 25.0 4.4 0.0 29.1 3.8 0.0 52.2 0.0 0.0 SF 5.0 25.0 2.8 12.5 22.1 2.4 16.3 38.9 3.5 .3
*Data values divided by 10. Table 60. Contnd
FVI 100- Csl. 100- Cal. 100- COP. x FVI 13 x AAEL 16 x WAEL 19 x HAEL Resc. Resc. Co1.2 Resc. Resc. Ca1.2 Resc. Resc. Co1.2 (11) (121 (13) (14) (15) (16) (17) (18) (19) (20) Large CF 88.0 16.3 0.0 0.0 0.0 0.0 0.0 0.0 86.6 43.3 PF 525.5 97.3 19.3 9.6 1.0 33.2 16.6 2.0 1.1 5.5 Med i urn CF 539.6 100.0 18.0 9.0 .9 49.6 24.8 3,0 0.0 0.0 PF 141.1 26.1 73.4 73.4 7.6 80.1 80.1 9.7 19.4 19.4 MDF 187.9 34.8 68.3 204.9 21.3 83.3 249.9 30.2 9.9 29.7 SF 190.1 35.2 65.8 131.6 13.7 72.0 144.0 17.4 15.5 31.0 Sma 1L CF 9.2 1.7 94.0 564.0 58.6 71.6 429.6 51.9 76.9 461.4 PF 11.2 2.1 94.8 616.2 64.0 79.5 516.8 62.4 67.6 439.4 MDF 9.5 1.8 96.3 963.0 100.0 82.8 828.0 100.0 61.1 611.6 NF 12.8 2.4 96.2 240.5 25.0 100.0 250.0 30.2 48.8 122.0 SF 6.2 1.1 97.6 244.0 25.0 83.7 209.3 25.3 61.1 152.8 Table 60. Contnd Residual Indices
Trampled HAEL FVE x AAEL FVI % HAEL FVI Residual Residual Residual Trampled Residual Index Index Index Data Index ------_---_-_-_------*------100- 100- FVIx Col. FVIx Col. Raw Col. 13 AAEL 22 x HAEL 25 x Trampled x Col. Resc, Resc. Co1.2 Resc. Resc. Cs1.2 Resc. Data Resc. 29 (21) (22) (23) (24) (25) (26) (27) (28) (29) (30)
I,argt3 CF 7.1 0.0 0.0 0.0'83.7 41.9 4.3 0.0 0.0 0.0 PF 0.1 46.1 23.1 2.3 2.7 1.4 0.1 1.0 5.3 10.2 Medium CF 0.0 58.6 29.3 2.9 0.0 0.0 0.0 1.0 5.3 95.4 PF 3.2 94.7 94.7 9.5 73.9 73.9 7.5 4.0 21.1 154.9 MDF 4.9 94.7 284.1 28.6 65.2 195.6 19.9 6.0 31.6 215.8 SF 5.2 90.4 180.8 18.2 64.8 129.6 13.2 4.0 21.1 138.8 Small CF 75.5 98.3 589.8 59.3 98.3 589.8 60.1 7.0 36.0 345.9 PF 71.9 99.0 643.5 64.7 97.9 636.4 64.8 16.0 84.2 790.2 MDF 100-o 99.4 994.0 100.0 98.2 982.0 100.0 19.0 100.0 963.0 NF 19.9 100.0 250.0 25.1 97.6 244.0 24,8 2.0 10.5 101.0 SF 25.0 99.7 249.3 25.1 98.9 247.3 25.1 3.0 15.0 154.2
091. . 8c0 -w OCO. . 04 m OW. . mcoom.... 00 COCOOPI P m N'O hO 4 OW 0 Oao N om. . om 4
O?. . om Ln om. 00
OW. . om m or-. . ON -3' Table 61. Stage 3 Biface reduction utility indices (CF=Complete Flake, PF=Proxlmal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake; Resc,=Rescaled),
FVI xAAEL Utility Indices Index
FVI Raw FVI AAEL MAEL x Data Resc. FVI AAEL HAEL Resc. Resc. Resc. AAEL Resc. (1) (2) (31 (4) (5) (6 (7) (8 (9)* (10) Med i urn CF PF MDF Srna 11 CF PF MDF SF Table 61. Contnd
Residual Indices
Med i urn CF PF MDF Small CF 'PF MDF SF Table 61. Contnd
Residual I nd ices Trampled HAEL FVI x AAEL FVI X HAEL FVI Residual Residual Residual Trampled Residual Index Index Index Data Index
------1------I------__I 100- 100- FVIx Col. FVIx Col. Raw Col.13 AAEL 22 x HAEL 25 x Trampled x Col. Resc. Resc. Co1.2 Resc. Resc. Co1.2 Resc. Data Resc. 29 (21) (22) (23) (24) (25) (26) (27) (28) (29) (30)
Med 1urn CF 7.8 0.0 0.0 0.0 0.0 0.0 0.0 1.0 2.8 0.0 PF 8.5 27.0 20.8 2.1 40.6 31.3 3.4 5.0 13.9 40.4 MDF 0.0 88.0 67.8 6.9 18.6 14.3 3.5 2.0 5.6 38.3 Sma 1 P CF 4.3 96.6 25.1 2.6 99.7 25.9 2.8 0.0 0.0 0.0 PF 46.2 97.0 397.7 40.5 90.5371.0 39.9 15.0 41.7 369.0 MDF 100.0 98.3 983.0 100.0 92.9 929.0 100.0 36.0 100.0 930.0 SF 2.1 97.7 15.6 1.6 63.4 iO.l 1.1 3.0 8.3 77.4
Table 61. Contnd,
Co'h . Csl . 7 x 8 x Cal . Col . Resc. 29 Besc. 29 Resc. -(dl) (42) (43) (44). (459 tried P urn CP 28.4 27-2 11.3 10.9 2.8 PF 100.0 139.0 57.7. 45.4 11.5 HDF 18.0 20.5 8.5 85.0 21.5 ZSmalf CF 0.0 0.0 0.0 8.0 0.0 PF 48.7 106.8 44.3 141.3 35.7 MDF 71.0 241.0 100.0 396.0 100.0 SF 5.6 29.2 11.3 17.6 4.4 Table 61. Contxd.
Trampled Utility Indices
Csl. Col . 10 x 12 x Col Col 29 Wesc. -29 Resc. ------(46) 447) C481 (49) Hed i urn CF 28.0 27-6 28.0 33.9 PF 181.5 980.0 82.6 100.0 MDF 6,7 6-6 455.8 55.2 Smll CF 0.0 0.0 0.0 0.0 BF 14.2 14.0 39.6 48.0 HDF P'7.Q 16-8 71.0 86.0 SF 1.9 1.9 30.4 37.0
------s______I__s______I__s______I__s______I__s______I__
Table 62, Contnd Residual Index
Trampled HAEE FVI x AAEL FVI x HAEL FVI Resrfdual Residual Residual Trampled Residual Index Index Index Data Index --a------I------_-_ EOO- 180- FVIx Col. FVIx Col. Raw Col .I3 AAEL 22 X HAEL 25 x Trampled x Col. Resc, Resc. Co1.2 Resc: Resc. Co1.2 Resc. Data Resc, 29 (21) (22) (231 (24) (25) (26) (291 (281 (29) (38) Med i urn PF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 9.1 0.8 MDF 1.1 3.3 3.0 ,3 56.9 51.8 5.4 1.0 9,E 48.4 Small CF 49.6 79.8 434.9 49.0 94.4 514.5 53.6 1.0 9.1 78.4 PF 100.0 87.9 878.0 99,O 95.9 959.0 100.0 10.0 91.0 806.3 MDF 62.0 97.6 887.2 100.0 97.5 886.3 92.4 11.0 100.0 955,O SF 15.3 96.0 174.7 19.7 96.3 175.3 18.3 2.0 18.2 167.3 ------.---.-----_--- 0-. dm el* om. . om
OcD. . 04 rn om. 0
00. . 00 m
ON O. 09-4
ON. . OF ma. . mo mo. @a w' . . Qrn
ON. . C)r-WrnN..... OCV WWCV-I'O 0 CV 4 rl
Table 63. Cantnd.
Trampled Utility Indices Trampled Trampled Trampled Trampled Trampled FVI AAEL HAEL FVI x AAEL FVI x WAEL Index Index Y ndex Index Index
Col. Col. Col, Col . 7 x 8 x 10 x . 12 x Col . Col . CoL. Col. Resc. 29 Resc, 29 Resc. 23 Resc. 29 Resc, (41) (42) (43) (44) (45) (46) (47) (48) (49) ------___-__------.------.------Med l urn CF 100.0 64.9 17.7 125.0 41.3 125.0 61.6 125.0 100.0 PF 84.4 250.Q 68.2 11?.8 38.9 203.0 100.0 49.5 39.6 Small CF 37.8 366.8 100.0 117.0 38.7 44.3 21.8 75.0 6.0 PF 69.6261.0 71.2 300.0 99.2 43,O 21.2 26,O 20.8 MQF 16.8 203.3 55.4276.0 91.2 10.5 5.2 7.5 6.0 PIF 48.4 35.7 9.7 195.8 64.7 16.5 8.1 47.3 37.8 SF 51.Q 161.9 44.1 302.5 100.0 31.9 15.7 30.6 24.5 --1----___-_1_1_1____---I------4------.------I - I 1 4 I u--I ot-m I WXI mo I . . I a41 OOWI 2; i &--I om- I XCI HW I > I k.4 I I
I I I I I i I I I I I U11 a, i U I 4 I .a1 C H I I h 1 U I --i I rl I 4 I L, I 3 1 I I 1 I I I I I I
I OOQ 1
Table 66. Cantnd
Trampled HAEL FVI x AAEE FVI x HAEL FVI Residual Residual Residual Trampled Residual Index Index Index Data Index
100- 100- . FVIx Col, FVIx Col. Raw Co1.1.3 AAEL 22 x HAEL 25 x Trampled x Col. Resc. Resc. Co1.2 Resc. Resc. Co1.2 Resc. Data Resc. 29 (21) (22) (23) (24) (25) (26) (27) (28) (29) (30) Large CF 0.0 100.0 111.0 i1.4 0.0 0.0 0.0 1.0 6.7 0.0 Medium CF 51.6 28.2 156.8 16.1 95.0 528.2 53.0 3.0 20.0 169.8 PF 32.2 0.0 0.0 0.0 84.4 374.7 37.6 6.0 40.0 208.8 MDF 5.1 56.9 126.3 13.0 76.6 170.3 17.1 3.0 20.0 143.8 SF 4.4 91.6 203.4 20.9 97.4 216.2 21.7 1.0 6.7 60.8 Small CF 15.4 96.0 106.6 11.0 100.0 111.0 11.1 0.0 0.0 0.0 PF 100.0 99.0 880.1 90.5 99.9 888.1 89.1 7.0 46.7 462.3 MDF 96.3 97.2 972.0 2.00.0 99.7 997.0 100.0 15.0 100.0 990.0 NF '11.1 99.0 109.9 11.3 99.8 110.8 11.1 1.0 6.7 66.3 SF 72.2 98.2 655.0 67.4 99.7 665.0 66.7 1.0 6.7 66.0
Table 65. Contnd.
Col. Col. Col. Col . 7 x 8 x 10 x 12 x Col. Col . Col. Col* Resc. 29 Resc. 29 Resc. 29 Resc. 29 Resc. (41) (42) (43) (44) (451 (46) (47) (486 (49) Large CF 51.2 0.0 0.0 67.0 22.0 0.0 0.0 67.0 100.0 Med i um CF 23.1 188.8 32.6 66.0 21.6 143.6 35.9 10.0 14.9 PF 100.0 243.2 41.9 130.8 62.6 400.0 100.0 62.4 93.1. MDF 43.0 61.2 10.6167.0 54.8 86.2 21.6 46.8 69.9 SF 4.7 57.3 9.9 19.0 6.2 5.6 1.4 1.7 2.0 Small CF 0.0131.2 0.0 0.0 0,0 0.0 Q.0 0.0 Q.0 Pi? 3.6 580.0 22.6 87.8 28.8 4.7 1.2 0.5 .7 MDF 7.6 11.4 100.0 305.0 100.0 28.0 7.0 3.0 4.5 NF .5 0.0 2.0 18.6 6.1 0.7 0.2 .1 .2 SF 12.4 16.3 2.8 314.6 4.8 1.0 .3 .2 .3 H--I P-FcUQ LnQm >m i .... s.. CLYl am0391 4glm I mdva emrp, I 'QIvtWrl N Table 67. Contnd.
Residual Indices
------I-* FVI xHAEL FVI Residual AAEL Residual HAEL Residual Index Index Index Indcx ------I------.-I------Y-l------_-______FVI 100- Col. 100- Col. 100- Col. x FVI 13 x AAEL 16 x HAEL 19 x HAEL Resc. Resc. Col.2 Resc. Resc. Co1.2 Resc. Resc. Co1.2 (11)*(12) (13) (14) (15) (16) (17) (18) (19) (20) Large CF 4694.5 10Q.O 0.0 0.0 0.0 100.0 910.0 40.1 25'3 23,G PF 4622.0 98.5 4.7 4.3 .4 31.2 28.4 12.5 22.9 20.8 MDF 3487.7 74.3 44,s 40.5 4.1 100.0 91.0 40.1 0.0 0.0 S? 1257.4 26.8 76.2 69.3 7.0100.0 91.0 40.1 16.0 14.6 Med i urn CF 1825.6 38.9 53.9 49.1 5.0 100.0 91.0 40.1 37.L 33.8 PF 205.1 4.4 91.1 165.8 16.8 52.0 94.6 41.7 63.5 115.6 MDF 445.0 9.5 86.4 235.9 23.9 39.3 107.3 47.2 48.1 131.3 SF 422.2 9.0 86.0 156.5 15.9 100.0 182.0 80.1 52.0 94.6 Small MDF 11.5 .2 98.5 985.0 10Oe8 0.0 0,0 a.0 87.9 879.0 NF 41,7 .9 96.8 176.2 17.9 100.0 182.0 80.1 79.4 144.5 SF 37.5 .8 97.3 265.6 27.0 83.2 227.1 100.0 78*1 213.2 or-Q'io.... 8Nmw (Vw OrnOlQ a*.. oaao Fi aJ af-! f t? E-c
I I c-mmo I .... I mr-00 I rtwr- I I 1 OdOO I **.. I qm-8 I NmQI I I I m-3'00 1 .... i NNOO I bn I I I Po00 I .... I CDrlOO I w i rl I I Porno I .... I NOIIO I morn rl
" &I ~a~rr-ii Irc mabeh oc~a~b~~rcho~a .~UsXmZ&XzrnmUaZzm I U I u In- I 3 alw I a* fLdS-- r arlrlu I U.cr a4 I JJ Eri 1 Q) 34a I L4 Xv) I 0-0 II I UaJUlt I xELm I am1 I arpk I - I L.iWBrU I rn r: I aa -6, al a\@ E w ux w aaa rg Cf4d u 3-Ezth HZI I I =+u1 4-1 nm mwcncn owe I C& 11-4 ad1 .... e. a1 13x441 aa. dmrnt Ib =--I me dm-N I PW C
I ??'-is f ag'o I ad I cw 1 I moos t '*..I -om0 l cam I 7.4 I 1 OOCnO l . . .I mmU>C3 1 W'mcD I Cnm 1 I morn0 1 .'.a1 --Q096) 1 dPm I rl f I Cn6C00 1 ..a .I -acX)o 8 -am t mm I I a9C%aR8 1 a*. .I s*0-C9 1 am I 4 I 1 S1mO6 1 . .I U3V)CVQ I mew 1 * I I I -ON0 1 ....I marno I 8P 4 I I mOW8 1 0. .I aBricnc3 i mu3 I m9-I 1 I mCSW8 1 . .I WBQ)C> I 04 1 d I I wow0 l . . 'J rnFeOc08 1 rlmm D rn I I 4a3cVa l s.. .l
I * I I U- l I ERm 3 I am l I a- I I f ! I I -see -1 trl prme I I om ome3 I I UdU -I . I I I I La- l 1 tnm 1 1 Q)m I I ffiw 1 I I I I I -X -1 Id dGIW I 9 OLD omm I Ud-IC) -1 I
Table 72. Construction of basic split flake indices from trampled unprepared core reduction (large flake) (Table 69). (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Dlstal Fragment, SF=Spiit Flake).
FVI xAAEL Utility Indices f ndex
FVIxHZBL f ndex ------100- 1130- FVI 100- 100- 100- FVI X FVI X x FVI AAEL HaEL aAEL HAEL HAEL Resc. Resc. Resc. Resc. Resc. Resc. (3.1)------(121 (13) (16)* [19)* 122)* (23)* 96.4------.1 99.3 47.7 85.7 11.3 3.6 *Column numbers correspond to appropriate columns on model construction tables. as the fundamental buildlng block of the modelling process.
It allows the numerical characterization of the utility of a given flake class, provided by the utility index, to be directly converted into some anticipation of that flake class' proportionate representation in some hypothetical archaeological assemblage.. All utility index data are offered as homologs rather than analogs to the archaeological record k e a 1989:8). In othez words. if I have been confounded by problems in the archaeologlcal record regarding variation in debitage assemblages, then 1 can seek to better understand the archaeological record th~oughthe creation of models reflecting actual behavioral variability. I can then explain some aspects of the relationship between behavior and the archaeological record. Having done this, theoretically, I can move back to the archaeological record to better understand the conditions standing behind its formation, This is the process of working "back and forth between experimental work and the archaeological recorde (Amick et al. 1989:6) In order to learn something about the past.
Binford (1978:106) described his modelling process as anticipating the structure of faunal assemblages. My research relies on the same basic technique to quantify a range of variation in lithic dehttage assemblages. To accomplish this, I first.develop the baseline MSRT and utility index data fcols. 1-5, Tables 60-721, The rescaled MSRT data fcsl. 2) provide an indicator of the proportional contribution of each flake class to the assemblage as a whole. 1 have demonstrated that the profiles of HSRT flake types are relatively unique for each reduction type employed. Thus, independent flake
utility index sequences are also required for each reduction technique under consideration.
Three flake utility indices have been created to measure
different aspects of potential flake utility. The Flake Volume Index (FVP) measures overall flake size for each flake
class. The Acnte Angle Edge Length index (BaEL) measures
variation edge lengths with edge angles less than 45 degrees.
The high angle edge length index (HWEL) measures variation in
edge lengths with edge angles greater than 45 degrees (up to
and including obtuse edge angles). when rescaled (cob. 6-81 the indices provide simple predictions of assemblage composition where each flake class is scaled against one another in relation to its size or acute or high angle edge length. They provide an ideal indication of what a profile of flakes, chosen for use as tools on those parameters only, might look like.
Anticipating the appearance in the archaeological record of flake culling also requires a consideration of the joint effects of size and edge morphology. This is accomplished by multiplying the PVI and AAEL and HAEL raw data columns [cols 9 and 111 and rescaling the resulting values such that the
results *are comparable to that of the basic converted utility
indices (cols. 6-81. The basic and combined utility indices sesve to anticipate some potential variability in distributions of flakes chosen specifically for use as tools. These data sets serve best in evaluating the formation of either actual flake tool assemblages or of cached groups of flakes awaiting msdlficatfon. A more difficult problem lies In discerning what has been eulled from groups of reduction flakes left about as unused waste material, This problem is considered through the creation of a series of utility fndex residual distributions (eols. 13-27). The construction of each msidual distribution requires three steps. First, the resealed utility index data kg. col. 6) must be subtracted from a score of 100, creating an exact inverse sf the utility index (eol. 13). This serves as a scale for modifying the original reduction data profile to reflect the removal of certain classes sf flakes. The modification process then is simulated by multiplying the original rescaled nSRT data distribution (col. 2) by the inverse data (col. 13). The nesulting data set (col. 14) is then rescaled simulating the potential appearance of a assemblage of waste flakes, culled for usable item. This mathematical procedure is necessary to maintain the same utility index scale in the residual models as in the orlglnal utillty indices. It also retains important components of the distinctive HSRT distribution enabling the researcher to identify reduction behavior even though culling has occurred.
Researchers in the future may want to explore %he effects sf variability in culling strategies by comparing subjectively culled assemblages to residual models based on utlllty indices, as is presented here. The final problem to be overcome, if Ifthie debitage on a housepdt floor is to be any utility in recognizing strategies
of flake p~~ductionand use, is that of trampling. Trampling
can be expected to have modified Both flake tool and unused debitage asseanbfages. Evidence Esr trampling can be a very useful tool for studying the spatial structure of housepit floors, Thus, the effects of tramplfng were considered. Raw and rescaled tramplinq MSRT data from each ~eduction experiment are provided fn columns 28 and 29. Trampled residual distributions (cols. 30-39) are created by multiplying the flake utility inverse distribution kegecop.
13) by the converted trampling MSRT data (eol. 29) for each reductfon type. The resulting data distribution fcol. 30) is then resealed. This bs the same process as described above, only the effects of trampling have been considered. Trampled utflity index distributions (cols. 40-491 ame constructed by multiplying the rescaled utility index data sets (e.g. csl. 6) by the rescaled trampbed MSRT data (csl. 299. The results of this process (COP. 48) are then resealed such that all data sets are comparable.
UTILITY INDEX SEQUENCES
Inqthfs section f will describe and explain the results of each utP%Pty index sequence. Ts accsmplfsh this, f will start with a discussion sf the resealed HSRT distribution. % will then review each utility fndex sequence, moving from Basic and residual distributfons to trampled distributfxms (Figures 19-1189. The goal of this Bfscussisn will be %a
highlight the processes behind each transfozmatlon. Through this analysis, I am seeking to learn more about the processes which pmo&uce vaniatlon in MSRT data dfstrfbutfons. I: am
particularly interested in the effects sf %lake culling on
originally unmodified MSRT distributions. 1 define culling as the remavaP sf flakes fxom either unmodNied or trampled debitage assemblages for use as tools accoxdfng to criteria set by the utility index scales.
The stage two biface was produced using a msoftwkaxd hammer, with the intention sf making medium sized flakes. The resulting M8WT distxibutfon (Table QQ, COP. 2; Figure 891 reflects hard hamem input with numerous medium and smkl split flakes and smll nonomientabhe fragments. Platfokm preparation is indicated by extremely high small complete flake and proximal fragment scones. While the small flake cateqsry appears to be quite similar to that af prepared core xeduction, the medium flake category contains ind8eats~sof bfface psoduetisn. Complete flake and proximal fragment scores are low, in ceqp~fsonto mediai/dista$ fragments.
This fna?rates that thin medium flakes often snapped producing reduced complete flake scores and increased medlalldfstal fragment scones. Figure 19. Comparison of untrampled and trampled, MSRT stage 2 biface heducti~ndata.
309 Figure 20. Comparison of rescaled FYI, AAEL and HAEL data from stage 2 biface reduction. Figure 21. Comparison of rescaled FVfxWL and FVI data from stage 2 biface reducti on. Figure 22- Comparison of PVI, WLand HAEL %esiduab distributions for stage 2 biface reduct ion.
312 Figure 23. Comparison of FVIXAAEL and FVIxHABL residual distributions from stage 2 bifaee reduction.
3 13 Figure 24. Comparison of trampfed FYI, W%and HWE residual distributions Erom stage 2 biface reduction. 314 Figure 25. Comparison sf trampled FVIx=L and FVIxHAEE residual distributions from stage 2 biface reduet i on Figure 26. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from stage 2 biface reduction.
316 Figure 27. Comparfsun of trampled rescaled FVIxAAEL and FVIxHAEL distributions from stage 2 biface reduction. Figure 28, Comparison of untrampled and trampled MSRT stage 3 biface reduction data. Flgure 29. Comparison of rescaled FVI, AAEL and HAEL data from stage 3 biface reduction. Figure 30. Comparison of rescaled FVIxAAEt and PVIxHnEL data from stage 3 biface reduction. Figure 31. Comparison of FVI, AABL and HAEL residual distributions for stage 3 bigace reduction.
321 Figure 32. Comparison of FVIxAAEL and FVIxMAEL residual distributions from stage 3 biface reduction. - r-i C1
Figure 33. Comparison of trampled FVI, AAEL and HAEL residual distributions from stage 3 bifaee reduction. Figure 34. Comparison of trampled FVIxAAEL and FVSxHAEL residual distributions from stage 3 biface reduction Comparison of tnampled rescaled FvP, WLand HAEL distributions from stage 3 biface reduction. Flgure 36. Comparison of trampled rescaled FVlxAAEL and FVfxHAEZ distributions from stage 3 biface reduction. Figure 3?, Comparison of untrampled and tranpled soft hammer flake reduction data.
327 Figure 38. Comparison of rescaled FVI, AAeL and HAEL data from soft hammer flake xeduction. Figure 39. Comparison cf rescaled EVIxAAEL and FVIxHAEL data from soft hammer flake reduction. Figure 40. Comparison of FVI, AAGL and HAEL residual distributions for soft hammer flake reduction. figure 41. Cumparison of FVIXAABL and FVlxHAEL residual distributions from soft hammer flake reduction- Figure 42. Comparison of trampled FVI, AAEL and HAEL residual distributions from soft hammer flake reduction. Figure 43. Comparison sf trampled FVIxAAEL and FVIxHWEL residual distributions from soft hammer flake reduction Figure 44, @ornpasi.son of trampled- rescaled FVI, WAEL and HAEE distributions from soft hammer flake reduct ion. Figure 45. Comparison of trampled rescaled FVIxAPlEL and FVIxHaEL distxibutions from soft hammer flake reduction. 335 Figure 46. Comparison of untrampled and trampled MSRT hard hammer flake reduction data. Figure 47. Comparison of rescaled FVI, WLand HM3L data from hard hammer flake reduction. igure 48. Comparison of rescaled FVlxAAEL and FVZxHAEL data from hard hammer flake reduction. Figure 49. Comparison of FVI, AAEL and HWEE residual distributions for hard hamner flake reduction. Figure 50. Comparison of FVIxAAEL and FVlxMML residual distributions from hard hamrner flake reduction.
340
Figure 52. Comparison of trampled FVlxAAeL and FVIxHAEL residual dlstributlons from hard hammer flake meduct ion 342 Figure 53. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from hard ham me^ flake reduction. 343 Figure 54. Comparison of tramgled rescaled FVIxkAEL and FVIXHAEL distributions from hard hammer flake reduction. 344 Figure 55. Untrampled MSRT blface pressure flaking data. Figure 56. Comparison of rescaled FVI, AAEL and HAEL data from biface pressure flaking. Figure 57. Comparison of rescaled FVIxAAEL and FVIXHUL data from biface pressure flaking. Figure 58. Comparison of FVI, AAEL and HAEL residual distributions for biface pressure flaking. Figure 59* Comparison of FVIxAAEL and FVIxAAEL residual distributions 5x0s biface pressure flaking. Figure 60. Untrampled MSRT flake edge pressure flaking data, Figure 61. Comparison of rescaled FVI, nAEL and HnEL data from flake edge pressure flaking. Figure 62- ComparPssn of rescaled FVIxAaEL and FVIxHAEL data from flake edge pressure flaking, Figure 63. Comparison of FVI, AAEL and HAEL residual distributions for flake edge pressure flaking. gure 64. Comparison of FVIxA24EL and FVXxHAEL residual distributions from flake edge pressure flaking, Figure 65. Comparison of untrampled and trampled MSRT medium flake, prepared block core reduction. Figure 66. Comparison of rescaled FVI, AAEL and HAEL Oata from medium flake, prepared block core reduction. Figure 67. Comparison of rescaled FVIxAAEL and FVIxHAEL data from medium flake, prepared block core reduct ion, Figure 68. Comparison of FVI, AAEL and HAEL residual distributions for medium flake, prepared block core reduction. Figure 69. Comparison of FVIxAAEE and FVIxNAEL residual distributions from medium flake, prepared block care zeductisn. Figure 70. Comparison of trampled FVI, AkEL and HAEL residual distributions from medium flake, prepared block core reduction. 360 Figure 71. Comparison of trampled FVIxAAEt and FVIXHAEL residuzl distributions from medium flake, prepared block core reduction. Figure 72. Comparison of trampled rescaled FVI, AAEL and distributions from medium flake, prepared HAEL 3 block core reduction.
362 Figure 73. Comparison of trampled rescaled FVIxAAEL and FVIxHAEL distxibutions from medium flake, prepared block core reduction. Figure 74. Comparison of untrampled and trampled KSRT medium flake, unprepared core reduction data. Figure 75- Comparison of rescaled FVI, AXEL and HAEL data from medium flake, unprepared core reduction. Figure 76. Compa~lsonof rescaled FVIxAAEL and PVIxHAEL data fxon medium flake, unprepared core reduction, 366 Figure 77. Comparison of FYI, mLand HAEL residual distributions for medium flake, unprepared core reduction. Figure 78. Comparison of FarIxaAEL and PVIxH-3E;L residual distributions from medium flake, unprepaxed core reduction, Figure 79. Comparison of trampled FVI, AaEL and HAEL residual distributions from medium flake, unprapared core reduction. Flgure 80. Comparison of trampled FVIxWL and FVIxHAEL residual distributions from medium flake, unprepared core reduction. FIgure 81. Comparison of trampled rescaled FVI, AAEL and HAEL distributions from medium flake, unprepared core reduction. Figure 82. Comparison of trampled rescaled FVIwAAEL and FVIxHAEL distributions from medium flake, unprepared core reduction. Figure- 83. Comparison of untnampled and trampled la~geflake, prepared block core reductf on. Figure 84. Cornpaxison 05 rescaled FVX, WWEL and HAEL data from large flake, prepared Slock core reduction. 574 Figure 05. Comparison of rescaled FVIxNUSL and FVIxHAEL data from large flake, prepared block core reduction. Figure 86. Comparison of FVI, APUEL and XAEL residual distributions for large flake, prepaxed block core reduction, 376 Figure 87. C~mparissnof FVIxAAEL and FVXxHAEL residual distributions from large flake, prepaxed block core reduction. 397 Figure 88. Comparison of trampled FVI, AAEL and HAEL residual distributions from large flake, prepared block care reduction. 378 Figure 89. Comparison of trampled FVfxAAEL and FVXXH-~EL residual distributions from large flake, prepared block core reduction. 3 7 9 lgure 90. Comparison of trampled rescaled FYI, and HAEL distributions from large flake, prepared Block core reduction. Figure 91. Comparison of trampled rescaled FVIxRAEL and FVIxHAZL distributions from large %lake, prepared block core reduction. Figure 92. Comparison of untram~,ledand trampled MSRT reduction data. Figure 93. Comparison of rescaled FVI, AA@L and data from large flake, unprepared core reduction. Figure 94, Comparison of resealed FVIxAAEL and FVIxHAEL data from large flake, unprepared core reduction. Figure 95. Comparison OF FVI, AAEL and HAEL residual distributions for large flake, unprepared core reduction.
385 Figure 96. Comparison of FVIx=EL end FVIxHAEL residual distributions from large flake, unprepared core reduction. Figure 97. Cdmparison of trampled FVI, AAEL and residual distributions from farqe flake, unprepared core reduction. 387 Figure 98. Comparison of trampled FVIxWL and PVIxHAEL xesidual distributions from large flake, unprepared core reduction. 388 Figure 99. Comparison of trampled rescaled FVI, AAEL and HmL distributions fram barge flake, unprepared core reduction. Figure 100, Comparison sf trampled rescaled FVIxAREL FVIxNAEL distributions from large flake, unprepared core reduction. Figure 101. Comparison of untrampled and trampled MGRT bipolar core reduction data. Figure 102. Comparison of rescaled FVI, AAEL and HmL data from bipolar case reduction. Pfgure 103. Comparison of rescaled FVIxAaEL and FVIxHAEL data from bipolar c~rereduction.
393 Figure 104. Comparison of FVI, WLand HAEL residual distributions •’on bipolar core reduction. Figure 105. Comparison of FVlxAAeL and FVXxHAEL residual distributions from bipolar core reduction, Figure 106, Comparison a•’trampled FVI, AABL and HAEL residual distributions from bipolar core reduction. 396 Figure 107. Comparison sf trampled FVIxML and FVIxHAEL residual distributions from bipolar core reduction. 397 Figure 108. Comparison of trampled rescaled FVI, AABL and WAEL distributions from bipolar core reduct ior,. Ffgure 189. Comparfson of trampled rescaled FVlxAaEL and FVIxHAEE distributions from bipolar core reduction. 399 Figure 110. Comparison of untrampled and trampled MSRT bipolar core reduction data (bifacial core) . Figure 111. Comparison of rescaled FVI, AAEL and HAEL data from bipolar core reduction (bifacial core). Eiguxe 112. Comparison of rescaled FVfxAAEL and FVIxMAEL data from bipolar core reduction (bifacial cone 1. 402 Figure 113. Comparison of FVI, AAEL and HAEL residual distributions for bipolar core reduction (bifacial core). Figure 114. Comparison of FVIxAAeL and FVIxHAEL residual distributions from bipolar core reduction (bifacial core). Figure 115. Comparison of trampled FVI, AAEL and HAEL residual distributions from bipolar core reduction (bifaeial come!.
405 Figure 116. Comparison of trampled FVIxPLAEL and residual distributions from bipolar reduction fbifacial core). Figure 117. Comparison of trampled rescaled FVI, WLand MaEL distributions from bipolar core reduction (bifaciah core). 409 Figure 118. Comparison of trampled rescaled FVIxAML and FVfxHAXL distributfsns from bipolar core reduction (bifacial esre), 408 Utility indices (Table 60, Cols. 5-8 (Figure 20) exhibit profiles slmiiar to those from the validlty analysis. FYI
scsres are highest in the large flake category as well as in
the medium complete flake category. AiU3L scores are highest fox each size category in the complete flake categories.
Large proximal fragments also contain a high score. HAEE scores are highest in the broken flake categories. The m3sr exception in the medium complete flake score which is the result of a single flake which contained very high edge angles along its circumference. These patterns indicate that all indices are behaving in a manner predicted and explained in the validity analyses (Chapter 3).
The combined FVIxAMZ (Col. 10) and FVXxHWEL (Cox. PI) indices accentuate earlier trends in the basic utility indices
(Figure 21). POI example, the FVIxklhlEL model contains a predictably high large complete flake score, somewhat reduced medium complete flake and large proximal fragments and drastically reduced scores in all sthem categories. This pattern is not surprising given that the intent is to provide a look at the results sf culling decisions focussed on larger flakes with either acute or high edge angles.
The residual distributions (CoEs, 13-29) provide indications of what. stage two biface reduction assemblages could look like having lost ce~tainclasses of flakes through culling (Figures 22 and 24). Most interestingly, the basic shapes of the distributions remain unaltered. The small categories are modified very little. The large category is typically eliminated on at least heavily modifled. This
varies according to utility index. For example, the FVI and
AXEX, residual distributions contain very few of either large
complete flakes or pxoxfrnaf fragments, while the Hm% residual distribution contains an exceptionally high large complete
flake scone and very few large proximal fragments. Large
proximal fragments are removed from bath the AAEL and HWEE
residual dlstributlon because they contain both long acute
angle edges and relatively lengthy broken portions scoring
highly on the HWEC index* The medfun category fu modifled in a similar way. Complete Slakes axe generally removed. RePativefy high scores are maintained for medium proximal and
medial/distal fragments. One exception is the HAEE residual distribution, where med%al/distab fragments are also substantially removed. Trampling has a numben of effects on the MSRT distribution (Col. 29 1. First, it reduces the complete flake and inflates the proximal and medfal/distal fragment scores
(Figures 19, 24-27). It also drastfsa1I.y reduces the nonorlentable fragment score, This pattern has been noted previously in trampled debitage assemblages (Prentlss and Romanski 1989) and ft appears to have been produced by the crushing of many of the nonorientable fragments causing them to faf%*through the 1/4 inch screen. This results in a highlighting effect on the fact that this a bifaee reduction assemblage, despite the use of a hard hammer. The presence of splf % flakes and nonor ientable fragments separates this assemblage from soft hammer bfface production (Table 611, while the inflated proximal and medfal/distal fragment scores
Indicate a difference from other forms of core reduction. The trampled residual fCols, 30-39) and utility index (Col. 46-49) models are much like that of the nontrampled versions, only they reflect the effects of trampling in their distributions. Though, in places, these experimental trampling data are not very robust, they do fit an extremely consistent pattern noted in previous studies (Prentiss and Romanski 1989) and expected from trampied archaeological contexts.
STAGE 3 BLFACE
The stage three biface was produced using a sodt hammer (small hard hammer used to grind platforms only), with the intention of making medium sized flakes. The resulting MSRT dlstributisn (Tabbe 61, Col. 2; Figure 28) reflects soft hammer input with few split flakes and no nonsrientable fragments. Platform preparation is indicated by extremely high small proximal fragment scsmes. Small complete flakes are not common due to breakage during their production. The medium flake category contains equal numbers of complete flakes, and proximal and medial/distaf fragments. This my be the result of the fact that even though flakes produced were thin and often broke, forming the proximal and rnedial/distal categories, enough control was exerted during the reduction process that some complete flakes also survived. This differs from hard hamen biface reduction, where control was reduced and fewer complete flakes were recorded.
Utility indices (Table 61, Cofs. 6-8; Figure 29) are
explained in refezence to the validity analysis. PVI scores are highest in the medium complete flake and proximal fragment
categonies, AAEL scores are high for each size category in the complete flake categories. Medium proximal fragments also score highly here. HAEL scores are highest in the medium medial/distal category.
The combined FVIxML (Col. 10) and FVIxHaEb (Col. 11)
indices accentuqte some earlier trends in the basic utility indices while also creating some new patterning (Figure 30).
The FVIxaAEL model is quite similar to the FVI model, except that it substantially reduces scores from all categories except the medium complete flakes and proximal fragments. The
FVXxHmL index contains high scores in all of the medium categories. It also contains a high small split flake score.
This indicates that where size and high edge angles are involved, there axe more categories with highly useful flakes than that predicted by the FVIxAA3ZL index,
The residual distributions (Cols. 13-27) provide lndlcatfons of what stage three biface reduction assemblages could look like having lost certain classes of flakes through culling (Figures 31 and 32). As occurred with stage 2 bifaces, the small categories are modified very little. The medium complete flake category is typically removed. The medium proxfmal flake category is generally reduced, except in the case sf the HAEL residual distribution (Col. 213 where it is inflated. Not surprisingly, medium medial/distal fragments
are reduced substantially only in the HAEL culls (C~ls,21 and
27). Trampling has several effects on the MSRT distribution (Table 61, Col. 29; Figures 28, 33-36). First, it has the typical effect of reducing the complete flake and inflating the proximal fragment scores. Medium medial/distal fragments appear to have been broken enough to fall into the small size class. Interestingly, the small split flakes increase due to breakage of small complete flakes. The overall MSRT pattern is not greatby modified and the profile maintains its standard appearance. The trampled residual (Cols. 30-39) and utility
index (Col. 40-49) distributions are similar to the nontrampfed versions, only they reflect the effects of trampling in their distributions.
SOFT WANMER FLAKE RETOUCH
No flake size goal was used during the modification of a flake with a soft halmer. The MSRT distribution (Table 62, Col. 2; Figure 37) reflects soft hammer input with numerous mediaf/distal a~dproximal fragments and no nonorlentable fragments. A surprisingly high number of small split flakes are present. This may be due to the fact that reduction was carried out using the flat ventral face of the flake as a platform, thereby inzurinu splitti~igin some of the flakes. The primary difference between this form of reduction and other forms of tool production or core reduction is the high
small proximal fragment score. This pattern also shows up in pressure flaked assemblages. The soft hammer flake retouch assemblage differs from pressure flaked assemblages with the
presence of split flakes as well as medium sized flakes. Utility indices (Table 62, Cols. 6-8; Figure 38) continue to •’Itthe patterns predicted in the validity analysis. FVI scores are highest in the medium proximal fragment category. aAEL scores are high in the small complete flake and proximal fragment categories. The AAEL score is excessively high in the medium medial-distal fragment category. HAEL scores are predictably high (respectively) in the small and medium mediaf/distal categories. The medium proximal fragment category has an unexpectedly high score. Variability in the medium size categories is primarily due to the accidental removal of two larger than expected flakes. It is critical to note here that while the indices used in the modelling process discussed in this chapter may be reliable and valid in the statistical sense, idiosyncratic actions during the experimental reduction process will cause some variability in Individual sequences. This is one such idiosyncratic data set, This i3 not a bad thing, however, as it provides some indication sf the variability which can occur in archaeological contexts.
The primary focus of the FVIxAAEL (Col. 10) and FVIxHaEL
(Cof. 11) Indices is on the medium size flake categories
414 (Figure 39). FVIxAAEL scores are equally high, while the small flake scores are relatively low. The FVIxHWEL index highlights the medium proximal flake over the medium mediaf/distal. These distinctions are not surprising given the high FVI score (Cols, 3 and 6).
The residual distributions (Cols. 13-27) provide indications of what soft hammer flake retouch could look like having lost cextaPn classes of flakes through culling (Figures
40 and 41). Culling of retouch (and pressure) flakes may not have been common in many contexts during prehistory, except when constructing tools with micro-lithic components.
Regardless, I consider it an important excercise in recognizing debitage assemblage variability to explore the effects of culling on these types of reduction assemblages. Only in this way can we hope to recognize the full range of archaeological variation possible. The small flake categories are modified very little, with the significant exceptions of the AaEL distributions (Gols. 18 and 24). Here, in both cases, some culling removes enough small proximal fraqments that small medial/distal fragments assume the highest score. There is an accompanying decrease also in small complete flakes. This pattern nay be explained in reference to the
AAEL scores in columns 4 and 7, where small proximal fragments and complete flakes score quite highly. Thus, in a culling context, the result is an altered small flake profile wherein the scores of the proximal and medial/distal fragments are reversed. All medium flake residual indices vary in direct 415 relationship with utility index values such that where utility
indices are high fi.e. above a raw score of 30.01, that flake category is removed and vise versa.
Trampling has one primary effect on the MSRT distribution
(Table 62, Csl, 29; Figures 37, 42-45). It causes a drastic reduction in complete flakes and a reversal in the small proximal and medial/distal f ragmen& scores such that the mediakfdistal fragments assum the highest score* otherwise,
MSRT pattern is not greatly modified and the profile maintains its standard appearance, As usual, the trampled residual fCols. 30-39) and utility index (Col. 40-49) distributions are similar to the nontrarnpled versions, only they reflect the effects of trampling in their distributions.
HARD HMMER FLAKE RETOUCH
As in the ease of soft hammer flake retouch, no flake size goal was attempted during hard hammer flake retouch f Table 63, Col, 2; Figure 46 1 . The use of a hard hammer f s reflected in the appearance of nonorientable fragments and xelatively high numbens of split and complete flakes. The distribution is emi in is cent of that produced by the stage 2 biface, which was also produced with a hard hammer. It differs-from this assemblage, however, with its high split flake count which is the result of hard hammer contact with the flat ventral flake surface. Complete flakes are fan too numerous fox any resemblances to core reduction assemblages. Utility indices (Table 63, Cols. 6-8; Figure 49) fit some of the patterns predicted in the validity analysis in addition
to providing some interesting new patterning. FVI scares are
highest in the medium complete flake category, followed by medium proximal fragments and small nonorlentable fragments.
Nunorientable fragments often receive relatively high FVI scones due to their thickness. AAEL scores are highest in the medium complete flake and proximal fragment and small cornple te
flake categories. The AAEL score is highest in the medium proximal fragment category. With one exception, HAEL scores are highest in the categories where flake breakage is most extensive such as the rnedial/dfs%al and nonoxieatable
fragments. Medium complete flakes score highest however. This Is due to the presence of one particularly thick flake with high edge angles. This kind of patterning is to be expected where hard hammer reduction is involved, as Larger flakes are often thicker (Hayden and Mutchings 1989).
The primary focus of the FVfxML (Col, 101 and FVPxHAEL (Col. 11) indices is on the medium complete flake category
(Figure 48). The FYIxAAEL score 1% also high for the medium proximal fragment category. The FVfxHAEL scores axe all low with the exception of medium complete %lakes. This patterning
Is due to the substantial size of the medium complete f%akes
in comparison to all other flakes produced. 1% fs a pattern which may well be common in hand hammer retouch assemblages where the comon pattern is to produce small flakes mixed with the occasional larger flake. The ~esidualdistributions (CoSs. 13-27] provide indications of what hard hammer flake retouch could look like havlng lost certain classes 0% flakes through culling (Figures
49 and 50). Without exception, the small flake categories are modified very little. This is due to the comparatively high index scores in the medium f Hake class, compared to the relatively low scores in the small size class. Culling variability ks more directly reflected In the medium flakes. With the exception of the MLresidual distzibution (Col.
181, the medium complete flake category is nrcmsved. As the
AAEL focusses on acute edge angle, the thicker: medium complete flakes are left relatively undistu~bedin the AAEL residual distribution.
Tra~plinghas two major effects on the MSRT distribution
(Table 63 Col. 29; Figures 45, 51-54), It causes a drastic reduetfon in complete flakes and a reversal in the small proximal and medial/distal fragment scores such that the proximal fragments assume the highest score. This is due to breakage in the small complete flakes contnibuting to the small proximal fragments. AddPtisnaE breakage in small medfal/distal fragments produces reductions fn the numbers of flakes in this category. The trampled residual (Cols. 38-39) and utility index (Ccf. 40-49) distributions are similar to the nontssmgled versfsnsi only they refleet the effects sf trampling Ln their distributions. Host notably, trampling reduces the contributfon of the medium complete flakes, which produces lnc~easedutility index scores in many of the smaller categories CCols. 41-49).
STAGE 3 BIFACE PRESSURE FLAKING
Pres%urd-flaking was conducted without any flake size
goal in mind other than to pwoduce the largest flakes possible
(Table 64, Col. 2; F~~uE€!55). The pressure flaking technique
used was to grind the edge thonoughly and then to prepare a
single platform at one end of a bong ground edge. The
platform was prepared by first rem~vinga small flake by
pressing down from one face near the edge in a manner similar
to that of notching. This set up the first of a series sf platforms. After this, pressure flakes were removed by
applying Ssnce at the ground flake edge inward into the body sf the bfface at an angle parallel to the face of the biface
from which the flake was to be removed. Small complete flakes and proximal and medial/dista%
fragments wene recovered. Proximal and complete flakes are
the most common. MedialIdfstaP fragments are relatively rare. This differs from any other f~rmsof reduction where small medialddistal fragments are typically the most common type sf artifact. The csmoa appearance of platform bearing flakes may be the result of three major factors. First, the edge sf a biface is thin enough that pressure flakes can be readily removed. Yet, it is thick enough that platforms are extremely strong and resilient to shatteafng. Second, the high degree oE control and more gradual application of force applied in 419 pressure flaking Lends more towards the surv%v%fof platforms dumPmj the reduction process. Three, when platforms and bulbs of force do shatter, the debris is so small that it rarely is captured by the 114 inch screen. Much of it would n~tsurvive even a 1/8 inch screen.
Utility indices (Table 64, Csls. 6-8; Figures 56 and 57) fit the patteras predicted in the validity analysis. The FV% 5c~reis highest fn the complete flake category followed by proximal and medial/distall fragments with equal scores. The
AAEL score is illso highest in the comgllete flake category.
The HAEL index scores predictably highest fn the medial/distaB category. As all flake sizes are relatively close, %Re
FVIxWL (COP, 18) and FVPXHAEZ (COX. 11) indices are similar to the AAEL and HAEL counterparts. The residual distributions (601s. 13-27 also closely mtch the utility index values (Figures 58 and 59). All WAEk residual distributions are missing medial/dfstal fragments, while the WLdistributions are missing complete flakes. No trampling experiments were conducted with the pressure flake assemblages. Trampling of small flakes results in losses of a11 flake types (Prentiss and Womanski 1989). In pressure flake assemblages, very little could be expected to be collected by a 1/4 inch screen after intensive trampling. Small biface and core reduction flakes are teically thicker and survive the trampling process better. These assemblages also produce new small flakes through breakage of larger flake types . FLAKE EDGE PRESSURE FLAKING
The sane techniques were used to pressure flake the margin of a flake with an acute edge angle (approximately 30 degrees) as were applied to the pxessure flaked biface (Table
65, Col. 2; Figure 601. As before, small complete flakes and proximal and medial/distal fragments were recovered. Proximal and medial/dlstal fragments are the most common. Complete flakes are relatively rare. This differs from the biface pressure flake pattern which contained numerous complete flakes and few mediaf/distal fragments. The reduction in platform bearing flakes Is primarily the result of the thin flake edge often collapsing under pressure. This results in fewer platform bearing flakes and more numerous rnediak/df stal fragments.
Utility indices (Table 65, CoPs. 6-8; Figuse 61) are little different from those described for the bifaclal pressure flaking. FVI scores are highest in the complete flake category followed by proximal and medial/distal fragments, The AAEL score is also highest in the complete flake category. The H-AEL index scares predictably highest Ln the rnedial/distal category, As all flake sizes are relatively close, the FVfxlViEL (CoL. 10) and FVIxHAEL ICol. 11) Indices are similar to the aAEL and HAEL counterparts (Figure 62).
The residual distributions (CoZs. 13-27) also closely match the utility index values (Figures 63 and 64). A11 HmI; residual distxiSutions are missing mediaPJdistal fragments,
while the AAEL models are rnissknq complete flakes. No
trampling experiments were conducted.
PREPARED BLOCK CORE REDUCTION (MEDIUH SIZE FLAKE PRODUCTION
GOAL f
Prepared block core reduction was carried out with the
intention of producing medium sized flakes (Table 66, CoE. 2;
Figure 65). The use of a hard hammer is reflected Ba the appearance of nonorlentable fragments and relatively high numbers of split flakes. Small complete flakes are rare while small proximal fragments are common. Small proximal fragments axe produced in prepared core reduction by shaping platform
for major flake removals. Since the platform angle 1s high, ranging from 75 to 90 degrees, Pt is cornan that the small edge shaping flakes terminate in step fractures. Thus, the flakes are classified as proxianal fragments. The use sf edge preparation in core reduction allows for greatex control over the p~oductionof %lakes. In other words there 1% less flake breakage during reduction. This gattern 1s demonstrated in the medium sized flake prof ble. Mere, complete flakes are most cormon, followed by proximal fragments. One large complete flake was also produced.
Utility indices (Table 66, Cols. 6-8; Figure 56) fit most sf the patterns predicted In the validity analysis. FVI scores are highest in the large complete flake category, followed by medium proximal fragments. AAEL scores are highest In the medium and small ~smpleteflake and medium proximal and split flake categories* 1% is interesting to note here that FVI scores can be very low as in the case of the small complete flakes, AAEL scores can be exceptionally high. The magnitude of the converted score depends Pn part on the structure of the entire distribution. In this case the category with the largest FVI contains no edge angles less than 45 degrees. This leaves small complete flakes as the category with the highest WLscore. Correspondingly, the highest HAEE score Ps found in the Barge complete flake category. Other than this high score H&EL scoxes are highest in flake categories with high breakage rates, such as the medium and small media%/distal fragments.
The primary focus of the FVPxAM3L (Csl. 10) index is on the medium flake category, focussing on proximal fragments and complete flakes (Figure 67). The PVIxHAEL index (Csl. 12) scores highest on large complete flakes. his is no surprise as both the FVI and HAEL indices are highest on this category.
The residual models (Cols. 13-21) provide a variety sf modifications to the original MSRT distribution (Figure 68).
The FVI residual distribution (Coi. 15) leaves the ofiglnal distribution little changed with the exception of the removal of the large complete flake. The aBlEE residual distribution (Col. 18) is culled heavily for medium and small complete flakes as well as medium split flakes and smhl medial/distal fragments. This results Ln inflated numbers of small proximal fragments and deflated numbers of small rnedial/d%stal
Eragments, The H2bEL residual dfstxibutisn varies fsom the
original MSRT distribution in two major areas: first, the large complete flake has been removed, and second, small medhal/distaP fragments have been culled enough $0 drop their score and inflate the proximal fragment scone,
As anticipated by the FVlxAAEL and FVIxHAEL utility indices, less variability is found Pn the FVIxML or FVIxHML residual distributions (Figure 69). Here, the former is culled only for medium complete flakes and proximal fragments.
The latter modei indicates culling only of %he Pange complete flake category.
Trampling has two interesting effects on the NSWT distribution (Figures 65, 70-73). Finst, Pt causes xeductisns in all flake types in the small flake categories, with the exception of medial/distal fragments. HedhaP distal fracgments become Inflated through the addition sf new pieces by way of breakage in the larger categsnbes. Other small flake types axe generally $noken beyond the size tha% they are typically seeoverad in f/4 inch screens, Second, medium complete and split flakes are reduced with P%$%Pe corresponding modification ts the other medium sfzed flakes types. These effects extend through the traapled residual
(Cols. 30-39) and utility index (Col. 40-49) distributions-
UNPREPARED BLOCK CORE REDUCTION (MEDIUM SIZE FLAKE PRODUCTION Unprepared block core reduction vas carried out wlth the
intention of producing medium sized flakes (Table 62, Col. 2;
Figure 74)- The use of a hand Aamer is reflected in the appearance of no~srlentable fragments and relatively high numbers of split flakes. Small complete flakes and proximal fragments are not present. This reflects the lack sf edge preparation used to produce larger flakes. Corresponding to
the lack of edge preparation is a near lack of medium complete flakes. There are, haweven, numerous proximal and medial/distal fragments resulting from breakage during reduction. This is largely, again, due to a lack of prepared platEorms. Another biproduct sf unprepared platform was reduced precision on the part of the knapper In achieving the appropriate size goal. Thus, the Barge category contains four flakes (each of a different type).
Utility indices (Table 67, COPS. 6-8; Figure 75) generally fit most of the patterns predicted in the validity analysis. Some variability, however, is present. FOE each size "tier," FVI scores are highest in the complete flake category, With the exception of the large proximal fragment,
FVI scores drop off rapidly below the complete category for each size group. High $&EL scores are found in the prsxlml and medial/distaf fragments. This my indicate that wheme thinner flakes were produced, breakage accugred resulting in proxinml and mediaf/dfstal fragments with high AAEL indices,
This conclusion is supported to some degree by the HAEL distribution which contains high scores on complete and split flakes as well as some medial/distal fragments,
The FVIxaaEE (Col, 10) index is qufte simple, with a high score only on large proximal fxagments (Figure 76)- The
FVIxHAEL index [Col. 12) is more complex, grading down from large complete flakes to medium complete flakes. This is a
direct reflection of similar dis%nlbut%ons in the FVI and HAEL indices.
The residual distributhons (Cols. 13-27) provide lndicatfans nE culling primarily in the Larger flake class
(Figures 77 and 78). This is not to say that eullfng cannot Be seen in the medium or small categories. Rather, it occurs more subtly through slight reductions in certain C~~SS~Sof flakes. The MEresidual index distribution fs the only model which contains drastic differences fn the snna3.P and medium size categories from the original MSRT. In this case intense culling for small complete flakes and medium medial/distaP and proximal fragments has reorganized the ddstributfons to a great degree. Thfs is because small rnediai/dfstal fragments contain the lengthiest edges with acute angles. In all other cases sf culled prepared core reductfun assemblages, culling does not prevent the easy recognition of unprepared core reduction. Culling variability
in the PVfxWL and FVIxHAEL residuaP indices (Cols. 24 and 276 only appears in the large categories, The primary effect of trampling is the removal of medium and large complete flakes with slight consequent increases in
426 large proximal fragments and medium split flakes (Figures 74, 79-82). These effects extend through the trampled residual
(Cols. 30-39) and utility index iCal. 40-491 distributions.
PREPARED BLOCK CORE REDUCTION (LARGE SIZE FLAKE PRODUCTION
GOAL )
Prepared block core reauction was carried out with the
intention of producing large sized flakes (Table 68, Col. 2;
Figure 83). The use of a hard hammer is reflected in the appearance sf numerous nonorientable fragments and relatively high numbers of split flakes. The inflated numbers of
nonorientable fragments may be the result of increased force used to detach large flakes. The presence of small complete
flakes and proximal fragments reflects edge preparation. The medium sized flake class contains fragments of originally
larger flakes (primarily proximal and medial/distal fragments). The large flake category is dominated by complete
and proximal fragments f though some rnedial/distal fragments are also present). This fits the pattern described for
prepared core reduction where there is a relatively high
success rate in flake production.
Utility indices (Table 68, Cols. 6-8; Figure 84) fit the patterns predicted in the validity analysis. The large flake categor Fes dominate the FVI indices. Not surpzisingly, the
large complete flake category scores highest, AAEL scores are
generally low (Col. 4). With the predictable exception of the small complete flake category, -?EL scores are highest in the
medium and large flake Oategories- Highest NAEL scores fall in thosg catgories with maximum breakage. The large complete
flake category also contains a high HAEL score, as these are
relatively thick flakes with high edge angles.
The FVIxAAEL (Col. 18) index is quite simple, with a high
score only on large proximal fragments (Figure 85). The
FVIXHAEL index (col. 12) is only slightly more complex, with a
high score on large rnedial/distal fragments and an addltfonally high score on large complete flakes. Edge angles on these flake types axe almost entirely obtuse.
As was the case for the medium flake production cores, the residual distributions (Cols. 13-27) provide indications of culling primarily in the larger flake class (Figures 86 and
87). Again, this is not to say that culling cannot be seen in the medium or small categories. It appears more subtly through minor reductions in certain classes of flakes. All large flakes are heavily culled from the FVI residual distxibution, The AXEL residual distribution is cuffed only in the large proximal fragment category. The HAEL residual distribution is culled from the large medial/distal fragments.
The FVfxdlAEL and FVIxHAEL residual distributions closely resemble the basic AAEL and HAEL residual models respectively.
The primary effect of trampling is in the removal of small and medium nonorientable fragments and large complete flakes with slight consequent increase in medium rnedial/distal fragments (Figure 83, 88-31) . During trampling, nonorientable fragments tend to be crushed to a size which is not recoverable in 1/4 inch screens. Though medial/distal
fragments are often broken to a size less than I/$ inch, other more conplete flake types contribute new medial/distal
fragments (but not nonorientable fragments in sizes large enough not to pass through the f/4 inch screen). The result is that during trampling, the medial/distal fragment category is often little modified. These effects extend through the
trampled residual fCols. 30-39) and utility index (Cof. 40-49) distributions.
UNPREPARED BLOCK CORE REDUCTL ON f LARGE SI ZE FLAKE PRODUCT1 ON
GOAL
Unprepared block core reduction was carried out with the intention of producing large sized flakes (Table 69, Col. 2; Figure 92). The core used was a flattish cobble which produced a distribution of large flakes dominated by large complete flakes, This pattern is typical of cobbles of this shape {Prentiss et al. 1988:Table 7, Col. C2) since the thinness of the cobble allows easy production of complete
flakes without much edge preparation. Were the core thicker, production of complete flakes without edge preparation would have been much more difficult. The use of a hard hammer is reflected in the appearance of numerous nonorientable fragments and split flakes. The inflated numbers of nonoxientable fragments are certainly the result of increased
4 29 force used to detach large flakes, Small complete flakes and proximal fragments are all but lacking, indicating that edge preparation was not practiced. As In prepared large flake core reducti~n, the medium sized flake class contains residual flakes, or those fragments of originally larger flakes (primarily proximal and medialddistal fragments). Utility indices (Table 69, Cols. 6-8; Figure 93) again fit the patterns predicted In the validity analysis. The large flake categories dominate the FVf indices. Not surprisingly, the iarge complete flake category scores highest. AABL scores are generally low (Col. 4). AAEL scores are highest in the medium proximal and medial/distal categories. As indicated by the FVI and HAEL scores, these are somewhat smaller and thinner flakes with lower edge angles around their circumferences. Highest HAEX scores fall Pn the large size classes and bn those medium categories with maximum breakage (medial/distal and nonorientable). Due to the extreme size of the large complete flakes, the
FVIxAAEL (Coi. 10) and FVIxMAEL indices (Col. 12) are dominated by the large complete flake category (Figure 94). W relatively high FVI/HAEL score is found also in the large proximal fragments.
As was the case for the medium %lake ,DEQ~uc~.~@~cores; the residual distributions !Cob. 13-27] provide indications of culling primarily in the larger flake class (Figures 95 and 96). However, due to the high AAEL indices, the AAEL residual distribution contains strong indicators of culling in the
430 medium sized flakes. Most large flakes are heavily culled from the FVI zesidua? distribution. The HAEL residual distribution is culled from the large flake category. The
FVI/WL residual and FVI/HWEL residual distrlbutians closely resemble the FVI residual distribution. Again, size seems to be the critical factor.
Trampling results in the removal of complete Efakes and nonorientable fragments (Figures 92, 94-100). Often trampling causes increases on the numbers of medial/distal fragments and proximal fragments. As in this case, damage to all classes sometimes prevents these increases. Thus, the only flake type with dramatic increases in numbers are the small medial/distal fragments. Medium split flakes slide down into the small split flake category due to distal fracturing. These patterns extend through the trampled residual (COPS. 38-39) and utility index (Cal. 40-49) distributions.
BIPOLAR CORE REDUCTION (MEDIUM SIZE FLAKE PRQDUCTION GOAL)
Bipolar core reduction typically produces a Sullivan and
Rszen typology pattern dominated by medialtdistal and nonorientable fragments (Prentiss and Kui jt n.d. ; Xuljt et al. n.d.1. This is the result of intense crushing and fragmentation at the proximal and distal ends of the flakes resulting in few recognizable platforms. Also, a fair amount of debris hs created during the crushing and flake initiation process that can only be defined as nonorientable, as no 431 single ventral surface can be recognized. As the flake size production goal was medium flakes, nunerous medium sized
flakes and several large sized flakes were produced (Table 76, Col. 2; Figure 1013. All are classified as medfalPdista1
fragments. In the small size class, two flakes retained enough platform that a fracture initiation could be recognized
(proximal fragment and split flake), The most eomon flake
Utility indices (Table 76, Cols. 6-8; Figure 102) fit some sf the patterns predicted in the validity analysis. Same variability is introduced due to the unique strategy sf flake
production used FA bipolar technology. The la~ge medial/distaf fragment category contains the dominant the FVI
index. By comparison, all other categories score quite low.
AAEE scores are generally low (Colt, 1) with highest scores in the mediuzi and large medialldistal fragments and small split flakes. It is interesting that the small split flake WL
index is highest as this flake does contain one of the most intact margins. Highest HAEL scores Ball fn the large and medium size class and in those small categories with maximum breakage (rnedial/dlstal and nonorlentable). Due to the extreme size of the large medtal/distal fragments the FVIxAAEL
(Col. 10) and FVIxHAEL indices ((201. 12) are dominated by the large medial/distal fragment category (Figure 103).
As was the case for the medium and large flake production cores, the residual distributions (Cols. 13-27) provide indications of culling primarily in the larger flake class
432 (Figure 104). However, due to the high kAEL indices, the AM3L residual model contains strong indfcators of culling of small split flakes. Most large flakes are heavily culled fxom the
FVI residual model. The HAEL residual model is culled for large flakes. The FVIxAAEL residual and FVIxHAEL residual distributions closely resemble the FVZ residual distribution (Figure 105). The primary effect of trampling is the removal of nonorientable fragments (Figures 101, 106-109). Otherwise the distribution is little modified, as there are no complete flakes to begin with, The trampled residual (Cols, 30-39) and utility index (Col. 40-49) distributions reflect this patterning.
BIPOLAR REDUCTION ON BIFACIaL CORE (MEDIUM SIZE FLAKE PRODUCTION GOAL)
Bipolar reduction of a bifacial core produced a fairly typical MSRT distribution for bipolar reduction, although it is somewhat Pow in small nonorientable fragments (Table 71,
COP. 2; Figure 1101. This is primarily due to the fact that the core was thin at both ends with subsequent increased control over fracture inhtiations and propagations. This resulted in reduced nonoxientable fragments and increased medfal/distal fragments. As the cores were relatively thin, it was difficult to detach as many medium size flakes. Thus the number of medium or large flakes is far less than that of
433 the pzevlous bipolar experiment. Utility indices (Table 71, Cols. 6-8; Figure 111) are
somewhat different from the previous >*-miar experiment. Though the FVI scores are fairly typical, with the highest
Eound on the large medial/distal fragment, the M%scores
more closely resemble bifaee production than sore r%duction. This should come as no surprise since the core here is a
biface. his indicates that bipolar reduction can produce
useful acute edged flakes from bifaees. HAEL scores are highest among the broken flake types, especially the
medlal/dlstal and nsnorientable fragments, Due to the extreme size of the single large medial/distal fragment the FVIxAAEE (Col. 10) and FVIxHAEL indices (601. 12) are dominated by the large medial/distal fragment category (Figure 112).
The residual dlstributlons (Cols. 13-27) provide indications of culling in the large and medium flake classes
(Figures 113 and 114). The major focus, however is fn the large class, whish is removed in all cases except the HAEL residual distribution. The small flakes are not modified in any of the models. The primary effect of trampling is the removal of the large medial/distal fragment, the medium complete flake and most of the nonorientable fragments (Figure 110, lb5-lf8j, The distribution still maintains is characteristic shapz. This is maintained in the residual (Cols. 30-39) and utility index (col. 40-49) distributions. SUMMARY
The intent of this chapter has ken to discuss %he vitreous trachydacite experimental research. The research was designed to explore the effects of core types, technological behavior, flake culling behavlsr and trampling on distributions of Modified Sullivan and Rozen Typology flaie types. This involved, first, constructing an experimental design which produced data sn technological variability, while eliminating bias from such factors as time of day when reduction was conCucted or the order in which individual flakes and cores were reduced Following the production of the basic MSRT data, utility indices were constructed and utility index based distributions were produced. There are a number of specific conclusions which can be drawn from the discussion of these data sets. From a technological perspective, the MSRT functioned well in segregating assemblages in several areas: reduction technique, hammer type, use of edge preparation and flake size goal. I review reduction technique last, as the identification of reduction technique used, largely depends upon input from the three other areas.
The use of hard versus soft hammers showed up well in the MSRT data. Soft hammer reduction was consistently indicated by reduced numbers of small, medium and large complete flakes and numerous large, medium and small proximal and medial/distal fragments. Conversely, hard hammer reduction produced numerous complete flakes of all size elasses. Soft hammer reduction rarely produced nsns~ientable fragments, whfle hard hammer reduction always pzodueed this Pfake type, Fxessure flaking was distinctive with its foeus on smlB complete flakes and g~aximal.and medial/dista% fragments . Edge preparation was also indicated in these data. Any time edge preparation was used, the small complete and proximal ffaka types were relatively abundant. when ht was not employed, these flake types were not typically present. Some varfab%lfty in their proportionate occurrence occurred dependf ng on kamer type. Flake size goal was represented by variation in size classes as well as breakage dfstsibutfons in each size elass. . For example, where the flake size goal was oriented towards large flakes, the medium flake class contained broken flakes from the Pargen category. This also oeeunred when medium sized flakes were the goal of reductisn, except that the residual broken medium sized flakes were found in the small slze category. Given the success of these three areas, it was easy to Isolate technological variation. Bi face reduction varied, depending on haamer type used. Generally however, proxfmaf and medbaf/distal fragments were most common. Although similar-to bigace reduction, flake netouch was distinguishable due to the relatively high numbers of complete flakes and proximal fragments produced compared to the lower numbers of medial/dfstal fragments. Pressure flaking was distinguished due to the exceptionally high numbers of proximal fragments produced, as well as by its narrow production focus on small
complete flakes and proximal and rnedial/dfstal fragments. Prepared and unprepared core reduction were easily recognized due to high numbers sf nonorientable fragments and split
flakes and very low numbers of %ma13 complete flakes. Cone reduction was also indicated by the presence of a wide range of larger flake types. Bipolar cove reduction continued to
fit patterns explored previously (e.g. Prentiss and Kuijt n.d.), containing exceptionally numerous medialddistal and non~rientablefragments and few other flake types. Utility index data were useful for anticipating the effects sf culling or scavenging behavior. Most notably, the independent technological origins of each model were generally prese~vedthroughout each utility index sequence. This was due primarily to the fact that %he structure of the MSRT distribution canditioned the general structure of each distxfbutfon within each sequence. Xn other words, as each distributlsn was derived from a distribution of broken flakes typical of a reduction strategy, the resulting utility profile also reflected that unique reduction strategy. P will now review how thds affected ehch index between the different reduction techniques.
The FVI was designed to measure variation in flake size alone. Its raw and converted application to actual assemblages also reflected the technological origins of those assemblages. This is demonstrated in the scale of variation across the MSRT flake types. For example, pressure flake
assemblages and soft hammer reduced bifaces and flakes contain less varfatlon in scores than do core reduction assemblages.
This is the kind of flake size variability described by mlex (1989:Ffgure 1). Naturally, ea~lierstage hiface production
and core reduction produce a substantially greater variety sf
flake sizes than does latex stage biface production or edge retouching,
The AAEL index was designed to measure variation in
length oE flake edge containing acute edge angles. This index varies also with different neduction techniques. Essentially, techniques pmoduePng thin flakes tend to produce high ME& scores. Conversely, those same techniques produce more broken flakes thus reducing the AAEE scores in some categories. Technological variation can be monitored By examining relationships between converted WEscores and and the actual numbers of each flake type present, High AAEL scores in complete categories associated with reduced AFlEE scores in proximal and medial/distal categories generally
indicates soft hammer tool production. Low WLscores in the complete flakes and higher &APL scores in the proximal and rnedial/dlstal fragments indicates some form of core reduction. These patterns occur because flake breakage tends t~ occur in the thinner flakes. Thus, where entire assemblages produce thin flakes, MI, scores will scale from highest to lowest depending on the degree of breakage. Where assemblages contain more thicker flakes with associated high edge angles, broken flakes will also be the thinner flakes with acute edge angles and higher AAEL scores.
The HAEL index wsrks in exactly the same fashion as that of the AAEL, only that it reflects variability in thicker flakes wfth high edge angles. Thus to recognize technological variability from the HAEL, one must look for the opposite paternings as demonstrated for the ABEL, Low HAEL scores on complete flakes associated wfth higher scores on proximal and medial/distal fragments indicates reduction oriented towards tool production (thinning core tools or flakes) ox maintenance. High HABL scores on complete flakes and low scores on proximal and medial distal fragments indicates a greater focus on core reduction (production of flake blanks). Residual assemblages, or those with flakes removed from them through culling on scavenging are also technologically recognizable. Mere two patterns are preserved. First, the original technological structure typically remains intact.
Second, the structure the cull readily recognizable given the utility parameters discussed. Trampling produces a distinctive pattern across all assemblages. First, it reduces complete flakes and nonor ientable fragments sf a1 h sizes . Second, 1 t produces increases in the proportions of medial/distai fragments in the medium and small size classes. Thus, non-trampling variation can be expected to be recognizable through the wscreen't of trampling. Likewise, trampling can be recognized despite the effects of the other variables.
439 Given, this patterning, it will be possible to study a prehistoric assemblage of both flake tools and unused debitage and recognize both the technological inputs and the economic decisions made, resulting in flake removal from reduction assemblages. Trampling can be screened out due to the consistently distinctive pattern it produces. Chapter 5 contains an example of the application of this approach, CHAPTER 5
ANALYSIS OF THE DEBITAGE AND FLAKE TOOLS FROM A HOUSEPIT FLOOR
In this chapter, I will discuss the analysis of the debitage and flake tools from the floor of a pithouse excavated at the Keatley Creek site in south-central British Columbia. There are three major goals to this analysis. First, I am interested in demonstrating the effectiveness of the ideas and data developed in chapter 4 for recognizing
variability in the formation of assemblages from a housepit f losr. Second, I will produce the interpretations of technological and taphonomic variability which will be used to
assess the some of the effects of occupational history of different portions of the housepit floor. Third, this
analysis will provide the information necessary to assess some of the economics of stone tool pr~duceionand lithic raw material use by attempting to identify the flow of vitreous trachydacite into the house and flake tool production, use and discard systems.
I begin the chapter with an overview of the Keatley Creek site, with a review of its geographic location, excavations, prehistoric sequence, general site research goals and the role
of this study, The next portion of the chapter is devoted to a disc~ssisnof the analytical techniques used to undezstand debitage and flake tool variability on the floor of the housepit. This is followed by an in-depth discussion of the results of these analyses. The analyses result in a complex set of interpretations of spatial variability in lithic reduction and culling strategies and trampling. Implications Eoz the interp~etationof nousepit floor occupational history are discussed in the conclusions. This chapter serves two functions. It provides the archaeological half of the "back and forthF1process of understanding the archaeological record through experimental research (Amick et al. 1989). In other words it provides an opportunity to assess the usefulness of the methods developed experimentally. The second function is to provide a new look at a nousepit floor, by assessing the relationships between lithic reduction and flake culling to explore the effects of housepit floor use and the general economics behind lithic raw material use.
THE KEATLEY CREEK SITE
Archaeological research at the Keatley Creek Site, located in the Middle Zraser Canyon of south-central British Columbia (Figure 119) was begun in 1986 under the project title Fraser Rlvex Investigations in Corporate Group
Archaeology (spaf ford The primary goals of the project were to identify and explain the existence of prehistoric residential corporate groups in the Middle Fraser Canyon and to use this information to address the larger problem of the evolution of complex hunter-gatherers and socioeconomic differentiation. To work towards addressing these issues, Figure 119. Keatley Creek Archaeological Site location in the Middle Fraser Canyon of south-central British Cof umbia. 443 three housepits oE different sizes were excavated* Mousepits
12, 3 and 7 were chosen since they ilustrated a range of size
(3 m,, 14 m. and 19 m. from rim crest to rim crest, respectively) which may have been related to the social complexity of the groups inhabiting them (Spafford 1991:5-6). Also, each of these housepits contained one clearly identifiable floor, undisturbed by later construction events. Thus, they were ideal for studying intra-structural spatial organization. All date to the Kamloops Horizon of the Canadian Plateau Pithouse Tradition (Richards and Rousseau
This research focusses on housepit 7, dated at 1080+70 B.P., placing its last occupation very close to the time when much of the Middle Fraser Canyon area was abandoned during the Late Prehistoric period (Hayden and Ryder 1991). Excavations at housepit 7 have defined a distinctive compact floor containing numerous post-holes, hearths and storage pits
(Figure 3.28). Covering the floor was a relatively uncompacted set of roof deposits containing charred beams and other materials (Spafford 1991:9). Excavation strategy at housepit 7 has been discussed in detail by Spafford (1992). Essentially, a grid of 2 meter squares was superimposed on the housepit. Excavation proceeded in 50 cltl. squares numbered 1-16 for each 2 m. square. All subsquares were excavated to sterile glacial till following natural stratigraphy (Spafford 1991:9-13). Research in housepit 7 has focussed on identifying the Figure 120. Housepit 7 floor map, Keatley Creek Archaeological Site. locations of d~rnesticand gender specific work areas on the
house floor (Spa2ford 1391). To date, a faix degree of
success has been achieved, Spafford (1991) has defined a minimum of three primary domestic areas (located in the west-northwest, south and east-northeast portions of the flocr) and two probable gender specific work areas (female oriented activities in the central portions and male activities around the margins). Additional variability in artifact contents through these areas has led Spafford to argue that houvepit 7 was more complex in its internal arrangement than the other two housepits. He has suggested that it may have been occupied by a multi-family residential corporate group as opposed to a lower status extended family. My research at housepit 7 is complimentary to that of
Spaf ford. I fccus on the economics of vitreous trachydacite core reduction and tcol production, use and reuse systems. Though I am interested primarily in the role of lithic technology as a component of risk management strategies in this Late Prehistoric context, I draw conclusions on spatial organization which have implications for further understanding the spatis1 organization of the housepit floor. Before addressing problems of economics and spatial organization I address one additional important consideration, the effects of housepit floor reoccupations, It is difficult to draw c~nciusionsabout econemics or socio-political orgjnization from housepit floor without addressing first, the potential confounding effects of floor reoccupation. If the
446 data from the excavated floor of housepit 7 were produced by reoccupations by people who did not clean out the old lithie debris from the previous occupation, then the archaeological result would be a complex jumble of occupation debris not clearly reflecting a single group's dynamics. Rather, it would reflect the scale of activity area reuse from different sets of dynamics. Clearly, even during a single winter's occupation, any single place on the floor of a densely populated housepit will receive multiple usages. The key is to determine whether the scale of debris accumulation is more consistent with a model of activity area reuse from a single group's occupation extending over a number of years or from multiple occupations from different groups and periods. I rely on an assessment of technological behavior and taphonomic patterning in the lithic debitage and flake tools to aid in identifying the effects of occupational history.
Two possible questions can be asked here: Was the floor used during more than one winter; and was the floor used by one or more than one group of people? One potentially useful indicator of whether the housepit floor was reused in more than one winter is patterning in bipolar reduction. First, it is possible that bipolar reduction was an activity practiced during late winter to salvage lithic raw material from exhausted cores and tools. If bipolar core reduction was practiced most comonfy at this time, then bipolar debitage and flake tools will not be trampled to the degree that other lithic items are. Housepit floor use from multiple winters will produce a pattern of trampled bipolar artifacts not much different from other types. If wintcz occupations were preceded by intensive floor clea<~ug,annual Late winter bipolar reduction would appear le5s trampled regardless of the number of winters during which tnat floor was occupied. It is likely that this may have been the case (Hayden 1992, pers comm, 1 .
A mre eritlcal problem for eva?uating the effects of cultural organization on tk:e formation of the housepit floor llthic assemblage is the potential effects of different groups inhabiting th2 housepit. It is possible that bipolar reduction was an activity practiced by specialists who may have used this technique tc produce specialized flakes in spatially bounded areas. All things being equal, reoccupations by the szme group of people could be expected to produce little variance in the spatial locations of bipolar reduction. If the floor artifact assemblages were produced by occupations of very different groups of people, then bipolar reduction (and probably other types as wellj could be expected to have been less spatially distinct. Most importantly, it is probably not the locus of a single form of reduction which is most likely to inform us about occupational history- Rather, the pattern of co-association between different artifact classes, will allow us to first, evaluate the movement of raw materials through reduction strategies to tool use a~ddiscard strategies and second, to indentify the effects of social organization and
448 economic strategies on assemblage composition (c-f.Boisrnier 1991). If a series of spatially distinctive reduction and tool use and discard strategies can be defined, then it is more likely that the floor was used by only one group of people. Whereas, if multiple strategies overlap one another without few clear organizing principles, then it is more likely that the floor artifact assemblages were produced by different groups of people with different ways of organizing the occupation of a housepit Eloor.
ANALYTI CU METHODS
In this section, I review the methods used to identify debitage and flake tool assemblage formation processes from the floor of hausepit 7 at the Keatley Creek site- I begin with a review of the techniques used to divide the data into analytical units. I then review the statistical techniques used to assess underlyfng patterning in the data distributions. Finally, I discuss the process of giving neaning to these data through the use of experimental utility index data.
For analytical purposes a distinction was made between subsquares, analytical units and analytical sectors. Ideally, each of'the 16 subsquares per excavation square would have been considered independently in the analysis. Unfortunately, flakes were not common enough on the floor to allow each subsquare to be considered in this manner. Thus, a grouping
449 strategy was used. First, the floor was divided into 1'16
analytical units and 13 sectors, using SpaffordPs (1991)
significant density sectors (Figures 121 and 122). I add two additional sectors on the east side of the floor to segregate the bench from the floor areas. With oaly a few exception-
each analytical unit was defined as four subsquarzs (% square . Occasionally, three or Eive subsquares had to be used as an analytical unit to fit that unit within each sector. The purpose of this is to examine the contents of areas of significantly different artifact densities independently, assuming that different processes may have affected their formtion. Both unmodified flakes and flake tools were sorted into Modified SulPivar? and Rozen Typology flake types. Flake tools were defined as those flakes with evidence for use and/or modification in the form of retouched edges. Formal tools such as bifaces and end scsaperq were not considered. The key to sorting flake tools into Sullivan and Rozenfs
(1985; Sullivan 1987) flake types was to look closely at margin characteristics. Minimally retouched edges without
evidence for fracturing were considered to be intact margins. Heavy abruptly retouched edges were considered not to be intact. For example, flakes with lightly retouched distal
margins; which were clearly intact before modification were defined as complete, Flakes with invasive retouch, with margins which appeared intact before retouch were also defined as complete. Flakes with intensive abrupt distal margins were
450 Figure 121. Housepit 7 floor artifact density distribution (adapted from Spafford 1992). Figure 122. Division of analytical units on the of housepit 7. defined as proximal, as they may have started with broken edges. Occasionally, platforms were partially removed to produce more working edge or to facilitate hafting. Where it was clear that the platform had been removed after the production of the flake itself, the flake was defined as platform bearing and categorized into the complete or split flake or proximal fragment types, depending on margin characteristics. Approximately 95% of the flake type identifications from flake tools were accomplished unambiguously, as narginal retouch was typically minimal. In the more difficult cases, strict adherence to these typological rules was followed.
All raw MSRT data were rescaled to facilitate multivariate analyses. I have found this scale tc be extremely useful in allowing a close look at differences in proportions of flake types present, while eliminating problems resulting from assemblages with different sized flake counts.
It also allows direct comparisons between archaeological MSRT distributions and experimental and modelled MSRT distributio~s. These distributions also require less data tranformation for multivariate analysis than do chi-square scores (Binford 1989), or log transformations (Draper 1985) and the analyst remains closer to the raw data. In other words, by looking at rescaled data, the analyst is actually interpreting distributions which are much closer to the raw data than that of heavily transformed data sets. This produces fewer errors in interpretation (Nance 1990, pers.
453 comm. ) .
The cmplex data matrices %ere analyzed using princip components analysis. This technique is useful for finding
underlying dimensians of variability in complex data sets. As this study is oriented towards discerning variability of this nature, principal components analysis was considered to be the optimal technique.
~llprit~cipal cofipenents analysez were run in the "Rgl mode (Runnel 1970:193) using the SYSTAT program for the IBM PC (Systat Inc.). Ail cases were analytical units and all variables were MSRT types. An eigenvalue cut-off criterion of 1.00 was used to eliminate consideration of insignificant factors (Kim and Mueller 1978). All extracted loading matrices were rotated with varirnas rotations. A significant loadings criterion was set at .4 to eliminate insignificant loadings from consideration (Rumme1 1970:477). Factor scores were produced for each case to allow an interpretation of the contribution of each case to the overall factor solution, The interpretation of the meaning of each analytical unit (represented as eases in the principal components analysis) requires first a consideration of the basic dimensions of variability in the data set as a whole. This is provided by the interpretation of the rotated loadings matrix (Rummel
2.9701, Tc provide a hypothetical example, a factor loadings matrix might include five factors, each interpreted as significant for different reduction strategies, such that factor one might be a bipolar core reduction factor; factor
454 two, a biface reduction factor and so-on. Factor scores allow an assessment of the contrfbutfon of each case to each factor
(Rummel 1970). This can be in a positive or 3 negative dimension, such that variability in the nature of that
contribution can be measured. Very low positive or negative
factor scores on each case indicate minimal contribution, In other words that case has little to do with that factor, In
this situation, the factor might be measuring bipolar reduction, while a low scoring case might be the result of biface production. High negative or positive factor scores
indicate a substantial contribution of that case to that
factor. While each cofitributes, there is variability In that contribution, indicated by the negative or positive signs.
For example, we could expect, on a bipolar factor, positive high factor scores from cases resultinq from bipolar reduction without any associated trampling effects. High negative factor s'cones could also be found on this factor, associated with cases produced by bipolar reduction associated with intezse trampling and related modification in the data distribution. It is very useful, therefore to look at plotted factor score distributions associated with the analysis of loadings matrices. I employ debitage utility index data to enhance the
interpretation of lithic assemblage variability in a similar fashion to that of Binford (13811, Speth (1983) and Todd
119871, who have used Binford's (1978) utility indices to enhance interpretations of faunal data. This is accomplished 455 using bivariate data plots. I do not attempt to demonstrate any statistical relationship between the experimental indices and the archaeological data sets as 1 am more interested in the shape of the relationship between the data sets.
In faunal studies, Binford (1978:81! has demonstrated two major curvilinear patterns in the relationships between
Minumum Number of Individuals (MNI) and his Modified General
Utility Index (MGUI). These include the "bulk curveffand the "gourmet curve. " The bulk curve anticipates situations where butchers selected for many animal parts of both medium and high utility. The gourmet curve anticipates selection for high utility alone. When archaeological MNI data are plotted in relation to the MGUI, it is ideally possible to make interpretations regarding some of the possible processes conditioning those archaeological assemblages, such as differential butchering and animal part culling. This interpretation comes not from the degree of correlation between the two data sets, but pattern exhibited by the bivariate distribution of the two data sets. Utility curves for debitage distrib~tionsdepend on a range of factors. FVI and WLdistributions typically resemble Binford's (1978:81) gourmet curve in the sense that flakes of high utility are few (Figures 123 and 124: flake types scoring greater than 50 on the x-axis). HAEL distributions more closely resemble Binford's bulk curve (Figure 125) since a much broader range of flakes contain high utility index scores. The analogy to Binford's gourmet and Column 2
Figure 123. Relationship between resealed MSRT debitage distribution and FVP cull index for early stage biface production (from Table 60; lower case=small flakes, upper case=medium flakes, darkened upper case=large flakes). Column 7
40
- mdf m
0 20 40 60 80 100
Column 2
Figure 124. Relationship between rescaled MSRT debitage distribution and mLcull index for early stage biface production (from Table 60; Power case=smll flakes, upper case=medium flakes, darkened upper case=laxge flakes). Column 8
40
0 20 40 63 8 0 100
Column 2
Figure 125. Relationship between rescaled HSRT debitage distribution and HAEL cull index for early stage biface production (from Table 60; lower case=small flakes, upper case=medium flakes, darkened upper case=large flakes). bulk curves is not entirely appropriate as his curves were produced by comparing different culling decisions tu a ~ingle
index scale (MGUI). However, the idea is similar in that flake culling may also be approached from "bulk" and "gourmet" perspectives. The broader range of choices for flakes predicted by the HAEL is clover to a bulk than a gourmet model. I could anticipate, however, that gourmet decisions could be made entirely within the HAEL predicted context. The same could be said for the ,AAF,L predictions. This discussion is designed only to introduce some of the ways the utility lndex distributions could be used to anticipate prehistoric culling decisions, using plotted comparisons between archaeological data and experimental utility index models.
Archaeological MSRT distributions are compared to two types of utility index distributions: utility indices and utility index residual models. Debitage distributions are compared to trampled and untrampled residual models in order to identify reduction strategies and culling decisions. Where debitage assemblages are complex and appear to be the result of different lithic reduction techniques mixed together, mathematically derived sequences are undertaken to better understand the sequence of processes responsible for that patterning. Flake tool distributions are compared to trampled and untrampled utility indices in order to evaluate the origins of these assemblages through reduction strategies and culling decisions.
The formation of flake tool assemblages is a complex process, It depends on the sequential effects of a number of
processes including flake production technique, culling decisions, use, breakage, discard, trampling, scavenging and reuse. The goal of this analysis is to identify, as closely as possible, the sequence of processes affecting the formation of flake tool assemblages on the floor of housepit seven. This is accamplished through the construc~ionof additional mathematical sequences designed t~ demonstrate the effects of multiple processes on basic utility index data sets.
The use of experimentally derived data to aid in the interpretation of archaeological data is a technique now widely in use by archaeologists :e.g. Speth 1983). Following Einford (19781, my experimental data have not been offered as analogies to the archaeological record. Rather, they are used to quantitatively demonstrate the causal effects sf different processes on the archaeological record. They can be used to support "warranting argumentst' (Binford 1981) about the nature of past processes and how they produced the archaeological patterning we recognize today (Speth 1983; Todd 1987). I offer comparisons between experimental and archaeological data as an aid to the construction of theoretical arguments about the formation of the archaeological record.
ANALYSIS
For both the debitage and flake tools, I first discuss the raw and converted data sets to identify some basic
461 patterning. Next, I present the results of the principal components analyses. Here, I focus on underlying sources of variability and how this is reflected in the individual cases repr~sented. This requires extensive discuss ion using comparisons between converted archaeological data distributions and experimental utility index distributions. Following the debitage and flake tool analyses, I draw conclusions on the relationships between the independent results of these analyses. These conclusions allow an integr3tion between both analyses and demonstrate continuity in lithic reduction, culling and tool me and discard on the hsusepit f locr .
DEBITAGE ANALYSIS
Six anslyticah units were dropped from the analysis due to low artifact counts (< 12 flakes in each unit) (locations shown in Figure 121). Data used in this analysis are presented in Tables 73 and 74. Assemblage 95 is the result of the qrouping together of all data from the analytical units in area 12 (raw data in Table 73). Most of the analytical units from area 12 have low numbers of flakes, even though the actual distributions are very similar between units. Thus, rather 'than drop this entire portion of the floor from the analysis, I decided to collapse it into a single unit. One unit with a small sample size (#49) was retained in the analysis due to its location in a boundary zone between two Table 73. Debitage data matrix (flake coants) from the floor of hsusepit 7 (CF=Cornplete Flake, PF=Prcximal Fragment, MDF=Me3?al/Distal Fragment, SF=Split Flake).
I Assemblage 1 5 8 ------2 3 4 6 7 9 Large CF 0 0 0 0 Ca 0 0 0 0 I PF' 0 0 0 0 0 1 0 0 0 HDF 0 1 0 0 1 0 0 0 0 Med i urn CF 1 0 5 1 1 0 0 1 0 PF 0 1 3 2 1 1 4 1 2 MDF 4 0 2 6 7 9 0 3 0 PIF 0 1 0 1 0 0 0 0 8 SF 0 0 0 0 2 0 0 1 1 Small CF 0 2 3 7 6 8 f 3 2 PF 3 2 18 45 7 19 4 7 7 MBF 7 11 24 81 15 22 34 38 28 NF 0 0 3 1 3 0 5 0 0 SF ------1 0 0 3 3 2 I 2 4
Assemblage 10 11 12 13 14 15 16 17 18 19 Large CF 0 0 0 0 1 0 0 0 0 1 PF 8 0 0 0 0 h 0 0 0 0 MDF 1 1 0 0 0 0 0 0 0 0 Hed 1urn CF 1 2 0 0 2 5 1 1 0 0 PF 0 2 2 1 4 8 2 0 3 3 MDF 4 3 4 1 10 6 4 2 5 3 FF 1 0 0 0 0 0 0 0 0 0 EF 0 0 0 1 0 0 0 0 0 0 Small CF 3. 2 4 2 6 11. 2 1 2 4 PF 9 15 11 5 21 26 7 19 7 5 HDF 25 30 22 26 45 59 22 18 19 26 NF 1 0 2 2 5 8 1 f 3 0 SF 2 I 0 2 2 4 1 2 a I ------~~-~~~~~~~~~~~~~-~~~~~~~~~~~~______~~~~~~~~~~~ Asseablage 20 21 22 23 24 25 26 27 28 29 Large CF 0 0 0 0 0 0 0 0 0 0 PI? 0 0 0 0 0 1 0 0 0 0 MDF 0 0 0 0 0 0 0 0 0 0 Med i urn CF PF MDF NF SF Sma 1 l CF PF MDF PVP SF
Assemblage 30 31 32 33 34 35 36 37 38 39
Large CF PF MDF Med i urn CF PF XDF NF SF Small CF PF MDP NF SF Table 73. Contnd. Assemblage 40 41 42 4 3 44 45 46 47 48
Laxge CF 6 0 0 0 0 0 0 0 0 PF 0 0 0 0 0 0 0 0 0 MDF 0 0 1 8 0 0 0 0 0 Medium CF 0 0 0 0 1 0 1 0 2 PF 0 0 2 1 2 0 1 I 1 14DF 1 0 1 1 0 3 2 2 3 NF 0 0 0 0 0 0 0 0 0 SF 0 0 1 0 0 0 0 0 1 Small CF 0 1 2 3 5 2 3 0 4 PF 8 2 14 3 3 1 4 5 4 MDF 5 5 39 10 13 4 12 13 16 NF 0 2 2 0 5 0 1 0 0 SF ------1 1 2 0 1 1 0 0 2 Assemblage 49 50 51 52 53 54 55 56 57
Large CF 0 0 0 0 0 0 1 0 0 PF 0 0 0 0 0 0 0 0 0 MDF 0 0 0 0 0 8 1 1 0 Med i urn CF o o 0 o a I 2 0 3 PF 0 1 2 3 3 I. 4 1 0 MDF 0 0 2 4 1 5 2 2 6 NF 0 0 0 0 0 0 0 0 0 SF 0 1 1 1 1 0 0 f 0 Small CF 1 2 3 2 3 5 6 8 5 PF 1 11 il 16 6 18 16 7 15 HDF 7 28 26 15 33 30 38 17 23 MF 0 0 1 0 2 3 3 2 1 SF ------1 3 3 1 4) 4 0 0 3 Table 73. Contnd. Assemblage 58 59 60 61 62 63 64 65 66 Large CF PF MDF Xed i urn CF PF MDF NF SF Small CF PF MDF NF SF
Assemblage 67 68 69 70 71 72 73 74 75 76 __---_-_---___-_-----______------Large CF PF MDF Med i urn CF PF MDF NF SF Small CF PF MDF NF SF Table 73. Contnd, Assemblage 77 78 79 80 81 82 83 84 85
Large CF 0 0 0 0 0 0 0 0 0 PF 0 0 I 0 0 0 0 0 0 MDF 0 1 0 0 0 0 0 0 0 Med i urn CF 0 2 0 1 I 1 4 0 0 PF 3 5 0 3 2 1 0 0 1 MDF 8 10 2 4 0 3 7 2 1 EIF 0 0 0 0 0 0 0 0 1 SF f. 8 1 0 0 0 0 0 0 Small CF 4 10 1 6 4 1 10 0 2 PF 14 23 7 13 7 4 15 10 8 MDF 43 71 9 40 27 6 48 18 16 NF 0 18 1 3 2 I 6 0 4 SF 4 o 3 2 1 n 3 1 2
Assemblage ------86 87 88 89 90 91 92 93 9 4 Large CF 0 0 0 0 0 0 0 0 0 PF 0 0 0 0 0 8 0 0 0 MDF 6 0 0 0 0 0 0 0 0 Med i urn CF 0 0 0 0 0 0 0 0 0 PF 2 1 0 5 2 0 0 0 2 MDF 4 2 0 3 1 2 2 0 3 NF 1 0 0 0 0 0 0 0 0 SF 8 0 0 0 6 0 0 0 0 Small CF 3 1 3 3 1 2 0 1 3 PF 8 6 12 17 4 8 7 13 24 MDF 10 10 12 4 2 12 10 18 30 51 MF 4 2 0 3 0 1 2 1 0 SF 2 2 1 0 1 2 1 0 3 Table 73. Contnd. Assemblage
Large CF PF MDF Pfed i urn CF PI? MDF NF SF Small CF PF MDF NF SF Table 74. Rescafed dehitage data mtrix from the floor of housepit 7 fCF=Complete Flake, PF=Proximal F~agment,MDF=Medial/Distal Fragment, SF=Split Flake). Assemblage 1 2 3 4 5 6 7 8 3 Large CF 0.0 0.0 0-0 0.0 0.0 0.9 0.0 0.0 0.0 PF 0.0 0.0 0.0 0.0 0.0 4.5 0.0 0.0 0.0 MDF 0.0 9-1 0.0 0.0 6-7 0.0 0.0 0.0 0.0 Med i urn CF 14.3 0.0 20.8 1.2 6.7 8.0 0.0 2.5 0.0 PF 0.0 9.1 12.5 2.5 6.7 4.5 11.8 2.6 7.1 MDF 57.1 0.0 8.3 7-4 46.7 40.9 0.0 7.9 0.0 NF 0.0 9.1 0.0 1.2 0.0 0.0 0.0 0.0 0.0 SF 0.0 0.0 0.0 0.0 13.3 0.0 0.0 2.6 3.6 Small CF 0.0 18.2 12.5 8.6 40.0 36.4 2.9 7.9 7.1 PF 42.9 18.2 75.0 55.6 46.7 86.4 11.8 18.4 25.0 HBF 100.0 100.0 100.8 100.0 100.0 100.0 100-0 100.0 100,O rJF 0.0 0.0 12.5 1.2 20.0 0.0 14.7 0.0 0.0 SF 14.3 0.0 0.0 3.7 20.0 9.1 2.9 5.3 14.3 ----_----__------
Assemblage ______-_____------I0 3.1 12 13 14 15 16 67 18 19 Large CF 0.0 0.0 0.0 0.0 2.2 0.0 0.0 0.0 0.0 3.8 PF 0.0 0.0 0.0 0.0 0.0 1.7 0.0 0.0 0.0 0.0 MDF 4.0 3.3 0-0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Med i urn CF 4.0 6.7 0.0 0.0 4.4 8.4 4.5 5.3 0.0 0.0 PF 0-0 6.7 9.1 3.8 8.9 13.6 9.1 0.0 15.0 11.5 MDF 16.0 10.0 18.2 3.8 22-2 10-2 18.2 10.5 26.3 11.5 NF 4.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0:O 0.0 SF 0.0 0.0 0.0 3.8 0.0 0.0 0.0 0.0 0.0 0.0 Small CF 4.0 6.7 18.2 7.7 13.3 18.6 9.1 5.3 10.5 15.4 FF 36.0 50.0 50.0 19.2 46.7 44.1 31.8 100.0 36.8 19.2 MDP 100.0 100.0 100.0 100.0 100.0 100.0 100.0 94.7 100.0 100.0 SF- 4.0 0.0 9.1 7.7 11.1 13.6 4.5 5.4 15.8 0.0 SF ---______------8.0 3.3 0.0 7.7 4.4 6.8 4.5 10.5 5.3 3.8 Table 34. Contnd
Assemblage 20 21 22 23 2 4 25 26 27 2 8 29 Large CF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 PF 0.0 0.0 0.0 0.0 0.0 1.8 0.0 0.0 0.0 0.0 MDF 0-0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Med l urn CP 2.2 18.8 1.9 7.6 0.0 1.8 0.0 2.3 4.8 0.0 PF 2.2 6.3 11.3 0.0 5.6 3.5 11.1 9.1 6.5 0.0 MDF 4.3 37-5 3.8 0.0 22.2 8.8 11.1 0.0 11.3 40.0 NF 0.0 12.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 SF 0.0 0.0 0.0 0.0 0.0 8.0 0.0 0.0 0.0 0.0 Small CF- - 8.7 25.0 15.1 15.4 5.6 8.8 11.1 9.1 6.4 0.0 PF 26.1 62.5 32.1 46.2 27.8 29.8 44.4 33.6 32.3 20.0 MDF 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 NF 26.1 18.8 3.8 0.0 0,O 7.0 0.0 11.4 1.6 0.0 SF 4.3 18.8 3.8 7.6 0.0 10.5 22.2 9.1 4.8 20.0
Assemblage 30 33. 32 33 34 35 36 37 38 39 Large CF 0.0 0-0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 PF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 6.0 MDF 0.0 0.0 0.Q 0.0 0.0 0.0 2.8 0.0 0.0 0.0 Med i urn CF 2.3 0.0 0.0 0.0 3.6 0.0 5.6 0.0 6.5 0.0 PF 2.3 0.0 0.0 3.4 0.0 0.0 0.0 0.0 6.5 7.7 MDF 4.7 7.1 12.5 6.9 14.3 9.6 5.6 13.3 8.7 7.7 NF 0.0 0.0 0.0 0.0 0,O 0.0 0.0 0.0 0.0 0.0 SF 0.0 0.0 0.0 0.0 3.6 0.0 0.0 0.0 0.0 0.0 Small CF 4.7 7.1 6.3 6.9 10.7 5.7 5.6 6.7 6.5 0.0 PF 39.5 0.0 37.5 31.0 46.4 21.2 27.8 60.0 26.1 69.2 MDF 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 NF 0.0 0.0 0.0 0.0 14.3 3.8 19.4 6.7 8.7 0.0 SF 0.0 0.0 6.3 10-3 3.6 9.6 5.6 6.7 2.2 15.4 Table 74. Contnd. Assemblage 40 41 42 43 44 45 46 47 48 Large CF 0.0 0.0 0.0 0.0 0.0 8.0 0,O 0.0 0.0 PF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 MDF 0.0 0.0 2.6 0-0 0.0 6.0 0.0 0.0 0.0 Med ium CF 0.0 0.0 0.0 0.0 7-7 0.0 8.3 0.0 12.5 PF 0.0 0.0 5.1 10.0 15.4 0.0 8.3 7.7 6.3 MDF 12.5 0-0 2.6 10.0 0.0 75.0 16.7 15.4 18.8 NF 0.0 0.0 0.0 0.0 0.0 8.8 0.0 0.0 0.0 SF 0.0 0.0 2.6 0.0 0.0 0.0 0.0 0.0 6.3 Small CF 0.0 20.0 5.1 30.0 38.5 50.0 25.0 0.0 25-0 PF 100.0 40.0 35.9 30.0 23.1 25.0 33.3 38.5 25.0 MDF 62.5 100.0 100.0 100.0 100.0 100.0 100.0 180.6 100.0 EIF 0.0 40.0 5.1 0.0 38.5 0.0 8.3 0.0 0.0 SF 12.5 20.0 5.1 0.0 3.7 25.0 0.0 0.0 12.5
Assemblage 49 50 51 52 53 54 55 56 57 Large CF 0.0 0.0 0.0 0.0 0.0 0.0 2.6 0.0 0.0 PF 0.0 0.0 0.0 8.0 0.0 0.0 8.0 0.0 0.0 MDF 0.0 0-0 0.0 6.0 0.6 0.0 2.6 5.9 0.0 Med i urn CF 0.0 0.0 0.0 0.0 3.0 3.3 5.3 0.0 13.0 PF 0.0 3,6 9.7 18.8 9.1 3.3 lo,•˜ 5.9 0.0 MBF 0.3 0.0 7.7 25.0 3.0 10.0 5.3 11-8 25.1 NF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 SF 0.0 3.6 3.8 6.3 3.0 0.8 0,O 5.9 0.0 Smll CF 14.3 7.1 11.5 12.5 9.1 10.0 15.8 0.0 21.7 PI? 14.3 39.3 42.3 100.0 18.2 60.0 42.1 41.2 65.2 MDF 100.0 100.0 100.0 93.8 100.0 100.0 100.0 100.0 106.0 MF 0.0 0.0 3.8 0.0 6.1 18.0 9.9 11.8 4.3 SF 14.3 10.7 11.5 6.3 0,O 13-3 0-0 0.0 13.0 Table 74. Contnd. Assemblage
Large CF 5.3 0.0 0-0 0.0 0.0 0.0 0.0 0.0 0.0 PF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 PiDF 0.0 0.0 0.0 0.0 0.0 0.0 2.2 0.0 0.0 Medium CF 5.3 5.9 4.2 9.1 9.1 4.4 0.0 8.3 0.8 PF 15.8 11.8 4.2 9.1 9.1 2.2 15.2 16.7 25.0 MDF 10-5 0.0 8.3 0.0 27,3 8.8 8.6 41.7 0.0 NF 0.0 0.0 0.0 9-1 0.0 0.0 0.0 0.0 12,5 SF 0.0 0.0 0.0 0.0 0.0 0.0 2.2 0.0 0.0 Small CF 15.8 5.9 2.1 0.0 9.1 13.3 4.3 25.0 0.0 PF 47.4 35.3 50.0 45.4 81.8 42.2 56.5 66.7 37,s MBF 100.0 100.0 100.0 100.0 180.0 100.0 100.0 100.0 100.0 NF 0.0 11.8 8.3 0.0 18.2 6.7 6.5 8.3 0.0 SF 0.0 11,8 8-3 0-0 18.2 5.6 4.3 8.3 0.0
Assemblage 67 68 69 70 7 1 72 73 74 75 76 Large CF 0,O 4.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 PF 0.0 0.0 0.0 0.0 0.0 0.0 8.8 0.0 0.0 0.0 MBF 0.0 4.5 8.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Med i urn CF 5.9 6.0 15-4 9.5 0.0 0-0 0.0 0.0 7.7 0.0 PF 29,4 4.5 15.4 9.5 29.4 0.0 5.6 8.3 7.7 0.0 MDF 23.5 18.2 23.1 14.3 17.6 10.7 16.7 0.0 7.7 16.7 NF 5.9 0.0 0.0 0.0 6.0 0.0 0.0 0.0 0.0 0.0 SF 0.0 0.0 0.0 0.0 11.8 0-0 16.7 0.0 0.0 0.0 Small CF 17.6 36 3.8 4.8 35.3 3.6 11.1 4.2 7.7 11.1 PF 47.1 31.8 53,8 47.6 100.0 25.0 27.8 45.8 53.8 38.9 MDF 100.0 100.0 100.0 100.0 82.4 100.0 160.0 100.0 100.0 100.0 NF 0.0 0.0 7.7 0.0 17.6 7.1 0.0 0.0 15.4 5.6 SF 11.8 9.1 11.5 4.8 0,O 3.6 0.0 16.7 7.7 16.7 ------Table 74. Contnd. Assemblage 77 78 79 80 81 82 83 8 4 85
Large CF PF MDF Med turn CF PF MDF NF SF Small CF- - 9.3 14.1 11.1 15.0 14.8 16.7 20.8 0.0 12.5 PF 32.6 32.4 77.8 32.5 25.3 66.7 31.3 55.6 50.0 MDF 100.0 100.0 100.0 100.0 100.0 100.0 100,0 10Q.0 100.0 NF 0.0 25.3 11.1 3.5 7.4 16.7 12.5 0.0 25.0 0.0 5.0 3.7 0.0 6.3 5.6 12.5 SF ------9.3 33-3
Large CF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 PI? 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 HDF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Med i urn CF 20.0 10.0 0.0 0.0 0.0 0.0 0.0 0-0 0.0 PF 40.0 20.0 0.0 11.9 16.7 0.0 0.0 0.0 3.9 MDF 90.0 0.0 0.0 7.1 8.3 20.0 11.1 0.Q 5.9 NF 0.0 0.0 0.0 0,0 0.0 0.0 8.0 0.0 0.0 SF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 8.0 Small CF 30=0 10.0 25.0 7.1 8.3 20.0 0.Q 3.3 5.9 PF 80.0 60.0 100.0 40.5 33.3 e0.0 38.9 43.3 47,1 MDF 100.0 100.0 100.0 100.0 100.8 100.0 180.0 100,Q 980.0 NF 40.0 20.0 0.0 7.1 0.0 10.0 11.1 3.3 0.0 SF 20.0 20.0 8.3 0.0 8.3 20.0 12.5 0.0 5.9 Table 74. Contnd. Assemblage 95 9 6 97 98 99 100 3.01 102 Large CF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 PF 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 MDF 0.0 0.0 0.0 0.0 0.8 0.0 3.1 0.0 Med ium CF 0.0 0.0 0.0 11.1 Q.0 0.0 3.1 5-8 PF 3.1 0.0 0.0 11.1 4.5 33.3 12.5 0.0 MDF 12.5 5. 15.4 33.3 9.1 13.3 18.8 29.4 NF 0.0 5.0 7.7 0.0 0.0 0.0 0.0 0.0 SF 1.6 0,O 0.0 11.1 0.0 0,O 0.0 5.8 Small CF 9.4 20.0 15.4 22.2 9.1 0.0 15.6 29.4 PF 34.4 55.0 84.6 77.8 45.5 100.0 81.3 76.5 MDF 100.0 lOO.0 100.0 100,O 100.0 86.7 100,0 100,O NF 3.1 5.0 7.6 11.1 9.1 0.0 12.5 5.8 SF 4.7 5.0 7.6 22.2 4.5 20.0 6.3 11.8 analytical sectors (Figure 121). As this is a critical place
on the fioor (see Spaffoxd 1992), the importance of assessing
the data from this location outweighed any potential error input into the analytical matrix. However, conclusions on the formation of this assemblage should be considered with
caution.
1 review patterning in debitage data (Tables 73 and 74) by size class. A number of interesting patterns are evident in the small MSRT distributions. First, medial/distal fragments are by far the most common flake type. This is to be expected if much of the lithie reduction is from cores or bifaces and if trampling is present. Second, a few assemblages are dominated by proximal fragments, which may indicate some evidence for tool maintenance or resharpening. Third, nonoxientable fragments are relatively common in most assemblages, indicating that hard hammers may have played an important much the lithic reduction present. Fourth, there are very few complete flakes, indicating that, with few exceptions, trampling and possibly culling may have played a ma!c~r role in the formation of these distributions. In general, there are indicators of core reduction in the form of assemblages with high medial/distal counts associated with moderate numbers of nonorientable and proximal fragments and indicators of tool production in the form of assemblages with almost equally high numbers of medial/distal and proximal fragments and low nonor ientable fragments. There are two major patterns in the medium size class flakes from housepit 7. The most common pattern occurs when assemblages are dominated by medial,/dlutal fragments, with reduced numbers of proximal fragments and split and complete flakes. This pattern could be indicative of either core reduction or sone form of biface production with associated trampling and/or culling. Determining the actual origins of the pattern will depend largely on the associated small artifact distributions. The other major pattern Is where proximal fragments and complete flakes are nearly as common as that of rnedial/distal fragments. This is probably the result of core reduction, where trampling and flake culling are affecting the data differently from the high rnedial/distal pattern. Large flakes are extremely rare in the floor assemblages and are not considered further here. Lithic
~eductionappears to be oriented towards the pxoduction of medium sized flakes (4 to 16 square cm.). In no assemblage is there a pattern in the medium flake size class clearly reflecting residual breakage from reduction oriented towards large (16 to 64 square cm.) flake production. All interpretation of reduction strategies therefore assumes the reduction of cores designed to produce medium and small f.64 to 4 square cm.) sized flakes.
The converted data matrix (Table 74) was subjected to a principal components analysis. The correlation matrix is shown in Table 75. Using the eigenvalue cutoff of 1.0, six factors were extracted and rotated with the result accounting for 68.8% of the total variance (Tables 76 and 77). Factor
476 Table 75. Wousepit 7 debitage analysis correfatfott matrix (<.'=Complete Flake PF=Proximl Fragment, MDF=Medial/Bistal Fragment, SF=Spl%t Flake 1 . Large Large Large Medium Medium CF PF PIDF CF PF Large CF 1.060 PF -0,043 MDF 0.3132 Medi urn CF -0 050 PF 0.184 MDF -0.020 NF -0.070 SF -0.092 Small CF 0.045 PF -0.096 MDF 0.043 NF -0.116 SF -0.171 Table 75. Contnd.
Small Sma l l Table 76, Housepit 7 debitage analysis initial statistics, Variance Percentage Explained by sf Total Variance Factors Ei genvalues Rotated Components ExpPa ined 1 2.348 1.746 13.428 2 1.536 1.589 12.222 3 l.40@ 1.381 10.620 4 1.361 1.262 9.787 5 1.253 . 1.244 9.566 6 1.033 1.718 13.212 Table 77. Housepit 7 debitage analysis rotated loadings (CF=Cah:plete Flake, PF=Proximal Fragment, MDF=Mediaf/DPstaL Fragment, SF=Spllt Flake).
1 2 3 4 5 6 Large CF PF MBF Med i urn CF PF MDF NF SF Small CF PF MDF NF SF Table 78. Housepit 7 debitage analysis factor scores. Factor Factor Factor Factor Factor Factor: 1 2 3 4 5 6 _--_---_------__-_------.------Case 3. 2.361 0.576 -0.117 0.226 -1.596 -0.535 Case 2 -1.565 1.417 2,309 1.665 3.470 0.297 Case 3 -0.176 -0.353 -0.168 0.378 -1.119 2.139 Case 4 -0.403 0.007 0,043 -0 ,004 -6 ,465 -0.601 Case 5 2.530 0.555 -0.448 -0.371 5.071 0.934 Case 6 2.387 -0.484 1.222 0.180 -0.108 -0.863 Case 7 -1.605 0,547 -0.343 -0.374 -0.032 0.350 Case 8 -8.365 0.625 -0.337 -0.201 0.064 -0.768 ease 3 -0.439 C.195 -0.213 -0.678 0,283 -0.826 Case 10 -0.272 0.896 0. 832 0.302 0.870 -0.362 case I1 -0.470 0.219 -0,014 0.615 0.539 -0.042 case 12 -0.144 0.022 -0.329 0.255 -0.035 0.121 Case 13 -0.656 0.526 -0.432 -0,698 0.543 0. 505 Case 14 8.484 0.021 -0.365 1.795 -0.395 0,239 Case 15 -0.081 0.074 0.291 -0.162 -0.316 1.104 Case 16 -0.081 0.330 -0.211 0.192 -0.496 -0.232 Case 17 0.087 -1.473 -0,307 -0.432 -0.837 -0.390 Case 18 -0,124 0.055 -0.244 0.013 -0.107 0.515 Case 19 0.225 0.219 -0.317 3.343 -0.192 -0.640 Case 26 -1.082 0.793 -0.657 -0.941 0.075 0.804 Case 21 2.002 0.690 2.728 0.035 -1.124 2.191 Case 22 -0.649 0.200 -0.221 0.141 -0.180 0.668 Case 23 -0.090 0.354 -0.308 -0.171 -0.614 -0.195 Case 24 -0.206 0.429 -0.237 0.389 -0.395 -0.856 Case 25 -0.187 0.477 0.261 -0.611 -0.318 -0 490 Case 26 0.457 -0.157 -0.026 -0.518 -0.532 -0.628 Case 27 -0.824 0.143 -0.284 -0.512 -0.254 -0.309 Case 28 -0.324 0.346 -0,203 0.092 -0,594 -0,281 Case 29 1.374 0.753 -0.191 -0.470 -0.805 -1.613 Case 30 -0.800 0.277 -0.271 8.098 -0,487 -0.665 Case 31 -8.817 1.153 -0.420 0.059 -0.191 -1,003 Case 32 -0.157 0.390 -0.292 -0.162 -0.452 -1.067 Case 93 -0.296 0.374 -0.230 -0.309 -0.425 -0,908 Case 34 -0.143 0.303 -0,550 -0,613 0-394 0.040 Case 35 -0.299 0.728 -0.359 -0.479 -0.355 -0.904 Case 36 -0.942 0.996 -0.453 -0.472 0.792 0.601 Case 37 -0.126 -0.002 -0.333 -8.390 -0.420 -0.650 Case 38 -Q.673 0.556 -0.326 -0,039 -0.478 0.298 Case 39 -0.163. -0.621 0.005 -0.468 -0.767 -0.827 Case 48 -0.882 -5.871 -0.970 -0.188 -0.373 -1.898 Case 43. -0.499 0.714 -0.845 -2.137 0.445 1. $96 Case 42 -0.991 0.320 -0.215 -0.131 1.205 -0.453 Case 43 -0.856 0.384 -8.313 0,597 0.163 -0.108 Case 44 -0.566 0.785 -0,746 -0.812 0 * 664 3.069 Case 45 4.673 1.193 -0.436 0.260 -0.039 -1.085 Case 46 0.217 0.559 -0.392 0.443 -0.222 0,805 0.305 -0.513 -8.779 Case 47 0.667 ------0.067 -0.146 ------481 Table 7%. Csntnd.
Case 48 1,266 Case 49 -0,186 Case 50 -0.467 Case 51 -0.107 Case 52 0,403 Case 53 -1.017 Case 54 0.095 Case 55 -0.419 Case 56 -1.243 Case 57 1.518 Case 58 0.438 Case 59 -0.776 Case 60 -0.453 Case 61 -1,215 Case 62 1.012 Case 63 -0.163 Case 64 -0.922 Case 65 1.625 Case 66 -1.505 Case 67 0.427 Case 68 0,741 Case 69 0.501 Case 70 -0.072 Case 71 -0.295 Case 72 -0.638 Case 73 0,042 Case 74 -0,419 Case 75 -0.430 Case 76 0.539 Case 77 0.212 Case 78 -0.809 Case 79 2.145 Case 80 -0.309 Case 81 -0,795 Case 82 1.420 Case 83 0,341 Case 84 -0,363 Case 85 -0,633 Case 86 0.231 Case 87 -0.382 Case 88 0.254 Case 83 -8.651 Case 94 -0.549 Case 91 1,230 Case 92 -0.347 Case 93 -1,113 Case 94 -0.508 Case 95 -0,243 Case 96 0.211 Case 97 0.239 Contnd,
Case Case Case Case Case 2.5
Factor 1
1.0 2.5
Factor 2
Figure 126. Housepit 7 floor debitage an?l.ysis factoz score plot: factors 1 and 2 (Factor scores are tabulzfed in Table 78). 2.5
Factor 3
Factor 4
Figure 127. Housepit 7 flo~rdebitage analysis factor score plot: factors 3 and 4 (Factor scores are tabulated in Table 78). Factor 5 1.0
1.0 2.5 4.0
Factor 6
Figure ,128. Housepft 7 flour debitage analysis factor score plot: factors 5 and 6 (Factox scores are tabulated in Table 78). scores are presented in Table 78 and Figures 126-128. I will now describe and interpret these results on a factor by factor bas is.
Factor 1
Factor 1 (Table 77) contains high positive loadings on rnedlum medlal/distal fragments and small complete and split flakes. No significant negative loadings are present. A total of fifteen cases contain positive factor scores greater than 1.0. These assemblages all contain extremely high numbers of medium rnedial/distal fragments. They typically also contain relatively high numbers of small complete and split flakes. There is some variability however in these categories. Although not a strong contributor to factor 1, small proximal fragments show up strongly in many of these cases. Two primary patterns appear to be present. First, the most common pattern is one where small complete and split flakes are relatively common and small proximal and rnedial/dlstal fragments are dominant. Small nonor ientable fragments are infrequent. Medium medial/distal fragments are common, but other medium size flake types are more rare. The second,pattern is similar to the first, except that small proximal fragments and complete flakes are much reduced and there are occasionally more nonorientable fragments. The pattern sf high numbers of small proximal and medial/distal fragments, relatively high small complete and split flakes and iow nonorientable fragments (cases 6, 45, 57,
62, 65, 79, 82, 91, 98 and 102), 1s most reminlvcent of hard hammer produced bifaces where some trampling has occurred. Coinmon small platform bearing flakes indicates extensive edge preparation. The focus on complete and split flakes in the factor loading matrix (Table 771 indicates hard hammer use. Consistently slight to heavy reductions in complete flakes and nonor ientable fragments indicates some trampling. This may slso be indicative of the presence of soft hammer blface reduction, as it is common in biface production to work with both soft and hard hammers. Case 45 may have been affected to a greater degree by hard hammer blface production due to the presence of high numbers of small complete flakes. Where small proximal fragments and complete flakes are more reduced and nonsrientable fragments and/or split flakes are inflated (cases 1, 5, 21, 29 and 48), hard hammer reduction of prepared cores may be a more realistic interpretation. This process will produce numerous small proximal fragments and complete flakes, but not the same quantity as produced during biface manufacture. It also tends to produce more nonorientable fragments and split flakes. Trampling, however, will remove complete flakes and nonoripntabbe fragments. Thus, the second pattern here appears to be attributable to prepared core reduction where trampling has occurred. Both patterns contain high numbers of medium medial/distal fragments and relatively few other medium flake types, This pattenn ~~~formsnicely to what could be expected were larger acute edge angle flakes regularly culled from these assemblages, followed by some degree of trampling. This overall pattern of hard hammer biface reduction, trampling and
acute edge angle flake culling can be seen in a plotted comparison of the archaeological data (using case 6 as an example) with data from Table 60, (201. 33 (experimental stage
2 biface reduction trampled AAEL residual distribution) (Figure 129). I interpret all assemblages with positive factor scores from factor 1, greater than 1.0 to be the result of acute edge angle flake culling and trampling. One significant exception is case 45 which contains extremely numerous small complete flakes and may not have been trampled as extensively as the other units.
Eight cases have strong negative factor score% (less than -1.00) which can be interpreted differently from the above. High negative factor scores indicate those cases on a factor which contribute in the opposite manner to those with high positive factor scores. Here, small complete and split flakes are rare, as are medium medlal/distal fragments. Small proximal fragments are typically reduced and small or medium nonorientabfe fragments are present. Two slightly different patterns can be seen across the eight cases. Cases 20, 56, and 93 are somewhat similar to those cases from factor 1 with high positive factor scores identified as trampled prepared core reduction with acute edge angle flake culling. Culling in the negative factor appears to be more intense and oriented Table 74 Case 6
Table 40 Column 33
Figure 129. Comparison of case 6 with the stage 2 biface trampled AAEL residual index (Sower case=smll flakes, upper case=medium flakes, darkened upper case-large flakes). towards both flake size and acute edge angles. This would account for the greater reductions in all medium sized flake types and the small complete and split flakes which are typically Larger than small proximal, medial/distal and nonorientable flake types. This pattern can be seen in a plotted comparison between case 20 and Table 66, Col. 37 (experimental prepared, medium flake production, block core, trampled FVIxAAEL residual distribution) (Figure 130) where most flake types axe reduced with the exceptions of small proximal anc~ medial/distal fragments. Small nonorientable fragnents jcore highly on case 20 compared to the experimental values, although they are much reduced in the other cases of this type. The primary pattern is that of small nonorientable fragments being present, but reduced in numbers, comparable to that of typical trampled assemblages. Cases 7 and 53 are identical-in most respects to cases
20, 56 and 93 with one significant exception. The pattern in medium flake types is far more reminiscent of that which occurs in trampled culled assemblages favoring the removal of high edge angle flakes. The resulting patterning is one of reduced medium complete flakes and medial/distal fragments and comparatively inflated numbers of proximal fragments. Cases 7 and 53 are interpreted to be the result of trampled prepared core reduction with high edge angle flake cuiling.
Three cases (2, 61 and 66) appear to have been affected by bipolar reduction as well as prepared core reduction as indicated by high numbers of medium medial/distal fragments. 0 20 40 60 80 100 Table 66 Column 37
Figure 130. Comparison of case 20 with the medium flake, prepared core reduction, trampled, FVIxAAEL
I residual index (lower case=small flakes, upper case=medfum flakes, darkened I upper case=large dlakes). These cases are highlighted more strongly on factor 3 and are
considered in greater depth below.
Factor 2
Significant positive loadings from factor 2 are found only on small medial/distal fragments. Significant negative
leadlnys are found on amall, and to a lesser degree, medium proximal fragments. This appears to mark two significantly different form of lithic reduction on the housepit floor.
The most common forms of lithic reduction produce vast quantities of small medial/distal fragments. Both biface and
bipolar and prepared core reduction result in this pattern. ' Assemblages containing small proximal fragments as the dominant flake type result primarily from retouching of tool edges through pressure or percussion flaking. 1 focus on the negative dimension of factor 2, as it appears to have isolated an important form of lithic reduction compared to the other forms which are more adequately covered on other factors. six cases (17, 40, 52, 71, 86 and 100) contain strong neyatlve factor scores, With the exception of case 86, all are dominated by small proximal fragments. Small medial/distal fragments are typically numerous, while small complete and split flakes and nonomientable fraginznts are minimal. This could indicate that two factors are involved. First, lithic reduction is almost entirely oriented towards producing small edge shaping or resharpening flakes. This process results in as many or more platform bearing flakes than rnedial/distal fragments. Second, trampling reduces the numbers of complete flakes present and further inflates the number of proximal fragments. Some cases (17 and 71) have more numerous small nonorientable fragments than might be expected. This my indicate some cvemlap with hard hammer biface production. Case 86 (factor score slightly over 1.0) appears to represent a prepared cone reduction assemblage where edge preparation was very extensive. Thus it contains both high numbers of proximal as well as nonorientable fragments. Medium sized flakes form two patterns. The first (cases
17, 40 and 52) is a typical acute edge angle flake culling pattern with numerous medial/distal fragments and reduced proximal fragments and complete and split flakes. The second (cases 71 and 1001 is a pattern more similar to the expected high edge angle flake culling pattern with a high proximal fragment score and reduced medial/distal Cragments. This pattern can also occur, however, in trampled retouch flake assemblages. The presence of small flake distributions with indications of edge retouching/resharpening activities and medium flake distributions indicating culling of acute and possibly high edge angle flakes indicates that two general form of'lithic reduction my be occurring: edqe retouch and bifaee reducti~n, Comparison of assemblage 52 with data from Table 63, Col. 29 (hard hammer flake modification trampled MSRT distribution) [Figure 131) indicates a primary focus on Table 74 Case 52
0 20 40 60 80 100 Table 63 Column 29
Figure 131. Comparison of ease 52 with the trampled, hazd hammer flake retouch data set (lower case=srnall flakes, upper case=rnediurn flakes, darkened uppen case=larqe flakes). edge retouch. Small complete and split flakes are outliers on
Figure 19 probably due to the greater effects of trampling in the archaeological data than in the experimental. Medium medial/distal fragments are common in the archaeological assemblage due to the presence of biface reduction, not present in the experimental data illustrated in Figure 20. The presence of small nonorientable fragments and complete flakes indicates the use of hard hammers. Reductions in complete flake counts in the small and medium size classes indicates trampling.
Factor 3
Significant positive factor loadings are found on large proximal fragments and medium nonorientable fragments (Table
773. No significant negative loadings were produced. Fact ox scores are strong only in the positive dimension (Table 78) with eight cases (2, 6, 21, 61, 66, 67, 79 and 97) containing scores greater than 1.0. These are considered to be the primary contributors to this factor. The primary producer of medium nonorientable fragments is bipolar core reduction. However, substantial variation in other flake types across the eight cases indicates that Sipolar reduction may be mixed with other forms of reduction. Cases 6 and 79 are not considered further here as they appear on this factor due to the presence of one large proximal fragment each. The origins of these assemblages are better explained on factor 1. Three principal patterns are present in the six remaining
cases. Cases 21 and 67 contain fairly standard prepared core ireduction assemblages with some indications of acute edge angle flake culling and trampling. The only difference is that medium nonorientable fragments are far more common than is normally expected in this context. This may indicate bipolar mixing in these assemblages.
Cases 2, 61 and 66 contain high numbers of small rnedial/distal fragments as well as relatively numerous proximal fragments. Case 2 has fewer proximal fragments and comparatively higher numbers of complete flakes. None of these cases contains any medium medial/distal fragments, although medium nonorientable and proximal fragments are present. These appear to be assemblages similar to those of cases 21 and 67 which have been Surther modified by culling and trampling.
Case 97 is different from any of the five other cases in that it contains very few small nonorientable fragments and many small proximal fragments along with the usual unexpectedly high medium nonorientable fragment category. This appears to indicate a new form of assemblage with hard hammer biface production mixed with bipolar core reduction. To construct arguments regarding the origins of these assemblages f will first construct two new experimental sequences designed to demonstrate the effects of different sets of processes on single assemblages. Raw prepared core (from Table 66, Col. 1) and bipolar core reduction (from Table
71, Col. 1) MSRT data was combined to anticipate the effects
of mixing debitage of both reduction types (Table 79, Col.1). This was rescaied in Column 2. To assess the effects of flake
culling in a mixed assemblage of this nature, raw FVIxAAEL
data from both reduction types were combined (Col. 3) and
rescaled (Col. 4). FVIxAAGL utility index data best explains
the patterning in a malurity of cases from the houseplt 7 floor where culling has occurred. Combining the index data in this way provides a utility index scale for evaluating Elake culling where two different reduction strategies are combined
in equal proportions in the same assemblage. A mixed FVIxAAEL residual assemblage is produced by subtracting column 4 from
100 (Col. 5), multiplying this result against the data from
column 2 (Col. 6) and rescaling (Col. 7). This is the same procedure for predicting the appearance of a residual assemblage as was discussed in Chapter 4. These data do not
anticipate the effects of trampling. This is reflected by the inflated numbers of medium complete flakes. With trampling. I would expect reduced numbers of medium complete flakes and inflated medium medial/dlstal fragments. The pattern
produced, however, does reflect culling for large acute edge angle flakes with comparatively reduced medium complete flakes and proximal fragments and little modification to the small flakes or medium nonorientable and split flakes.
Comparison between case 21 and Table 79, column 7, indicates both some similarities and differences (Figure 132). Table 74 Case 21
Table 79 Column 7
Figure 132. Comparison of case 21 with the mixed prepared and bipolar core reduction FQfxAaEL residual index (lower case=sllltall flakes, upper case=medium flakes, darkened upper case=large flakes ) . The data sets are quite similar in all areas except medium and small complete flakes, medium medlal/dlutal fragments and small proximal fragments. Complete flakes are under-represented by the experimental data. This may be due to the fact that the experimental data weight bipolar and prepared core reduct ion equally, while the archaeologleal example is more heavily weighted towards prepared core reduction. It may also over-reflect culling In the rnedfum complete flake category. Medium medial/distal fragments arc heavily culled in the experimental sequence and this does not appear to be the case in the archaeological example. Culling may have centered more on complete and proximal fragments.
Medium proximal fragments are especially reduced in case 21. Finally, small proximal fragments are under-represented in the experimental data set due to equal input from bipolar and prepared core reduction. This does not appear to be the case in the archaeological example where the form of reduction producing the most debitage was prepared core reduction, Although trampling may have contributed to some degree to this assemblage, especially through inflation of the medium medial/distal category, it does not appear to have been a major factor. Case 67 is quite similar to that of 21 with two major exceptions. First, it appears to have been trampled to a greater degree with substantial reductions in complete flakes and increased numbers of proximal and medial/distal fragments.
Second, it does not appear to have been culled to any recognizable degree, The pattern clearly reflects mixed prepared and bipolar core reduction with substantial trampling. As in case 21 reduction is weighted towards p~eparedcores. Further construction of utility index sequences is required to understand cases 2, 61 and 66. The FYIxASaEL mixed residual assemblage (Table 79, Col. 7) was further modified following the production of a new set of values on mfxed HaEL culling (Cols. 8-10). W mixed prepared and bipolar core reduction HAEL index was produced combining resealed data sets from Tables 66 and 71, column 8 (Table 79, Col. 89 and then rescaling these data (Col. 9). To anticipate the effects of
HZaEL culling on a previously FVHxAAEL culled lor residual) assemblage, the data in column 9 was subtracted from 3.00 (Col.
19) and then multiplied against the data from column 7 (COP. II) and rescaled (Cole 12).
A comparison between case 61 and the data from Table 99, column 12 indicates that the modelled sequence anticipates much of the variation from the archaeological cases (Figure
133). Some variance between the two appears in the small nonorientable fragments and small and medium split flakes. Small nonorientable fragments and split flakes are over-represented by the experimental data. I expect this to be the tesult of equal input from bipolar reducti~nin the experimental example whereas it is less strongly represented arckaeologically. Pt also may reflect the effects of trampling, The archaeological cases appear to be quite I Table 79. Housepit 7 debitage analysis Factor 3 experimental sequence A. (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake; Rest.-Rescaled). Mixed Prepared Block Core and Bipolar Core FVIx-L Residual Index Table 66 Table 66 Col* I. Col* 9 f f CoLL. 9 Table 31 Table 71 100 - X COP. 1 Reuc. Col. 9 Resc. Col. 4 C0l. 2 Resc, (11 (2) (3)* (4) (5) (61% (7) Large CF MB Med i urn CF PF MDF NF SF Small CF 1.0 3.2 7.1 1.7 98.3 31.5 3,2 PF 8.0 25.8 1.7 0.4 99.6 257.Q 26.2 MDF 31.0 180.0 7-9 1.9 98.1 981.0 100.0 NF 6.0 18.4 3.4 0.8 99.2 192.5 19.6 SF 6.0 19.4 3.2 0.8 99.2 192.5 19.6
*All values divided by 10 Table 79. Contnd.
Secondary HAE;L Resfdual Index
Table 66 Col. 8 -b Csl* 10 Table 71 100 - X Col. 8 Resc. Col. 9 Cox. 7 Resee ($1 (a) (10) (II)* (1.2) Large CF MD Fled i urn CF PF MDF NF SF Small CF PF MDF NF SF Table 74 Case 6 B
Table 79 Column 12
Figure.133. Comparison of case 61 with mixed prepared core reduction (FVIxAA9L residual) and bipolar core reduction (MAEL residual) index (hawer case=small flakes, upper ease=medhum flakes, darkened upper case=large flakes).
504 heavily trampled, while the experimental data set is not trampled. The experimental data set over-represents medium split flakes. Medium split flakes and medium complete flakes
(cases 2 and 66) appear to have been removed through more extensive culling and possibly trampling, The high numbers of medium nonorientable fragments does not reflect trampling, though trampling is clearly indicated by the proportions of other flake types. This may indicate that bipolar reduction was a reduction strategy occurring after most trampling was completed. Thus, it may have occurred immediately before abandonment of the structure.
Case 97 appears to be unique in the sense that it combines bipolar reduction and biface production. To better understand this assemblage, a second experimental sequence was undertaken (Table 80). Since the primary form of flake culling in biface reduction contexts on the floor of housepit 7 appears to be oriented towards acute edge angled flakes
(AAEL) thls sequence evaluates the effects of W&culling in a mixed context. As trampling was important in the previously considered cases on factor 3, trampling was included in this experimental sequence. First, raw trampled HSRT data from hard hammer biface production and bipolar core reduction were combined (Table 80, Col. 1) and rescaled (Col. 29. Then, raw
AAEL data from both reduction types were combined (Col. 31 and rescaled (Col. 4). This created the basis mixed AAEL index. The residual index was created by subtxacting column 4 from 100 (Col. 51, multiplying column 5 by column 2 (Csl. 6) and
50s Table 80. Housepit 7 debitage analysis Factor 3 experimental sequence B. (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake; Ress.=Rescaled).
Mixed Trampled Hard Hammer Eiface and Bipolar Core AAEL Residual Index Table 60 Table 60 Col. 28 Col. 4 t + Col. 5 Table 71 Table 71 100 - X Col. 28 Resc. Col. 4 Resc. Csl. 4 Cal. 2 Wesc. (1) (2; C 3 1 (41 (5) (GI* (71 Large CF 0 .0 0.0 76.7 POO.0 0.0 0.0 0.0 PF 1.0 2.4 51.2 66.8 33.2 8.0 1.0 MD 1.0 2.4 74,6 97.3 2.7 6.5 0.1 Med i urn CF 2.0 4.9 38.7 50.4 49.6 24.3 2.9 PF 4.0 9-8 15.3 19.9 80.1 78.5 9.5 MDF 8.0 19.5 12,8 16.7 83.3 162.4 19.6 NF 2.0 4.9 0.0 0.0 100.0 49.0 5.9 SF 4,O 9.8 21.5 28.8 72.0 70.6 8.5 Small CF 7.0 17.1 21.8 28.4 71.6 122.4 14.8 PF 16.0 39.0 15.7 28.5 79.5 310.5 37.4 MDF 41.0 100.0 13.2 17.2 82.8 828.0 100*Q NF 7.0 17.1 0.0 0.0 100.0 70.0 8.5 SF 3.0 7.3 12.5 16.3 83.7 61.1 7.4
*All values divided by 10 rescaling these data (Col. 7).
A comparison between the archaeological data of case 97 and the experimental va?ues from Table $0, column 7 (Figure 134) illustrates a substantial degree of similarity between
these data sets. Two flake types vary between the data sets. Small proximal fragments are under-represented in the experimental data. This is, again, the result sf equal mixing of bipolar and, in this case, bifacial data, his causes increases in numbers of medialidfstal fragments and decreases In proximal fragments. With more emphasis on biface reduction, more proximal fragments would be expected, as is apparent in the archaeological assemblage. Medium proximal fragments are lacking from the case 97 data set. This appears to be due to culling togethex with extensive trampling. When medium proximal fragments are trampled they reduce to the small size xange and they produce both medium and small medial/distal fragments. This could also partially explain the exceptionally large number of small proximal fragments in this assemblage.
TO summarize, factor 3 consists of several different assemblages each containing mixes of bipolar core reduction and either biface or prepared core reduction debitage. A11 have been culled and most trampled. Culling has focussed primarily on acute edge angle flakes of any size as well as larger acute edge angle flakes in some contexts. Three assemblages were secondarily culled for high edge angle flakes. Trampling in these latter cases may have occurred Table 74 Case 97
Table 80 Column 7
Figu~e134. Comparison of case 37 with the mixed trampled biface and bipolar core reduction WLresidual index [ Power case=small flakes, upper case=medium flakes, darkened upper case=large flakes), primarilly before bipolar flakes were deposited.
Factor 4 contains significant positive loadings on large complete flakes and significant negative loadings on small split flakes (Table 77). The presence of large complete flakes in Pow quantities in medium flake production prepared
core reduction assemblages was predicted experimentally (Table
66, Col. 2). The presence of these in archaeological contexts indicates culling narrowly oriented towards acute edge angle flakes or a lack of culling, as these flakes are typically large enough to have many uses, especially for tasks requiring obtuse edge angles.
Six eases (2, 14, 19, 55, 68 and 89) contain positive factor scores above 1.0 and are considered to be critical contributors to the positive dimension of this factor (Table
789. Case 2 is not considered further here as it was discussed under factor 3. There is, in general, little variation between these cases. All are dominated by small medial/distal fragments, followed by small proximal fragments and medium medial/distal fragments. Complete flakes are present as are small nonorientable fragments. Split flakes are extremely uncommon. Some variation does occur in the medium size class. Cases 14, 19 and 68 are dominated by mediali'distal fragments, while cases 55 and 89 are dominated by proximal fragments. The primary small flake type pattern appears to be that of prepared core reduction. Small proximal fragments are too few to be representative of biface production. Trampling is clearly present also in the form of reductions in complete flakes and nonor ientable fragments . A comparison between case 14 and Table 66, column 37 (prepared, medium flake production, block core, trampled
FVIxAAEL residual di=trlbutlon) indicates that it 1s llkaly that culling of some assemblages (14, 19 and 68) may have been oriented towards removing large acute edge angle flakes (Figure 135). This type of cull leaves medium medial/distal fragments most common and proximal and complete flakes reduced by comparison. Cases 55 and 89 are patterned differently and although they appear to have been trampled, pxoxlmal fragnents are most common. The most parsimonious explanation of this is a lack of culling. When prepared core reduction assemblages are trampled but not culled a pattern of high numbers of proximal fragments appears through reductions in the formerly high complete flake category. The pattern present is not strong enough to support arguments for any other forms of culling,
Three cases (41, 79 and 98) have ;ligh negative factor scores* This is ?ri%arily due te the fact that each has an exceptionally high number of small split flakes, Two general patterns appear to be present. Case 41 is a classlc prey3red core assemblage, with numerous nonorientable fragments, an equal number of proximal fragments and reduced complete and Table 74 Case 14
40 60 Table 66 Column 37
Comparison of case 14 with the medium flake, prepared core reduction, trampled FVIxkAEL residual index (lower case=s~mailflakes, upper case=medium flakes, darkened upper case=large flakes 1. split flakes. No flakes outside the small size class are present. This cannot be interpreted as anything other than a size sorted assemblage. The cause of this size sorting is difficult to pin down and may include some combination of culling, trampling and/or larger debris clear ing . Cases 79 and 98 are well explained on factor 1 as trampled hard hammer biface reduction assemblages culled for acute edge angle flakes.
Factor 5
Significant positive loadings are found in factor 5 on large medial/distal fragments and medium split flakes (Table
77). Eight cases (2, 5, 42, 56, 68, 71, 73 and 98) have high positive factor scores (Table 78). With the exception of cases 71 and 98, all of these appear to have been produced through prepared core reduction. Only two of the eight cases contributing to this factor are unique to it (cases 42 and
73). All others have been adequately dealt with elsewhere.
Case 42 is patterned in much the same manner as cases 55 and 89 which were interpreted to be most likely the result of prepared core reduction and trampling without any significant culling. It is Likely that case 42 is the result sf a similar process, It is different from the latter two assemblages only in the presence of one large medial/distal fragment.
Case 73 is a typical trampled prepared core zeduction assemblage culled for larger acute edge angle flakes. This pattern shows up repeatedly throughout this analysis and is strongly represented on this factor. It is described on
factor 1 (negative dimension) and factors 3 and 4 (positive dimensions ) .
Factor 6
Significant positive loadings appear on rnedLum complete flakes and proximal fragments and small complete flakes and nonorientable fragments (Table 77). These are the flake types typically most modified during trampling which may indicate that this is a trampling factor. Thus, assemblages with high quantities of these flake types cannot be expected to have been significantly altered by trampling. Ten cases (3, 15,
21, 44, 69, 71, 78, 82, 86 and 87) have factor scores beyond the 1.0 mark and are primary contributors to this factor
(Table 78). Two cases have been well described on other factors.
Case 21 was documented on factor 3 and appears to be the result of mixed prepared core reduction and bipolar core reduction. Although trampling may be present, it appears to be minimal, given the presence of numerous medium complete flakes and nonorientable fragments. Case 71 has been described on factor 2 (negative dimension) and appears to be the result of mixed tool edge maintenance and hard hammer biface reduction.
The critical pattern isolated by this factor appears in 513 cases 3, 15, 44, 86 and 87. Here, both medium and small complete flakes are relatively common as are small nonorientable fragments. Medium and small proximal fragments
- are also quite common. This patterning appears to indicate a lack of trampling as we31 as a high degree of effort expended
in platform preparation. Comparison between case 86 and Table 66, c~lurnn2 (prepared, medium flake pnoduction, block core MSRT distribution) reveals some interesting similarities and
differences (Figure 116). Medium complete flakes and small split flakes axe heavily over-represented in the experimental data. Variation In small split flakes is probably the result of variation in hammer weights used to reduce cores. Prehistoric hammerstones may have simply been less hard than those used in this experimental study, thus resulting in fewer small split flakes. Medium complete flakes do not appear to
have been removed through trampling. Had this been the ease, small complete flakes and nonorientable fragments would also have been reduced. This has not occurred. These flake types are actually under-represented in the experimental data. The presence of high numbers of small complete flakes and nonorientable fragments is the result of hard hammer reduction without any effects from trampling. It is more likely that reductions in medium complete flakes as well as medial/distal
fragments from the housepit ? data, are the result of culling
behaviors oriented towards large high edge angle flakes. Cases 87 and 44 also fit this pattern and can be explained in
this manner. Cases 3 and IS do not have the same ratio of Table 66
Figure 136. Comparison of case 86 and the medium flake, prepaxed core reduction data set (Iswer case=small flakes, upper case=medium flakes, darkened upper case=lazge flakes). medium complete flakes to proximal fragments and although they are similar in all other respects, they appear not to have been culled.
Cases 69 and 78 are relatively minor coneributors to this factor and contain distributions indicative of (better explained on factor 5) typical prepared core reduction with some trampling and a definite pattern of culling for large acute edge angle flakes. ~Lthoughhaving some attributes of an untrampled assemblage (high small nonorientable fragments) case 82 fits the expected profile for a txampled hard hammer biface production assemblage culled for larger acute edge angle flakes.
Eight cases (29, 31, 32, 40, 45, 73, 79 and 84) contain strong negative factor scores on factor 6. In each case, the medium category is totally dominated by medial/distal fragments. Other flake types are'either entirely lacking or selatively uncommon. Small flake distributions in cases 29, 31, 32, 73 and 84 indicate prepared core ~eductionwith Tow proximal fragments and variable numbers of nonorientable fragments. In general these assemblages appear to heavily trampled as well as culled for larger acute edge angle flakes.
Case 40 is better explained on factor 2. Cases 45 and 79 appear to xepresent biface reduction assemblages also heavily cuIEed for acute edge angle flakes. The presence of numerous small complete flakes and nonorientable fragments argues against excessive trampling as does appear to be the case in the other assemblages. Summary of Basic Pattern Recognition Criteria For All Other
Gases
A number of cases analyzed in the principal components analysis were not formally discussed above due to the presence of positive or negative factors which did not rise or drop, respectively, above 1.0 or -1.0. All of these assemblages contain important data which has contributed to the multivariate solution. I assessed each of the remaining cases for the effects of reduction technique, culling and trampling using criteria derived from the multivariate analysis. Prepared core reduction was indicated by small proximal fragment scores of less than 60.0 (generally 20.0 to 50.0 in unambiquous cases! and unless trampled heavily, nonorientable fragment scores greater than 6.0. Small proximal fragment scores greater than 60.0 were considered to be the result of some form of tool production (thinning or edge shaping/modification core or flake tools). Any assemblage with small nonorientable fragments scoring greater that 40.0 was defined as the primarily the result of block cone reduction. Untrampled assemblages contained relatively substantial numbers of small complete flakes (greater than 10.0 when nonorientables are between 5.0 and 10.0) and nonorientable fragments (greater than 10.0 when complete flakes are between 5.0 and 10.0). However, when medium size class complete flakes are common (i.e. 10.0 ox greater) trampling is not used to explain reductions in small complete flakes or nonorientable fragments, Likewise, when medium complete
flakes are reduced (0 to 5.0) and medium medial/distal fragments and proximal fragments have roughly equivalent scores (i.e. 5.0 each), Pack of trampling cannot be used to explain increased small complete flake and nonorientable fragment scores. Flake assemblages which have not been culled are dominated in the medium size class by proximal and complete flakes, though some medial/distal fragments are also present
(generally remaining above 5.0). Flake assemblages culled fox
AAEL or FVIxAAEL flakes are often heavily reduced for complete, proximal or split flakes. Thus, medial/dlstal
fragments axe dominant. This pattern is reversed with HAEL culls, where med ial/d istal fragments are highly reduced (below
5.0) or eliminated from the assemblages.
Spatial Patterning
All cases not formally discussed were assessed •’ow the effects of reducti on techniques, culling and trampl ing using the criteria discussed above (Table 811. I will now describe the spatial patterning resulting from this analysis.
Implications for occupational history will be discussed in the final section of this chapter. Cases interpreted to be the result of tool maintenance Table 81. General summary of debitage analysis conclusions on a case by case basis (Tr=trampled, MT=Not Trampled, Prep= Prepared PPatform, Resharp=Maintenance/ resharpening of tools, SS=size sorted). Case Interpretation 1 Tr ., Prep. Core Reduction, FVIxAAEL type cull 2 Tr., Bipolar and Prep. Core Reduction, HAEL type cull 3 NT., Frep Core Reduction, No cull 4 Tr., Prep. Core Reduction, FVIxPldlEL type cull 5 NT., Prep. Core Reduction, FVIxAAEL type cull 6 Tr., Biface Reduction, AZBL type cull 7 Tr., Prep. Core Reduction, HAXL type cull 8 TE., Prep. Core Reduction, FVIxAAEL type cull 9 Tr ., Prep. Core Reduction, HML type cull 18 Tr., Prep. Core Reduction, FVIxAAEE type cull I1 Tr., Prep. Core Reduction, FVIXAAEL type cull 12 Tr., Prep. Core Reduction, FVHxAASL type cull 13 Tr., Prep. Core Reduction, FVKxAAEL type cull 14 Tx., Prep. Core Reduction, FVIxWL type cull 15 NT., Prep. Core Reduction, Mo cull 16 Tr., Prep. Core Reduction, FVIxAAEL type cull 17 Tr., Resharp. and Biface Reduction, AAEL type cull 18 Tr., Prep. Core Reduction, FVPxAAEL type cull 19 Tr., Prep. Core Reduction, FVIxAAEE type cull 20 Tr,, Prep. Core Reduction, FVKxWL type cull 21 Tr,, Bipolar and Prep. Core Reduction, FVIxWL cull 22 MT., Prep. Core Reduction, HAEL type cull 23 NT., Prep. Core Reduction, HAEL type cull 24 Tr., Prep. Core Reduction, FVPxaAEL type cull 25 Tn., Prep, Core Reduction, FVIxAAEL type cull 26 Tr., Prep. Core Reduction, FVfxWL type cull 27 NT., Prep. Core Reduction, HWlEL type cull 28 Tr,, Prep. Core Reduction, FVfxlbAEL type cull 29 Tr., Prep. Core Reduction, FVPxaAEL type cull 30 Tr., Prep. Core Reduction, FVXxWL type cull 31 Tr., Prep. Core Reduction, FVIxWL type cull 32 Tr., Prep. Core Reduction, FVIxAAEE type cull 33 Tr., Prep. Core Reduction, FVIxAML type cull 34 Tr., Prep. Core Reduction, FVIxAWEE type cull 35 Tr., Prep. Core Reduction, FVfwaAEL type cull 36 Tr., Prep. Csxe Reductfon, FVXxAAEL type cull 37 Tr,, Biface Reduction, Fk#EL type cull 38 Tr., Prep. Cone Reduction, PVZXAAEL type cull 39 Tr., Biface Reduction, No 6~11 40 Tr., Reshaxp. No cull. 41 TE., Prep. Core Reduction, SS or WAEL+AAEL cull
------*-<4------Table 81. Contnd.
42 Tr., Prep. Core Reduction, No cull 43 Tr., Prep. Core Reduction, PVIxAAEE type cull 44 NT,, Prep. Csre Reduction, HAEE type cull 45 Tn. (minor 1, Biface Reduction, AAEt type cull 46 Tr ., Prep. Core Reduction, FVPxPlAEL type cull 47 Tn., Prep. Csre Reduction, FVIxML type cull 48 Tr., Prep. Core Reduction, FVIxWL type cull 49 Tr., Prep. Core Reduction, SS or HAEL+UE& cull 50 Tr., Prep. Core Reduction, HAEE type cull 51 Tr., Prep. Core Reduction, FVIxRAEE type cull 52 Tr., Resharp. and BBface Reduction, No cull 53 %re, Prep. Core Reduction, HAEL type cull 54 TE., Bfface Reduction, AAEL type cull 55 NT., Prep. Core Reduction, No cull 56 TE., Prep. Core Reduction, FVIxAAEL type cull 57 Tr., (rnPnor1 Biface Reduction, Mo cull 58 Tr., Prep. Core Reduction, No cull 59 MT., Prep. C~reReducti~n, WAEL type cull 6Q Tr., Prep, Core Reduction, FVIxBVlEL type cull 61 Tr., Prep. (FVIxAAEk cull), Bipolar (MAEL) reduct. 62 Tr., Biface Reduction, AAEL type cull 63 Tr., Prep. Core Reduction, FVIxAPlEL type cull 64 Tr., Prep. Core Reduction, No cull 65 Tr., (minor) Biface Reduction, AAEL type cull 66 Tr., Prep. (FVIxABEL cull), Bipolar (HAEL) reduct, 57 Tr., Bipolar and Prep. Core Reduction, No cull 68 TK., Prep. Core Reduction, PVIxAAEL type cull 69 Tr., Prep. Core Reductfon, FVIxAAEL type cull 70 Tm., Prep. Core Reduction, FVIxdlAEE type cull 71 NT., Resharp. and Biface Reduction, HML cull 72 Tr., Prep. Core Reductian, FVIxWL type cull 73 Tr., Prep. Core Reduction, FVIxRAEL type cull 74 Tr., Prep. Core Reduction, HAEL type cull 75 Tr., Prep. Core Reduction, FVIxdaAEL type cull 76 Tr., Prep, Core Reduction, FVIxABaEL type cull 77 Tr., Prep. Core Reduction, FVIxaaEL type cull 78 Tr., Prep. Core Reduction, FVIxPIWESL type cull 79 Tr., (minor) Biface Reduction, WLtype cull 80 Tr., Prep. Core Reduction, FVkxkAEE type cull 81 NT., Prep. Core Reduction, HAEL type cull 82 NT., Biface Reduction, Ne cull 83 Tr., Prep, Core Reduction, FVIxAaEL type cull 84 Tr,, Prep. Core Reductim, BVExAAEL type cull 85 Tr., Prep. Core Reduction, FVIxAkLEL type cull 86 NT., Prep. Core Reduction, No cull Table 81. contnd. ------87 NT., Prep. Core Reduction, HAEL cull 88 Tr., Resharp, No cull 89 Tr., Prep. Core Reduction, No cull 90 Tr., Prep. Core Reduction, No cull 91 Tr., (minor) Biface Reduction, WAEL type cull 92 Tr., Prep. Core Reduction, FVLxWL type cull 93 TE., Prep. Core Reduction, SS or HAEL+&%EL cull 94 Tr., Prep. Core Reduction, FVHxAdlEL type cull 95 Tr., Prep. Core Reduction, FVIxAAEE type eull 96 Tr., (minor) Biface Reduction, AAEL type cull 97 Tn., Bipolar and Biface Reduction, AAEL cull 98 Tr., (minor) Biface Reduction, AAEL type cull 99 Tr., Prep. Come Reduction, FVPxAAEL type eull 100 Tr., Resharp, No eull 101 Ta., (minor) Biface Reduction, No cull 102 Tr ., Biface Reduction, aAEL eull and resharpening (edge modification of bifaces and flake tools) are located, with one significant exception, adjacent
to hearths (Figure 137). Most notably they locate primarily on the wall sides of the hearths. One case is located roughly in the center sf the floor. It is possible that this is a place where light shown through the smoke hole in the roof at a certain time of day allowing better visibility for detailed tool maintenance activities. Similar practices have been documented in other contexts (c.f. Binford 1983:181). Cases interpreted to be the result of biface reduetion alss cluster in association with hearths (Figure 138). With the exception of a large cluster on the west central side of the floor, biface reduction is alss located on the wall sides of the hearths. This central cluster area may also have been partially the result of the location below a good light source. These patterns of tool reduction and edge modification are replicated in an independent analysis conducted by Spafford (1991). Spafford's analysfv shows that. biface thinning flakes cluster around the wall sides of hearths, most densely on the west side of the housepit Zloor. Prepared core reduction is ubiquitous on the floor (Figure 139). Several areas adjacent to hearths have been less affected by prepared core reduction. It Is important to realize here that some core reduction (and other reduction type) overlaps may occur in units not identified as the result of this technology. This is inevitable, as interpretations are made on the basis of overall flake size and breakage Figure 137. Distribution of analytical units interpreted to be associated with tool mintenance/resharpening. - ,edge of bench
Figure 138. Distribution of analytical units interpreted to be associated with blface reduction. ff ,- ,edge - of bench /";'., pits used 11 4 -edge of floor
FLgure 139. Dlstributlon of analytical units interpreted to be associated with prepared core reduction. Figure 140. Distribution of analytical units interpreted to be associated with bipolar core reduction. Figure 141. Distribution of analytical units not interpreted to be associated with trampling. fire-reddening
- ,edge of bench pi, wed & -edge of fbor oiclcnapatiisa
Flgure 142. Dlstrlbutlon of analytical units interpreted to be associated with acute edge angle flake culling. distributional patterning. Minute inputs %rom other reductlsn types will not be recognized. In general, prepared core reduction appears to have been carried out in almost all areas of the floor. Bipolar core reduction occurs intensely in two restricted areas, the northeast corner and the west-central zide sf the floor (Figure 140). In the northeast area, it overlaps with @ prepared core reduction. On the west side, Pt overlaps with both blface and prepared core reduction. Trampling is common to most areas of the floor with some significant exceptions (Figure 141). Untrampled areas tend to be located where post-holes are dense, in some hearths and along some walls. This is to be expected as these are the types of places least likely to receive foot-traffic. Lack of trampling in some hearth areas may indicate continuous use or designation of that location as a'pface not to be stepped on. Presence of trampling in other hearth areas may indicate discontinuous reuse 02 those places. Culling for acute edge angle flakes from biface and prepared core reduction is found throughout the floor (Figure 142). Culling for high edge angle flakes is found in the southwest, northeast and northwest corners of the house floor
(Figure 1431. Cases witkeut indications of cufllng behavior axe found primarilly against walls or in clusters of post-holes (Figure 144). Exceptions to this pattern are found in the southeast corner and in the west central portion of the floor. One of these is a maintenance assemblage which may - ,edge of BexB pi% xsed d in latest edge of floor ocrcnpatba
Figure 143. Distribution of analytical units interpreted to be associated with high edge angle flake cufllng. 530 Ffgure 144. Dist~ibutionof analytical units interpreted not to be associated with any form of flake cuiling. have been of little value to prehistoric inhabitants of the housepft ilsor due to the small size of the productu. The other (southeast corner) is harder to explain as it is the result of biface reduction. In this case, the primary goal of reduction may have been the blface, not the flakes.
TO summarize, this analysis has provlded interpretations of debitage assemblage patterning where tool maintenance and biface reduction centers primarily around the wall sides of hearths (conclusions also arrived at by Spafford using different techniquesf, prepared core reduction occurs almost everywhere and bipolar core reduction appears to be concentrated in two specialized locations. Trampling is common, although variable on top of hearth features, which may indicate different hearth reuse histories. Flake culling for acute edge angle flakes appears to occur over most of the floor while high edge angle flakes are culled in three specialized types of areas. Assemblages without indicators of culling appear primarily around floor edges or in clusters of post-holes. Clearly, flakes were produced and used as tools preventing them from appearing in the unmodified debitage data set. The next step is to evaluate the flake tool assemblage. Primary questions, stemming from the debitage analysis, regarding the flake tool assemblage are as follows:
I. Are the flakes culled from the debitage assernb~ages on the housepit floor being used and discarded back onto the floor or are they leaving the house?
2. if flake tools are being discarded back onto the floor, are they discarded in the same areas as the original flakes were culled from?
3. To what degree are flake tools subjected to trampling and reuse?
4. what are the effects on the flake tool assemblages of long term use of some flake tools versus short term use of others?
These questions are addressed in the following analysis of flake tools.
FLAKE TOOL ANALYSIS
Flake tools (N=S87) were sorted into MSRT flake types to facilitate pattern recognition using utility indices.
Analytical suitability of each tool depended on the presence of attributes of the original flake from which the tool was produced. Tools without clear original flake attributes were not used. These included blfaces and some heavily modified unifacial scraping tools. The same MSRT criteria was used as in the case of the unmodified debitage analysis. Edges were considered to be intact regardless of modification if retouch was either invasive on an otherwise clearly intact edge, or semi-abrupt extending less than 3 millimeters from an otherwise intact edge. Edges were not considered to he intact if abrupt or semi-abrupt retouch (beyond 3 mm.) was present. In this situation it is impossible to tell whether an edge had been intact before modification. The goal here was to attempt to identify the original beakage condition of the flake before its modification and use as a tool. In this way knowledge derived from experimental utility indices could be used to interpret technological origins of flake tool asuemblayeu. This also facilitated the recognition of trampling and tool re-use or scavenging, Flake tools were grouped into 13 cases based on the previously defined analytical sectors on the housepit floor
(Tables 82 and 83). All except one sector have substantial numbers of artifacts [ranging from 21 to 82). Sector twelve contains a predictably low number of artifacts (PI), but was retained for the analysis as the low numbers of artifacts may indeed be significant to the use history of that part of the floor. My interpretation of the formational history of this assemblage will attempt to consider sample size as a critical variable. Before presenting the multivariate analysis, I first discuss the data distributions. Two principal patterns are found in the small flake tool category. Most comiiion is a pattern cf dominance by small medial /distal tool fragments (cases 1-5 and 10-13) . Proximal toof fragments are in most cases, the next most common, although they are almost or entirely missing from cases 1-3. Table 82. MSRT flake tool data matrix from the floor of housepit 7 QCF=Complete Flake, PF=Proximl Fragment, MDF=Medial/DistaE Fragment, SF=Split Flake).
Assemblage 1 2 3 4 5 6 7 8 9
Extra Large MDF Large CF PF MDF SF Med i urn CF Pi? MDF NF SF Small CF PF MDF NF SF Tabhe 82. Contnd.
Assemblage 10 11 12 13 Extra Large MBF 0 0 0 0 Large CF 0 0 0 0 PF 0 0 0 0 MDF 2 0 1 2 SF a 0 o o Medium CF PF MDF NF SF Sma b l CF PI? MDF NF SF Table 83, Rescaled MSRT tool data matrix from the floor of housepit 7 (CF=Cornplete Flake, PF=Praximl Fragment, MDF=Med ial/Bis tal Fragment, SF=SpPit Flake). Assemblage 1 2 3 4 5 6 7 8 9 Extra Large HDF 0.0 0.0 0-9 0.0 0.0 0.0 3.4 0.0 0.0 Large CF 10.8 8.0 8.0 0,0 0.6 6.0 0.0 5.2 0.0 PF 30.0 0.0 $1.1 21.9 8.3 1 0.0 5.2 27.3 MDF 10.0 28.6 5.6 0.0 33.3 11,1 6.9 5.2 8.7 SF 10.0 0,O 0.0 0.0 8.3 0.0 0.0 0.6 4.3 Medium CF 28.0 0-0 16.7 6.3 $.3 11.1 13.8 15.8 17.4 PF 50.0 71.4 50.0 28.1 66.7 77.8 51.7 84.2 100.0 MDF 60.0 85.7 100.0 28.1 75.0 100.0 106.0 100.8 95-7 MF 0-0 0-0 5.6 0.0 8.3 0.0 0.0 0.0 0.0 SF 0.0 0.0 5.6 3.1 0,O 0.0 0.0 15.8 4.3 Small CF 0.6 14.3 0.0 3.1 0.0 8.0 3.4 0.0 4.3 PF 10.0 0.0 0.0 21.9 25.0 11.1 27.6 0.0 26.1 MDF 100,O 100.0 108.0 100.0 100.0 66.7 44.8 84.2 69.6 PIP 0,O 0.0 0.0 0.0 8.3 8.0 0-0 0.0 0.0 SF 0.0 0-0 0.0 3.1 0,O 0.0 3.4 0.0 0,O Table 83. Contnd.
As sernblage 10 11 12 13 ...... Extra Large MDF 0.0 0.0 0.0 0.0 Large CF 6.0 0.0 0.0 0.0 PF 0.0 0.0 0.0 0.0 MDF 25.0 0.0 33.3 10.5 SF 12.5 0.0 0.0 0.0 Med i urn CF 0.0 71.0 0.0 0.0 PF 50.0 17.9 66.6 31.6 MDF 50.0 25.0 180.0 42.1 NF 12.5 3.6 33.3 10.5 SF 25.0 0.0 0.0 0.0 Small CF 0.0 3.6 0.0 0.0 'PF 37.5 17.9 33.3 21.1 MDF 109.0 100.0 100.0 100.0 PJF 0.0 3.6 0.0 10.5 SF 0.0 0.0 5.3 ------3.6 Complete and split flake tools and nonorientable tool fragments are extremely rare or nonexistant. The other pattern Is that of comparatively reduced medial/distal tool fragments (cases 6-91. Case 8 is also missing all small proximal tool fragments. As all of these flake types are also flake tools, it msy be that these are primarily pieces of larger tools which have broken through use and/or trampling.
within the medium size class, most cases are also dominated to a greater or lesser degree by medfaf/distal tool fragments. case 9 deviates from this pattern with a maximized couct of proximal tool fragments. Cases 4 and 10 contain equal numbers of medial/distal and proximal tool fragments.
Cases 4 and 11 have comparatively low numbers of medial/distal tool fragments, 8% before, split flake tools and nonorientable tool fragments are quite rare. Complete flake
Large artifacts are present in reduced numbers and their patterning is best dezilt with in the multivariate analysis. Extra-large artifacts are extxemely rare, appearing only in one case (7).
A correlation matrix (Table 84) was produced from the converted data matrix (Table 83). This was subjected to a principal coii;iponents analysis from vhich 5 factors were extracted and varimx rotated (Tables 85 and 86) capturing 80.7% of the total variance. Factor scores were produced to measure the contribution of each ease to the factor solution
[Table 8?). Table 84. Housepit 7 flake tool analysis correlation matrix (CF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake).
Extra- Large Large Large Large Large PF MDF SF ------MDF CF Extra- Large MDF 1.000 Large CF -0.116 PF -0.2 41 MDF -0.170 SF -0.178 Pfedium CF 0.196 PF -0.043 MDF 0.267 NF -0.194 SF -0.161 Small CF 0.089 PF 0.232 MDF -0.744 NF -6.145 SF 0,348
Medium Medium Med i urn Medium Med i urn CF PF MDF NF SF ------Medium CF PF MDF MF SF Small CF PF MDF NF SF Table 84. Contnd.
small Small Small Small Small. CF PF HDF MF SF
Small CF 1.000 PF -0.282 1.000 MDF -0.024 -0.105 1.000 biF -0 212 0.174 0.299 1.000 SF 0.001 0.219 -0.068 0.512 1.000 Table 85, Housepit 7 flake tool analysis rotated loadings (CF=Complete Flake, PF=Proximal Fragment, MDF=Hedial/Distal Fragment, SF=Splft Flake).
1 2 3 4 5 ------Extra- Large MDF -0.062 0.114 -0.089 0.030 -0.039 Large CF 0.065 -0.751 0.205 0.050 0.143 PF 0.103 -0.816 0. LO6 0.067 0.007 MDF 0.529 0.568 0.390 0.303 0.053 SF 0.067 -0.217 0.290 0.526 0,585 Med i urn CF 0,231 -0,873 -0.337 -0.037 -0.039 PF 0.905 -0.109 -0.100 -0.048 0.059 MDF 0.887 -0.020 -0.349 -0.080 -0.137 NF 0.224 0.552 0.267 0.587 -0.084 SF 0.035 0.050 0.052 0.088 0.882 Small CF 0,084 0,322 0 020 -0.761 -0.019 PF -0- 157 0.422 -0.218 0.732 0.210 MDF -0 262 0,112 0.921 0.034 0.057 ESF -0.409 0.211 0.265 0,370 -0.495 SF -0.840 0 ,191 -0.301 0.013 -0.348 Table 86. Housepit 7 flake tool analysis initial statistics. -
Table 87. Housepit 7 flake tool analysis factor scores. Factor Factor Factor Factor Factor i 2 3 4 5
Case 1 -0.087 -2.293 0.803 0.548 0.177 Case 2 0.770 1.220 0.870 -2.403 -0.028 Case. 3 0.327 -0.570 0.218 -0.547 -0.259 Case 4 -1.466 -0.326 0.128 -0.527 0.229 Case5 0.514 0.204 0.789 1.307 -0.712 Case 6 0.813 -0,324 -0.515 -0.238 -0.527 Case 7 -0.207 0.380 -2.967 0.103 -0.131 Case 8 0.621 -0-818 -0.113 -0.705 0.657 Case 9 1.032 -0.707 -0.570 0.202 0.068 Case 10 -0.333 0.993 0.296 0.926 2.864 Case 11 -1.906 8.228 0.184 -0.531 -0.294 Case 12 1.215 1.391 0.478 1.207 -0 -684 Case 13 -1.492 0.622 0.398 0.662 -1.361 Factor 1
Factor 1 contains significant positive loadings on large medial/distal tool fragments and medium proximal and medial/distal tool fragments (Table 85). Positive factor scores are high for cases 9 and 12 and relatively high for
cases 2 and 6 (Tabhe 87). The patterning displayed in these assemblages is actually fairly generic to most assemblages. The small, medium and large size classes are dominated by medial/dfstal tool fragments (with the minor exception of case 9) and medium proximal tool fragments are also quite common. The positive dimension of factor 1 appears to have identified the primary commonality among all 13 cases in this analysis. It is of limited usefulness for understanding the variability in the matrix. I, therefore, Focus first on the negative dimension of this factor. significant negative loadings are found on small nonorientable tool fragments and split flake tools (Table 85).
Negative factor scores are high on cases 4, 11 and 13. These cases contain flake tool type profiles reminiscent of unmodified debitage assemblages. The presence of small nonorientable t~oLfragments and split flake tools indicates some input from hard havmer reduction. Medium complete flake tools are found in cases 4 and 11 which may also be indicative of hard hammer reduction. Unfortunately, the origins of these assemblages are not resolved this easily. Almost all 3ard
545 hammer reduction in the unmodified debitage assernblaqes was accompanied by culling for larger acute edge angled flakes.
If this is the case then the resulting tool profile should match this expectation. As it does not, I suspect that a more complex process of tool assemblage formation is operative, This problem is best understood through the construction of a new experimental sequence designed to illustrate the effects of trmplir~y,culling and use-breakaye an original cull populations of flakes chosen for use as flake tools.
To anticipate the appearance of cases 4, 11 and 13, I begin with a distribution of iarger acute edge angle culled flakes from prepared core reduction (Table 88, Col. 1). This dist~ibutionhighlights medium sized proximal fragments as well as complete flakes and medial/distal fragments. ~ll other flake types make negligible contributions. As these flake tools were used and discarded they may have begun to form a sort of a flake tool resource pool lying on the floor of the housepit. With this as the case, scavenging or culling of previously discarded tools may have occurred over time. The effects of this on assemblage compssition are demonstrated in columns 2-5 (Table 88). First, the small flake tool categories (especially the small proximal and medlal/distal tool fragment categories) expand tremendously because of the expected effects of trampling and use-related attrition of larger flake tools (and perhaps purposeful flake tool breakage). Medium proximal tool fragments are removed in their entirety and medium medial/distal tool fragments
546 Table 88. Housepit 7 tool anarysis Factor 1 sequence A (CF=Cornpleke Flake, PP=Pnoximl Fragment, MDF=Medial/Distal Fragment, SF=Split Flake; Resc.=Rescaled).
Large CF Medium CF PF MDF SF Small PF MDF NF SF increase substantially. This highlights the effects of repeated culling of previously discarded tools, thus reducing the numbers of available medium sized tools in discard contexts on the floor. This is the expected profile before final discard of tools culled from previously discarded populations of tools. After final discard, a different pattern emerges. This is due to the effects of use-related breakage as well as trampling. Thus, while the initial processes involved culling and deposition of acute edge angle flakes, final deposition contributes high edge angle flake tools (Table 88, Cols. 6 and 7). The final profile lu dominated by small and medium sized medial/distal tool fragments, followed to a much lesser degree by small and medium proximal tool fragments and medium complete flake tools. The profile also predicts the presence of large complete flake tools. A comparison of case 11 with the experimental data (Table
88, Col. 7) illustrates a great deal of congruity (Flgure 145)- Archaeological data vary primarily in the complete flake tool categories and in the medium medlal/distal and proximal tool fragment categories. The medium medial/distal and proximal tool fragments are distributionally similar between the experimental and the archaeological data though the experimental distribution over-anticipates the absolute numhexs of these tool types, It is likely that trampling, culling, purposeful breakage, and use-attrition of complete flake tools as well as medium medial/distal and proximal tool Table 87 Column 7
Table 83 Case I 1
Figure 145, Comparison of flake tool analysis case 11 with factor 1 trampled flake tool discard index A (lower case=small flakes, upper case=medium flakes, darkened upper case=large flakes). fragments may have been more extreme in the archaeological case than in the experimental sequence. Otherwise, I argue that the experimental sequence anticipates much 0% the variability of the portion of the archaeological record contained on cases 4, 11 and 13. Case 13 has some additional sources of variability which are discussed under factor 2. The flake tool assemblage formation processes identified above may be summarized as follows: It iu likely that the original tools were produced on flakes from prepared cores, culled for their size and acute edge angles. ~oll3winga sh~rtperiod of use, they were discarded and (as indicated in the comparison of the archaeological flake tool size and breakage data to the experimental distribution) subsequently culled from their discard contexts for additional use. Final discard and burial found the original tools in much more damaged states due to intensive use, modification, and trampling. From this analysis it is not clear that any single flake tool was used for a long term. Tools appear to have been produced for short term use only (based upon the dfstlnctive size and breakage distribution), resulting in a dynamic cycle of discard, trampling and reuse. Thus any intensive retouch present (see Spafford 1991) may have accumu%ated th~oughdiscontinuous episodes of reuse rather than any single period of intensive long term use. A critical component to factor 1 also is the emphasis in the negative dimension on small split flake tools and nonorlentable tool fragments. This means that these flake tool types are not contributors in the positive dimension and although this pattern is replicated In part on other factors
it is important to further discuss it here. Cases 2, 6 and 8 have relatively high positive factor scores on factor 1.
Cases 1 and 3 have very similar patterns to these 3 cases and may also be considered here. All have ext~emelynumerous small medial/distal tool fragments and practically no other small flake tools. This is indicative of intense trampling of small flake tools only, Mad larger flake tools been trampled together with small flake tools, then more small split and proximal tool fragments would have been contributed. With intensive trampling of small flakes, all categories are reduced to the point that only the one starting out with the most common (such as small medial/distal tool fragments) retains any at all. This process can only occur where larger flake tools have been removed or retained for longer term use. This process can be simulated with a second experimental sequence (Table 89 1. I begin the sequence with the result of the initial sequence (Table 88, Col. 7) representing the archaeological result of a system of tool use and discard (discussed above) where curation has not been a factor. Trampling of the small category only is demonstrated in columns 2-4, by eliminating the medium and large flake tools. The predictable results are a drastic reduction in all flake types except the small medial/distal tool fragments. Then, larger flake tools are reintroduced (Col. 5). This simulates the process of flake Table 89. Housepit 7 tool analysis Factox 1 experimental sequence 9. fGF=Complete Flake, PF=Proximal Fragment, MDF=Medial/Distal Fragment, SF=Split Flake; Resc.=Rescaled).
Large CF Med ium CF PF MDF SF Smail
PF ' 26.8 26.8 125.2 12.5 12.5 MBF 100.0 100.0 1000.0 100.0 100 .0 NF 4.5 4.5 3.8 0.3 .3 SF 4.6 4.6 3.8 0.3 .3 "Divided by 10 tool curation whereby larger flake tools are used and re-used for comparatively longer periods of time than si'fialler flake tools, which are quickly discarded and trampled, Fragments of larger flake tools continually enter the archaeological record as these tools wear out. Thus, the medium size category is no different between Table 88 (Co1.7) and Table 89 ICol. 5) while the small size class flake tool distribution is quite different . The similarity of the utility index based data to the archaeological is shown in a comparison of archaeological data
(using case 3 as a representative example) and Table 89, column 5 (Figure 146). The similarity between the t~ois great and it is likely that similar processes to those simulated in the experimental data may have produced the data from cases 1, 2 and 3. Most critical here, is the recognition of longer term use of larger flake tools. Smaller flake tools appear to have been consistently used for only short periods of time. Cases 6 and 8 also follow this pattern but are best discussed under factor 3.
Factor 2
Significant positive loadings are found on large medial/distal tool fragments, medium nonorientable tool fragments and small proximal tool fragments (Table 85). High positive factor scores are found on cases 2, 10, 12 and 13.
Case 2 has been best explained under factor 1 and is not considered further here. It appears on this factor only due to the presence of a high score on large medialidiutal tool fragments. The other three cases have a consistent pattern of relatively common small and large medial/distal tool fragments, somewhat reduced medium medial/distal tool
fragments (10 and 13) (compared to most other cases) and relatively common medium nonorientable tool fragments and small and medium proximal tool fx-.gments. A11 other flake types make negligible contributions. The presence of numerous medium nonorientable fragments may indicate the presence of bipolax reduction as a contributing factor to these assemblages. Otherwise, the distribution is somewhat reminiscent of that described in the negative dimension of factor 1. The effects of mixed technological input are demonstrated in a third experimental sequence (Table 90)- Columns 1-9 (Table 90) follow the same sequence as that of Table 88 although the rate of culling for previously discarded tools 1s increased. This has ths effect of decreasing the number of medium medial/distal tool fragments and complete flake tools present. Final reintroduction of broken and worn out tools (col. 9) provides a profile very similar to that of Table 88, column 7, only the numbers of flakes in each sf the medium categories have been reduced. The only reduction technique to consistently produce medium nonorientable frag'nents is bipolar reduction, If the bipolar
core is relatively thin (as in the case of a biface) many medium nonorientable fragments will have substantially long Table 90, Housepat 7 tool analysis Factor 2 experimental sequence (CF=Complete Flake, PF=Proximal Fragment,MDF=Medial/Distal Fragment, SF=Split Flake, PBC=Prepared Block Core; Resc.=Rescaled). Trampled Flake Cull Discard Population Index Culling Indices ------...... Table 66 Col. 3 Col. 5 Col. 47 X X Table 66 X Table 66 Table 66 Col. 47 Col. 22 Resc. Col. 22 Resc. Coh, 22 Resc, (1) (2) (3) (4) (5) (61 (71 Large CF 0.0 0.0 0.0 0.0 0.0 0.0 0,O Med i urn CF PI? MDP SF Small PF MDF NF SF
Trampled PBC Tool Discard Trampled Final Index Discard Index
Large CF Med ium CF 34.1 17.1 17.1 8.6 PF 62-6 31e3 3'1.3 Is.? MDF 116.7 58.4 117.7 58.9 NF 38.2 19.1 SF 22.8 11.4 11.4 5.7 Small PF 46.8 23.4 23.4 11.7 MDF 200.0 1OQ.O 200.0 100.0 NF 9.1 4.6 16.6 8.3 SF 9.2 4.6 4.6 2.3 edges with high edge angles- Thus, culling of a bipolar assemblage for high edge angle flakes will naturally include some medium nonorientable fragments. Mixing of the former assemblage with a trampled bipolar high edge angle cull assemblage is demonstrated in columns 10 and 11. The primary effects are reductions in all complete and split flake tools and proximal tool fragments and a concomitant increase in medial/distal tool fragments of all size clasueu , The experimental distribution (Table 90, Col. 11) anticipates much of the patterning from cases PO, 12 and 13 (Figure 147) (eases 5 and I1 do not score strongly on this factor but have many of the attributes discussed and may be similar in origins). It tends to slightly under-represent the proximal tool fragments. This may be a result of a slightly increased use of acute edge angle tools from bipolar reduction flakes in the archaeological record compared to that of the experimental distribution which anticipates the discard of high edge angle (primarily due to snapped edges) bipolar flake tools. It is evident however, that the archaeological data has been produced through a process of mixed flake tool discard, some originating in prepared core reduction Elakes and others from bipolar flakes. All assemblages appear to have been heavily affected by trampling.
Factor 3
A significant positive loading from factor 3 is found 557 0 20 - 40 60 80 100 Table 82 Case 13
Figure 149. Comparison of case 13 with the factor 2 trampled final discard index (lower ease=small flakes, upper case=medium flakes, darkened upper case=large flakes ) . only on small medial/distal tool fragments (Table 85). NO significant negative loadings are present. All caves (1-5 and
10-13) with small medial/distal tool fragments scoring 100.0 produce positive factor scores (Table 87). As discussed under factors 1 and 2, these assemblages are the result of prepared core medium sized flake production with occasionah mixes of bipolar flake production. The negative dimension of the factor 3 factor scores are most critical here as they appear to isolate a new type of flake production in this analysis. Cases 6-9 have comparatively reduced numbers of small medial/distal tool fragments and increased numbers of medium medial/distal tool fragments. Medium proximal tool fragments are common, as are medium complete flake tools, Significantly, nonorientable tool fragments are not present In any size class, The lack of nonorientable tool fragments, reduced small coreplete flake tools and increased m~dium proximal and medial/distal tool fragments indicates that although prepared core reduction could be contributing flakes, bifaee reduction may also be contributing flakes to these assemblages. To explore this possibility, a fourth experimental sequence was constructed (Table 91). The reduction in small medial/distal tool fragments indicates that discarded flake to01 culling and reuse may not have been as intense as In other areas of the floor. However, given the presence of both acute and high edge angle culling in parts of the floor, it is possible that a strategy of more intense useage of the original flake assemblages rather than 559 Table 91. Housepit 7 tool analysis Factor 3 exper Fmental sequence. (CF=Complete Flake, PF=Proximl Fragment, MDF=Me Trampled Trampled PBC Reduction BF + PBC Reduction Cull Index Cull Index ------Table 66 Col. 2 COP. 47 + f Table 60 Col. 45 Resc. Col. 43 Resc. (1) (2) (3) (4 1 Large CF PF Med i urn CF PI? MDF SF Small PF MDF NF SF the discarded tool assemblages may have occurred in these areas. Columns 1 and 2 (Table 91) demonstrate the effects of mixed high and acute edge angle cull assemblages, The result is a maximizing of the medium proximal tool fragment category followed by small medial/distal tool fragments and medium medial/distal tool fragments. Mixing with acute edge angle culled flakes from biface reduction is produced in columns 3 and 4. The result Is a pattern not unlike those of cases 7 and 9 (Figure 148). Cases 6 and 8 were mentioned under factor 1 as having indicators of mixed long term flake use as well as short term use and discard. A final sequence was constructed to demonstrate this (Table 92). This is designed to demonstrate the effects of differential trampling on the cull population produced in Table 91, column 4 (Table 92, Col. 1). First, in the same fashion as Table 89, all medium and large flakes were removed and a series of trampling events simulated (Table 92, Cols. 2-61, This has the effect of removing most small flake tools, with the exception of the small medlal/dlstal tool fragments. With the re-addition of the larger flake tool types a pattern is produced whereby the larger flake tools have been minimally trampled and the smaller heavily trampled. I consider it likely that cases 6 and 8 have undergone a similar process (see Figure 149). Thus, although flake tool re-culling or scavenging appears to have been less Intense, reduction assemblage culling may have been more variable and individual larger tools used more intensively. Table 92. Housepit 7 tool analysis Factor 3 experimental sequence B. fCF=Complete Flake, PF=Proximaf Fragment, MDF=Medial/Distal Fragment, SF=Split Flake; Resc.=Rescaled). Factor 3 Index A Trampled Curation Index ------I------Col. 2 Col. 4 Col. 1 X X Coi. 6 Table 91 Size Table 66 Table 66 Re-Size Col. 4 Sorted Cole 29 Resc. Col. 29 Resc. Sorted ------(1) (2) (3)* (4) 15)* (6) (7) Large CF 9.2 0.0 0.0 0.0 0.0 0.0 9.2 PF 0.6 0.0 0.0 0.0 0.0 0.0 0.6 Medium CF 27.6 0.0 0.0 0.0 0.0 0.0 27.6 PF 73.5 0.0 0.0 0.0 0.0 0.0 73.5 MDF 100.0 0 I0 0.0 0.0 0.0 0.0 100.0 SF 0.0 0.0 0.0 0.0 0.0 0.0 10.8 Small PF 34.8 34.8 162.5 30.7 143.4 14.3 14.3 MDF 52.9 52.9 529.0 100.0 1000.0 100.0 100.0 NF 2.7 2.7 9.8 0.3 2.0 0 .1** 0.1 SF 2.9 2.9 1.9 .O. 4 2.7 0.1 0.1 *A11 values divided by 10 **A11 values < .1 rounded to .1 Table 90 Column 4 Table 82 Case 7 Figure,l48. Comparison of case 7 with the mixed biface and prepared core reduction flake, trampled cull index (lower case=small flakes, uppen case=medium flakes, darkened upper case=large flakes), m 0 md f MDF Table 91 Column 7 &'F &'F , MDF 0 20 40 60 80 100 Table 82 Case 8 Figure, 149. Comparison of case 8 with the factor 3 trampled curation index (lower case=small flakes, upper case=medium flakes, darkened upper case=large flakes 1. Factors 4 and 5 Factors 4 and 5 primarily highlight aspects of factors 1-3. Factor 4 contains significant positive loadings on large split flake tools, medium nonorientable tool fragments and small proximal tool fragments (Table 85). Case 5 contains an extremely high positive factor score (Table 87) on this factor. Its distribution is in most ways identical to those discussed under factor 2. Thus, it is probably the result of mixed prepared core and bipolar core flake production, with both acute and high edge angle culling, tool recycling and intensive tramp1 ing. Factor 5 loads heavily on medium and large split flake tools. Case 10 contributes most heavily to this factor. Case 10 has been well explained under factor 2, but the inflated split flake tool scores do provide an interesting variation in this case, compared to the others considered on this factor. As the absolute numbers of flake tools considered here are relatively smsll, the inflatian of the split flake tools may be more the result insignificant variation resulting from small sample size. However, they my also indicate some patterned variability, as they lie in an area of intense trampling and splitting can be caused by trampling. Regardless of the explanation for more numerous split flake tools, the interpretation of case 10 remains as stated in factor 2. Spatial Patterning I will now present the spatial patterning resulting from the analyses of the cases discussed above. Implications for occupational histories are considered in the final section of this chapter. cases (case numbers equivalent to housepit floor sectors) 1, 2, 3 and 4 have been interpreted to be the result of prepared core, flake production followed by culling for larger acute edge angle flakes for use as flake tools, expedient tool use, discard and trampling, culling and recycling of previously discarded tools reuse and final discard. There is some variation, however, in rates of tool use and intensities of trampling. Figure 150 shows the locations of all of the cases thought to be primarily the result of prepared core flake product ion. Flake tool assemblages resulting from mixed prepared core and bipolar core flake production occur in the central and southeastern part of the floor (Figure 1511, In the original debitage assemblages, prepared core flakes have been culled for larger size and the presence of acute edge angles, while bipolar flakes were culled for acute and high edge angles. The flake tool assemblages produced through prepared core reduction have been heavily culled following initial tool discard, while the bipolar flake tools do not appear to have been intensively culled in this manner. All have been 566 metres - ,edge of bench pits wed in hiest edge of floor occ8pation I Figure 150. Distribution on the housepit 7 floor of areas with flake tool assemblages interpreted to be derived from prepared core reduction flakes only* 567 Ir - ,cage of bench pits used d edge of floor occupation Figure 151. Distribution on the housepit 7 floor of areas with flake tool assemblages inter~retedto be derived from prepared and bipolar-core reduction ' flakes. 558 crampled, Mixed biface and prepared core, flake production assemblages occur around the western side of the floor (Figure 152). Culling of previously discarded tools does not appear to have been a major factor here, although culling of the original flake assemblages appears to have been variable (focussing on acute and obtuse edge angle flakes) and intensive. Trampling is, as usual, a factor. Cases 1, 2, 3, 6 and 8 distribute roughly around the edges of the floor. Only the southwest and northwest corners are not covered. These areas contain patterns of intense use of larger flakes, so much so that small flakes are trampled far more intensely than are the larger flakes (Figure 153). At the conclusion of the unmodified debitage analysis I posed four critical questions regarding the flake tool assemblages. I will now address these using the interpretations from the flake tool analysis, MY first question considered the final deposition of flake tools. Did they end up back on the floor or were they removed from the house? All indicators from this analysis show that very little in the way of flake tools appears to be leaving the house during the last occupation(s) of the housepit. Dense lithics found In the some rim deposfts (Prentiss 1991) may be largely the result of reoccupations of the housepit which excavated old floors and collapsed roofs, depositing debris on the rim of the housepit* Indeed, distributions of tool types are very similar between the floor 569 fire-reddening - ,edge of bench pi- used * in hes% edge of -or occupatio~ Figure 152. Distribution on the housepit 7 floor of areas with flake tool assemblages interpreted to be derived from prepared core and biface reduction flakes. 570 rrVetres - ,edge of bemh # edge of floor P Figure 153. Distribution on the housepit 7 floor of areas with flake tool assemblages interpreted ta be derived from some degree of larger flake tool curation. 571 and the rim deposits at housepit 7 (Prentiss 1991). There appears to be an intensive recycling industry leaving most flake tools in conditions of limited further usefulness. My answer, therefore, is that I can find little indication of any focus on flake tool production for export beyond the confines of the housepit floor. This conclusion is further supported by a ratio of flakes (larger than 1/4 inch square) to tools of approximately 411, Even considering intense tool breakage through use, trampling, and possibly purposeful action, this is a high number of tools compared to waste flakes. Certainly some flake tools were removed from the house. However, there does not appear to have been enough for this to have had a significant effect on overall patterning within the house. My next question concerned the locations of flake tool discard. It appears that flake tools were discarded in areas close to that of the original flake production process. Thus, biface reduction produced flakes which were used as tools and discarded in the same localities of that original flake production process. This points to a very fine grained activity structure where there may have been specific activities consistently requiring specific flake tool types which were produced and used as needed. My third question concerned the effects of trampling and re-culling. Did either vary on the floor of the house? The answer is no and yes, respectively. Trampling appears to have been ubiquitous on the floor. It is clearly most intense in the center of the house and flake tools resulting from prepared core reduction have been affected the worst. Bipolar reduction flake tool assemblages may have been least affected, although some trampling is present. A similar pattern was also noted in the unmodified debitage analysis. Culling of previously discarded tools is most intense spatially in the center of the floor. It tends to be associated fairly specifically with prepared core reduction assemblages. Other flake production strategies appear to have been more specialized and initially culled more specifically. There is little indication of intensive re-culling and tool recycling among bipolar or biface reduction flake tools. My final question concerned the recognition of tool use strategies. Did some classes of tools appear to have longer use-lives than others? The answer is yes. This analysis has isolated two basic patterns of tool use. Around the perimeters of the housepit floor, larger tools have been selectively used (and probably re-used) for far greater periods of time than that of smaller flake tools. Thlv has created a pattern of differential damage in which smaller tool assemblages have been trampled to a far greater degree than the larger flake tool aqsemblages. The other major pattern is that of very short term use of all flake tools. This occurs throughout the rest of the floor. This does not mean that each tool was not ultimately well used. They were, probably due to multiple use episodes, each followed by discard, trampling, culling and further use. The pattern of longer term flake tool use around the perimeter of the house and 573 short-term use of tools throughout other portions of the floor has also been recognized in an independent analysis (Spaf ford SUMMARY AND CONCLUSI OMS In this chapter, I have been concerned with demonstrating the effectiveness of the Modified Sullivan and Rozen Typology (MSRT) as well as the utility indices and experimental debitage assemblage distributions for recognizing patterning in the archaeological record. To accomplish this goal, I have conducted an analysis of the debitage and flake tools from the floor of a large housepit, excavated at the Keatley Creek Site in south-central British Columbia. I used principal components analysis to determine the principal sources of variability in the MSRT data matrices. The following interpretations have been made: 1. Reduction Strateqies. Results of the debitage and tool analyses indicate that prepared core reduction was practiced on almost all parts of the housepit floor. Reduction appears to have been most intense in the north half and around the perimeter of the southern half of the house. Core reduction appears to have been oriented primarily towards the production of medium (16-64 square cm.) flakes with relatively acute edge angles. Biface reduction occurred in the vicinity of hearth features, clustering most densely on the west side of the house. Biface reduction appears to have been oriented, like prepared core reduction, towards the production of medium acute edge angle flakes. Tool maintenance and edge shaping activities were conducted in close proximity to hearths, generally on the wall sides. Bipolar core reduction was conducted on the west and northeast sides of the house floor. 2. Flake Cullins and Tool Production. Both the tool and debitage analyses indicate that the most common form of flake culling was that which was oriented towards acute edge angle flakes. Prepared core reduction assemblages were culled for larger sized flakes, while size appears to have been less important in the case of biface reduction. Some high edge angle flakes were culled from prepared core reduction assemblages on the southwest and northeast sides of the house and from biface reduction assemblages on the northwest side of the house. Bipolar assemblages appear to have been culled for acute and high edge angle flakes. 3. Flake Tool Recycllny. Flake tools produced Erom prepared core reduction flakes appear to have been intensely culled following initial tool discard and reused throughout the floor. Some variability is notable between the edges of the floor and the interior of the floor. The Elake tools from the edges of the floor show a pattern of longer term use of individual larger tools, while smaller toois have zeceived much more short term usages. On thc interior sf the floor all tools have been intensely used although no single size class appears to have been used for any longer periods of time than 575 any others. Biface and bipolar derived flake tools appear not to have been heavily recycled. 4, Tramplinq. Tzampling is ubiquitous on the housepit floor. The only places it is not recognizable is in some areas of dense postho1.e patterning, the centers of some hearths and along some walls. Trampling has been so consistent in its representation, that it has been relatively easy to contuol for its effects and continue to draw more detailed interpretations on reduction techniques and culling. Most interestingly, bipolar assemblages do not appear to have been affected to nearly the extent that other reduction asemblages have. This may indicate that bipolar reduct ion occurred late in the occupation of the floor, perhaps quite close to the time of abandonment. The spatial distribution of reduction techniques and culling have some interesting implications for the occupational history of the housepit floor. I define occupational history as the "purposes for which a place was used in the past and the number of separate uses of that place ..." (Camilli 1983:2). 1 con side^ the distribution of reduction techniques first. Sectors 4, 10, 11 and 12 appear to have been used almost excluslvely for prepared core reduction, with little Input from other reduction strategies. Sectors 2 and 5 have some indications of biface reduction within a pattern of relatively continuous prepared core reduction. Sectors 1, 3, 6, 7, 8 and 9 have substantial amounts of prepared core reduction present, but also notable contributions of other forms of reduction such as biface reduction and tool maintenance. Finally , sector 13 appears have relatively even representation from all reduction strategies, including prepared core, biface and bipolar core reduction and tool maintenance. culling distributions are quite similar. Sectors 2, 5, * 10, 11 and 12 all contain culling of larger acute edge angle flakes for use as tools. Sectors 10, 11 and 12 also contain strong Indications of relatively minimal use compared to other areas around the floor (low numbers of artifacts and numerous areas classified as unculled). Interestingly, these areas are probably the most trampled of any areas on the housepit floor. Sectors 1, 3, 6, 7, 8 and 9 have a consistent pattern of mixed culling for both acute and high edge angle flakes. Sector 4 has an exceptionally large amount of high edge angle flake culling. Sector 13 appears to have relatively even quantities of cases with acute and high edge angle culling and nonculled I interpret these distributions as the result of three primary processes. First, sectors 4, 10, 11 and 12 appear to be focussed use places on the floor, or places where only limited activities were conducted. This may be due to the fact that they served primarily as traveling areas, or avenues for human movement between different more critical areas of the housepit floor. Second, sectors 1, 3, 6, 7, 8 and 9 have similar patterning in a variety of lithic xeduction, culling and tool use and discard patterns. They appear to be locations of redundantly practiced multiple activities. There is also s consistent assGciation between hearth features and these areas. Thus, they appear to be the most likely candidates for domestic use. Sector 5 may have served as a temporary or brief area of domestic use, but its overall pattern of reduction and zulling does not indicate a strong domestic type pattern. Third, sector 13 has all the indlcationv of an overlapped multi-use place (Can11111 1988 1. By this, I mean that the even representation of all reduction strategies and culling strategies indicates that this part of the floor received a variety of different and often specialized uses, unlike the rest of the floor where activities were much more spatially discrete. What are the implications for reuse of the housepit floor? The trampling distribution indicates that some hearths may have been used more recently or intensely than others. Hearths in areas 7 and 9 appear not to contain the intense trampling that those of areas 3, 5, 6, and 8 contain. There are two possibilities. First, these hearths may have been intensely used during the last occupation of the floor, while the others were not, thus producing a pattern of trampling over some hearths areas while those of areas 7 and 9 were only minimally affected. Second, it is possible that even if all or most hearths were used during the final occupation, those of areas 7 and 9 were simply used more intensively leading to reduced trampling of these places. In general, the patterning throughout the floor is 578 exceptionally crisp. There is little indication of variable reuse of the floor. In other words, it does not appear llkely that reoccupations resulted in a reorganizing of the use of space on the housepit floor. The patterning is so clear that it is possible that the artifact distributions from this floor are the result of a limited number of occupations. Only one sector (13) contains indications of intense, variable, reuse. I suggest that this is not due to reocc~?patiooalvariability but more likely to the situational availability of this place (sector 13) for conducting specialized activities, perhaps related to hierarchical ordering of families within the house (Spafford 1991). Use of space in this sector may have been constrained by large posts and even a roof ladder nearby. It may also have been well lit by the sun shining through the smoke hole in the roof and the hearths immediately to the east and possibly to the west and north. Patterning in these assemblages is too distinct to define sector 13 as a dump. I expect dump patterning to be less easlly identifiable and rmre complex. Elsewhere there appears to be little indication that the floor was swept or cleaned during the last occupation, other than the culling activities focussing on larger flakes and tools. Bipolar reduction assemblages appear not to have been heavily trampled, even though they 1Ze in areas of trampling. This indicates that bipolar reduction, may indeed have occurred late in the occupation of the floor, perhaps immediately before abandonment. I conclude that the floor of housepit 7 contains a record of the final occupation of the house in which the archaeological assemblages have been little altered by cleanup other than that which probably occurred when the house was first entered at the beginning of the final period of occupation. Thus, any discussions of technological organization and variability in spatial structure may be conducted with minimal worry over the effects of earlier reoccupations or cleanup during the last occupation. I consider the impllcatlons of these data and lnterpretation~ for the organization of technology and risk management strategies in the Middle Fraser Canyon in the final chapter. CHAPTER 6 SUMMARY AND CONCLUSIONS The primary intent of this thesis has hen to develop and test the Modified Sullivan and Rozen typology and three utility indices for their effectiveness in predicting and explaining flake culling and flake tool uselreuse systems. These instruments have been developed in response to the growing needs of archaeologists for explaining technological behavior in prehistoric cultural systems. More specifically, I have argued that these methods are required for us to be able to explain economic decision making from the study of lithic artifacts from prehistoric contexts. The concept of risk management has been used as a general explanation of a wide range of economically oriented decision making processes and behaviors. I sought to demonstrate that the identification of flake culling strategies is critical to understanding prehistoric economics and risk managenent from a technological perspective, as it is at this point that lithic reduction of ten shifts lithic tool use behavior. Lithic tool use, in turn, is often a dynamic process of use, resharpening, recycling, tool discard, tool scavenging ox culling of previously disca~dedtools and final discard. Identification of different tool use/reuse systems works together with the recognition of reduction technology, culling and actual tool use (i.e. use-wear) in producing an archaeological *windowo to past systems of technological organization, economics and risk management. In Chapter 2, I presented risk management as a general concept for explaining a wide range of economic strategies. I defined risk as the potential of suffering a loss of access to critical resources. Risk management was defined as the means by which the potential for loss can be reduced or eliminated. Hunter-gatherers accomplish this by means of preventing loss through territorial defence, technological organization, and mobility; transfer of risk through warfare or feasting and give-away ceremonies (i.e. potlatching); storage; and pooling of risk through simple and elaborate systems of resource sharing (Wiessner 1982). Lithic technology serves the crucial role of providing tools to be used in procuring resources (arrow and spear tips) for processing resources (butchery knives, scrapers, etc.) and for making other tools for accomplishing these ends (variety of cutting, boring, piercing, planing, sawing and scraping tools). Managing risk from a lithics perspective then requires insuring access to lithic raw materials and for translating raw material into efficient tools in a manner which is concordant in energy/time expenditure with other constraints in that system. This is the realm of technological organization (Nelson 1991). For archaeologists to recognize and explain different forms of technolsgical organization, they need to be able to recognize the flow of lithic raw materials through archaeological contexts, the transformation of those raw materials into tools and the systems of tool use/reuse. In 582 Chapter 3, I argued that two critical components of this process included flake cuffing for tool use and flake tool use/reuse/discard. I noted that until now there were no techniques available for recognizing culling strategies. Many archaeologists have noted the removal of flakes, but it has been difficult to predict the forms of the flakes removed and the reasons why they were removed. Likewise, few have attempted to understand the effects of flake tool use, discard and reuse, especially where trampling has also had effects on assemblage composition. I presented and tested a method tor recognizing flake culling strategies from debitage assemblages, which does not depend on comparison tc flake tool assemblages. Independent analysis of flake tool assemblages, however, can be used to corroborate conclusions drawn from debitage assemblages. The method requires the use of a modified version of Sullivan and Rozen's (1985; Sullivan 1987) debitage typol~gy(MSRT), which measures variation in flake size and breakage distributions. The reliability and validity analyses demonstrated that the MSRT measures variation in debitage assemblages consistently where technological behavior is consistent and variably (In a very predictable manner) where technological behavior is variable. Given this result, I argued that the typology would be useful as a framework for the application of Elake utility indices, developed for predicting the 8teconomicftcomposition of debitage assemblages in a manner analogous to Binford's 1 (1978) neconomic anatomyw of caribou. 583 Three flake utility Indices were developed based on criteria for Elake choice noted in ethnographic descriptions of flake production and flake tool use. The three indices included the Flake Volume Index (FVI), the Acute Angle Edge Length index (AAEL) and the High Angle Edge Length index (HAEL). The reliability and validity analyses of the indices confirmed that each index behaved relatively consistently where technological behavior was consistent and that each produced very predictable results where technological behavior was varied. Thus, the indices were useful for predicting the potential utility of technologically distinct debitage assemblages. The next step was to develop the means for recognizing this type of behavior in a "real worldw context. This depended on the use of the indices and NSRT to develop a aeries of experimental, mathematically derived distributions predicting a variety of culling situations in different technological and taphonornic contexts (Chapter 4). Distributions produced in Chapter 4 were assessed in relation to the results of the analyses in Chapter 3 and it was concluded that they were valid and that the method would be useful in recognizing archaeological variability. A pattern recognition study was undertaken of the lithic (vitreous traehydacite only) debitage and flake tools from a large and complex housepit floor excavated at the Keatley Creek archaeological site, a large pithouse village from the Middle Fraser Canyon of south-central British Columbia. This analysis demonstrated a number of interesting patterns. Technologically, reduction of medium sized prepared cores was most common throughout the housepit floor. Biface reduction centered primarilly around the south, west and northwest sides of the floor. Tool resharpening/edge modification activities were identified as associated with the major hearths on the floor. Bipolar reduction was recognized on the west-central and northeastern portions of the floor. Trampling was fdentifled throughout the f losr, The only exceptions noted were in some areas around the margins, in areas of dense posthole patterns and in some hearths. Bipolar assemblages did not appear to be extremely damaged through tramp1 ing. The primary flake culling strategy identif ied was that of larger acute edge angled flakes from core reduction. Culling of biface reduction debitage was oriented more towards acute edges with less concern for flake- size. Bipolar debitage was probably culled for acute and high angle edges (pxsbably broken edges). Some culling in csre and biface reduction assemblages was oriented towards high edge angles. Three clusters were noted (in the southwest, northeast and northwest corners f . Two fundamental flake tool use/discard/reuse patterns were identified. The first, located around the edges of the floor, appears to be the result of larger flake tool curation and subsequently reduced trampling damage and small flake tool expedient use resulting in Intensive trampling damage. The second, found in the middle portions of the floor indicates continuous expedient flake tool use, discard, trampling, culling of discarded tools and tool reuse. The key lesson here is that the final form of the discarded tools often had little resemblance to their original form (Figure 154). The flake tool analysis corroborated much of the conclusions from the debitage analysis. Flake tools from core reduction flakes are discarded throughout the floor. Flake tools from biface reduction flakes are discarded most commonly on the south, west, and northwest sides of the floor. Bipolar flake tools are discarded in the center of the floor, roughly in between and to the south of the areas of most intense bipolar reduction. The lack of trampling in some hearths and among bipolar debitage and flake tool assemblages, coupled with the crispness of the patterns of reduction, culling and tool reuse and discard indicates that although the floor was probably used for more than one winter, archaeological patterns noted in this study are likely the result of the final period of housepit occupation. If this is the case, then we are viewing the results of a single set of adaptive strategies and group dynamics. I consider the larger issues of housepit floor organization and risk management in the Middle Fraser Canyon in the following sections. Figure i54. Example of flake tool assemblage formation based on the reduction and breakage of a single flake tool. 587 EXPLAINING THE SPATIAL STRUCTURE OF A HOUSEPIT FLOOR LITHIC ASSEMBLAGE To facilitate a discussion of explanations for the organization of space on the floor of housepit 7, I first review the results of another study (Spafford 1991) of the same llthic artifacts. f comment on any similarities and differences between these two sets of research. Finally, I use this discussion as a point of departure for further considering the role of risk management in the Late Prehistoric cultural system of the Middle Fraser Canyon. An independent analysis of the housepit 7 lithic artifacts has been conducted by Spafford (1991) which has produced a number of both well founded and tentative conclusions on the processes responsible for the formation of these assemblages. Technologically, Spafford has identified core reduction activities throughout the housepit floor, commenting that perhaps they were most intense in the southern portion (on evidence of discarded cores). He notes that billet flakes (defined on flake morphological criteria independent of those used in this study) are most common on the west and southwest sides of the floor. He identifies bipolar reduction (defined on the basis of the presence of bipolar cores and flakes using criteria independent of that used in this study) as occuring primarily in the northwest portion of the floor. Spafford (1991) identifies a number of specialized "activity areas" in different portions of the floor. He concludes (1991:141), largely on the basis of cached spa11 scraper tools, that a portion of the southeast side (sector 5) is possibly a hide working area. Similarly, he identifies the northwest corner (sector 9) as a special activity area within a domestic space on the basis of concentrations of heavily retouched scrapers, utilized flakes and fire cracked rock (Spafford 1991:140). Spafford (1991:145) identifies a third special activity area in the east-central (secctor 11) portion of the floor, primarily on the basis of dense concentrations of fire-cracked rock. The south-central portion of the floor (sector 12) is classified as a possible corridor area. Most apparent from Spaffordls (1991:146-147) study is a distinction between the central and outer zones of the housepit floor. He notes that the outer perimeter contains high numbers of heavily retouched tools and far fewer numbers of minimally retouched tools. The inner zone is characterized by high numbers of minimally retouched, possibly expedient tools, biface fragments, spall tools, fire cracked rock and numerous hearths. On the basis of these distinctions, he argues that the inner zone was possibly the focus of mast female activities such as food and hide processing, while the outer zone was more commonly associated with male activities such as equipment repair and lithic reduction activities. Finally, Spafford (1991:146) identifies at least three and possibly more primary domestic areas lacated on the northwest (sectors 8, 9, 13 and west half of lo), the 589 northeast (sectors 1, 3, 4, east halves of 10 and 11) and southern (sectors 2, 5, 6 and 7). He argues that each area potentially contained a multi-family group belonging to the larger household social unit and that each domestic unit used several hearths. Some hearths were used more as domestic hearths (large hearths in sectors 6, 7, 8 and 11) while others were perhaps more often used for warmth and light during special activities (hearths in sectors 3 and 5). The hearth in sector 9 appears to have been used for both domestic and special activities. On the basis of the tripartite domestic area distinction and dual distinction between male and female work areas, Spafford (1991) argues that the house was inhabitated by a single hierarchically organized coresidential corporate group. He suggests that the highest status family group inhabited the southern domestic area, while others of lesser status inhabited the other two areas. Spaffordts (1991) technological conclusions are largely in concordance with mine concerning the dominant pattrns present. Both studies suggest that core zeduction was practiced throughout the housepit, although most intensely in the peripheral zone where debitage concentrations are located, and that biface reduction was practiced primarily on the western side of the floor. We disagree on the spatial locality of bipolar reduction. This may be, in part, the result of the difficulty in identifying bipolar flakes on morphological criteria alone due to usual platform shattering as well as the apparent low frequency of bipolar reduction. Bipolar cores may also have been displaced by cleanup or other activities. During my analysis of the housepit 7 debitage, I maintained an independent count of recognizable individual bipolar flakes (recognizable on the basis of proximal and/or distal crushing and the presence of a sheared ventral morphology). I found clusters of bipolar flakes in sectors 1, 3, 8, 9 and 13. My analysis using the MSRT was able to identify the effects of bipolar reduction in areas 1, 0 and 13 (Spaffordfs bipolar concentration area was in sector 9). Thus, my study of debitage variation on the floor identified two of the three clusters of bipolar reduction from the floor. Bipolar debitage from the northwest corner (sector 9) are more sparse than in the west or northeast sides and it is not surprising that it blended in with the core reduction and bifacial reduction debris when viewed through the MSRT. Spa•’fordfsidentification of special activity areas are largely based on criteria not considered in this study. However, results of this study reflect on his conclusions to some degree. From the perspective of vitreous trachydacite use, sectors 4, 10, 11 and 12 appear to be places where a fair degree of human movement occuzred, as might be expected from the central location of three of these sectors. Tool praduction and use appear to have been consistently of an expedient nature, focussing on hand-held core reduction and bipolar reduction based flake tool production, use, and discard. Sector 12 Is particularly sparse in artifacts, though the general pattern is no different from that of the central and eastern sides of the floor with core reduction and larger acute edge angle flake culling. This corresponds with Spaffordls identification of sector 12 as a corridor. I add that sectoxs 10, 11 and 4 may also have received a fair amount of foot-traffic. Sector 5, identified by Spafford as a hide working area, contains some elements of hearth oriented patterning found around other hearths such as associated core and biface reduction (very minimal in this case, however) debris, both culled for acute edge angled flakes. This sector does not contain any strong indicator of tool edge retouch/modification and it has been heavily trampled. Flake tools discarded here are primarily a combination of core reduction flakes with some bipolar reduction flakes. A11 have been used and reused expediently and heavily trampled. Thus, this area appears little different from areas 4, 10, 11 and 12 other than the presence of minimal biface reduction. Spafiord's argument, based on available space and the presence of spa11 tools, may be a useful explanation for this sector. Identification of this sector as a female oriented activity area is concordant with my identification of this area as generally more similar to the central portions than to any other portions of the floor. The presence of biface reduction in sector 5 indicates, however, that some male oriented activities may also have been conducted in this area. This is likely given its proximity to the edge of the housepit floor. MY analysis has identified a consistent hearth centered pattern in sectors 1, 3, 6, 7, 8 and 9. This pattern 1s one of continuous core reduction and large acute edge angled flake culling associated with clusters of biface reduction and acute edge angle flake culling and tool edge maintenance/resharpening clustering immediately adjacent to the hearths. Flake tool use/reuse/discard strategies of mixed core reduction and biface flake use and discard are found in sectors 6-9. Sector 1 and 3 flake tools are primarily the result of core reduction flake culling and use. Biface reduction certainly appears to be far less intense In these sectors than in the others. This repeated patterning around hearths appears to be indicative of regular domestic activities requiring flake tools and blfacial preparation. This generally supports Spafford's identification of the northeastern, northwestern and southern sectors as domestic areas. It also indicates that other hearths may have been the locus of additional independent domestic units. Further support comes from the distribution of high edge angle flake culling, which occurs in clusters on the southwest, northeast and the northwest sides of the floor. The identification of domestic activities on the northeast corner (sectors 1 and 3) is still somewhat problematical as the hearth is small, food storage pits are few and although smaP1 amounts of tool maintenance and biface reduction are present, core reduction is by far the dominant lithic reduction activity. Further, the hearth in the northeastern area, defined by Spafford, is separated from the central part of the floor by a row of postholes, Debitage assemblages from this area of dense posthole patterning have been only minimally trampled and in places, not culled for any flakes. This suggests that some form of barrier existed between the two sub-areas. Thus , one must keep open the possibility that this area may have been a place where a variety of special activithes occurred, rather than any exclusively domestic occupation. On the other hand, if there is any place on the floor where low status people my have lived, it is here. Lithic reduction and tool use activities certainly appear little different from other potential domestic areas, only in somewhat different proportions. A third possibility is that this area was used domestically by lowest status people such as slaves and because of this, the area was also used as an activity area (possibly wood working) by slaves and possibly others, as indicated by the intense focus on culling of high edge angle flakes and the presence of numerous cut beaver teeth from this area (1989 field notes). My identification of flake tool use/reuse variation between the perimeter and interior portions of the floor is concordant with Spafford's identification of outside and central qender-related areas. Spafford argued that use of outer areas focussed on tool curation while central areas saw more expedient toof use. My results indicated a system of intense trampling of small artifacts and very minimal trampling of larger artifacts around the perimeters. Interior floor flake tools were all heavily trampled regardless of size. Certainly there is variation, on a spatial basis, In flake tool use/reuse strategy which may well have been gender related to some degree. MY study has not been particularly concerned with status organization on the housepit floor. However, it does serve to corroborate some of the arguments made by Spafford regarding social organization on the housepit floor. I have identified potential domestic areas on the south, west, northwest and possibly northeast sides of the floor. Much of the central and east central floor accumulated artifacts through processing operations requiring less complex technologicai behavior than those activities conducted around the major hearth areas or on the far eastern perimeter. Likewise these areas were most trampled due to foot traffic. One sector (13) appears to have been something of an overlap area in which a variety of possibly unrelated activltles were conducted. Given these results, I turn now to the role of lithic technology in ;isk management strategies of the inhabitants of the Middle Fraser Canyon, from the perspective of vitreous trachydacite use on the floor of housepit 7. LITHICS AND RISK MANAGEMENT AT KEATLEY CREEK I now compare ethnographic descriptions and predictions to archaeological results to facilitate a discussion of Late Prehistoric risk management strategies. I review first the role of lithic technological organization and second, the role of mobility. Though social organization has been noted as important also to risk management, it has been dealt with in greater depth by Spafford (1991). Thus, my treatment of social organization is relatively limited. A number of predictions were drawn from the ethnographic record regarding strategies of lithic raw material use by people in the ~iddleFraser canyon area concerned with managing risk. Briefly, I predicted that late fall lithlic procurement was embedded into subsistence pursuits, especially through movements of hunters and their families, moving down through the mountain forests and grasslands Eollowing populations of deer. Lithic raw materials were expected to have been stockpiled in winter housepits in the form of small prepared cores and blfaces. This was a form of storage designed to ensure a ready supply.of lithic raw materials during a time when access to the sources was often impossible and lithic raw material needs were high. Lithic reduction was then centered around producing flakes for use as tools primarily for processing hides for clothing, food items, bone and antler and wood for tools, and miscellaneous fibres and bark for basketry, clothing and footwear. Manufacture and repair of msny of these items was seen as critical for survival during the rest cf the year. Managing risk through lithdc technology in the Middle Fraser Canyon during the winter months depended on ensuring access to lithic resources through storage and using those 596 lithic resources in an efficient manner to produce nonlithic tools for hunting, gathering and processing operations in the spring and summer, when time for complex tool manufacture would be more limited. With these theoretical concerns in mind and from Teit's (1900, 1906, 1909) stone tool descriptions, I predicted that flake tools would be produced from prepared cores and often curated during the winter to extend their use-lives. I also predicted that bipolar core reduction would be a late winter activity designed to extend the uselife of exhausted cores and tools, by creating new flake tools from bipolar flakes. Obtaining access to llthic resources in the spring would be an activity probably again embedded into more important subsistence pursuits. The analysis of the housepit 7 debitage and flake tools both supports and adds additional complexity to the predictions derived frtm the ethnographic record. The production of medium size class core reduction flakes from prepared cores throughout the floor of housepit 7 indicates that the storage and use of prepared cores throughout the winter was probably common. Bifaces were also reduced to make bifacial tools and to provide smaller, more specialized flake tools. The lack of trampling among bipolar flakes suggests that bipolar technology may have indeed come at a point close to housepit abandonment. Thus, bipolar reduction may have been a device for extending raw material use-life. Flake product ion for flake tools came from prepared block cores as well as from bifaces. There is no indication that unprepared randomly reduced cores and resulting raw material wasteage occurred. This supports the predictions drawn from the ethnographic record. On the other hand, the use of flake tools was far more complex than predicted. Two primary systems of flake tool production, use, reuse and discard were present on the floor. First, a system of larger, acute edge angle, flake production and curation was practiced (around the margins of the floor) fitting more closely with that expected from the ethnographic descriptions. Second, a complex system of primarily acute edge angle (though speclallzed, high edge angle flakes were also culldd) flake production, short-term use, discard, trampling and eventual culling or scavenging of discarded tools and reuse followed by discard operated throughout the housepit. In the central portlons of the floor this process affected all sizes of flakes, while on the margins it was apparently more confined to smaller flake tools. It would seem that lithic reduction debris and discarded expedient tools were a continuously available resource for the creation of new flake tools. This process is not intultivaly surprising given previous archaeological (Ode11 1981; Yerkes 1987) and ethnographic (Gould et al. 1971; Hayden 1977 1 descriptions of similar processes. Thus, although many flake tools were not curate6 personal gear (Binfsrd 1979 1, raw material was intensively uses. I can find little indication that flakes or flake tools were intensively exported from the housepit. Culling and flake discard appear to match each other in both spatial locality and in quantity. Though this canclusi~nshould be further tested, it is appears that much of the llthic reduction, tool zse and discard system was entirely contained within the floor of the housepit. Undoubtedly, some tools or flakes were removed upon abandonment. However, I can find no indication that this occurred at any other than a limited scale. It is likely that lithic procurement was a planned activity In spring movements of these people following wanter housepit abandonment. Though somewhat more complex, lithic technological organization at housepit 7 fits closely with expectations drawn from the theoretical lithics literature and from the ethnographic record of the Middle Fraser Canyon. Some implications can also be drawn for the effects sf mobility strategy on these lithic assemblages. The ethnographic record of the Middle Fraser Canyon (Alexander 1992b; Teit 1906; 1909) has indicated a biseasonal pattern of winter residential sedentism and occasional logistical mobility and spring through fall residential and logistical mobility. Mobility strategies were used solve problems created by inequalities in resource accessibility. The data discussed in this study support strongly the contenthon of winter residential stability. The system of lithic raw material movement and flake tool production and use indicates a pattern of high effort expended in maintenance and manufacture of clothing and tools for aiding in resource procurement and process ing. This corresponds to Bleed (1986) and Binford's (1979; 1980) predictions on the critical role of down-time among Logistically organized hunter-gatherers. I conclude that the lithic assemblages from the floor of housepit 7 formed during the period of winter sedentism used as down-time far the production of anticipatory gear (Binford 1979) critical to survival in the winter and throughout the year. Though I have not focussed my research on recognizing social organization from these lithic assemblages, my research has corroborated some conclusions drawn by Spafford (1991). If the Keatley Creek site did form as the result of a large population regularly inhabiting many of its housepit structures and given the fact that highly productive salmon fishing locales were limited, it is not surprising that some degree of socio-political complexity did develop (Binford 1990; Hayden et al. 1985). The floor of housepit 7 is organized differently from that of housepits 3 and 12 (Spafford 1991). Some components of these data from hsusepit 7 could he interpreted as indicative of the behavior of members of a hierarchically organized political group (spafford 1991). This would match descriptions made by Telt (1906; 1909) of Lillooet and Shuswap socio-political organization during the contact period where some family groups did gain higher status than other families around them through control of access to salmon. Further assessment of hypotheses regarding socio-political complexity in the Middle Fraser Canyon area and the adjacent northwest Plateau will 600 require continued research into housepit organization and burial patterning. This research has demonstrated that it is possible to construct utility indices for predicting the composition of lithic debitage and flake to~lassemblages under diEferent technological and organizational conditions. A method was tested experimentally and applied to an archaeological case. The data from the archaeological study were complex but interpretable. Culling strategies and flake tool production and use systems were identified. Further archaeological and experimental research would benefit from the application of these types of methods. The primary benefit of using utility index models is that research problems associated with prehistoric economieu can be better approached given background methodological research anticipating the character of archaeological assemblages under different economic conditions. Thus, the primary area of future research is continued experimentation. I suggest that the next focus of l.ithic experimentation (other than development of utility indices for other Llthic raw saterial types) should be on gaining a better understanding sf the relationships between flake culling and the organizati~nof flake tool use and discard. It may be useful in the future to divide the HAEL index further into two indices measuring flake 601 edges between 45 to 79 degrees and 80 t~ 119 degrees. Basic experiments in the production of lithic assemblages and the removal of flakes should be carried out to explore variability in utility index derived predictions regarding culling strategies. Likewise, experiments in the evolution of flake tool assemblages through flake culling and use, tool discard and trampling, and tool scavenging and reuse are required. Then, modelled flake tool assemblage formation sequences (as presented in this study) could be more realistically evaluated and expanded. Archaeological research into culling strategies and flake tool assemblage formation processes is only beginning. Hayden (1992, personal communication) has suggested that it will be productive to view entire llthic assemblages, graphically, by combining tradltlonal flake type categorizations with utility index related distributions. I illustrate this with a breakdown of housepit 7 analytical unit 96 and 97 (area 13) data (Figures 155 and 156), combining debitage data from this study with those from Spafford (1991). With further experimentation, this approach may become extremely useful for quickly assessing variability in complex assemblages. Another promising avenue will involve combining refitting studies with utility indices. fn this way conclusiens regarding the presence and form of culling present could be cross-checked and more detailed interpretations on the formation of llthic assemblages in general developed. As anthropology, lithics research should strive to understand and explain the behavior of prehistoric peoples. The use of flake utility indices and the Modified Sullivan and Rozen Typology should contribute to this process. Percentage 40 36.2 Cores Figure 155. Debitage and tool assemblage distribution, analytical unit 96 (area 13) (SC=small complete, SP=small proximal, SS=small split, MMD=medium medlal/distal MN=medium nonorientable, SN=small nonorientable; SMD=small medial/distal, FT=flake tools, FORTI=formal tools; all data expressed as percentages of total count of artifacts). 30 Percentage 29.7 24.3 R 25 x * * * 15 * * 13.5 * * * 10 * 8.1 R x * * * 5.4" 5 2.7 * 2.7 2.7 * 2.7 2.7 2.7 * * * 2.7 * * * * * * * * * * * * 0* * * * * * * * * * * * ------_------SC SF SS SC SP MN SMD MMD SMD SN FT BPCR Eiiliei: Hard Bipolar Unknown Hanimer ------Debitage Tools/ Cores Figure 156. Debitage and tool assemblage distribution, analytical unit 97 (area 13; BPCR=Bipolar Core; also see Fig. 15 ). 604 REFERENCES CITED Ahler, S.A. 1989 Mass Analysis of Flaking Debris: Studying the Forest Rather than the Tree. In Alternative Ap~xoaehesto Lithic Analysis, edited by D.O. Henry and G.H. Odell, pp. 85-118. Archeological paper of the American Anthropological Association Number 1, Ahler, S.A. and J. Vannest 1985 Temporal Change in Knife River Flint Reduction Strategies. In Lithic Resource Procurement: Proceedinss from the Second Conference on Prehistoric Chert Exploitation, edited by S.C. Vehik, pp. 183-198. Southern Illinois University at Carbondale, Center for Archaeological Investigatians Occasional Paper No. 4. Alexander, D. 1992a Environment. In A Complex Culture of the British Columbia Plateau, edited by B. Hayden, Vancouver: University of British Columbia Press. 1992b A Model of Prehistoric Land Use in the Mid- Fraser River Area Based on Ethnographic Data. In .A Corn~lexCulture of the British Columbia Plateau, edited by B. Hayden, Vancouver: University of British Columbia Press. Allchin, B. 1957 Australian Stone Industries, Past and Present. Journal of the Royal Anthropolosical Institute of -Great Britain and Ireland 87:115-136. Ames, K.M. 1985 Hierarchies, Stress and Logistical Strategies Among Hunter-Gatherers in Northwestern North America. In Prehistoric Hunter-Gatherers: The Ernersence of Cultural Complexity, edited by T.D. Price and J.A. Brown, pp. 155-198, Academic Press, New York. Amick, B.S. 1986 Calculating Artifact Planview Area. Lithic Technolssv 15!3!r90-95. Amick, D.S. and R.P. Mauldin 1989 Comments on Sullivan and Rozen's "Debitage Analysis and Archaeological Interpretation.'' American Antiquity 54(1):166-168. Amick, D.S., R.P. Mauldin and L.R. Binford 1989 The Potential of Experiments in Lithic Technology. In Experiments in Lithic Technolaqy, edited by D.S. Amick 605 and R.P. Mauldin, pp. 1-14. BAR International Series 528. Amick, D.Se8 R.B. Mauldin and S.A. Tomka 1988 An Evaluation of Debitage Produced by Bifacial Core reduction of a Georgetown Chert Nodule- Lithic Technoloqy 17(1):26-36. Arima, E. and J. Dewhirst 1990 Nootkans of Vancouver Island. In Handbook of North American Indians, Volume 7, the Northwest Coast, edited by W. Suttles, pp. 391- 411. Smithsonian Institution, Washington. Armor, D.J. 1974 Theta Reliability and Factor Scaling. In Socioloqical Methodoloqy edited by H.L. Costner, pp. 17-50. Jossey-Bass, San Francisco. Audouze, F. 1987 The Paris Basin in Magdalenian Times. In The Pleistcene Old World, edited by 0. Soffer, pp. 183-200. New York: Plenum. Bamforth, D.B. 1986 Technological Efficiency and Tool Curation. American Antiuuity 51:38-50. 1988 Ecoloqv and Hunter-Gatherer Orqanization on the Great Plains. New York: Plenum. 1990 Settlement, Raw Material and Lithic Procurement in the Central Mojave Desert. Jsunnal of Anthro~olosicalArchaeolosv 9:70-104. 1991 Technological Orqanization and Hunter-Gatherer- Land-Use: A ~aliforniaExample. American Antiquity 56(2):216-235. Barton, C.M. 1990 Stone Tools and the Paleolithic Settlement of the Iberian Peninsula. Proceedinqs of the Prehistoric Society 56:15-32, Baumler, M.F. and C.E. Downum 1989 Between Micro and Macro: A Study in the Interpretation of Small Sized Lithic Debitage. In Experiments in Lithic Technolosy, edited by 0,s. Amick, pp. 101-116. BAR International Series 528 Beil, C.E., R.L. Taylor and G.A. Guppy 1976 The Bioqeoclimatic Zones of British Columbia. 606 I Davidsonia 7(4):44-55. Binford, L.R. 1973 Interassemblage Variability: The Mousterian and the "Functional Argument." In The Explanation of Culture Chanse: Models in Prehistory. edited by C. Renfrew, pp. 227-254. Duckworth, London. 1977 Forty-seven Trips: A Case Study in the Character of Archaeological Formation Processes. In Stone Tools as Cultural Markers: Chanqe, Evolution and Complexity, edited by R.V.S. Wright, pp. 162-168, Australian Institute for Aboriginal Studies, Canberra. 1978 Nunamiut Ethnoarchaeoloqy. New York: Academic Press. 1979 Orqa~izationand Formation Processes: Lookina at ~uratedTechnologies. Journal of ~nthropoloqicil Research 35(3):255-273). 1980 Willow Smoke and Dog's Tails: Hunter-Gatherer Settlement Systems and Archaeological Site Formation. American Antiquity 45(1):4-20. 1981 Bones: Ancient Men and Modern Myths. New York: Academic Press. 1982 The Archaeology of Place. Journal of Anthro~oloqical Archa~olosy1(1):5-31. 1983 In Pursuit of the Past: Decodins the Archaeoloqical Record. New York: Thames and Hudson. 1986 An Alyawara Day: Making Men's Knives and Beyond. American Antiquity 51(3):547-562. 1987 Searching for Camps and Missing the Evidence? Another Look at the Lower Paleolithic. In The Pleistcene Old World, edited by 0. Soffer, pp. 17-31. New York: Plenum. 1989 Technology of Early Man: An Organizational Approach to the Oldowan. In Debatins Archaeolosy edited by E.R. Binford, pp. 437-463. Academic Press, San Diego. 1990 Mobility, Housing and Environment: A Comparative study. ~ournalof Anthro~oloqical Research 46(2):119-152. 1991 [Then the Going Gets Tough the Tough Get Going. In Fthnoarchaeoloqical kp~roachesto Mobile Campsites. edited by C.S. Gamble and W.A. Boismier, pp. 25-138. International Monographs in Prehistory, Ethnoarchaeological Series I, Ann Arbor. Binfard, L.R. and J.F. OfConnell 1984 An Alyawara Day: The Stone Quarry. Journal of Anthno~oloqicalResearch 40(3):406-432. Binford, L.R. and G.I. Quimby 1963 Indian Sites and Chipped Stone Materials in the Northern Lake Michigan Area. Fieldiana Anthropoloqy 36:277-307. Bleed, P. 1986 The Optimal Design of Hunting Weapons: Maintainability or Reliability. American Antiauity 51:737-747, Boismier, W.A. 1991 Site Formation Among Sub-Arctic Peoples: An Ethnohistorical Approach. In Ethnoarshaeoloqicaf A~proachesto Mobile Carn~sites. edited by C.S. Gamble and W.A. Boismier, pp. 189-214. International Monographs in Prehistory, Ethn~archaeological Series 1, Ann Arbor. Bum, H.T., J.W.K. Harris, G. Isaac, Z. Kaufulu, E. Kroll, K. Schick, N. Toth and A.K. Behrensmeyer 1980 FxJjSO: An Early Pleistocene Site in Northern Kenya. World Archaeolosy 12(2): 109-136. Burgess, R.J. and K.L. Kvamme 1978 A New Technique for Measurement of Artifact Angles. American Antiquity 43: Burton, J. 1980 Makinq f3ems~auk sf Watittz Flakes: New Methods for Investigating the technology and Economics Behind Chipped Stone Assemblages. Journal of Archaeoloqical Science 7:131-148. Callahan, E. 1979 The Basics of Bifaee Knapping in the Eastern Ploted Point Tradition: A Manual for Flintknappers and Lithic Analysts. Archaeolosy of Eastern North America 7(1):1-180. Camiili, E. 1983 Slte Occopatlonal History and Lithlc Assemblage Structure: An Example from Southeastern Utah. Ph.D. 608 Dissertation, Department of Anthropology, University of New Mexico. 1988 Interpreting Long-Term Land-Use Patterns from Archaeological Landscapes. American Archaeoloqy 7(1):57-66. Cannon, A. 1992 Conflict and Salmon on the Interior Plateau of British Columbia. In A Complex Culture of the British Columbia Plateau, edited by B. Hayden, Vancouver: University of British Columbia Press (in press). Caraco, T. 1981 Risk-Sensitivity and Foraging Groups. Ecolosv 62(3):527-531. Carmines, E.G. and R.A. Zeller 1979 Reliability and Validity Assessment. Beverly Hills: Sage University Paper 17. Cashdan, E. 1983 Territoriality among Human Foragers: Ecological Models and an Application to Four Bushman Groups. Current Anthro~olosv24(1):47-66. Cohen, J. 1960 A Coefficient of Agreement for Nominal Scales. Educational and Psycholosical Measurement 20(1): 37-46. Cotterell, B, and J. Kamminsa 1987 The Formation of Fiakes. American Antiquity 52(4):675-708. Cowgill, G.L. 1970 Some Reliability and Validity Problems in Archaeology. In Archeolosie et Calcateurs: Problemes Semiolosiaues et Mathematiques, pp.161-175. Editions du Centre National de la Recherche Scientifique. Crabtree, D.E. 1982 An Introduction to Flintworkinq. Occasional Papers of the Idaho Museum of Natural History, Number 28. Cronbach, L. 1951 Coefficient Alpha and the Internal Structure of Tests. Psychometrika 16:297-334. Deal M. and B. Hayden 1987 Persistence of Pre-Columbian Lithic Technology 609 in the Form of Glassworking- In Lithic Studies Amonq the Contemporarv Hiqhland Maya, edited by B. Hayden, pp. 235-331. University of Arizona Press, Tucson. Dibble, H. 1987 Reduction Sequences in the Manufacture of Mousterian Implements of France. In The Pleistocene Old World, edited by 0. Soffer, pp. 33-45. New York: Plenum. 1991 Mousterian Assemblage Variability on an Interregional Scale. Journal of Anthro~olosical Research 47(4):239-257. Dibble, H.L. and M.C. Bernard 1980 A Comparative Study of Basic Edge Anqle Measurement Techniques. American Antiquity 45:877-885. Dibble, H.L. and J-C. Whittaker 1981 New Experimental Evidence on the Relation Between Percussion Flaking and Flake Variation. Journal of Archaeoloqical Science 8:283-296. Draper, N. 1985 Back to the Drawing Board: A Simplified Approach to Assemblage Variability in the Early Paleolithic. world Archaeoloqy 17(1):3-18. Earl, T. 1989 The Evolution of Chiefdoms. Current Anthro~olosy30(1):84-88. Ensor, H. B. and E. Roemer JL. 1989 Comments on Sullivan and Rozen's Debitage Analysis and Archaeological Interpretation. American Antiquity 54(1):175-178. Ericson, J.E. and B.A. Purdy 1984 Prehistoric Quarries and Lithic Production, Cambridge: University of Cambridge Press. Ferr ing, C.F. 1976 Sde Divskon: An Upper Paleolithic Site on the Divshon Plain. In Prehistory and Paleoenvironments in the Central Negev, Israel Vol. 1. edited by A.E. Marks. pp. 199-226. SMU Press, Dallas. Flsh, P.R. 1978 Consistency in Archaeological Measurement and Classification: A Pilot Study. American Antiauity 43(1):86-89. 610 1981 Beyond Tools: Middle Paleolithic Debitage Analysis and Cultural Inference. Journal of Anthropoloqical Research 37(4):374-386, Fisher, R.A. 1960 The Desiqn of Ex~eriments.New York: Hafner Publishing Co. Fladmark, K.R. 1984 Mountain of Glass: The archaeology of the Mount Edziza Obsidian Source, British Columbia, Canada. World Archaeolosy 16(2):139-156. Ford, S. 1987 Flint Seattexs and Prehistoric Settlement Patterns in South Oxon and East Berks. In Llthic Analysis and Eater British Prehistory edited by A.G. brown and M.R. Edmonds, pp. 101-136. BAK British Series 162. Fortuine, R. 1585 Lancets of Stone: Traditional Methods of Surgery Among Alaska Natives. Arctic Anthropolosv 22(1):23-45. Francis, J. 1983 Lithic Procurement Strategies in the Big Morn Basin and Mountains of North-Central Wyoming. Ph.D. Dissertation, .Department of Anthropology, Arizona State University. Frison, G.C. 1978 Prehistoric Hunters of the Hiqh Plains. New York: Academic Press. Frison, G.C- and B.A. Bradley 1980 Folsom Tools and Technolosy at the Hanson Site, Wvominq. Albuquerque: University of New Mexico Press. Gallagher, J.P. 1977 Contemporary Stone Tools in Ethiopia: Implications for Archaeology. Journal of Field Archaeolosy 4:407-414. Gif ford-Gtnzalez, D.?., D. 3. Darnsosch, 9.R. Damrosch 9. Pryor and R.L. Thunen 1985 The Third Dimension in Site Structure: An Experiment in Trampling and Vertical Dispersal. American Anticruitv 50(4):803-818. Gilman, P. 1987 Architecture as Artifact: Pit Structures and Pueblos in the American Southwest. 611 Goodyear, A.C. 1979 A Hypothesis for the Use of Cryptocry- stalline Raw Materials of North America. University of South Carolina, Institute of Archaeology and Anthropology, University of South Carolina. Gould, R.A. 1968 Living Archaeology: The Ngatatjara of Western Aus-zralia . Southwestern Journal- of Anthropo:Loqy 2482):101-122. 1971 The A~chaeologistau Ethnographer: A . Case from the Western Desert of Australia. World Archaeolosy 3 (2):130-177. 1977 Ethno-archaeology; or, Where do Models Come From? A Closer Look at Aboriginal Lithic Technology. In Stone Tools as Cultural Markers: Chanqe, Evolution and Com~lexity, edited by ReV,S. Wright, pp. 162-168, Australian Institute for Aboriginal Studies, Canberra. 1980 Livins Archaeolosv. Cambridge: Cambridge University Press. Gould, R.A., D.A. Koster and A.H.L. Sontz 1971 The Lithic Assemblage of the Western Desert Aborigines of Australia. American Antiquity 36(2):149-169. Greene, V.L. and E.G. Carmines 1980 Assessing the Reliability of Linear Composites. In Sociolosical Methodolosv edited by K.F. Schuessler, pp. 160-175. Jossey-Bass, San Francisco. Guilford, J.P. and B. Fruchter 1973 Fundamental Statistics in Ps~cholosy and Education, New York: McGraw Hill, Hayden, B. 1949 Stone Tool Functions in the Western Desert. In Stone Tools as Cultural Markers: Chancre, Evoiution and Complexity, edited by R.V.S. Wright, pp. 178-188, Australian Institute for Aboriginal Studies, Canberra. 1979a Paleolithic Reflecti~ns:Lithie Technolosv and Ethnoqxauhic Excavations amons the Australian Abor icrines. Canberra: Australian Institute of 612 Aboriginal Studies. 1979b Litbic Use-Wear Analysis- New York: Academic Press. I981 Research and Development in the Stone Age: Technological Transitions Among Hunter-Gatherers. Current Anthropoloqy 22:519-548. 1987 Lithic Studies Amonq the Contemporary Hiqhland Mayao Tucson: University of Arizona Press. 1988 From Chopper to Celt: The Evolution of Resharpening Techniques. Lithic Technolosv 16 (2-3) :33-43. 1989 From Chopper to Celt: The Evolution of Resharpening Techniques. In Time, Enerqy and Stone Tools, edited by R. Terrence, pp. 7-16. Cambridge University Press, Cambridge. 1990 The Right Rub: Hide Working in High Ranking Households. In The Interpretive Possibilities of Microwear Studies, edited by Bo Graslund, pp. 89-102. AUN 14, Societas Archaeological Upsaliensis: Upsala. 1992a Introduction: Ecology and Culture. In A Com~lexCulture of the British Columbia Plateau edited by 8. Hayden University of British Columbia Press. 1992b Conclusions: Ecology and Complex Hunter/ Gatherers. In A Complex Culture of the British Columbia Plateau edited by B. Hayden University of British Columbia Press (in press). Hayden, B., M. Eldridge, A. Eldridge and A. Cannon 1985 Complex Hunter-Gatherers in Interior British Columbia. In Prehistoric Hunter-Gatherers: The Emersence of Cultural Complexity, edited by T.D. Price and J.A. Brown, pp. 155-190, Academic Press, New Ysrk. Hayden B. and W.K. Hutehings 1989 Whither the Billet Flake? In Experiments in Lithic Technolosy, edited by D.S. Amick and R.P. Mauldin, pp. 235-257. BAR International Series 528. Hayden, B. and M. Nelson 1981 The Use of Chipped Lithic Material in the Contemporary Maya Highlands. American Antiaiuity 613 Hayden, B. and J. Ryder 1991 Prehistoric Cultural Collapse in the Lilfooet Area. Amer Fcan Antiauitv 56 (1): 50-65. Heffley, S. 1981 The Relationship between Northern Athapaskan Settlement Patterns and Resource Distribution: An Application of Horn's Model. In Hunter-Gatherer Foraqinq Strateqies, edited by B. Winterhalder and E.A. Smith, pp. 126-147. University of Chicago Press, Chicago. Henry, D.O. 1989 Correlations Between Reduction Patterns and Settlement Patterns. In Alternative Approaches tuJithic Analysis, edited by D.O. Henry and G.H. ddell, pp. 139-156 Archeological Paper of the American Anthropologicai Association Number 1. Holland, S. 1964 Landforms of British Columbia: A Physiographic Outline. British Columbia Dept. of Mines and Petroleum Resources, Bulletin 48. Horn, H.S. 1968 The Adaptive Significance of Colonial Nesting In the Brewer's Blackbird (Euphagus cyanocephalus). Ecoloqy 49:682-694. Isaac, G.L1. 1977 Olorsesailie: Archeoloqical Studies of a Middle Pleistocene Lake Basin in Kenya. Chicago: University of Chicago Press. Ingbar, E.E., M.L. Larson and B.A. Bradley 1989 A Nontypological Approach to Debftage Analysis. In Experiments in Lithic Technoloqy, edited by D.S. Amick and R.P. Mauldin, pp. 117-136. BAR International Series 528. Johnson, Jay K . 1981 Lithic Procurement and Utilization Trajectories: Analysis, Yellow Creek Nuclear Power Plant Site, Tishomingo County Mississippi. Tennessee Valley Authority Publications in Anthropology Number 28. Johnson, J.K. and C.A. Morrow 1987 The Orqanization of Core Technolosy. Denver: Westview. Keele L.H. 198g1 Experimental Determination of Stone Tool Uses. Chicago: University of Chicago Press. 1982 Hafting and Retooling: Effects on the Archaeological Record. American Antiquity 47~798-809. Keene, Arthur S . 1981 Optimal Foraging in a Nonmarginal Environment: A Model of Prehistoric Subsistence Strategies in Michigan. In Hunter-satherer Forasins Strateqies, edited by B. Winterhalder and E.A. Smith, pp. 171-193. University of Chicago Press, Chicago. Kelly, Robert L. 1983 Mobility Strategies. Journal of Anthropolosical Research 39(3) :277-306. 1988 The Three Sides of a Biface. American Antiauit~53(4): 717-734. Kennedy, D.I.D. and R. Bouchard 1992 Fraser River Lillooet Fishing. In A Complex Culture on the British Columbia Plateau, edited by B. Hayden, University of British Columbia Press, Vancouver. Kew, M. 1992 Salmon Availability, Technology and Cultural Adaptation on the Fraser River Watershed. 16 A Complex culture on the British Columbia Plateau, edited by B. Hayden, University of British Columbia Press, Vancouver. Kim, J-0 and C.W. Mueller 1978 Factor Analysis: Statistical Methods and Practical Issues. Beverly Hills: Sage University Paper No. 14. Knudson, Ruthann 1983 Orqanizational Variability in Late Pale~indianAssemblases . WSU Laboratory of Anthropology Reports of Investigations No. 60. Kuhn, S.L. 1991 "Unpacking" Reduction: Lithic raw Material Economy- in the Mousterian of West-. Central Italy. Journal of Anthropoloqical Archaeolosv 10:76-106. Kuijt, I., W.C. Prentiss and D.L. Pokotylo n.d, Broken Rocks and the Identification of Bipolar Industries: An Experimental Study of Debitage Variability. Ms. in preparation. Leach, H.M. 1984 Jigsaw: Reconstructive Lithic Technology. In Prehistoric Ouarries and Lithic Production, edited by J.A. Ericson and B.A. Purdy, pp. 107-118. University of Cambridge Press, Cambridge. Lewenstein, S.M. 1987 Stone Tool Use at Cerros: The Ethnoarchaeoloqical and Use-wear Evidence. Austin: University of Texas Press. Lothrop, J .C. 1989 The organization of Paleoindian Lfthic Technology at the Potts Site. In Eastern Paleoindian Lithic Resource Use edited by C.J. Ellis and J.C. Lokhrop, pp. 99-138. Lowie, R.H. 1954 Indiane of the Plains. New York: Mcgraw Hill. Loy, T. 1983 Bnehistoric Blood Residues: Detection on Tool Surfaces and Identification of Species Origin. Science 220:1269-1271. Luedtke, €3. E. 1976 Lithic material Distributions and Interaction Patterns During the Late Woodland Period in Michigan. Ph.D. Dissertation, Department of Anthropology, University of Michigan. Macdonald, G.P. 1985 Debert, A Paleo-Indian Site in Central Nova Scotia. Buffalo: Persimmon Press. Magne, M.P.R. 1988 Lithics and Livelihood: Stone Tool TechnoPoqies of Central and Southern Interior British Columbia. National Museum of Man, Mercury Series, Archaeoiogieai Survey of Canada Paper No. 133. 1989 Lithic Reduction Stages and Assemblage Formation Processes. In Experiments in Lithlc Technoloqy, edited by D.8. Amick and R.P. Mauldin, pp. 15-31. BAR International Series 528. 616 Magne. M. and D.L. Pokotylo 1981 A Pilot Study in Bifacial Lithic Reduction Sequences. Lithic Technoloqy 10: 34-37. Mauldin R.P. and D.S. Amick 1989 Investigating Patterning in Debitage from Experimental Bifacial Core Reduction. In Ex~erimentsin Lithic Technoloqy, edited by D.S. Amick and R.P. Mauldin, pp. 67-88. BAR International Series 528. McGuire, R,H.J., J. Whittaker, M. McGuire and R. Swain 1982 A Consideration of Observational Error in Lithic Use-Wear Analysis. Lithic Technoloqy 11:59-63. Miller, T.O. 1979 Stonework of the Xeta' Indians of Brazil. In Lithic Use-Wear Analysis, edited by B. Hayden, pp. 401-408. Academic Press, New York. Milner, G.R., E. Anderson and V.G. Smith 1991 Warfare in Late Prehistoric West-Central Illinois. American Antiquity 56(4):581-603. Mitchell W.R. and 'R.E. Green 1981 Identification and Interpretation of Ecosystems of the Western Kamloo~sForest Resion. Province of British Columbia, Ministry of Forests, Victoria. Montgomery, D. C. 1976 The Desiqn and Analysis of Experiments. New York: John Wiley and Sons. Morice, A. 1893 Notes Archaeological, Industrial and Sociological on the Western Denes. Transactions of the Royal Society of Canada, 1892-1893. Moss, E.2. 1986 Aspects of Site Comparison: Debitage Samples, Technology and Function. World Archaeoloqy 18(1):116-133. Muto, G.R. 1971 A Technological Analysis of the Early Stages of Lithic Implements. M.A. thesis, Department of Anthropology, Idaho State University. 617 Nance, J.D, 1987 Reliability, Validity, and Quantitative Methods in Archaeology. In Quantitative Methods in Archaeoloqy, edited by M. Aldenderfer, pp. 244-293. Sage, Beverly Hills. Nance, J.D. and B.F. Ball 1986 No Surprises? The Reliability and Validity of Test Pit Sampling. American Antiquity 51(3):457-483. 1989 A Shot in the Dark: Shott's Comments on Nance and Ball. American Antiauity 54(2):405-412. Nelson, M. 1991 The Study of Technological Orga'liization. In Anchaeoloqical Method and Theory Vol. 3. edited by M.B. Schiffer, pp. 5','-10p. Tucson: University of Arizona Press. Newcomer, M.H. and G, de G. Sieveking 1980 Experimental Flake Scatter Patterns: A New Interpretive Technique. Journal of Field Archaeolosy 7:345-352. Nielsen, A.E. 1991 Trampling the Archaeological Record: An ~x~erimentalStudy. American Antiauit~ 56(3):483-503. OtConnell, J.F. 197'7 Aspects of Variation in Central Australian Lithie Assemblages. In Stone Tools as Cultural Markers edited by R.V.S. Wright, pp. 269-281. Australian Institute for Aboriginal Studies, Canberra. 0de11, G.H. 1979 A Mew Improved System for the Retrieval of Functional Information from Microscopic Observations of Chipped Stone Tools. In Lithic Use-Wear Analysis, edited by 8- Hayden, pp. 329-344. Academic Press, New York. 1981 The Morphological Express at Function Junction: Searching for Meaning in Lithic Tool Types. Journal of Anthro~oloqical Research 37:319-340. 1989 Experiments in Lithic Reduction. In Experiments in Lithic Technofosy, 618 edited by D.S. Amick and R.P. Mauldin, pp. 163-198. BAR International Series 528. Parry, W.J. and R.L. Kelly 1987 Expedient Core Technology and Sedentism. In The Brqanization of Core Technolosy edited by J.K. Johnson and C.A. Moxrow, pp. 285-304. Westview, Denver. Patterson, L,W. 1982 Replication and Classification of Large sized Debitage. Lithie Technolow 11(3):50-59. 1990 characteristics of the Bifacial Flake . Size Distribution. American Antiquity 55(3):550-558. Peebles, C.S. and S.M. Kus 1977 Some Archaeological Correlates of Ranked Societies. American Antiquity 42(3):421-446. Peterson, N. 1975 Hunter-gatherer Territoriality: The Perspective from Australia. American Anthropoloqist 77:53-68. Piddocke, S. 1965 The Potlatch System of the Southern Kwakiutl: A New Perspective. Southwestern Journal of Anthropolosv 44:244-264. Pokotylo, D.L. and C.C. Hanks 1989 Measuring Assemblage variability in Curated Lithic Technologies: A Case Study from the Mckenzie Mountains, Northwest Territories. In Ex~erimentsin Lithic Technolosy, edited by D.S. &nick and R.P. Mauldin, pp. 49-66. BAR International Series 528. Prentiss, W.C. 1991 A Preliminary Analysis of the Lithic Artifacts from the Rim Deposits, Housepit 7, Keatley Zreek Archaeological Site (Eerl7J. Unnublished Report on File, Keatley Creek Archaeological Project, Department of Archaeology, Simon Fraser University, Burnaby. Prentiss, W.C. and I. Kuijt n.d. Sullivan and Rozenfs Debitage Typology and Archaeological Site Formation: Experimental and Archaeological Research. Ms. in possession of 619 Authors. Prentfss, W.C, and EIJ. Romanski 1989 Experimental- Evaluation of Sullivan and Rozen's Debitage Typology. In Experiments in Lithic Technolow, edited by D.S, Amick and R.P. Mauldin, pp. 89-100. BAR International Series 528. Prentiss, W.C., E.J. Romanski and M.L. Douthit 1988 Hunter-Gatherer Land Use and Lithic Procurement in the Central Big Horn Basin, Wyoming. The Wyominq Archaeoloqist 3l( 1-2): 33-53. Ray, V.F. 1939 Cultural Relations in the Plateau of Northwestern America. Los Angeles: Southwestern Museum. Richards, T.H. and M.K. Rousseau 1987 Late Prehistoric Cultural Horizons on the Canadian Plateau. Department of Archaeology, Sinon Fraser University Publication Number 16. Robinson, W.S. 1957 The Statistical Measurement of Agreement. American Sociolosical Review 22:17-25. Rolland, N. and H.L. Dibble 1990 A New Synthesis of Middle Paleolithic Variability. American Antiquity 55(3):480-499. Romanoff, S. 1986 Fraser Lillooet Salmon Fishing. Northwest Anthro~olosical Research Notes 19(2):119-160. 1990 The Cultural Ecology of Hunting and Potlatches Among the Lillsoet Indians. Northwest Anthro~oloqicalResearch Notes 23. 1992a Fraser Lillooet Salmon Fishing. In A Complex Culture of the British Columbia Plateau edited by B. Hayden, University of British Columbia Press, Vancouver. 1992b The Cultural Ecology of Hunting and Potlatches Among the Lillooet Indians. In A Com~lexCulture of the British Columbia Plateau edited by B. Hayden, University of British Columbia Press, Vancouver. Roulan, P.J. 1939 A Simplified procedure for Determining the 620 RePiabilit of a Test by Split Halves. Harvard ~ducationarReview 9 :99-lO3. Rozen, K.C, and A.P. sullivan 111 1989a Measurement, method and Meaning in Lithic Analysis: Problems with Amick and Mauldints Middle Range Approach. American Antiuuity 54(1):169-174. 1989b The Nature of Lithic Reduction and Lithic Analysis: Stage Typologies Revisited, American Antiquity 54(1):179-184. RummeP, R.J. 1970 Applied Factor Analysis. Evanston: Northwestern University Press. Ryder, J. 1978 Geomorpkology and Late Quaternary H istory of the Lillooet Area. In Re~ortsof the tillooet Archaeolosical Project, No. 1 Introduction and Settinq. National Museum of Man, Mercury Series, Archaeological Survey of Canada Paper No. 73. pp. 56-67. Sahlins, M. 1972 Stone Ase Economics. Chicago: Aldine. Schiffer, M.B. 1976 Behavioral Archaeoloqy. New York: Academic Press. 1987 Formation Proceses of the Archaeoloqlcal Record. Albuquerque: University of New Mexico Press. Schuessler, K. 1971 Analyzing Social Data: A Statistical Orientation. Boston: Houghton-Mifflin. Shelley, P .H. 1990 Variation in Lithic Assemblages: An Experiment. Journal of Field ArchaeoPoqy 17:187-193. Shott, M. 1987 Technological Organization and Settlement Mobility: An Ethnographic Examination. Journal of Anthropoloqical Research 42:15-41. 1989 On Tool-Class Use Lives and the Formation of Archaeological Assemblages. American Antiquity S4(1) :9-30. 621 Singer, C.A. 1984 The 63-Kilometer Fit. In Prehistoric Quarries and Lithic Production, edited by J.A, Ericson and B,A1 Purdy, pp. 35-48, University of Cambridge Press, Cambridge. Smith, E.A. 1979 Human Adaptation and Energetic Efficiency. Human Ecoloqy, 7(1):53-74. 1983 Anthropological Applications of Optimal Foraging Theory: A Critical Review. Current Anthropoloqy 24(5):625-651. Soffer, 0. 1989 Storage, Sedentism and the Eurasian Palaeolithic Record. Antiauit~63:719-732. Spafford, 3. 1991 Artifact Distributions on Housepit Floors and Social Organization Fn Housepits at Keatley Creek, M.A. Thesis, Department of Archaeology, Simon Fraser University. Spector, P.E. 1981 Research Desiqns. Beverly Hills: Sage University Papers on Quantitative Analysis in the Social Sciences No. 23. Speth, J.D. 1972 Mechanical Basis of Percussion Flaking. American Antiauity 37(1):34-60. 1983 Bison Kills and Bone Counts: Decision Makins BY Ancient Hunters. Chicago: University of Chicago Press, Stahle, D.W, and J.E. Dunn 1982 An Analysis and Application of Size Distribution of Waste Flakes from the Manufacture of ~ifacialStone Tools, World Archaeoloqy 14(1) :84-37. Stanley, J. 1971- Reliability. In Educational Measurement edited by R. Thorndike, pp. 356-442. American Council on Education, Washington. Stephens. D.W. and E.L. Charnov 1982 Optimal Foraging: Some Simple Stochastic Models. ~ehaviorarEcoloqy and Sociobiolosv 6:27-47. Stevenson, M. G. 1985 The Formation of Artifact Assemblages at Workshop/Habitation Sites: Models from Peace Point in Northern Alberta. Antiquity 50il): 63-81. Strathern, M. 1969 Stone Axes and Flake Tools: Evaluations from Two New Guinea Highlands Societies. Prsceedinqs of the Prehistoric Society 35:311-329. Sullivan, A.P. 111 1987 Probing the Sources of Variability: A Regional Case Study near the Homolovi Ruins, Arizona. North American Archaeoloqist 8(1):41-71. Sullivan, A.P. I11 and K.C. Rozen 1985 Debitage Analysis and Archaeological Inference. American Antiquity 50(41:755-779. Statistics Systat Inc. Evanskon: Systat Inc. Teit, J.A. 1900 The Thompson Indians of British Columbia. Memoirs of the American Museum of Natural History I( 4):163-392. 1906 The Lillooet Indians. Memoirs, American Museum of Natural History 2(5):193-300. 1909 The Shuswap Indians. Memoirs, American Museum of Natural History 4:443-758. Teltser, P.A. - 1991 Generalized Core Technology and Tool Use: A Mississippian Example. Journal of Field Archaeoloqy 18:363-375. Thomson, D .F. 1964 Some Wood and Stone Implements of the Bindibu Tribe of Central and Western Australia. Proceedinqs of the Prehistoric Society 30: 400-422. Todd, L.C. 1987 Taphonomy of the Horner I1 Bonebed. In The Horner Site: Type Site of the Cody Cultural Complex edited by G.C. Prison and L.C. Todd, Academic Press, Orlando . Terrence, R. 1983 Time Budgeting and Hunter-Gatherer Technology. In Hunter-Gatherer Economy in Prehistory, edited by 623 G. Bailey, pp. 11-22. Cambridge, Cambridge University Press. 1489 Retooling: Towards a Behavioral Theory of Stone Tools. In Tnedited by R. Terrence, pp. 57-66. Cambridge University Press, Cambridge . Tringham, R.G., G. Cooper, G.H. Odell, B. Voytek and A. Whitman 1974 Experimentation in the Formation of Edge Damage: A New Approach to Lithic Analysis. Journal of Field Archaeoloqv 1:171-196. 1918 Food Plants of British Columbia Indians. Part 11. Inter ior Peoples. British Columbia Provincial Museum Handbo~kNo. 36, Victoria. 1979 Plants in British Columbia Indian Technolouy. British Columbia Provincial Museum Handbook No. 38, Victoria. 1992 Plant Resources of the Fraser River Eillooet People: A Window into the Past. In A Complex Culture of the British Columbia Plateau edited by B. Hayden, University of British Columbia Press, Vancouver. Vaughaun, P.C. 1985 Use-Wear Analysis of Flaked Stone Tools. Tucson:University of Arizona Press. Vehik, S.C. 1985 Lithic Resource Procurement: Proceedinqs from the Second-Conference on Prehistoric Chert Exploitation. Center for Archaeological Investigations Occasional Paper No. 4. Carbondale. Webley, L. 1990 The Use of Stone "Scrapersw by Semi-sedentary Pastoralist Groups in Namaqualand, South Africa. South African Archaeolosical Bulletin 45:28-32. Weissner, P. 1982 Beyond "'"w~l~~ow Smoke and Dog's Taiis: A Zoii-ient on Binford's Analysis of Hunter-gatherer Settlement Systems. kserican Antiquity 4?:171-1?8. White, J.P. 1968 Ston Naip Bilong Tumbuna: The Living Stone Age in New Guinea. In La Prehistoire: Problemes et Tendances. edited by D. de Sonneville-~oudes, pp, 511-516, CNRS, Paris. 624 White, J.P., N. Modjeska and I. Hipuya 1977 Group definitions and Mental Templates: An Ethnographic Experiment. In Stone Tools as Cultural Markers, edited by R.V.S. Wright, pp. 380-390. Australian Institute for Aboriginal Studies, Canberra. White, J.P. and D.H. Thomas 1972 What Mean These Stones? Ethno-taxonomic Models and Archaeological Interpretations in the New Guinea Highlands. In Models in Archaeoloqy, edited by D.L. Clark pp. 275-308, Methuen, London. Whitelaw, T. 1991 Some Dimensions of Variability in the Social Organization of Community Space among Foragers. In Ethnoarchaeoloqical Approaches to Mobile Campsites edited by C. S . Gamble and W. A. Boismier, pp. 139-188. International Monographs in Prehistory, Ethnoarchaeological Series 1, Ann Arbor. Wiant H.D. and H. Hassen 1985 The Rale of Lithic Resource Availability and Accessibility in the Organization of Technology. In Lithic Resource Procurement: Proceedinqs from the Second Conference on Prenistoric Chert Exploitation, edited by S.V. Vehik, Center for Archaeological Investigations Occasional Paper No. 4. Carbondale. Wilmsen, E.N. and F.H.H. Roberts Jr. 1984 Lindenmeier, 1934-1974 Concludinq Report on investiqations. Smithsonian Contributions to Anthropology Number 24. Winterhalder, B. 1980 Environmental Analysis in Human Evolution Research. Human Ecoloqy 8(2):135-169. 1981 Foraging Strategies in the Boreal Forest: An Analysis of Cree Hunting and Gathering. In Hunter-qatherer Foraqinq Strateqies, edited by 3. Winterhalder and E.A. Smith, pp. 66-98. University of Chicago Press, Chicago. 1986 Diet Choice, Risk and Food Sharing in a Stochastic Environment, Journal oL Anthro~oloqicalArchaeoloqy 5:369-392. Yerkes, R.W. 1989 Prehistoric Life on the Mississippi Floodplain. Chicago: University of Chicago Press. 625 APPENDIX A RELIABILITY AND VALIDITY ANALYSIS CORE WEIGHTS VALIDITY DATA Post-Reduction Core Core Weight Debitage Weight Total ------(Grams) ( Grams ) ( Grams ) 1 146.7 868.6 1015.3 2 408.2 1127.1 1535.3 3 417.2 1451.2 1868.4 4 971.9 1567.6 2539.5 5 202.7 1263.1 1465.8 6 195.4 1185.0 1380.4 7 247.6 78.9 326.5 8 137.0 94.4 231.4 9 183.2 46.4 229.6 10 136.0 101.5 237.5 11 112 3 39.6 151.9 12 350.3 58.5 108.8 13 114.3 730.5 844.8 14 301.0 54.9 355.9 15 356.8 242.6 599.4 16 821.8 59.0 880.8 17 325.0 488.1 813.1 18 302.4 381.5 683.9 19 210.0 556.2 766.2 20 273.3 . 788.0 1061.3 21 * 201.1 22 586.4 104.4 690.8 23 543.8 47.7 591.5 2 4 859.9 48.5 908.4 25 47.3 4.4 51.7 26 51.7 6.5 58.2 27 37.3 3.4 40.7 28 64.4 5.2 69.6 29 65.0 5.9 70.9 30 38.0 1.0 39.0 31 80.1 2.9 83.0 32 34.5 1.0 35.5 33 48.5 2.1 50.6 34 55.5 12.9 68.4 35 59.8 12.2 72.0 36 40.4 22.5 58.9 37 65.7 9.3 '95.0 38 31.9 9-6 41.5 39 45.0 3.4 48.4 40 66.0 2.9 68.9 41 61.5 1.7 63.2 42 74.3 2.9 77.2 43 334.6 278.5 613.1. 44 284.1 102.6 386.7 ------627 Validity Data Continued * Removed from Laboratory RELIABILITY DATA Post-Reductf on Core Core Weight Cebitage Weight Total ( Grams 1 (Grams ) (Grams 1 1 124.9 34.2 159.1 2 133.2 23.0 156.2 3 105.5 46.5 152.0 4 133.4 30.4 163.9 5 k 40.4 6 144.5 24.2 168.9 7 95.3 33.6 128.9 8 107.8 13.2 121.0 9 71.8 18.8 90.6 10 117.3 19.0 136.3 11 77.3 44.1 121.4 12 130.2 35.3 165.5 13 96.8 30.0 126.8 14 61.3 19.0 80.3 15 45.8 16.9 62.7 16 101.7 25.0 126.7 17 90.3 40.9 131.2 f 8 57.4 14.8 72.2 19 95.0 15.5 110.5 20 79.9 29.7 109.6 21 73.9 30.0 103.9 22 62.3 18.3 80.6 23 * 68.4 24 67.9 25.0 92.9 25 76.0 10.9 86.7 26 71.6 26.6 98.2 27 75.2 25.0 100.2 28 67.6 30.0 97.6 29 11362 31.9 145.1 30 90.8 18.0 108.8 * Removed from Laboratory APPENDIX B DATA COLLECTION DESIGN: ORGANIZATION OF LITHIC REDUCTION EXPERIMENTS Validity Analysis: Randomized Reduction Order Unprepared Prepared Sizes Flake Biface Core Core Extra- 1. 17" 4. 50 Large 2. 37 5. 10 3. 13 6. 33 Large Hed ium 25, 15 34. 29 43. 41 52. 54 26. 49 35. 1 44. 16 53. 34 27. 38 36. 59 45. 39 54. 60 28. 40 37. 58 46. 56 55. 26 29. 12 38. 45 47. 30 56. 55 30. 3 39. 5 48. 32 57. 35 * Reduction Order 1. Core Reliability and Validity Analysis: Randomized Blank and Hammer Associations Validity Analysis Reiiabifity Analysis Validity Contnd. APPENDIX C VITREOUS TRACHYDACITE DATA Post-Reduction Core Weight Debitage Weight Total Core (Grams1 (Grams) (Grams ) la Bifaee 2 146.2 43.8 190.0 lb Biface 2 121.8 42.8 164.2 2a Biface 3 99.2 12.5 111.7 2b Biface 3 76.8 26.8 103.6 3a Biface 3 82.5 .9 83.4 3b Biface 3 46.8 1.0 47.8 4a Flake 79*9 .6 66,5 4b Flake 73.8 1.0 74.8 5a Flake 33.8 5.8 39.6 5b Flake 62.5 5.8 68.3 6a Flake 86.9 16.2 103.1 6b Flake 62.3 2.8 65.1 7 PrCore 179.6 103.8 283.4 8 UprCore 183.8 210.0 393.8 9 PrCore 154.4 575.5 729.9 10 UprCore 180.7 474.0 654.7 lla BpCore 73.2 41.0 114.2 llb BpCore 58.0 54.2 112.2 12a BpCore(BF) 28.7 40.0 68.7 12b BpCore(BF1 34.7 42.2 76,9