otnigtruhteclna eidaddw noerymdr ie seUtn2003; Urton (see times modern early into down and period colonial the through continuing hw,tesrneteiec o r-nacr-epn per oprant h Middle the to pertain to appears be cord-keeping with will pre-Inka end As 2006). for an (Covey evidence to highlands Peruvian strongest came southern the period the in shown, latter state The Inka direction; the 1000–1450). through of opposite (AD 600–1000) emergence the the (AD period Horizon Intermediate in Middle Late view the the the from to stretching concerns period article time This the specifically, 2010). Brokaw 2004; Salomon infiatavne aebe aei eetdcdsi h td fIk record-keeping Inka of study knotted-cord the in of decades means recent in by made been have advances Significant Introduction Urton Gary central the in khipus Inka of rise the to cord-keeping Horizon Middle From * D6010,khipus 600–1000, period AD Intermediate Late the system. Keywords: in decimal a early to or them Wari Horizon adapting among subsequently Middle khipus 1000–1450), five the (AD base in the encountered late have communities to descendant appear decimal a Inka around the instead organised Hence were system. khipus Inka place-value Later, system. number five base a with people ANTIQUITY " 0 C N niut ulctosLtd. Publications Antiquity eateto nhoooy avr nvriy 30MsahstsAeu,Cmrde A018 USA 02138, MA Cambridge, Avenue, Massachusetts 1350 [email protected]) University, (Email: Harvard Anthropology, of Department km 8(04:2521http://antiquity.ac.uk/ant/088/ant0880205.htm 205–221 (2014): 88 500 ot mrc,cnrlAds iaau na ai ideHrznperiod, Horizon Middle Wari, Inka, , Andes, central America, South ∗ khipus uigteLt oio eid( period Horizon Late the during 205 v,sgetn htte eepoue ya by produced of were groups they in that knots suggesting of five, clustering the is Horizon Middle khipus the of these feature A of khipus. four early of dating This AMS the Andes. reports central article the pre- of the culture Wari by Inka period used Horizon probably Middle 600–1000), are the (AD of part ancestry from dating later khipus the that eight of group a into by offered system insights documented. recording clearly origins Important less this much The however, of are, America. antecedents South and Andean Inka the of among administration imperial of cords, knotted as of known formed devices Recording khipus r elkonfeature well-known a are , c .AD1450–1532),

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes

Horizon Wari culture. In this article, I will describe and analyse what I will show to constitute acorpusof17MiddleHorizoncordsamples.Together,thesesamplesdisplayasufficiently complex and homogeneous set of features to support the argument that they belong to a tradition of cord-keeping from this period that was shared by a probably diverse group— of indeterminable size—of individuals who employed the cord technology for recording information in the Wari state. The Middle Horizon in central Andean prehistory was dominated by two large, expansive states: Wari, in the central Andes, and Tiwanaku, in the south-central Andes. Wari territory (or societies influenced by Wari) stretched from the northern highlands and northern coast of present-day Peru, southward to the Moquegua Valley and then eastward, into the highlands, to the region that would become the capital of the Inka empire, the Cuzco Valley. Tiwanaku stretched from the southern boundary just delineated to the south and south-east, through the Titicaca basin and on down through present-day Bolivia and into what is today central Chile. Wari and Tiwanaku co-existed for several centuries with little evidence of conflict between them but with a great deal of interaction and sharing of values, ideology and material cultural traditions (see Isbell 2008). Wari and Tiwanaku societies would have had need of administrative accounting technologies as these societies attained high levels of social, political and economic integration and complexity. Accounting in such societies would have involved the management and oversight of state resources, including censuses and other administrative matters relating to conquered/subordinate populations, labour and military recruitment, storage and the redistribution of state goods (see Schreiber 1992: 29). These were all resources that were managed in the Inka empire by a complex administrative structure for which the khipus served as the principal recording device. The question is: how were administrative accounts maintained in Middle Horizon times, in the Wari and Tiwanaku states? I must state at the outset that all of the evidence for pre-Inka accounting that is available to us pertains to sites within territory thought to have been controlled or influenced by the highland Wari polity. The cord-keeping terminology of the Tiwanaku region, which was primarily Puquina-speaking in the Middle Horizon period and Aymara-speaking in the colonial period (Torero 2002; Cerron-Palomino´ 2008), centred on the device known as the chino (see Platt 2002). Unfortunately, we have virtually no information about the characteristics of Tiwanaku chino, nor is it clear to what extent chino accounting technology might have been an ancestral, descendant or parallel tradition to that in use in Wari territory. It is the tradition of cord-keeping in Wari territory that is the focus of this article.

Previous studies of Middle Horizon khipus There has been a scant number of articles published to date pertaining to khipus thought to date to the Middle Horizon period. The most notable article is Conklin’s description of 11 examples of what he referred to as “wrapped ” (1982: 267). These include a set of nine examples in the Museo Amano, Lima, Peru, recovered by Yoshitaro Amano in 1968 at the site of Pampa Blanca, near the Hacienda Huayuri, in the Santa Cruz Valley, a tributary of the Nazca River, on the southern coast of Peru (Figure 1). These examples were recovered along with what are described as “Huari [Wari] pottery and a Huari mummy”. The pottery

C " Antiquity Publications Ltd.

206 a fatp sindt ideHrzn2 hc ugssadt of date a suggests which 2, Horizon Middle to assigned type Peru a Lima, of Museum, Amano was the in Now River. Nazca Valley, Cruz 6). Santa fig. Blanca, 1982: Pampa Conklin from (from examples Khipu 1. Figure nPafr fteHaaSnMro.Mtra on ttesm ee n eoi as deposit and level same the at found Material Marcos. San passageway Huaca a in the deposit of sealed a Maranga, 2 from of recovered Platform site was the in example at The Valley. constructions Rimac mound lower adobe the principal in the Sol of one Shady Marcos, San Sol Ruth Huaca Shady by study, this article In an (2000). is ancestry Horizon private the from example 275–79). the 1982: History date—of (Conklin valuable to highly Natural collection performed and one detailed of a only produced analysis—the Museum collection Conklin and private 2). Figure description a American see from 12; example the & other 11 one of figs. as 1982: well (Conklin as collections 7–9), figs. the 1982: Conklin threads from see such Z). (AMNH; more or examples S or (respectively, two three direction opposite and Huayuri the S), in or twisted (Z be direction will right lower which one ply, to in a left spun in lower upper united is from to be material obliquely right may run Raw upper threads S). from the an runs S-spinning, 1982: in threads in string (Conklin (as the while a direction” Z), of in a Z axis threads in oblique the oblique (as the left in twisted, material, the is Z-spun of cords In orientation cord. the patterned the or of to with refers plying wrapped direction the Spin/ply are simple, [c] 268). so-called cords and multiple present—only knots; pendant knots overhand the long or no of are there shanks [b] The thread; multi-coloured “[a] Blanca: Pampa from 267–68). 1982: h nyohrpbiainkont h uhrta ouet hp fMiddle of khipu a documents that author the to known publication other only The Hacienda the from khipus wrapped of collection the into incorporate to on went Conklin khipus small of collection the of characteristic were that features three described Conklin ´ ıs tal. et ecietercvr n19 fakiufo the from khipu a of 1999 in recovery the describe ayUrton Gary 207 ´ s Joaqu ıs, ´ nNarv ın " C niut ulctosLtd. Publications Antiquity e n oi L Sonia and aez ´ c D70(Conklin 700 AD . opez ´

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes

Figure 2. Khipu, private collection (from Conklin 1982: fig. 12). the khipu included sherds of the pottery styles known as Niever´ıa, Lima 9 and Pachacamac, which are generally dated to the Lima culture (AD 200–750; Shady Sol´ıs et al. 2000: 2–4; see Note 1 at the end of this article). The Huaca San Marcos khipu (Figures 3 & 4) measured 118mm in length and bore 12 pendant cords. Five of the pendants are Z-ply and seven are S-ply. Nine of the pendant cords are cotton (as is the primary cord), while three are of camelid fibre. The primary cord has sections wrapped with narrow bands of blue, red and brown threads in a technique similar to the cord wrapping described by Conklin (1982). While no 14C dates were obtained for this example, Shady Sol´ıs et al. (2000) conclude their discussion of the chronological placement of the Huaca San Marcos khipu with the suggestion that it probably dates to Epoch 2 of the Middle Horizon period (AD 650–750). The suggested (relative) dating places this example chronologically at approximately the same time period as the relative dating of the examples discussed by Conklin. To summarise what we have learned Figure 3. Khipu from Huaca San Marcos, Rimac Valley, about supposed Middle Horizon khipus— Lima, Peru (from Shady Sol´ıs et al. 2000: fig. 1). all of which were dated by relative means— up to this time, the following are the central features shared by these samples: a) cotton was the primary material of cord production; b) cords were wrapped in multiple, stacked bands of multi-coloured, camelid fibre threads; c) a predominance of Z-plying (the exception being the Huaca San Marcos example, with its mixed plying directions); and d) the frequent use of simple, overhand knots (with no long or figure-of-eight knots; see discussion of Inka knot types, below).

Expanding the corpus of Middle Horizon khipus The data and studies outlined above represent the bulk of the investigations and published commentary to date on pre-Inka cord-keeping in the Andes. Table1 contains information on

C " Antiquity Publications Ltd.

208 five Sol Shady (after Peru Lima, Valley, Rimac Marcos, San Huaca from khipu of Drawing 4. Figure ideHrznkiutpsdsusdblw ehp ihadtoa nunefo an from influence additional with perhaps below, discussed types khipu Horizon Middle corpus: Horizon Middle feature. cord the constructional include central up in types a make these University), as of khipu Marcos wrapping Both of khipus. San now types ‘loop-and-branch’-type basic by is and two khipus (owned which ‘pendant’-type that Temple-Radicati and below Museo of suggest Primeglio, the collection I di Lima. the of Radicati in collections Carlos of formerly the Museum khipus, was the in of that in student one examples Italian–Peruvian and two Sweden, below the discuss Gothenburg, also in I Cultures, 1982). World Conklin by reviewed (briefly apeLb oB age BP 1230 (Reimer IntCal09 1180 from data atmospheric Beta-311352 2009); Ramsey ∗ (Bronk 1210 v.4.2.2 OxCal Calibration: Beta-270951 no Lab. 1140 41.2/6740 Beta-270950 AMNH E) 1180 41.2/7681 AMNH Beta-270949 D) 41.2/7679 AMNH Beta-270948 C) 41.2/7678b AMNH B) 41.2/7678a AMNH A) khipus. Sample Horizon Middle of dates Radiocarbon 1. Table apetosalfrseparate for small too Sample yhptei sta nakiu a aeeegdfo h ovrec ftetwo the of convergence the from emerged have may khipus Inka that is hypothesis My 14 ae bandfo eea ideHrznkiu ntecleto fteAMNH the of collection the in khipus Horizon Middle several from obtained dates C 13 C: 12 CratioandAMSanalysis ayUrton Gary 209 − + − + − + − + − + 02. D7981A 716–971 AD 687–937 AD 779–891 779–987 AD AD 730–884 716–971 AD AD 23.3 830–975 AD 30 22.5 779–891 AD 40 22.8 40 22.1 40 40 δ 13 ∗ C1 D7385A 689–882 AD 713–865 AD σ " C ai.2 calib. tal. et niut ulctosLtd. Publications Antiquity ´ ıs 2009). tal. et 00 g 4). fig. 2000: σ calib.

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes as-yet-to-be-defined (but see Urton 2013) tradition that moved Andean cord-keeping in the direction of decimalisation (i.e. the recording of numerical values in a base-ten, place-value knotting tradition). In other words, Inka khipus evolved from the merging of the primary cord pendant cord (and simple subsidiary) tradition of the pendant khipu type with the multiple+ subsidiary loop-and-branch-type khipu. Both Middle Horizon types may bear a single type of knot (i.e. not the three knot types of the Inka standard). I return to the matter of the evolution of Late Horizon Inka khipus from Middle Horizon and Late Intermediate Period antecedents below.

Pendant-type khipus The basic characteristics of this type of khipu include a primary cord to which are attached avariablenumberofpendantcords;theoverallstructuralarrangementisnotdissimilarto that of Inka khipus. Middle Horizon pendant-type khipus are generally made of Z-plied cotton with between four and eight stacked bands of multi-coloured camelid fibre threads wrapped around the upper portion of the pendant cords. While one end of this khipu type may terminate in a loop, what distinguishes the loops in this type from the loop-and-branch- type khipu is that in the pendant type, the loop is attached to a primary cord to which the thread-wrapped pendant cords are attached, whereas in the loop-and-branch type, thread- wrapped pendant cords are attached directly to the loop. It is unclear to me whether or not—and if so, how—this difference may have played into differences in recording strategies between the two types of khipus. AMNH 41.2/7678 (Figure 5; Table 1 A & B) This is one of the examples illustrated and briefly discussed in Conklin’s article (1982: fig. 9). The sample lacks provenance. The khipu primary cord measures 1130mm in length and the 125 pendant cords measure between 150–180mm in length. The example is of cotton and the pendant cords are wrapped in stacked bands of camelid fibre threads of different colours. All but one of the pendant cords are Z-ply (as viewed in Figure 5, the fifty-fifth cord from the left is S-ply). None of the pendant cords bears knots. There are 21 distinct cord types (as determined by colour patterning, thread-wrapping patterns, etc.), composed of between 1 and 15 examples each, that make up the 125 cords of this khipu. In January 2010, samples from two cords of this khipu were radiocarbon dated by the Beta Analytic AMS laboratory in Miami, Florida. The results are shown in Table 1 (A and B). The calibrated dates (at 2σ) for the first sample from this khipu (Table 1 A) fall between AD 716 and 971, and for the second cord sample (Table 1 B) between AD 779 and 987. Both sets of dates fall toward the later end of the range commonly used for the Middle Horizon period (AD 600–1000). AMNH 41.2/7679 (Figure 6; Table 1 C) This khipu was illustrated and briefly discussed earlier by Conklin (1982: fig. 8). The length of the primary cord is 230mm. It should be noted that the loop at the left end of the primary cord is attached to the primary cord itself, not directly to thread-wrapped pendant cords (as clarified above). The primary cord and all pendant and subsidiary cords are of Z-plied cotton, and all 30 pendant cords bear stacked bands of multi-coloured camelid

C " Antiquity Publications Ltd.

210 iue6 edn-yekiu MH4./69(ht yGr Urton). Gary by (photo 41.2/7679 AMNH khipu; Pendant-type 6. Figure Urton). Gary by (photo 41.2/7678 AMNH khipu; Pendant-type 5. Figure nltci 00 h eut r hw nTbe1() h apeyeddacalibrated a yielded sample The (C). 1 Table 2 in at 687–937 shown AD are of results date The 2010. in the Analytic of none subsidiary (but single cords a subsidiary bear the knots. overhand of khipu bear Several this pendants) subsidiaries. of multiple cords or pendant cord, (secondary) the of Twenty-five wrapping. thread h rvossml,bt aln oadteedo h ideHrznperiod. Horizon Middle the of end the toward falling both sample, previous the apewstknfo n odo hp MH4./69frASdtn tBeta at dating AMS for 41.2/7679 AMNH khipu of cord one from taken was sample A σ .Thesampleappearstobemoreorlesscontemporaneouswith ayUrton Gary 211 " C niut ulctosLtd. Publications Antiquity

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes

Figure 7. Pendant-type khipu; AMNH 41.2/7681 (photo by Gary Urton). It is important, in regards to the matter of their contemporaneity, to note that 21 of the 30 cords of this khipu are identical to one or other of the 21 cord types found on the previous khipu (Figure 5), whereas nine cord patterns on the khipu in Figure 6 are unique to this khipu. This suggests that the two khipus in Figures 5 and 6 were products of a shared cord- keeping tradition, which (as we will see below) probably included the next example as well. AMNH 41.2/7681 (Figure 7; Table 1 D) The third khipu from the collections of the AMNH was again illustrated and briefly discussed by Conklin (1982: fig. 7). The length of the primary cord is 1140mm; the primary cord and all pendant cords are made of Z-plied cotton. The 18 pendant cords all bear stacked bands of camelid fibre thread wrapping, and all but one of the pendant cords bear one brown cotton Z-plied subsidiary cord each. All but two of the cord patterns of pendants on this example are shared with those on the khipus shown in Figures 5 and 6, which seems clearly to suggest that all three examples are products of the same cord-keeping tradition, perhaps even the work of the same cord keeper. AMS dating of this khipu by Beta Analytic in 2010 (see Table 1 D) yielded a calendar date, at 2σ calibration, between AD 716 and 971. This is precisely the same range of calibrated dates as those obtained for AMNH 41.2/7678 (Figure 5) and very close to those for AMNH 41.2/7679. The contemporaneity and overlap in cord patterning of these first three examples suggest that they made up at least part of the archive of a highly skilled Wari cord keeper, or perhaps a group of cord keepers who were in contact and sharing cord-making techniques and perhaps information with each other. Loop-and-branch-type khipus We turn now to consider what I term loop-and-branch-type khipus. One finds on these examples a looped cord configuration from which one or more thread-wrapped cords are

C " Antiquity Publications Ltd.

212 wdn a ag opt hc r tahd(npnatcr aho)fu pendant four fashion) pendant-cord (in attached are which to loop large a has Sweden, khipu The knots. khipus. overhand loop-and-branch-type five other and one the between provenance. on on bear without those subsidiaries those is and from, as cords distinct well the yet of as to, Several examples similar khipu patterns pendant in multi- complex, wrapping the bear thread all cords fibre pendant-type camelid seven pendant four The coloured bears cotton. itself Z-plied pendant- which of (in loop are secondary attached cords a All are as cords. which well to Carlos as loop cords, specialist, pendant large khipu seven a fashion) Italian-Peruvian has cord late khipu The the (2006). of Primeglio di collection Radicati original the from comes and iue8 opadbac-yekiu MH41.2/ AMNH Urton). Gary khipu; by (photo 6740 Loop-and-branch-type 8. Figure that one and cords transverse two are there say could we or, cords; three emerge which khipus. cords pendant-type on The as knots. cord the spaced along space closely the much although five as wrapping, cover of thread not does cluster bear generally often a wrapping khipus in loop-and-branch-type configured of subsidiaries cases knots, and overhand several single in carry multiple, may are bear subsidiaries cords the which and suspended cords the the the khipus); between Both pendant-type stands subsidiaries. in which branching cord, as primary cords, a pendant to the attached and not loop is loop the (i.e. directly suspended hseape ntecletoso h uemo ol utrs nGothenburg, in Cultures, World of Museum the of collections the in example, This 1 khipu Gothenburg G) Peru, Lima, Temple-Radicati, Museo the of collections the in currently is example This khipu Radicati F) hseapei ntecleto fteAN.Tekiuhsa litcllo from loop elliptical an has khipu The AMNH. the of collection the in is example This 41.2/6740 AMNH Fgr 9) (Figure Fgr ;Tbe1E) 1 Table 8; (Figure Fgr 10) (Figure ayUrton Gary 213 r ecie eo,a tm oH. to khipus. F these obtained for items been not have as dates these Radiocarbon below, 8; Figure described in one are the are to similar that quite khipus loop-and-branch-type of E). 1 (Table 689–882 AD of 2 calibration, at 2011. date, calendar in a produced dating sample AMS The for Analytic Beta to fibres. camelid colourful 8) with Figure wrapped in are viewed central (as cords the side of right and sections Z-plied upper of The and are cotton. example cords this the of of subsidiaries one but All and subsidiaries. sub-cords multi-coloured Each multiple branching, loop. bears cords the projecting three the of of construction cord into integrated the are The cords three loop. the of the bodies from away directly projects maaeo he te examples other three of aware am I submitted was 8 Figure in example The " C niut ulctosLtd. Publications Antiquity σ

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes

Figure 9. Loop-and-branch-type khipu; Museo Temple-Radicati, Lima, Peru (photo by Gary Urton). cords, two of which bear complex arrangements of subsidiaries. Three of the four pendant cords bear a cluster of five overhand knots while one cord bears six of these knots; many of the subsidiaries bear multiple overhand knots as well. The four pendant cords also carry camelid fibre thread-wrapping on the upper portion of the cords. The thread-wrapping pattern is distinct from any of those seen on the three pendant khipus discussed earlier. Both this and the following example in Gothenburg are without provenance, and there are no data that I was able to obtain concerning when or how the examples came into the museum’s collections. H) Gothenburg khipu 2 (Figure 11) This example is built around a large loop to which a single cord is tied, in standard pendant-cord attachment fashion. There are multiple subsidiaries of first and second order, most of which are of cotton (like the loop), but some of which are of dyed camelid fibres. All cords are Z-ply. There are two sections of thread wrapping, both in a pattern similar to that on example G (above) and distinct from the earlier examples. The two thread-wrapped sections on this khipu divide the branches/subsidiaries into two separate sections. A few cords carry clusters of up to five overhand knots. Discussion If we include the nine small examples from the Museo Amano illustrated in Figure 1, then I argue that the corpus of Middle Horizon khipus known to date comprises 17 examples. This would be the cord-keeping tradition—or ‘traditions’—that would

C " Antiquity Publications Ltd.

214 iue1.Lo-n-rnhtp hp (Gothenburg Urton). Gary Sweden by Gothenburg, (photo Cultures, khipu World of Museum 1); Loop-and-branch-type khipu 10. Figure yekiu,ec) swl sqatte ngopnsu o(omny v nt.Iwould I units. five (commonly) to up groupings in quantities cords as of well loop-and-branch- as sets of etc.), structures (i.e. range khipus, branching categories wide the type patterns; a arrayed different recording in hierarchically wrapping for multiple, thread employed bearing to been pertaining have information cluster. could a of khipus in Horizon knots Middle five (commonly) the to above, of one principle with the cords a d) of branching—as knotting samples; and the loop-and-branch-type cords subsidiaries e) and of among between, structure; form patterns cord few or the thread-wrapping a on, hierarchy—in in of not of or, sharing incorporation but upper the the khipus c) around pendant-type cords; bands) on of eight and sections four middle between cases, (generally bands stacked in aueo n ainl o h s fcr-epn nWr otxspooe ale by earlier proposed contexts Wari in cord-keeping the of regarding use argument the the (2010). for supports Brokaw for rationale substantially device and discussion of a of This collection as state. nature structure, this well Wari administrative the served of state-level as have presence hierarchical, such would a The within have khipus etc. could generated Horizon not knot ten, information Middle could each five, recording the tradition by, example, that knotting multiplied knots—for suggests this to and features value) of at, unit users valued (i.e. that been one think have than to powers reason higher no applied is there that add ymaso h aiu tutrl iul(..clu) n raiainlfaue outlined features organisational and colour), (e.g. visual structural, various the of means By ayUrton Gary 215 )tra-rpigi h omo fine, of wound threads form camelid-fibre dyed the brightly in white thread-wrapping article); this of b) Z-ply end the at 2 cord Note basic of (see cotton a) cord primarily follows: the as construction Horizon question, are devices Middle first recording of the components for As khipus Horizon Middle of characteristics principal The standard? Inka the into cord-keeping Horizon of Middle transformation the to might Andean led of that what prehistory course the second, over And transpired have Andes? cord- the of concern: in period features Horizon Middle central the central during keeping the of were are what point, questions this At record- empire. two Inka knotted-cord the of in keeping Late tradition the the emerging during as and ultimately ways Horizon periods, Intermediate Middle various late in the on and transformed innovated manipulated, been have " C niut ulctosLtd. Publications Antiquity

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes

The transformation or evolution of Wari khipus into Inka khipus

As for the changes and transformations that might, or must, have transpired in order for Inka-style khipus to have evolved out of a Middle Horizon Wari prototype through perhaps some intermediate steps, there are several things to take into account. Two of the principal changes would have been: a) alossofcolourfulthreadwrappingand its replacement in Inka examples by a proliferation of forms of incorporating colour differences on primary, pendant and subsidiary cords (Inka khipus display solid- colour, barber-pole and mottled cords; colour changes down the length of a single cord, etc.); and b) an explosion in the use of knots, from the exclusive use of single/overhand knots on Middle Horizon examples to the three Inka knot types: single/overhand, figure-of-eight and long knots. This would also have involved a transformation in the value of the knots themselves. This was related to the Inka practice of arranging knots in hierarchical Figure 11. Loop-and-branch-type khipu (Gothenburg tiers, each level signalling a higher power khipu 2); Museum of World Cultures, Gothenburg, of ten as one moves from the bottom of Sweden (photo by Gary Urton). the cords to the top. These latter features were the prerequisites for incorporating the principle of ‘decimalisation’—that is, a base 10 place-value numeration system—into Inka cord accounting. The emergence of decimalisation in the Inka khipus from what appears to have been a Wari foundation in what I would term ‘iteration’ (i.e. one knot one knot one thing one thing two) requires further clarification, beginning with a+ brief overview= of the main+ features of= Inka khipus (see Conklin 2002; Urton & Brezine 2011). Inka khipus are made of spun and plied cotton or camelid fibres; ply direction is overwhelmingly (96 per cent) S-ply. The colours displayed in Inka khipus are the natural colours of cotton or camelid fibres, or of the dyeing of camelid fibres with natural dyes. The ‘backbone’ of an Inka khipu is the so-called primary cord to which are attached a variable number of thinner strings, called pendant cords. The average number of pendants on some 450 samples studied by the Harvard Khipu Database project is 84 cords. About one quarter of all pendant cords have second-order cords, called subsidiaries, attached to them. Subsidiaries may themselves bear subsidiaries, and so on, to as many as six levels of subsidiaries. The majority of Inka khipus have knots tied into their pendant and subsidiary strings. The knots, which are of three distinct types (as described above), are usually tied in clusters on different levels in a decimal place system of numerical registry. I have sent some 20

C " Antiquity Publications Ltd.

216 aecm notetrioyo ucu-paigpolso h eta ihad and highlands imperial central their the as increasingly Quechua of adopted would peoples gradually Cerr Inkas have Quechua-speaking would the of expansion they north-west, territory and subsequent the coast, the their toward into In especially come language. Valley, have Cuzco common Inkas the the their peoples, Wari of as remnant local, out Aymara with interactions adopted their have In article). would this of still end the been at have 3 Critically, would Huaro. and who Pikillaqta as peoples such Cerr Wari centres, administrative remnant Wari old local the around with the resident of interacted ancestors of have Puquina-speaking around end the would period—roughly valley, the Inkas the Intermediate in from Late themselves Valley early Establishing Cuzco the 900–1200. AD the throughout of and in Inkas place Horizon players The Middle 1999). the prominent (Urton the Tiwanaku, period increasingly colonial at become the in have times recorded widely would Horizon myths in Middle Inkas the during Cerr Puquina, of 2012). origin spoken spoke Beresford-Jones originally language probably & rise Inkas principal Heggarty the the (see the that to period persuasively through the argued Horizon Horizon has in Late Middle (2008) the Palomino the dispersals in from and Inkas period evolution the the linguistic over of and Andes southern transformation and cultural central of processes the numeration? Aymara five) in devices—Quechua, base example, these for who of to, peoples opposed users the (as and by decimal producers on spoken Puquina—based to inventors, languages Wari and from the principal evolution were the the were may likely in occurred degree that most that what question transformation to critical a cord-keeping: a prompts of Inka heart observations were place-value the of decimal khipus to in us Inka pair values take of This numerical majority of arrangements. recording the hierarchical the hand, but around tiered examples, other degree Horizon high the Middle a the On to in recording. five organised to There decimal well). one for as from knots values evidence lone, in no or interest single, an arranged many of often are suggestion (there are a knots is these five of and no above, groups principle knots, clustered seen display single/overhand a closely have in bear we khipus as as examples Horizon numeration Rather, Horizon organisation. Middle decimal Middle numerical repeat: decimal of many To in lack, foundation samples. a or Horizon of presence, evidence Middle the in of recording account of take to critical is (for dates calibrated the case, one but between all press). fallen in have and, certainty) dating AMS 95% for samples Inka presumed hr r needn eee o h ubr n ofie h ubrsx( six Aymara, number for The five. As to numeration. one of numbers systems the decimal-based for lexemes are independent both are ten; there to one 2. Table Aymara numbers in and feature shown Quechua a are Puquina, numeration ten The to decimal above? one was mentioned such numbers languages is: of the three ask for presence the to names clear of question all the obvious or and the any former latter, of the the in in numeration principle decimal a for evidence of absence ic n ftepicplpit fdfeetainbtenWr n nakiu sthe is khipus Inka and Wari between differentiation of points principal the of one Since play at languages principal the been have to appear Puquina and Aymara Quechua, it khipus, Inka into evolved have might khipus Horizon Middle how considering In sse nti al,bt uun n ucu aeidpnetlxmsfrthe for lexemes independent have Quechua and Puquina both table, this in seen As nPlmn 08 ro 2012). Urton 2008; on-Palomino Note (see peoples Aymara-speaking were Wari the that argued has (2008) on-Palomino ´ ´ c D16 n 60(e hrisy&Utnin Urton & Cherkinsky (see 1650 and 1460 AD . ayUrton Gary 217 " C niut ulctosLtd. Publications Antiquity igafranca lingua suqta (see on- is ) ´

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes

Table 2. Number names in three Andean languages (spellings respectively after Torero 2002: 448–56; Cusihuaman´ 1976; Briggs 1993). Number Puquina (Cuzco) Quechua (Altiplano) Aymara

1HUKSTUHOQMAYA 2SOISKAYPAYA 3CAPPAKINSAKIMSA 4SPERTAWAPUSI 5TACPAPISQAPISQA[QALLQU]∗ 6 CHICHU SOQTA SUXTA 7 STU QANCHIS PAQALLQU 8 KINAS PUSAQ KIMSAQALLQU 9 CHIQA ISQON LLATUNKA 10 SCATA CHUNKA TUNKA

∗Paul Heggarty (2008) suggests that the original Aymara word for ‘five’ may have been qallqu. See online supplementary material in Heggarty 2008.

probably borrowed from Quechua (Cerron-Palomino´ pers. comm. 16 July 2009). As for the numbers seven, eight and nine, these are compound terms, the first two of which are based on additions to five (seven two/paya five/qallqu;eight three/kimsa = + = + five/qallqu). Nine is ‘almost’ (lla-) ten. Aymara appears to have originally had a quinary, or base five, system of numeration. As we have noted, there are numerous instances of the clustering of groups of five single knots in the Middle Horizon/Wari khipus, as illustrated above; this may be a sign or expression of the base value of the Aymara numbering system. From the information in Table 2, and in terms of our discussion of Middle Horizon/Wari (non-decimal/base five) and Inka (decimal) khipus, we could say that Puquina and Quechua numeration are consistent with, or accommodate, the decimal numeration principle evident in Inka khipus. On the other hand, if indeed the Middle Horizon khipus are Wari, and if Wari peoples spoke Aymara, then we could say that there is a coincidence in Wari cord technology of a base five terminological system and (perhaps) knotting technology. From the above observations, I argue for the following construction in linking language groups to types of khipus within the context of the succession of archaeological cultures in the central Andes in general, and within the Cuzco Valley in particular. First, I propose that the principal inventors of cord technologies for the recording of administrative information were Aymara-speaking Wari administrators from the Ayacucho Basin. The Wari records were based on wrapping colourful camelid threads around cords in stacked bands and knotting cords with single/overhand knots, often in groups of five knots. Recording of administrative information was performed by some means of signing values—still opaque to us—primarily by means of colour coding. During the Late Intermediate period, Puquina-speaking peoples (that is, the Inkas, or their ancestors) began moving into the Cuzco region, replacing—or merely following on from—the Wari. These Puquina speakers would have come into contact with Wari/Aymara (non-decimal/perhaps base five) cord-keeping technology. As the state apparatus of these Puquina-speaking Inkas became more formalised and complex, and as

C " Antiquity Publications Ltd.

218 epn noteIk tnad npriua,Iwudpitt h bec fa identity ( an period of Intermediate absence Late the the to of clear point cultures a would and/or of I cord- peoples particular, absence Horizon the In the Middle standard. for of is Inka evolution article the and this into transformation in keeping the articulated for process hypothesis and/or the mechanism of portion weakest The Conclusion destruction the of relatively keepers. result cord that the Inka be fact the may by The khipus records Inka administration. Wari to of compared Inka as of today cord-keeping exist feature khipus decimal-based Wari pervasive empire, few a Inka the became of and into life expanded the increasingly in numeration. come Inkas late decimal relatively have used the also and would who Subsequently, regions, As they north- these in expansion and ten). peoples Chinchaysuyu), west Quechua-speaking the of with the and contact toward Cuntisuyu powers accommodate movement their toward increasing to in (i.e. especially signalling desire west Basin, Cuzco (i.e. the the tiers of was out hierarchical expanded transformation in later expanded knotting Perhaps the an cords). of of of mottled signing for and favour practice barber-pole in and reason Wari solid, (e.g. fibres the organising construction the camelid cord abandoned dyed actual for in brightly Inkas colour colour of of the use bands in colour, stacked interest to in regard cords an wrapping In as categories. well and as identities structures cord hierarchical their to adapted have numeration. would decimal they which of quantification, tradition of own principle strong in Puquina a technology of the cord direction polity, Wari/Aymara/non-decimal/base-five the Inka modified and expanding adopted the have to would important increasingly became record-keeping sn epe hmevs u lota hybr eodn technology—alphanumeric recording a bore they that also but themselves, peoples using nain ti osalioyta h Spanish the Spanish that the systems region—until irony political the small the of in no in populations disjunctions is organisation Aymara decimal It of five) implant invasion. form (non-decimal/base to the the these efforts Nonetheless, in their of principles. conflict in and for persisted Collasuyu), practices Inkas conditions in cord-keeping the Titicaca, and the Lake (e.g. organisations of emerged administrative peoples south have the non-decimal/base-five-using would to and there with around populations contact Aymara expanding into came Intermediate/Late administrators Late be to hypothesised have diverse and I times. been rich what were Horizon have there during where may valleys, cord-keeping interaction Lurin of early and this Rimac traditions of Specifically, the of period. interaction region Intermediate the the in Late concentrated in highlands the and/or occurred is coast during central have 2013) Peru the (Urton of of may populations surmise Quechua-speaking transformation with My keepers this the accounting. cord Inka of Inka in dynamics of accounting tradition the non-decimal/base-five decimal that from the shift to the tradition Wari influencing for responsible were rmteWr rdto,teeryIk odkeeswudhv eandtecomplex, the retained have would keepers cord Inka early the tradition, Wari the From ial,tehptei eeoe eenipista hrvrdcmluigInka decimal-using wherever that implies herein developed hypothesis the Finally, ayUrton Gary 219 conquistadores eentol decimal- only not were " c C .AD1000–1450)who niut ulctosLtd. Publications Antiquity

Research From Middle Horizon cord-keeping to the rise of Inka khipus in the central Andes writing—that rapidly superseded the khipu as the primary instrument for record-keeping in the Andes.

Acknowledgements I express my appreciation to Carrie J. Brezine and Julia L. Meyerson for commenting on earlier drafts of this paper. I also thank Bill Sillar and for their reviews and helpful suggestions on a later draft, which greatly improved the final text. I alone am responsible for any errors that remain. I gratefully acknowledge the cooperation of the staff of the American Museum of Natural History, New York, for allowing me to take samples from four Middle Horizon khipus for radiocarbon dating. Thanks to Dr Alexander Cherkinsky, of the Center for Applied Isotope Studies at the University of Georgia, for his advice and collaboration on the interpretation of the AMS readings performed by the University of Georgia laboratory and Beta Analytic. The Khipu Database Project has been supported by grants from the National Science Foundation (2006-07-BCS# 0609719; 2011-12-BCS# 1111489). I express my deep appreciation to the NSF for its support of this project.

Notes 1) There is some confusion in the recounting of the circumstances of the recovery of this example in Shady Sol´ıs et al. (2000). It appears that the deposit in which the khipu was found was excavated and removed from its original location in the 1970s; however, no mention was made in the notes at that time of the discovery of a khipu. It was only in 1999, at the time of a re-excavation of the sector from which the deposit had earlier been removed, that the material from the earlier excavation was re-examined and the khipu was discovered. In their description of their later (1999) excavations, Shady Sol´ıs et al. state that the stratigraphic level from which the example had earlier been removed was, in fact, sealed by later deposits. 2) Cotton is also much more common as a construction material in the surviving Inka khipu corpus. However, as Brezine and I have argued (Urton & Brezine 2011), this is probably due to the greater preservation of coastal examples, where cotton predominated, than in the highlands, where camelid fibre was more common as a weaving material and where the preservation of fabrics was poor. 3) I would note that it is by no means universally accepted that the Wari spoke Aymara. The more common language attributed to Wari peoples by linguists today is Quechua (see the various contributions on this matter published in Heggarty & Beresford-Jones 2012). In my own analysis here and elsewhere (Urton 2012), I am following the suggestions by Torero (2002) and Cerron-Palomino´ (2004), who both argue that the Wari spoke Aymara. This remains a highly contested and controversial topic, which hopefully will gain greater clarity in the years to come.

References CHERKINSKY,A.&G.URTON. In press. Radiocarbon chronology of Andean khipus. Radiocarbon. BROKAW,G.2010.A history of the khipu. Cambridge: Cambridge University Press. CONKLIN,W.J.1982.Theinformationsystemofthe Middle Horizon , in A. Aveni & G. Urton BRIGGS,L.T.1993.El idioma Aymara. La Paz: ILCA. (ed.) Ethnoastronomy and archaeoastronomy in the BRONK RAMSEY,C.2009.Bayesiananalysisof American tropics (Annals of the New York Academy radiocarbon dates. Radiocarbon 51: 337–60. of Sciences 385): 261–81. New York: New York CERRON´ -PALOMINO,R.2004.ElAimaracomolengua Academy of Sciences. official de los Incas: identidad y transformacion´ en – 2002. A khipu information string theory, in J. Quilter el Tawantinsuyu y en los Andes coloniales, in P. &G.Urton(ed.)Narrative threads: accounting and Kaulicke, G. Urton & I. Farrington (ed.) recounting in Andean khipu:53–86.Austin: Perspectivas arqueologicas´ y etnohistoricas´ (Bolet´ın de University of Texas Press. ´ Arqueologıa PUCP 8): 9–21. Lima: Pontificia COVEY,R.A.2006.How the Incas built their heartland: Universidad Catolica´ del Peru.´ state formation and the innovation of imperial –2008.Voces del Ande: ensayos sobre onomastica´ andina. strategies in the Sacred Valley, Peru. Ann Arbor: Lima: Pontificia Universidad Catolica´ del Peru.´ University of Michigan Press.

C " Antiquity Publications Ltd.

220 H C R R P H I eevd aur 03 cetd ac 03 eie:2 ac 2013 March 22 Revised: 2013; March 5 Accepted: 2013; January 7 Received: SBELL LATT EIMER DI ADICATI USIHUAM EGGARTY EGGARTY acse uy2013). July 9 (accessed ∼ http://www.arch.cam.ac.uk/ at: available Journal Andes. the in case-study –000yasclBP. cal curves, years 0–50,000 calibration age radiocarbon Marine09 nvria ainlMyrd a Marcos. San de Mayor Nacional Universidad quipus. artv hed:acutn n eonigin Press. recounting and khipu Andean accounting (ed.) Urton threads: G. & Narrative Quilter J. in trial, tribute- restitution sixteenth-century a during readings Springer. archaeology American South (ed.) Isbell W.H. in & Horizon, Silverman Middle H. Andean central the in identities T Peruanos. Estudios de Instituto Cuzco-Collao ..R R.W. B H .K B. rts cdm xodUiest Press. University Oxford & London: Academy 173). British Academy British the of (Proceedings prehistory of a exploration Andes: cross-disciplinary the in language and Archaeology ..W C.E. B .M.G . M P. UCK ECK ALAMO .20.‘ihu eeto is:vral Chinu variable lies’: or deceit ‘Without 2002. T. , pah1003/loe/Eng/SupplInfo/2008Heggarty.htm EATON ,W.H.2008.WariandTiwanaku:international .J.,M.G.L.B . L . G . M , . J P. ROMER ,P.G.B ,G.S.B ,P.&D.B AN ,P.2008.Linguisticsforarchaeologists:a ´ ROOTES EIMER rnltdadeie yG ro.Lima: Urton. G. by edited and Translated EYHENMEYER 8 55 splmnaymaterial (supplementary 35–56 18: ,C.S.M.T ,A.G.H P ,A.1976. RIMEGLIO ,F.G.M ia iitrod Educaci de Ministerio Lima: . LACKWELL ,D.A.R :2 2 5 – 6 5 .A u s t i n :U n i v e r s i t yo fT e x a s URR ,T.P.G OGG ,R.L.E ERESFORD AILLIE URNEY ,C.2006. C Gram 09 nCl9and IntCal09 2009. . C ,K.A.H ICHARDS UILDERSON Radiocarbon ,C.B ORMAC ,E.B abig Archaeological Cambridge DWARDS tc Quechua atica ´ ,J.V :731–60.NewYork: -J RONK ONES ARD ,S.W.M ,J.R.S UGHEN suissbelos sobre Estudios NDER AN .H I. , ,M.F ,A.B adokof Handbook .2012. R 1 1111–50. 51: AMSEY OUTHON ,K.F.K P AJDAS ANNING AYLISS RIEDRICH LICHT n& on ´ ,C.E. ,T.J. AISER ,J.W. ,S. & , ayUrton Gary , , 221 S –2 0 1 3 .W a r ia n dC a j a m a r q u i l l aa d m i n i s t r a t i v e a as relationship herder-cultivator The 2012. – S U –2003. S U T HADY CHREIBER ALOMON ORERO RTON RTON rsne tteSceyfrAeia Archaeology 78 American for Society Paper the Peru. at of presented coast central the on cord-keeping University Press. Oxford & Academy 321–43. British 173): London: Academy British the of (Proceedings prehistory of exploration cross-disciplinary Beresford-Jones D. (ed.) & Heggarty P. in in Andes, record-keeping the of evolution the and dispersals linguistic origins, archaeological for paradigm Press. Texas records knotted-string Andean ueUiest Press. University Duke village Peruvian a in life asRsac Library. Research Oaks pre-Columbian America in pictographies and signs (ed.) scripts, Urton G. & Boone E.H. Andinos. historia ea Press. Texas uemo nhoooy8) n Arbor: Ann Michigan. 87). of University Anthropology of Museum Peru Horizon ainlMyrd a Marcos San de Mayor Nacional Arqueolog de Marcos. Museo San Huaca la de evidencias antig th S ,G.1999. ,G.&C.J.B ,A.2002. in fteIk hp:bnr oigi the in coding binary khipu: Inka the of Signs OL nulMeig oouu – pi 2013. April 3–7 Honolulu, Meeting, Annual ea e s e up ooeciua las escritura: como quipu del uso del uedad ¨ rhelg n agaei h ne:a Andes: the in language and Archaeology ,F.2004. ,K.J.1992. ´ IS ia nttt Franc Instituto Lima: . :3 1 9 – 5 2 .W a s h i n g t o n ,D . C . :D u m b a r t o n ,R.,J.N AtrplgclPpr fthe of Papers (Anthropological namyths Inca doa elsAds ling Andes: los de Idiomas h odkees hpsadcultural and khipus keepers: cord The REZINE ARV ´ ayAntropolog y ıa aiipraimi Middle in imperialism Wari " C AEZ ´ niut ulctosLtd. Publications Antiquity uhm(C London: & (NC) Durham . .2011.Khiputypologies,in .Austin:Universityof &S.L .Austin:Universityof sd Estudios de es ´ hi a fwriting: of way Their (0:2–23. 3(10): OPEZ ´ ´ a Universidad ıa, Bolet .2000.La u ¨ ´ siae ıstica ´ ndel ın

Research