Excavationand PreliminaryAnalysis of an Workshopin Tula, Hidalgo,

Dan M. Healan Janet M. Kerley George J. Bey III Tulane University New Orleans, Louisiana

Previoussurvey of the Toltec site of Tula, Hidalgo, Mexico, revealedwhat is believed to have been a zone of obsidianworkshops within the Early Post- classic city. TulaneUniversity conducted excavation of part of one topo- graphiccomplex within the zone, revealinga workshopcomplex consisting of a linear arrangementof residentialcompounds, peripheral refuse dumps, and open workareas, the latter tentativelyidentified by microscopicanalysis of soil samples. Over500,000 pieces of obsidianwere recovered,revealing an exclusivelycorelblade that importedpercussion macrocores from at least two differentsources and producedprismatic cores, blades, and certain bladeproducts. The reductionsequence and differencesin the processingof obsidianfrom differentsources are documented.Ceramic and stratigraphicdata suggest the locality was originallymarginal land set- tled relativelyearly in Tula's history,possibly by immigrantsfrom the Basin of Mexico.

Introduction the courseof prehispaniccultural development. In Cen- The study of prehistoric obsidian exploitation is a rel- tral Mexico there appearsto have been a successionof atively new and multifaceted area of research covering powerfulcenters that controlledsystems for the exploi- a variety of topics that include reconstruction of regional tationof resourcesthat included Middle Formative Chal- exchange networks, reduction , and archae- catzingo,'possibly Late Formative Cuicuilco, and Classic ological investigation of sites including periodTeotihuacan. Teotihuacan is of particularinterest both quarry and non-quarry workshops. Obsidian work- becauserecent researchhas demonstratedthat its econ- shops have been identifed in many prehistoric commu- omy was inextricablytied to obsidianexploitation.2 The nities largely on the basis of anomalous surface con- EarlyPostclassic period has been a notablegap in current centrations of obsidian and other artifacts. Their knowledgeabout this patternof succession, which has identification has played a major role in assessing the up to now been filled by speculationthat Tula, the dom- importance of obsidian exploitation in prehistoric eco- inantcenter in CentralMexico duringthe EarlyPostclas- nomic systems as well as in reconstructing such systems. Unfortunately, much of the archaeological investigation 1. David Grove, KennethHirth, David Buge, and Ann Cyphers, of obsidian workshops has been restricted to surface sur- "Settlementand Cultural Development at Chalcatzingo,"Science 192 vey or limited test-pitting, despite obvious benefits of (1976) 1203-1210; KennethHirth, "Interregional Trade and the For- large-scale excavation for technological, functional, and mationof PrehistoricGateway Communities," AtnAnt 43 (1978) 35- chronological study, as well as for providing information 45. about domestic, social, and other non-technological 2. MichaelSpence, "The ObsidianIndustry of Teotihuacan,"AtnAnt realms of workshop activity. 32 (1967) 507-514; idem, "ObsidianProduction and the State at Teotihuacan,"AtnAnt46 In , it is apparent from many lines of (1981) 769-788; ReneMillon, Urbanization at Teotihuacan, Mexico: The Teotihuacan Map l (University of Texas research that control of the exploitation of various re- Press;Austin 1973) 45-46; ThomasCharlton, ''Teotihuacan, Tepea- sources, including obsidian, played a key role in shaping pulco, and ObsidianExploitation," Science 200 (1978) 1227-1236. 128 Excavationof an ObsidianWorkshop in Tula, Mexico/Healan,Kerley, and Bey

Table 1. Provisional chronology of the Tula region.

PERIOD PEIASE A .C . 1600 Colonial Tesoro

1500 _ 1400 Late Postclassic Palacio (Second Intermediate phase 3 and Late Horizon) 1300 Fuego

1200

1100- Early Postclassic Tollan (Second Intermediate phase 2) 1000 Figure 1. Landform map of Central Mexico, showing the location of Tula with respect to the Classic period site of Teotihuacan and TerminalCorral three obsidian sources expoited by Tula: 1, Pachuca (Cruz del 900 Milagro), Hidalgo; 2, Otumba, Mexico; 3, Zinapecuaro, Michoacan. Corral Epiclassic sic, inherited the obsidian exploitation system previously 800 (Second Intermediate controlled by Teotihuacan.3 - phase l) This paper is a preliminary report of the excavation and analysis of an obsidian core/blade workshop located Prado in what was probably a larger workshop zone at Tula. 700 The excavation involved extensive exposure of a single workshop complex, and the analysis is unusual in the sheer quantity of workshop debitage plus other artifacts recovered in situ. The report is preliminary in nature, since much of the analysis and interpretation is ers and the moderntown of Tula de Allende (FIG. 2).5 still in progress. Ethnohistoricalresearch by W. Jiminez-Morenoand ex- tensive archaeologicalexploration by J. Acosta7 have Background

Tula is located in sw Hidalgo about 70 km. NW of 5. Sources: James Stoutamire, ''Trend Surface Analysis of Archae- Mexico City (FIG. 1).4 The site incorporates the alluvial ological Survey Data from Tula, Hidalgo, Mexico," unpublishedPh.D. bottoms and adjacent uplands of the Tula and Rosas Riv- dissertation, University of Missouri-Columbia ( 1975) figs. 5-10; Juan Yadeun Angulo, "E1 Estado y la Ciudad: E1 Caso de Tulae Hgo.," Coleccion Cientifica 25 (INAH 1975) figs. 17, 25-32; Alejandro Pas- 3. Michael Spence and Jeffrey Parsons, "Prehispanic Obsidian Ex- trana Cruz, "Produccion de Instrumentos en Obsidiana-Division del ploitation in Central Mexico: A Preliminary Synthesis,'' MichMus- Trabajo," unpublished thesis, Escuela Nacional de Antropologia e Anth 43 (1972) 29; Charlton, op. cit. (in note 2) 1235; Lee Parsons Historia (=E.N.A.H.) (Mexico D.F. 1977); Enrique Nalda and Ale- and Barbara Price, ''Mesoamerican Trade and its Role in the Emer- jandro Pastrana, 'Una Proposicion para la Investigacion de los "Tall- gence of Civilization,9' Contributions of the University of California eres de Litica" en Tula, Hgo.," in 'wProyectoTula, Segunda Parte,'' Archaeological Research Facility (CARFC) 11 (1971) 188; Robert EduardoMatos Moctezuma, ed., Coleccion Cientifica 33 (INAH 1976) Zeitlin, "Towards a More Comprehensive Model of Interregional 75-83. Commodity Distribution: Political Variables and Prehistoric Obsidian Procurement in Mesoamerica," AmAnt 47 (1982) 270. 6. Wigberto Jiminez-Moreno, ''Tula y los Toltecas Segun las Fuentes Historicos, " Revista Mexicana de Estudios Antropo/ogicos 5 ( 1941 ) 4. Adapted from Landforms of Mexicos Prepared for the GeographAt 79-83. Branch of the Office of Naval Researchs by Erwin Raisz (Cambridge 1959). 7. Jorge Acosta, ';InteIpretacion de Algunos Datos Obtenidos en Tula \\ C e *r r r o \ sM s< 0 s r9 ' o- ,> n s e; > o; . { j'r{ .- j { u ; s\v>l tL>;W o 4

Journalof Field ArchaeologylVol.10, 1983 129

Figure 2. The archaeological zone of Tula, Hidalgo, as determined by the tSo/// X\Sl"f"fovo *w-*@\*\ University of Missouri and INAH

concentration surface surveys. A more detailed map of ol C- V the area east of E1 Salitre (box) is continuous presented in Figure 3.

...,. . X . =

.. < =4

,.<-.<

,,,,,,,,,,'... B . !H..

A}IJ#ssulbJswwJLt8ttltst88C--;;t>Ns>/

{ / 1 M I0 n h a ) A II e n d e t

provided persuasive evidence that the site constitutes the of Missouri,'°which have greatlyamplified our knowl- ruins of Tollan, legendary capital of the Toltec empire edge of Tula's historyand internalstructure. Cobean has between the 10th and 12th centuries A.C.8Recently, Tula recentlyprovided a revised ceramicchronology for the was the subject of two extensive archaeological research Tula area (TABLE 1)ll based upon the Missouriproject's projects, conducted by the Instituto Nacional de Antro- urban survey, and the residentialexcavations and se- pologia e Historia (INAH) of Mexico9 and the University lected test pits of the INAH project. Data from both projects indicate that Tula was originally a relatively modest settlementsome 3-5 sq. km. in area that was Relativos a la Epoca Tolteca,'' Rerista Mexicana de Estudios Antro- centeredaround Tula Chico, a smallerversion of Tula pologicos 14, segunda parte (1956-1957) 75-110. Acosta also pub- lished detailed summaries of his field seasons at Tula, which are listed Grande, the political-religiouscenter of the later city in the bibliography of Supplement to the Alandbookof Middle American (FIG. 2). Duringthe Tollanphase of the EarlyPostclassic lndians, Volume 2, Victoria Bricker and Jeremy Sabloff, eds. (Uni- period(TABLE 1), Tula underwentexpansive growth, and versity of Texas Press: Austin 1981) . 8. Pedro Armillas, "Teotihuacan, Tula, y los Toltecas," Runa 3 (1950) 37-70; Lawrence Feldman, ''Tollan in Hidalgo: Native Accounts of 10. Diehl, ed., op. cit. (in note 8); ''Tula," in Bricker and Sabloff, the Central Mexican Tolteca," in ''Studies of Ancient Tollan," Rich- op. cit. (in note 7) 277-295. Diehl, ed., University of Missouri Monographs in Anthropology 11. Robert Cobean, ''The Pre-Aztec Ceramics of Tula, Hidalgo," 1 (1974) 130-149; Nigel Davies, The Toltecs Until the Fall of Tula unpublished Ph.D. dissertation, Harvard University (Cambridge 1978) (Oklahoma 1977). fig. 4. We have modified Cobean's period designations by including 9. Eduardo Matos Moctezuma, ed., "Proyecto Tula, Primera Parte," in parentheses an alternate scheme proposed by the 1972 Valley of Coleccion Cientifica 15 (INAH 1974); Matos Moctezuma, op. cit. (in Mexico Conference; cf. Eric Wolf, The Valley of Mexico (New Mex- note 5). ico 1976) fig. 16. 130 Excavationof an ObsidianWorkshop in Tula,MexicolHealan, Kerley, and Bey during its apogee encompassed an area of about 12 sq. north flank of El Salitre, anotheralong its east flank, km.l2 and a thirdimmediately east of CerroEl Cielito. During The importance of obsidian to Tula is evident in the the sumrnerof 1978, Healan made an informalrecon- frequency with which obsidian artifacts, chiefly pris- naissanceof the "east flank" concentrationand discov- matic blade fragments,l3 are encountered among the de- ered within it localized concentrations of obsidian, bris that litters the site's surface. This is no less evident utilitariansherds, and architecturalstone frequentlyas- in in situ deposits, given that over 25,000 obsidian ar- sociated with low topographicrises indicative of col- tifacts, again principally fragments of prismatic blades, lapsed structures.Thus the "east flank" concentration, were recovered by the Missouri project's excavation of and probablythe othersas well, seeminglyrepresents a a series of residental compounds. 14Prismatic blade cores series of distinctworkshop sites; in otherwords, a zone from Tula have a distinctive type of platform which has of obsidianworkshops within the city whereina number been ground or abraided to a flat surface with a texture of artificersboth lived and worked. like ground glass. 15Tula lies only about 70 km. west of Duringthe summerof 1980, Tulane Universitycon- the Pachuca or Cruz del Milagro obsidian source (FIG.1), ductedexcavations at a single locality within the "east an important source of obsidian for Mesoamerican corel flank" concentration(FIG. 3). The excavationswere di- blade industries and which was the major source ex- rected by Healan and sought to provide extensive ex- ploited by Teotihuacan during the Classic period. This posure of one workshopcomplex. The excavationand distinctive clear green obsidian is also the predominant subsequentanalysis was intendedto satisfy four goals: source represented at Tula.16 1) verificationof the existenceof an obsidianworkshop A significant result of the recent research at Tula was and delineationof its structure;2) recovery of a sub- the discovery of several areas of high surface concentra- stantialpart of the workshoplithic assemblagefor de- tion of obsidian artifact debris around a brackish marsh terminationof input, output, and reductiontechnology; (El Salitre) and an adjacent elongated hill (Cerro El Cie- 3) recoveryof datapertaining to domesticand other non- lito) in the eastern portion of the site. As shown in Figure technologicalrealms of workshopactivity; 4) diachronic 2,17 there are three major concentrations: one along the studyof the developmentof obsidianworkshop produc- tion at Tula throughstratigraphic excavation. 12. Stoutamire,op cit. (in note 5). 13. Prismaticblades, also called "fine blades," are unusuallyregu- Field Methods lar, parallel-sidedblades with a commonlytrapezoidal cross section The localityselected for excavationis locatednear the removedby pressurefrom equally distinctive, finely flutedcylindrical center of the "east flank" concentrationin a modern cores. For a discussionof ethnohistoricalaccounts of prismaticblade productionas well as modernreplicative experiments, cf Don Crab- agriculturalfield (FIGS. 2, 3). The localitycontains several tree, "MesoamericanPolyhedral Cores and PrismaticBlades," AmAnt low ridges or longitudinalmounds littered with stone 33 (1968) 446-478; CharlesFletcher, "Escapable Error in Employing typicalof collapsedstructures, but otherwiserather low Ethnohistoryin Mesoamerica,"AmAnt 35 (1970) 209-213; Payson frequenciesof surfaceartifacts. Conversely, little build- Sheetsand Guy Muto, ''PrismaticBlades and Total Cutting Edge: An ing stone but very high surfaceconcentrations of obsidi- Experimentin LithicTechnology," Science 175 (1972) 632-634; John Clark, "Manufactureof MesoamericanPrismatic Blades: An Alter- an and sherdswere found in the low areasbetween the nativeTechnique," AmAnt 47 (1982) 355-376 ridges. Healandecided to expose a continuousarea ex- tendingfrom the top of the largestridge (ridgeA) to the 14 Alice Benfer, "A PreliminaryAnalysis of the ObsidianArtifacts from Tula, Hidalgo," in Diehl, ed, op. cit. (in note 8) 56-87. adjacentinterridge area (FIGS. 3, 4) in orderto samplethe full range of topographicand surfaceartifact diversity. 15. At least some of these ''ground" platformsare actuallynatural cortex surfaces As seen in Figure4, we used 3-m. squaresexcavated in 10-cm. levels that were subdividedhorizontally and/ 16. Stoutamire,op. cit (in note 5) 41-42. or verticallywhen necessaryto separatedifferent depo- 17. Thereis disagreementamong the varioussurveys (cf. note 5) on sitional contexts. We plannedto expose only the first the natureof these surfaceconcentrations. Stoutamire's survey defined componentsof occupationencountered below the surface a single, largeconcentration that covers virtuallyall of the areanorth, in orderto maximizehorizontal rather south, and east of E1 Salitre. The alternativedefinition of distinct than verticalex- concentrationswas suggestedby Nalda and Pastrana'sinterpretation posure. At the same time, however, a series of 1-m. of Yadeun'ssurvey data and by Pastrana'sresurvey of the area;the squarepits normallyplaced in the NE corner of every lattersurveys, however, disagree on the precisenumber, location, and other squarewere excavatedto sterile deposits in order extentof theseconcentrations. Based on informalsurvey, Healan agrees to recoverdata pertaining to earliercomponents. thatdistinct concentrations rather than a continuousdistribution exists in the area. The concentrationsdepicted in Figure 2 are approxima- Considerableeffort was made to recoversoil samples tions intendedto reconcile the major differences among the latter from the various floors and other surfacesand deposi- surveys. tional featuresencountered in excavation, in the belief Journal of Field ArchaeologylVol. 10, 1983 131

Figure 3. Map of area of high surface concentration along the east flank of E1 Salitre, showing the location of the workshop complex excavated by Tulane University.

that microscopicmaterial, particularly obsidian micro- m. intervals (FIG. 5), and penetrated almost a full meter debitage, would be useful in identifyingthe actual ob- below the surfaceg Despite the obviously extensive dam- sidianwork sites and otheractivity areas. A preliminary age to the subsurface deposits, we decided to continue microscopicexamination of soil sampleswas undertaken excavation for two reasons: fortunately, the archaeolog- in the field using a Wild M-5 stereomicroscopewith ical deposits were sufficiently thick that at least part had magnificationof up to 50x. Furtherstudy is planned escaped the chisel plow, and below the tractorplow zone using better techniques of sample preparation and the chisel scars were easily detected and cleaned out to examination. avoid stratigraphic contamination of the lower levels. Unfortunately,much of the areaaround E1 Salitre has Hence we were able to overcome some of the major been underintensive tractor cultivation for over a dec- problems posed by the chisel plowing, but at consider- ade, and therehas been considerablygreater damage to able cost in time and resources, since it required exca- the archaeologicaldeposits than expected;much of the vating much deeper than planned in some areas (cf. FIG. damagewas found to be the resultof very recentchisel 8). It appears that none of the "east flank" concentration plowing. In 1979 Healan learned that a government- escaped the chisel plow, and it is not likely that the other sponsoredprogram was introducingchisel plows in the concentrations did either. We regret not having been able Tulaarea to breakup subsurfacedeposits of caliche, and to excavate at least one season earlier. the field seasonwas plannedwith this eventualityin mind. Chiselplowing in this area,however, occurred far ahead Delineation of the Workshop Complex of schedule(ironically, only about three weeks priorto the start of our excavationt). Chisel scars covered the GeneralComments entirefield in a crisscrosspattern, spaced at roughly 1- The artifacts and features encountered confirm that the 132 Excavationof an ObsidianWorkshop in Tula,MexicolHealan, Kerley, and Bey

Figure4. Plan of the obsidianworkshop complexexcavated by Tulane University.

locality had been an obsidianworkshop. The necessity andother features of weathering(FIG. 6) thatindicate the of deeperexcavation restricted the degree of lateralex- clay horizon had lain exposed before the locality was posure that had been planned;nevertheless, there was occupied.This is also indicatedby the fact that artifact- sufficientexposure to sample a large part of one ridge rich refuse soils and structuralremains lay directly on and interridgesystem and parts of adjacentareas. The and were partiallyembedded in the clay surface. complexwas a ratherdispersed affair, containing distinct habitation,obsidian-working, and refuse-dumpingareas (FIG. 4), each of which is describedbelow. Habitation Area Underlyingthe entireexcavation was a horizonof tan, Ridge A was an entirelyartificial accumulation of su- sterileclay (FIG. 5), which apparentlycomprised the sur- perposedstructures and fill (FIG. 5). The ridge was ap- face of the locality at the time of initialoccupation. The parentlyan accretionalfeature, which is to say it was surface of this clay was scarredby erosional channels not a purposefulterrace or platform,but ratherthe ac- , - 3 , lJ . = = . J . 1>/ . . ' . ,r . ' . e .. +

Journal of Field ArchaeologylVol. 10, 1983 133

o A STR I CORRIDOR STR 11 CIRCU B

2_ X KT,9ig;

B LAR PIT WORK AREA? ARROYO PITS C

J j

L X wrclwnscindyicicim

E<;<1, ccimpcict hrclwn Icicim plsIster f ccir

T >N 4 cidslhe ruhile , :- ccimpcicted cicly urfcice < M N D E

1 wslter Iclid sediments X chisel sscir

v 1 clhsidicin & clthe refuse ) stcine 0

S|lillVjTHIM| te ile Icin cicly i

Figure 5. Simplified stratigraphic transect of the workshop complex seen in Figure 4.

cumulation of localized structural renovation and col- lapse. The earliest structures were erected on the clay surface, over which artifact-rich fill, predominantly workshop and domestic refuse, was placed as a base for later structures. As many as three distinct, superposed structures were noted in some profiles. Invariably, only traces of the uppermost structures remained; we elected, therefore, to expose earlier structuralremains which we hoped had escaped much of the destruction of plowing (FIG. 7). Excavation to lower levels produced remains of three structures (FIGS. 4, 8) designated I-III. Both Structures I and II underlay traces of later struc- tures, and Structure III underlay a cobblestone surface that may have been a paved thoroughfare. Because of limitations of time and resources, all three of these struc- Figure6. The tan clay horizon, exposed immediatelynorth of StructureII, looking south. The extremelyeroded nature of the clay tures were only partially exposed. surfaceis seen in foreground;half of an apparentlycircular refuse StructuresI and II form part of a single complex com- pit is seen at top. parable to residential compounds excavated by the Mis- 134 Excavationof an ObsidianWorkshop in Tula,MexicolHealan, Kerley, and Bey

these structures, only nominal amounts of obsidian, mostly fragments of prismatic blades and cores, were recovered from floors. In a superficial microscopic ex- amination of floor samples from the structures and open areas, Healan did not detect any concentrations of mi- crodebitage that should have existed had obsidian been worked there with any regularity. Hence there is no evi- dence for obsidian working in the habitation area, though the debitage recovered from structuralfill must have come from nearby.

Refuse Dump In the area between ridges A and B (FIGS. 3, 4), high concentrations of obsidian and other artifacts were found Figure7. Habitationarea along ridge A, looking NE. The front(east that extended throughout the soil to the underlying clay end) of StructureI faced with small-stoneveneer is seen in foreground.The workersare digging throughthe badly damaged horizon (FIG. 5). Since very little loose building stone remainsof a later structure(note on floor) to expose the and no architecturalremains were encountered, we con- earlier,better-preserved Structure I. clude that this had been an open area; this circumstance is fortunate, since any structuralremains in the relatively thin soil here would have been destroyed. The clay sur- face was scarred by chisel plowing, but was otherwise souriprojectl8 at the Canallocality. Both arerectangular, well preserved. Its surface displayed numerous small, multiple-roomstructures erected over low platforms,and irregular channels and rills, including a localized arroyo they comparefavorably to the most elegant structuresof (FIG. 5), dating from when the clay horizon had lain ex- the Canal locality by virtue of their plaster floors and posed. In addition to these obvious erosional features, walls, their apparentspaciousness, and use of tabular several large pits, the largest over 6 m. in diameter, were small-stoneveneer as decorativefacing or lath for plas- encountered (FIGS. 4, 5). These were filled with the same ter. Unfortunately,both had sustainedconsiderable dam- dark, artifact-richsoil that overlay them. Throughoutthis age, StructureII fromchisel plowing(FIG. 4) andStructure soil, the densities of artifacts, particularlyobsidian, were I from later buildingactivity that destroyedits southern extremely high, in some cases averaging 60 pieces of margin.The two structureswere separatedby a narrow, obsidian per liter of earth removed. The majority of the open corridorwith a preparedclay floor that contained artifacts recovered in our excavation came from these an open-troughdrain. The clay floor extendedbeneath pits. Because of the artifact density here, we screened both structuresand into an open area to the east that levels from selected pits and squares, using quarter-inch probably comprised an open courtyardfor the com- screen, and, as elsewhere in the excavation, soil samples pound. StructureIII is of less elegantconstruction, con- were also taken for microscopic study. sisting of stone-and-mudwalls enclosing a compacted The pits were U-shaped or somewhat bell-shaped in earthfloor without an underlyingplatform. It probably cross section (FIG. 5), and measured a meter or more in predatesStructures I and II, given its lower elevation depth. Their internal stratification was a complex pattern and stratigraphiccontext (it was built directly on the of localized bands, lenses, and amorphous pockets of underlyingclay horizon), but was probablypart of an artifact-richsoil and yellow clay. Artifact concentrations earliercomplex that included previous versions of Struc- were highest in the lenses and pockets, which probably tures I and II, remains of which were found beneath comprise individual loads of refuse. This consisted pre- them. dominantly of obsidian, chiefly debitage from core/blade As yet no study has been made of the artifactsre- reduction (see below), but also large quantities of utili- covered from the floors of these three structuresor the tarian sherds, animal bone, and other domestic refuse. remainsof those thatoverlay them, but they were clearly It is importantto note that workshop and domestic refuse domesticin function,given the occurrenceof whole and occurred mixed in the same deposits, and while densities fragmentaryutilitarian vessels, , and hearthsin of obsidian were high, there were no pure layers of ob- one or more of them. Despite ratherlarge quantitiesof sidian comparable to waste piles seen at quarry or biface obsidiandebitage recovered from fill incorporatedinto workshops. 18. Dan M. Healan, "ArchitecturalImplications of Daily Life in The regularity of the pits suggests they were artificial, AncientTollan, Hidalgo, Mexico," WA9 (1977) 140-156. perhaps source pits for clay used in construction which Journal of Field ArchaeologylVol. 10, 1983 135

Figure 8. Panorama of the habitation area along ridge A, looking NE. Structure II is seen in left background (note chisel scars); Structure I is in foreground, and Structure III is at the extreme right. Chisel scars can be seen in many of the profiles.

were laterused to discardrefuse. It seems unlikely that x 3 m. squares(FIG. 4), but it was clearly an open area obsidian was worked at these dumps, though some of devoid of architecture.The underlyingclay horizondis- the bandeddeposits may representlocalized work sur- played the same small erosional features noted else- faces. More likely, the work areas were located else- where, but lackedlarge refuse pits, with one exception: where, probablynear the habitations,given the fact that a shallow, apparentlycircular pit, filled with obsidian domesticand workshoprefuse occur mixed in the same and otherartifact refuse, was foundprotruding from be- strata.During preliminary stratigraphic analysis, Healan neaththe northside of StructureII (FIGS. 4-6) along the notedthat some of the depositshad been disturbed(other south edge of this area. Aside from this pit, only mod- than by chisel plowing), and may representattempts to erate quantities of macroscopicobsidian debris were retrievewaste obsidianfor re-use. recoveredfrom this area; Healan, however, noted sig- nificantamounts of obsidianmicrodebitage in soil sam- ples from the top of the clay horizonand the overlying Obsidisn WorkAress soil in this area. The existence of the refuse pits dis- Where were the actual obsidian work areas located? cussedabove indicates that the workareas were regularly On the basis of preliminarymicroscopic examination of cleanedof debitageaccumulations, so that an areawith soil samples, Healanbelieves that they were located in moderatevisible debitage and disproportionatelylarge the area between the habitationsand the refuse dump. amountsof microdebitagewould very likely represent This area was only partiallyexposed by alternate3 m. the actual work area. Confirmationof this hypothesis 136 Execavcltionof canObsidism Workshopin Tuls, MexicolHecalcm,Kerley, cmd Bey awaits more systematicexamination of all of the floor ObsidianAnalysis and soil samplesrecovered from the excavation. GeneralComments With most of the tabulationcomplete, -we can report Discussion that we recoveredover 650 kg. of obsidiannumbering All three of these general activity areas maintained over 500,000 pieces, including over 375,000 macro- functionalintegrity throughout the prehispanicoccupa- scopic pieces. Kerley is conductinga two-stage tech- tion. No structuralremains were found outside the hab- nological analysis of the obsidian recoveredfrom our itation area, and continuous occupation in one area excavations.The first stage, now near completion,has contributedto the ridge buildup. Refuse dumpingwas involved a sorting of the entire collection into a series confined to the interridgerefuse dump, except for use of technologicalcategones (TAsLE<2);2°the second stage as structuralfill, and once the pits were filled, continued will involve more detailedstudy throughattribute anal- dumpingcovered the interridgearea with the artifact- ysis of a sample of each category. The categorieswere rich soil seen today. Neitherthe habitationarea nor the denved from previous studies of Mesoamericancorel refuse dump encroachedupon the interveningarea we blade technology,2lplus data obtainedduring Healans have tentativelyidentified as the obsidianwork area. The preliminaxysurface reconnaissance of the localityin 1979. integrityand continuityof activities in these areas indi- Inevitably,new categones were createdand some exist- cate a relativelyformal organizationthat was followed ing ones subdividedor modified in the course of the throughoutthe prehispanicoccupation. preliminarysorting,22 the results of which provide the The presence of residentialstructures and the mixed following generalizations. workshopand domesticrefuse depositsdemonstrate that obsidianworking took place in a domestic setting, as is 20. It should be noted that Table 2 is a tentativetabulation, since probably typical of preindustrialsocieties. We noted about 10%of the obsidianfrom our excavationshas not been sorted. above thatthe structurespartially exposed along ridge A We do not, however,expect the relativefrequencies of the categories are similar in layout to those excavated at the Canal in Table 2 to changesignificantly once the entirecollection has been locality by the Missouriproject, and in fact the degree analyzed We estimatethat the totalcollection will compriseover 700 kg. of obsidiannumbering over 560,000 pieces Some of the cate- of similarityis striking.The structuresof the Canal lo- gories listed in Table 2 have been modifiedfrom those used in the cality comprisedthree residential compounds, including actualsorting. a small temple platform,arranged side by side and con- 21. ThomasHester, Robert Jack, and RobertHeizer, '4TheObsidian nected by a public street as well as a closed system of of Tres Zapotes, Vera Cruz, Mexico,' CARFC13 (1971) 65- 131; passagewaysand alleys.l9 It is interestingto note that Payson Sheets, "BehavioralAnalysis and the Structureof a Prehis- these collapsed compoundsformed a low ridge similar toric Industry>'CA 16 (1975) 369-378; idem, "Part 1: Artifacts," to ridge A, though lower, and like ridge A, remainsof in The Prehisotryof Chalchuapa,El Salvador, Volume2, Payson earlierstructures were foundbeneath the excavatedcom- Sheets and Bruce Dahlin, eds. (Pennsylvania1978) 2-131; Thomas Hester, "The ObsidianIndustry of Beleh (ChinautlaViejo), Guate- pounds, which contributedto the ridge effect. Healanis mala," Actas del XLICongreso Internacional de Americanistas,Vol- certainthat ridge A constitutesa similarlinear arrange- umenI (MexicoD.F. 1975)473-488; Jay Johnson,"Chippped Stone mentof residentialcompounds, for severalreasons: scat- Artifactsfrom the WesternMaya Periphery," unpublished Ph.D. dis- teredarchitectural remains were encounteredin two test sertation,Southern Illinois University(Carbondale 1976). pits near the east end of the ridge, and we know that a 22. Both Healanand Kerley have had experiencein flintknapping, prominentmound, probablya temple platform,existed but we would not be overly modest in characterizingourselves as near the west end of ridge A before it was destroyed sophisticatednovices. Nevertheless,our understandingof the work- shop reductiontechnology has been greatly enhancedby our own abouta decadeago. Given the lengthof ridge A, at least replicativeexperiments in blade/corereduction, which are continuing. two juxtaposedcompounds plus the templeplatform had It is importantto note that this work has involved the fabricationof probablyexisted, and the paved surface that overlay prismaticblade cores "from scratch";that is, forming percussion StructureIII may have been part of a thoroughfarethat cores from raw obsidiannodules, which are then furtherreduced by connectedthem. pressureto becomeprismatic blade cores. Much of the previousrep- licativeexperimentation concerning prismatic blade core reduction has We are also certainthat ridge B, locatedimmediately involved the use of sawn blocks of obsidianrather than percussion northof ridge A (FIG. 3), was comparablein form, since cores for pressurereduction (cf. Crabtree,op. cit. [in note 13] and architecturalremains erected over artifact-richstructural Sheetsand Muto, op. cit. [in note 8]), whichdoes not yield the kinds fill were encounteredin two pits placed along this ridge, of debitagetypical of ourworkshop assemblage. John Clark first dem- and its surface was likewise littered with architectural onstratedto us the techniquesof fabricatingprismatic blade cores from scratch,and we gratefullyacknowledge the adviceand formal instruc- stone. tion in reducingpercussion macrocores by pressuregiven us by J. JeffreyPlenniken, director of the WashingtonState UniversityPlint- 19. Ibid. fig. 3. knappingField School, and Gene Titmus. Journal of Field ArchaeologylVol. 10, 1983 137

Table 2. Preliminarytabulation of obsidianartifacts from the workshopcomplex. Percentagesrefer only to the macroscopicartifact categories (A-W).

A. Macroblades 35 (0.01%) K. PrismaticBlade Cores 3,102 (0.83So) B . PlatformFaceting Flakes 56,988 ( l 5.19Wo) 1. whole 473 C. Percussion Blades 11 ,990 (3.19Wo) 2. frag. 2,629 1. whole 1,192 a. proximal 947 2. frag. 10,798 b. medial 765 D. IrregularPressure Blades 203,647 (54.26Wo) c . distal 917 1. primarydecortication 17 L. Miscellaneous Core Fragments 2,770 (0.74Wo) 2. secondarydecortication 6,661 1. rim fragments 312 3. first series- 10,401 2. distal flakes 974 a. whole 1,972 3. tablets- 318 b. frag. 8,425 4. miscl. fragments 1,166 4. general 186,568 M. Percussion Cores 177 (O.O5Wo) a. whole 2,824 1. whole 1 b. frag. 183,744 2. frag. 176 E. PrismaticBlades 45,701 (12.18io) N. Bifaces 35 (O.OlSo) 1. whole 128 1. whole 3 2. frag. 45,573 2. frag. 32 F. PrismaticBlade Errors 1,204 (0.32%) O. on Large Blades or Flakes 134 (0.04Wo) 1. plunging blades 983 P. AlternateFlakes 40 (0.01Wo) 2. bending fractures 190 Q. Thinning Flakes 3,272 (0.87Wo) 3. miscl. 31 R. PrimaryDecortication Flakes 47 (0.01Wo) G. PrimaryCrested Blades 1,155 (0.31Wo) S. Secondary DecorticationFlakes 2,493 (0.66No) H. SecondaryCrested Blades 515 (0.14Wo) T. EraillureFlakes 374 (O.l OSo) I. Blade Products 522 (0.14So) U. Unclassified Flakes with Platforms 34,628 (9.23Wo) 1. trilobal eccentrics 446 V . Macroflakes 14 (0.01So ) a. whole 109 W. Chunks 840 (0.22Wo) b. frag. 337 2. unifacial retouchedblades ("endscrapers") 18 TOTAL MACROSCOPIC SPECIMENS 375,288 3. miscl. 58 (lOOSo) J. Blade ProductDebitage 5,605 (1.49Wo) 1. unilateralnotched blade segments 5,363 X. Shatter 125,746 2. bilateralnotched blade segments 133 3. lateral flaked blade segments 109 TOTAL SPECIMENS 501,034

Source was handleddifferently in the workshop about which About 83% of the collection is green in color and more is said below. undoubtedly comes from the Cruz del Milagro (Pachuca) Input, Reduction Technology, Output source, given its proximity (FIG. 1) and its prior identi- fication at Tula.23 The remainder is predominantly grey About 75% of the tabulatedobsidian, or 375,288 and probably comes from a number of sources, including pieces, consistsof macroscopicpieces thatconstitute cat- Zinapecuaro, Michoacan (a major source of the grey ob- egoriesA-W in Table2; the remaindercomprise a single sidian recovered by the Missouri project), and Otumba, categorydesignated shatter, whichconsists of smallflakes Mexico (FIG. 1).24 Trace-element analysis will be under- andfragments that were mostlyrecovered from screened taken in the near future. Kerley performed separate tab- levels. We intendto examinethe shattercategory in more ulations of green and non-green obsidian in order to detail during the second stage of the analysis. In the determine if obsidian from Pachuca versus other sources following discussion, categories A-W will be collec- tively referredto as the macroscopic collection, to which 23. Thomas Hester, Robert Jack, and Alice Benfer, "Trace Element most of our remarkswill be directed. Because only a Analysis of Obsidian from Michoacan, Mexico," CARFC 18 (1973) smalland nonrandom part of the excavationwas screened, 167-176. we have omittedthe shattercategory from the percentage 24. Ibid. calculations. 138 Excavcationof an ObsidianWorkshop in Tula,MexicolHealan, Kerley, and Bey

Approximately 89% of the macroscopic collection be- TABLE 2). The remainder, which comprise about 54% of longs to categories that are clearly derived from core/ the macroscopic collection, are collectively called irreg- blade reduction (A-M, TABLE 2). Though the remainder ularpressure blades (D, TABLE 2). Besides their irregular include some flake categories that could have derived form, remnant percussion blade scars are frequently seen from biface manufacture, not a single unfinished biface over all or part of their dorsal surface. Our own expe- reject was recovered; the few whole and fragmentary rience in blade-core replication indicates that most of bifaces found were all finished products. Hence we are these comprise the first removals of pressure blades from almost certainly dealing with an exclusively core/blade a macrocore. During this process the macrocore is trans- workshop. formed from a generally conical form with wide, irreg- The virtual absence of macrodebitage associated with ular blade scars or flutes to a commonly bullet-shaped "roughing out" cores from raw obsidian indicates that form with narrow, parallel flutes that ensure subsequent some type of prefabricated core or preform constituted removal of typical prismatic blades. Given their predom- the workshop input. Our data indicate that these were inance in the collection, irregular pressure blades must percussion macrocores(M, TABLE 2),25 or large poly- have been considered waste and not utilized, though the hedral cores, usually with a conical or pyramidal shape, latter assumption awaits confirmation through wear anal- bearing percussion blade scars on their face. ysis. Within this category is a distinctive subcategory, A distinctive core preparationactivity at the workshop first seriespressure blades (D3, TABLE 2), which represent involved percussion flaking of the macrocore to produce the very first blade removals by pressure from percussion a multifaceted platform. The resulting plaMormfaceting macrocores. While Healan had suspected the existence flakes(B, TABLE 2) constitute about 15% of the macro- of such a subcategory on the basis of fragmentarysurface scopic collection. These flakes are easily distinguishable material examined during preliminary surface reconnais- by their perpendicular platform representing a remnant sance of the locality in 1979, it was first defined and blade scar on the macrocore face; their flat, thin cross named by John E. Clark of the New World Archaeolog- section; and the radial pattem of remnant flake scars on ical Foundation using data from replicative experiments their dorsal surface. Their high frequency in the collec- and analysis of core/blade debitage from Chiapas. Clark, tion is surprising, considering that cores and blades with who generously made his data available to us during multifacetedplatforms are extemely rare at Tula. Healan visits to our laboratory in 1980, defined this subcategory and Kerley believe that platform faceting was a prelim- on the basis of distinctive attributes of length, shape, inary step in producing the ground platform that typifies dorsal scars, and mode of termination. Similar blades the prismatic core/blade industry at Tula. Faceting and have been produced in our own replicative work. The subsequentgrinding of the macrocore platform must have first pressure blades do not run the full length of the constituted the first steps of the workshop reduction se- core, hence percussion blade scars will remain on the quence, since little core-derived debitage with other than distal portion of the core until later rinds of pressure ground platforms was recovered. blades are removed. These later pressure blades, there- Aside from platform faceting flakes, percussion deb- fore, possess remnant percussion blade scars at their dis- itage is in strikingly low frequency26in the macroscopic tal end, providing a means of further subdividing the collection, which implies that virtually all subsequent irregular pressure blade category according to order of core reduction involved the use of pressure rather than removal. percussion. This is not to say, however, that prismatic A large number of whole and fragmentary prismatic blade removal followed immediately. Of the 249,348 blade cores were recovered (K, TABLE 2). Most were ex- pressure blades so far tabulated in the collection, only hausted or in an advanced stage of reduction, and closer about 18% possess the distinctive parallel margins and examination will probably reveal the reasons for dis- dorsal ridges characteristic of typical prismatic blades (E, carding the unexhausted cores. We also identified sev- eral categories of core fragments which probably include 25. Thomas Hester, "Notes on Large Blade Cores and Core/Blade pieces derived from error recovery or core rejuvenation Technology in Mesoamerica," CARFC13 (1972) 9S-105; John Clark, (L, TABLE 2). It is likely that the category for irregular "A Macrocore in the Regional Museum in Tuxtla Guiterrez, Chiapas, pressureblades also contains specimens of errorrecovery. Mexico,'' LT 6 (1977) 30-32; Robin Torrence, "Macrocore Produc- Counting whole and proximal (platform) fragments, a tion at the Melos Obsidian Quarries," LT 8 (1979) S1-60. minimum of 1,420 individual prismatic blade cores are 26. The majority of the percussion blades in category C of Table 2 represented. Undoubtedly, additional cores are repre- are from levels examined at the very beginning of the sorting opera- tion. Kerley now believes that many of these so-called "percussion sented by the core fragments in category L of Table 2, blades" are in fact misidentified irregular pressure blades, and these perhaps as many as 300 or more. Hence we estimate that levels will be re-examined to determine whether or not they are. as many as 1,720, but not less than 1,420, prismatic Journal of Field ArchaeologylVol. 10, 1983 139 blade cores are represented in that part of the collection Table 3. Estimatednumber of individualpressure blades in that has been analyzed to date. the collectionand associatedblade:core ratios. About 2% of the macroscopic collection pertains to Estimated the manufacture of products from prismatic blades (I, J, mlnlmumand Estimated TABLE 2). Only two kinds of blade products and associ- maximumnumber mlnlmumand ated debitage occur with sufficient frequency to indicate of individual maximumnumber they were regularly manufactured at the workshop: tri- Category blades: of blades per core: lobal eccentrics and unifacially retouched blades, or D. Irregular ''endscrapers on blades". Trilobal eccentrics are small, PressureBlades ''3"-shaped forms common at Tula and Teotihuacan;27 1. primary the trilobal element is a common motif in bIesoamerican decortication 5-9 iconography believed to be part of Tlaloc symbolism 2. secondary decortication 1903-3331 1-2 representing drops of water, blood, or other fluid, but 3. first series 4379-6185 2-4 the function of these obsidian facsimilies is not known. 4. general 55322-94696 32-55 We recovered a sufficient number of rejects to permit a TOTAL 61609-104212 35-61 tentative reconstruction of the manufacturing sequence, E. Prlsmatlc which essentially involves pressure flaking of notched Blades 13149-22915 7-13 prismatic blade segments. It should be noted that the blade product debitage categories consist entirely of re- jects, which would account for only a small proportion of debitage from the production of blades; undoubtedly, inabilitysystematically to distinguishthese blades from the majority consists of very small pressure flakes that generalirregular pressure blades. Takenas a whole, the would escape recovery even in screened levels (though estimateof 35-61 irregularpressure blades per core seems some might be identified during microdebitage analysis to fall within the expected range based upon our own of soil samples). It is likely, therefore, that the magni- experience,though Healan believes that the lower end tude of the manufacture of blades in the workshop is far of thisrange is moreappropriate for a skilledblade maker. greater than the macroscopic collection would suggest. On the other hand, a strikingdiscrepancy exists in the In Table 3 we have attempted to estimate the total extremelylow ratio of prismaticblades to cores of 13:1 number of individual blades represented by the whole at most;even if the moreconservative estimation of 1,420 and fragmentary specimens.28 These estimates were then cores were used, this ratio would only increase to 16 divided by the estimated number of prismatic blade cores bladesper core at most. Ourown experienceand thatof in the collection (1,720) to derive a blade:core ratio for othersindicates that an averageprismatic blade core can each category. Assuming the collection is a reptesenta- produce 100 or more prismaticblades; hence at least tive sample of the workshop assemblage, if the workshop 85% of the expectedprismatic blade yield of these dis- were consuming all of its own output (a null hypothesis), cardedcores is missing, and presumablyleft the work- we would expect the blade:core ratio for each blade cat- shop as unmodifiedblades or blade products. egory to be consistent with ratios obtained during core We notedabove thatabout 83% of the obsidianin the reduction. The extremely low blade:core ratios for de- collection is green, presumablyfrom Pachuca, and in cortication blades probably reflect the degree of refine- fact green obsidian was predominantin all but one of ment of macrocores entering the workshop (but see the the categorieslisted in Table 2. There is considerable discussion of source differences below). The equally low variation,however, in the proportionof green obsidian ratio for first series pressure blades probably reflects an across the categories which ranges from 93% to 47% (TABLE 4); 27. Terrance Stocker and Michael Spence, "Trilobal Eccentrics at furthermoreas Table4 demonstrates,this var- Teotihuacan and Tula," AmAnt 38 (1973) 195-199. iation is clearly patterned.Categories with the highest proportionsof green obsidianare those that derive from 28. Two differrent methods were used to estimate the number of individual blades represented by the fragmentary specimens. In Table the final stages of the reductionsequence in the work- 3 the maximum estimates were obtained simply by assuming that at shop- the productionof prismatic blades and blade the least, two blade fragments equal one whole blade; this is almost products- and cluster around90%. On the otherhand, certainly too liberal an estimate, since many blades probably broke the lowest proportionsof green obsidianare those cate- into more than two macroscopic pieces. The second method, which gories thatderive from the initial stages of the reduction produced the minimum estimates in Table 3, involved calculating the average weight of whole blades for each category and dividing the sequence-macrocore trimmingand platformprepara- total weight of the fragmentary specimens by that figure; this worked tion. What,then, is the trueproportion of greenobsidian out to about 3.5 fragments per whole blade. processedat the workshop,and why is there such vari- 140 Excavationof an ObsidianWorkshop in Tula, MexicolHealan,Kerley, and Bey

Table4. Percentageof green obsidianamong the artifactcategories in Table 2, rankedfrom highestto lowest. Asteriskindicates categories in which whole and fragmentaryspecimens have been merged.

Percent Category Non-green Green green F3. Miscl. prismaticblade errors 2 29 93.5 I3. Miscl. blade products 4 54 93.1 I1 . Trilobal eccentrics* 34 412 92.4 L2. Distal core flakes 78 896 92.0 K1 . Whole prismaticblade cores 40 433 91.5 K2c. Distal prismaticblade core frags. 78 839 91.5 J1 . Unilateralnotched blade segments 487 4876 90.9 F2. Prismaticblade bending fractures 18 172 90.5 J2. Bilateral notched blade segments 13 120 90.2 E. Prismaticblades* 4507 41194 90.1 K2b. Medial prismaticblade core frags. 80 685 89.5 L4. Miscl. core frags. 122 1044 89.5 K2a. Proximal prismatic blade core frags. 100 847 89.4 F1 . Plunging prismatic blades 105 878 89.3 J3. Lateralflaked blade segments 12 97 89.0 I2. Unifacial retouchedblades 2 16 88.9 A. Macroblades 4 31 88.6 L3. Core tablets 41 277 87.1 C. Percussion blades* 1604 10386 86.6 O. Unifaces on large blades or flakes 20 114 85.1 D4. General irregularpressure blades* 28942 157626 84.5 X. Shatter 21055 104691 83.3 M2. Percussion core frags. 30 146 83 0 P. Alternateflakes 7 33 82.5 G. Primarycrested blades 205 949 82.2 U. Unclassified flakes wlplatforms 6719 27909 80.6 H. Secondary crested blades 101 414 80.4 D3. First series pressureblades* 2076 8325 80.0 T. Eraillureflakes 80 294 78.6 V. Macroflakes 3 11 78.6 Q- Thinning flakes 730 2542 77.7 W. Chunks 198 642 76.4 L1 . Core rim frags. 78 234 75.0 N. Bifaces* 9 26 74.3 B. Platformfaceting flakes 14695 42293 74.2 S. Secondary decorticationflakes 683 1810 72.6 D2. Secondary decorticationblades 1912 4749 71.3 R. Primarydecortication flakes 15 32 68.1 D1 . Primarydecoitication blades 9 8 47.1

ation in this proportionbetween categories?Given the a count of the number of pieces of debitage per core by coherentpattern of variationindicated and the size of color for each category. This is not intended to represent our collection, it is unlikely that this variationis due an estimation of number of individuals per core for each simply to chance. category, as was done for pressure blades in Table 3, In Table 5, the frequenciesof green and non-green simply because we are at present unable to estimate num- obsidianfor each categorylisted in Table 4 have been ber of individuals for many of the fragmentarynon-blade dividedby our minimalestimation of the numberof green categories. Rather, these counts of debitage per core are and non-greenprismatic blade cores representedin the intended to provide a means of comparing the amount collection (140 and 1,280 cores, respectively)to derive of debitage generated by green versus non-green core Journal of Field ArchaeologylVol. 10, 1983 141

Table5. Numberof pieces per core of green and non-greenobsidian among the core/bladeartifact categoriesof Table 4, rankedby per-corepredominance of non-greenobsidian from least to most.

Numberof piecesper Similarity prismaticblade core ratio (non- Category Non-green Green green:green) F3. Miscl. prismatic blade errors 0.01 0.02 0.6:1 I3. Miscl. blade products 0.03 0.04 0.7:1 I1 . Trilobal eccentrics 0.24 0.32 0.8:1 L2. Distal core flakes 0.56 0.70 0.8:1 J1 . Unilateralnotched blade segments 3.48 3.81 0.9:1 F2. Prismaticblade bending fractures 0.13 0.13 1.0:1 J2. Bilateral notched blade segments 0.09 0.09 1.0:1 E. Prismaticblades 32.19 32.18 1.0:1 L4. Miscl. core frags. 0.87 0.82 1.1:1 F1 . Plunging prismatic blades 0.75 0.69 1.1:1 J3. Lateralflaked blade segments 0.09 0.08 1.1:1 I2. Unifacial retouchedblades 0.01 0.01 1.1:1 A. Macroblades 0.03 0.02 1.2:1 L3. Core tablets 0.29 0.22 1.4:1 C. Percussion blades 11.46 8.11 1.4:1 D4. General irregularpressure blades 206.73 123.14 1.7:1 X. Shatter 150.39 81.79 1.8:1 M2. Percussion core frags. 0.21 0.11 1.9:1 P. Alternateflakes 0.05 0.03 1.9:1 G. Primarycrested blades 1.46 0.74 2.0:1 U. Unclassified flakes w/platforms 47.99 21.80 2.2:1 H. Secondary crested blades 0.72 0.32 2.2:1 D3. First series pressure blades 14.83 6.50 2.3:1 T. Eraillureflakes 0.57 0.23 2.5:1 V. Macroflakes 0.02 0.01 2.5:1 Q Thinning flakes 5.21 1.99 2.6:1 W. Chunks 1.41 0.50 2.8:1 L1 . Core rim frags. 0.56 0.18 3.0:1 B. Platformfaceting flakes 104.96 33.04 3.2:1 S. Secondary decorticationflakes 4.88 1.41 3.4:1 D2. Secondarydecortication blades 13.66 3.71 3.7:1 R. Primarydecortication flakes 0.11 0.02 4.3:1 D1 . Primarydecortication blades 0.06 0.01 10.3:1

reduction for each category, on the assumption that the What is immediately apparent in Table S is that the relative proportions of discarded green and non-green variation in the proportion of green obsidian is in fact prismatic blade cores in the collection approximate the caused by a disproportionately greater amount of non- true proportions of green and non-green cores that were green obsidian in all debitage categories except those that reduced at the workshop. We stress that these are mean- pertain to the final stages of the workshop reduction se- ingless quantities, except for comparison of green and quence. For most of these categories of "final stages", non-green debitage counts per core within the same cat- the quantity of debitage per core for non-green and green egory. This comparison is aided in Table S by the cal- culation of a ratio expressing the degree of similarity the absolute values of the piece per core counts themselves, but rather between the counts.29 the degree of similarity between the green and non-green counts for each category that is of interest here, in fact, any one or combination of the core categories (K and L) in Table 4, except for L1, could have 29. Our estimations of 140 non-green and 1,280 green prismatic blade been used as the denominator and would have produced counts having cores were obtained by counting the number of whole and proximal the same similarity ratios as those in Table 5. This is because all of fragments of prismatic blade cores of each color in Table 4; this is, these categories have about the same proportion of green obsidian of course, our minimal estimation of 1,420 cores. Given that it is not (roughly 90%), as noted in Table 4. 142 Excavationof an ObsidianWorkshop in Tula,MexicolHealan, Kerley, and Bey

obsidian is nearly identical, and in fact is perfectly so in of the bone implements, including fragments of rods and the case of prismatic blades. This would suggest, per- points, are slightly burned and polished, perhaps a hard- haps not surprisingly, that the average green and non- ening technique. Some of these points may have been green cores produce the same number of prismaticblades. used in the removal of blades by pressure, but this is at On the other hand, for the earlier stages of the workshop present only speculation. Stone implements include small reduction sequence, particularly the earliest stages, there and abraiders, but it is not yet certain that is a considerably greater quantity of non-green obsidian any of these were used to work obsidian. We hope that per core than green obsidian. This circumstance suggests microscopic examination for wear and residue will be of that macrocores arriving from sources other than Pa- value in making this determination. chuca required more preparatory reduction to become "ready" prismaticblade cores. If so, the quantities could Ceramic Chronology reflect differences in the inherent workability of different Bey has recently completed a typological classification types of obsidian, or perhaps lax production standards at of the from the excavations that provides a gen- quarry workshops other than Pachuca; Healan suggests eral chronological framework, pending submission of ra- it could as well reflect differences in the geological oc- diocarbon and obsidian hydration specimens for dating currence of obsidian at these sources, perhaps the fab- in the near future. Though tabulation is not complete, rication of macrocores from nodules instead of blocky over 133,000 sherds were recovered from our excava- flow fragments, the latter typical of raw material seen at tions. Bey's classification is based upon Cobean's ex- the Pachuca quarries today. In general, we believe that tensive study3l of the pre-Aztec pottery of Tula that was the proportion of green obsidian in the "final stages" based upon the Missouri project's residential excavations categories, roughly 90%,3° represents the true proportion and the urban survey, and selected test pits of the INAH of green obsidian entering the workshop, and that the project from near Tula Chico (FIG. 2). disproportionately greater amount of initial preparation Bey's classification places the entire workshop occu- of non-green macrocores accounts for the lower overall pation within the Tollan phase (TABLE 1): less than 200 proportion of green obsidian in the collection. Aztec sherds and less than 50 sherds of the Corral and We noted that at least 1,420 prismatic blade cores and earlier phases have so far been tabulated. During the concomitant blades and blade products were produced at classification process, however, Bey noted a rather high the workshop. This is obviously nowhere near the total frequency of Mazapa Red on Brown or "Wavy Line output, since only part of the workshop was excavated, Mazapan", a type previously found only rarely at Tula32 but we are not prepared to estimate total output at this and assigned by Cobean to the preceding Terminal Cor- time. It is important to note that only a small portion of ral phase (TABLE 1). In a trial seriation of pottery from a single complex was exposed; indeed, approximately selected levels of our excavation, Bey noted that Mazapa 55% of the 375,288 pieces that constitute the macro- sherds tend to occur in greatest frequency in lower strati- scopic collection came out of onlyfive 3-m. squares and graphic levels, but always as an integral part of an oth- two l-m. pits in the refuse dump! Given that this figure erwise essentially Tollan phase assemblage. In comparing constitutes a very small proportion of a refuse dump that these atypical levels with those having a more typical spans the entire length of the area between ridges A and Tollan phase assemblage, Bey noted two other differ- B, total workshop output would undoubtedly number in ences: 1) the "atypical'' levels had unusually low fre- the tens of thousands for prismatic blade cores alone. A quencies of the most common and diagnostic Tollan phase problem in defining the output of one workshop lies in type, JarraPolished Orange or "Naranja a Brochazos"; distinguishing the debitage boundaries of one workshop, 2) subtle but consistent modal differences occurred be- since only the habitation compounds would define dis- tween sherds of the same type from "atypical" and crete entities, and the refuse dump may have been used "typical" levels, most notably differences in surface fin- by workshops along both ridges. ish of certain cream-slipped wares. A number of as yet unstudied implements of bone, antler, and stone were recovered, some of which were 31. Cobean, op. cit. (in note 1 1). almost certainly used in obsidian working. We recovered 32. Ironically, Mazapa Red on Brown is commonly thought to be the numerous antler tines, for example, several of which diagnostic pottery type of the Early Postclassic period at Tula, hence have minute flakes of obsidian embedded in them. Some the frequent reference to Toltec Tula as ''Mazapan" or a "Mazapan'' period or phase at Tula. Cobean notes that this potteIy is indeed common in Early Postclassic and Terminal Classic sites in the Basin 30. This agrees with Spence's figures for Teotihuacan, as he notes of Mexico, but at Tula, the dominant Early Postclassic pottery consists that the frequency of green obsidian in core/blade workshops there of orange and cream, rather than red on brown, wares. Cf. Cobean, was "rarely less than 90%"; cf. Spence, op. cit. (in note 5) 777. op. cit. (in note 11) 396-399. Journal of Field ArchaeologylVol. 10, 1983 143

Bey hypothesized that these quantitative and qualita- ported raw obsidian obviously had considerable latitude tive differences are diachronic, and provide a means to in their reduction strategies; even in core/blade reduc- subdivide the Tollan phase into early and late subphases. tion, some of the macrodebitage derived from initial The ceramics of the "typical" levels represent the late stages of core formation make suitable blanks for bifacial or fully developed Tollan-phase ceramic assemblage that reduction.33Much of the debitage from initial core for- dominates the archaeological record at the site and rep- mation, however, is waste: decortication, irregularities, resents Tula's apogee and period of maximum expan- flaws, and the like, and the transportationof raw obsidi- sion. Conversely, the ceramics of the "atypical" levels an over long distances, as in the case of Tula, would be represent an early subphase marked by the initial ap- less efficient than performing these initial reduction steps pearance of the Tollan-phase ceramic assemblage along- at the quarries. side a continuation of the ceramic tradition of the The workshop was involved in the full range of core/ preceding Terminal Corral phase. Given the rarity of blade reduction activities except for the initial stages of Mazapa sherds at Tula, the Early Tollan subphase must core formation, and their absence has helped clarify sub- predate Tula's urban expansion. Bey has found support- sequent stages of core preparationprior to prismaticblade ing evidence for an early/late subdivision of the Tollan removal. Aside from platform preparation, virtually all phase in a study of ceramics from nine rural sites of the core reduction at our workshop involved the use of pres- Tollan phase in the Tula region, one of which displays sure, not percussion, but this should not be taken to mean a ceramic assemblage comparable to the "atypical" lev- that the imported macrocores were virtually ready for els of our excavation; i.e., the presence of significant prismatic blade removal. In our own experience with numbers of types of the Terminal Corral phase in an making prismatic blade cores from scratch, percussion otherwise Tollan-phase assemblage in which Jarra Po- was necessary to shape the raw material, which normally lished Orange is significantly low in frequency. includes establishing a platform, establishing the first The subdivision of the 250-year-long Tollan phase is blade scars or flutes on the core face, and removing an exciting prospect, for it would provide finer chrono- troublesomeirregularities or anomalies from the core face. logical control over the most spectacular period of Tula's Beyond this point, further percussion trimming accom- history. This would have importantramifications for our plishes little except the waste of potentialprismatic blades; workshop site, since it would therefore have been settled the shift, therefore, from percussion to pressure should at the beginning of the Tollan phase prior to the city's occur as early as possible to maximize prismatic blade expansion, and was probably continuously occupied for yield. Percussion is not used to establish the straight, the remainderof the life of the Toltec city. It is important parallel ridges on the core face that make prismatic blade to note that the ceramic sequence at the workshop so far removal possible; this is done entirely by pressure, as shows continuous rather than discontinuous change from evidenced by the predominance of irregular pressure the early to late subphases, which is also suggested by blades in our collection that are derived from this pro- the stratit,raphicdata discussed above. The presence of cess. We believe that the significance of irregular pres- Mazapa Red on Brown in significant quantities may have sure blades as defining a stage of core/blade reduction further ramifications, as discussed below. (rather than simply representing "duds" or substandard blades produced during prismatic blade removal) has not previously been appreciated, and we caution that the ab- Preliminary Assessment sence of percussion blades at a workshop site does not We must stress the tentative nature of our findings, necessarily mean that "ready" prismaticblade cores were based as they are upon incomplete analysis. A number being imported. of significant points have been raised, however, that have John Clark34has argued that the study of Mesoamer- important ramifications for other research. In conclu- ican obsidian technology must consider source differ- sion, we want to discuss the implications of our prelim- ences as an agent of technological variability, and our inary findings for the study of Mesoamerican core/blade data supportthis. Our data demonstrate that considerably technology and the study of obsidian workshops at Tula more preparationof non-green than green cores was nec- and elsewhere. essary, which included decortication, platform faceting, The exclusively core/blade industry at our workshop and preliminary (irregular) blade removal by pressure. is in part a reflection of the dominance of prismaticblades and blade products among obsidian artifact assemblages 33. Sheets, 1975 op. cit. (in note 21) fig. 3. in Mesoamerica, but is also a function of the relatively 34. JohnClark, "A SpecializedObsidian Quarry at Otumba,Mexico: specialized item or workshop input, percussion macro- Implicationsfor the Studyof MesoamericanTechnology and Trade," cores, rather than raw obsidian. Workshops that im- LT 8 ( 1979) 46-49. 144 Excavationof an ObsidianWorkshop in Tula,MexicolHealan, Kerley, and Bey

We have not yet establishedwhether this was because pattern of topographic rises and high surface concentra- of differencesin the source materialor in activities of tions of obsidian and other artifacts that probably define quarryworkshops at the differentsources. a continuous zone of workshop complexes like that which We stressthat our appraisalof the workshopindustry we partially exposed. Survey data from the Missouri and is basedupon a preliminarycategorization, and that more INAH projects revealed three other characteristics com- detailedattribute analysis of these categoriesis now un- mon to the "east flank" concentration, all of which were derway.We anticipatethat the resultof this second anal- observed in our excavation: 1) a predominance of core/ ysis will greatlyexpand our knowledgeof variousparts blade debitage; 2) a predominance of green obsidian (a of the reductionsequence, includingplatform grinding, higher proportion than observed in the other concentra- error recovery, and the manufactureof various blade tions); 3) relatively large quantities of Mazapa Red on products. Brown, a pottery type that is rare throughout much of The area sampledby excavationhas been interpreted the urban zone. as a linear arrangementof residentialcompounds com- Finally, we would like to consider what the available prisingtwo low ridges (FIG. 3, A, B) which flank a (com- data indicate about the occupation history of the locality mon?) refuse dumpingarea; if the results of Healan's and perhaps the entire 44eastflank" concentration. Our preliminarymicroscopic examination of soil samplesare ceramic data suggest that the locality was settled around accurate,the actual places of manufacturewere in the the beginning of the Tollan phase, prior to Tula's major open areabetween the habitationcompound and the re- expansion. Stratigraphicdata indicate that at the time of fuse dump. Each compoundprobably housed a series of initial occupation, the locality was marginal land, a bar- related nuclear families that were probablyrelated to ren, eroded terrain. Perhaps its marginal character was those of adjacentcompounds as well. It is likely that an inducementto locating obsidian workshops there, since such a configurationrepresents a corporateresidential it would have had little or no agriculturalvalue.35 Given entity or barrio,one which was heavily engaged in ob- the present estimation of Tula's extent before expan- sidian working. The presenceof a ceremonialstructure sion,36 the east flank of E1 Salitre lay outside the early at the end of ridge A, presumablya neighborhoodtem- community, though it was incorporatedinto the later city, ple, also suggests a relativelyself-contained residential as seen in Figure 2. entity. We anticipatethat furtherstudy of the domestic We noted that the basis of the early Tollan-phase dat- artifacts, including utilitarianpottery, figurines, and ing of the initial settlement of our workshop is the pres- faunalremains, will shed furtherlight uponthe domestic ence of large amountsof Mazapa Red on Brown, a pottery realmof the workshop. type that is rare throughout all except the southeastern How far beyond the immediateexcavation area can extreme of the urban site;37no other pottery type has this this patternbe extended?Based upon informalsurface distibution. Cobean has noted that Mazapa Red on Brown survey, a similarconfiguration almost certainly exists in is most common in Terminal Classic and Early Postclas- the next field to the north, where at least three ridges sic sites in the Basin of Mexico, particularly the Teoti- and artifact-rich,interridge areas are evident. Further huacan Valley,38and suggested that "it is at least possible north, surface obsidian diminishesin density, forming that the sections of Tula's urban zone with the highest the northernlimits of the "east flank" concentration. amounts of Mazapa Red on Brown were inhabited by a South of the excavationarea, the terrainhas been con- different social or ethnic group from the rest of the urban siderablydisturbed by the constructionof roads, irriga- zone, or that these sections had some different economic tion ditches, pipelines, a quarry,and a new highway, function."39 It is tempting to speculate that the workshop and it is possible that the concentrationeast of CerroE1 zone may have been settled by newcomers from the Basin Cielito is a continuationof the "east flank" concentra- of Mexico and the Teotihuacan Valley, which would tion beyond the disturbedzone (FIG. 2). East of the ex- cavationarea, the high surfaceconcentration of obsidian 35. Ironically, today this is one of the most agriculturally productive continuesfor at least half a kilometer,and includes as- areas in Tula, in large part because of the thick, organically rich refuse and construction fill that resulted from the prehispanic occupation. sociated topographicfeatures indicative of collapsed structures;in fact, we had consideredan alternativeex- 36. Stoutamire, op. cit. (in note S) fig. 12; Diehl, op. cit. 1981 (in cavation site in this area that consisted of an obsidian note 10) 282. concentrationalongside a moundthat had been cut by a 37. Stoutamire, op. cit. (in note 5) 68-69, fig. 13. ditch, revealingsuperposed floors and structuralwalls. 38. Cobean, op. cit. (in note 11) 393. This type was named for San In summary,our informalsurface survey suggests that Francisco Mazapa, a locality within the site of Teotihuacan, where it the "east flank" concentrationcomprises an areaat least was first reported. 0.5 km. in diameter,within which is found a consistent 39. Cobean, op. cit. (in note 11) 391-392. Journal of Field ArchaeologylVol. 10, 1983 145 indicate a continuity with the earlier Teotihuacan obsi- dian exploitation system far more direct than previously imagined. The fact that this pottery occurred alongside, and gradually gave way to, more common types of the Tollan phase suggests an eventual absorption of these peoples into Tula's cultural mainstream. At present this is speculation, but it poses a fascinating question that merits further exploration, if the origin and development of Tula's obsidian industry is to be placed in its proper regional and historical perspective.40

40. The Tulane University excavations at Tula were funded by a major grant from the National Science Foundation (BNS 79-24754), with the authorization of the Consejo de Arqueologia del Instituto Nacional de Antropologia e Historia de Mexico. In addition to those persons previously cited, we wish to acknowledge the advice and moral support provided by Robert H. Cobean of the University of Missouri Research Reactor Facility, the technical expertise of Terry Stocker, who is analyzing the ceramic figurines recovered from our excavation, and the amistad of the people of Tula de Allende, Hi- dalgo, particularly Sr. and Sra. Cuauhtemoc Baptista Cravioto. Fig- ures 1-5 were prepared by Dan M. Healan.

Dan M. Healan is a memberof thefaculty of the Departmentof Anthropologyand ResearchAssociate of the MiddleAmerican Research Institute, Tulane University.He is Director of Tulane'sarchaeological project at Tula, and his specific interestsand responsibilitiesfor the project includelithic technology,architectural and stratigraphicanalysis, and quantitativemethods and data processing. Janet M. Kerley is a graduatestudent in the Departmentof Anthropology,Tulane University. She has served as a field crew supervisorand as lithics analystfor the Tulaneproject at Tula, and is currently conductinga technologicalanalysis of the obsidian artifactsfrom Tula as part of her doctoraldissertation. GeorgeJ. Bey III is a graduatestudent in the Departmentof Anthropology,Tulane University. He has served as a field crew supervisorand as ceramics analystfor the Tulaneproject at Tula. His doctoral dissertationresearch is concernedwith prehispanic ceramiccontinuity and change in the Tula region.