You are on page 1of 28

Archaeological and Anthropological Sciences

https://doi.org/10.1007/s12520-018-0619-5

ORIGINAL PAPER

Petrographic and XRF analyses of andesitic cut stone blocks


at Teotihuacan, Mexico: implications for the organization
of urban construction
Tatsuya Murakami 1 & Matthew T. Boulanger 2 & Michael D. Glascock 3

Received: 5 December 2017 / Accepted: 18 February 2018


# Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract
This study presents preliminary results of petrographic and X-ray fluorescence analyses of cut stone blocks used for urban
construction at Teotihuacan, the capital of a regional state in Central Mexico (ca. AD 150–650). Cut stone blocks were concen-
trated in the civic-ceremonial core of the city and were probably prestigious architectural elements due to their higher costs of
procurement and transportation compared to alternative materials (boulders and clay amalgam). This suggests that the organi-
zation of stone block procurement and distribution was likely embedded in power relations between commissioners and mining
groups. By combining multiple analytical methods that complement one another, this study was able to discriminate local (within
10–15 km radius) from non-local materials. The results suggest that the majority (> 80%) of andesitic cut stone blocks were
brought from non-local sources. This paper discusses procurement organization and suggests that most rocks were quarried by
specialized groups and brought to the city through a tribute system and/or patron-client relations. This has implications for
understanding the nature of the urban-hinterland relationship and expansion of the Teotihuacan state.

Keywords Cut stone blocks . Procurement . Petrography . XRF . Urbanism . Teotihuacan . Mesoamerica

Introduction residences, pyramids, and other civic-ceremonial structures


(Millon 1973). The Teotihuacan Valley is a sub-valley in the
Stone masonry was an essential component of urban architec- northeastern portion of the Basin of Mexico (Fig. 1), which is
ture in Prehispanic Mesoamerica. Rocks were used for fill and mostly of volcanic origin and includes Tertiary and Quaternary
facing of buildings and were often shaped into square or rect- volcanoes. The valley offers different types of extrusive igne-
angular blocks (cut stone blocks). At Teotihuacan, the capital of ous rocks, such as basalt (both compact and scoriaceous ba-
a regional state in Central Mexico (ca. AD 150–650), various salts) and andesitic rocks (Murakami 2015, pp. 268–270). Of
types of rocks were brought to the city for constructing these rocks, tezontle (the local name for scoriaceous basalt) and
andesitic rocks compose the majority of cut stone blocks, which
were used for staircases, walls, entryway steps, and delimiting
Electronic supplementary material The online version of this article
(https://doi.org/10.1007/s12520-018-0619-5) contains supplementary
courtyards. Tezontle is easy to work with given its porosity and
material, which is available to authorized users. lightweight (Margáin 1967, p. 171), making it the most abun-
dantly used material for urban construction (Díaz Lozano 1979,
* Tatsuya Murakami pp. 60–61). Barba Pingarrón (2005) and Barba Pingarrón and
tmurakam@tulane.edu Córdova Frunz (2010) demonstrate that caves and depressions
found in the northwestern city and east of the Sun Pyramid
1
Department of Anthropology, Tulane University, 6823 St. Charles (Fig. 2) were quarrying loci for tezontle. Andesitic rocks in-
Ave., Dinwiddie Hall 101, New Orleans, LA 70118, USA clude andesite, dacite, rhyolite, and trachyte and are widely
2
Department of Anthropology, Southern Methodist University, PO available in the valley and adjacent areas.
Box 750235, Dallas, TX 75275-0235, USA In this study, we focus on cut stone blocks made of andesitic
3
Archaeometry Laboratory, University of Missouri Research Reactor rocks rather than tezontle given the greater labor costs of pro-
Center, Columbia, MO 65211, USA curement, preparation, and transportation of the former. The
Archaeol Anthropol Sci

Fig. 1 Map of Central Mexico


showing archaeological sites and
sources of obsidian mentioned in
the text

procurement and unequal distribution of such expensive mate- which time the major pyramids (the Moon and Sun Pyramids
rials likely constituted some dimensions of power relations, and the Feathered Serpent Pyramid) were built along the central
making the present provenance study an important basis for avenue called the Street of the Dead (Cowgill 2000; Millon
exploring sociopolitical aspects of urban formation and trans- 1981; Murakami 2015; Sugiyama 2005). During the
formation. Based on petrographic and X-ray fluorescence Tlamimilolpa to Early Xolalpan phases (ca. AD 250–450), the
(XRF) analyses of geological and archaeological samples, we city of Teotihuacan underwent urban renewal as over 2000 apart-
examine the source areas of andesitic cut stone blocks from ment compounds (walled enclosures that consist of multiple
Teotihuacan and discuss the organization of their procurement, courtyard units) were built across the city (Millon 1973, 1981).
specifically whether it was organized centrally by the state or The use of cut stone blocks was highly concentrated at
individually by smaller social segments. Although our geolog- large-scale monumental structures visible today along the
ical sampling was limited in scale, the combined use of petrog- Street of the Dead (the central precinct hereafter) during the
raphy and XRF successfully discriminated some local (within Tlamimilolpa and Early Xolalpan phases, which suggests that
10–15 km radius) and non-local (beyond 10–15 km radius) procurement was organized by the state (Murakami 2010).
source areas. The preliminary results suggest that rocks from Yet, a small number of cut stone blocks were used in apart-
certain source areas, specifically non-local sources, were re- ment compounds. During the Late Xolalpan and Metepec
peatedly exploited for urban construction. We suggest that the phases (ca. AD 450–650), construction activities in the central
procurement of the majority of rocks was centrally organized, precinct were reduced substantially, and the use of cut stone
probably as part of a tribute system and/or through patron-client blocks was concentrated in some intermediate elite
relations, and that intermediate- and small-scale procurement compounds.
and distribution systems coexisted to some extent. The use of cut stone blocks was likely determined by their
functional, and probably aesthetic, superiority over shaped
rocks covered with clay amalgam, common building materials
Cut stone blocks and the quality for walls and staircases. The architectural elements in which
of architecture at Teotihuacan cut stone blocks were used, such as staircases and steps in
porticoes, rooms, and courtyards (Fig. 3), were subject to fre-
Urbanization of Teotihuacan began around 100 BC, and large- quent foot traffic, and cut stone blocks would have been an
scale construction programs started around AD 200, during ideal construction material given their durability. In contrast,
Archaeol Anthropol Sci

Fig. 2 Location of architectural complexes mentioned in the text (redrawn with modification after Millon 1973). Caves and depressions identified by
Barba Pingarrón (2005) are inserted in the map

the edges of stairways and steps made of shaped rocks and blocks was not a structural requirement, with the exception of
clay amalgam were more susceptible to attrition (the likely large three-dimensional sculptures; the same functions could
reason why their edges are often slightly rounded), and so have been met by using alternative materials. This provides
cut stone blocks were likely used to reduce maintenance costs. the basis to consider the use of cut stone blocks as a type of
It is also possible that the sharper edges of cut stone blocks architectural conspicuous consumption and thus indicative of
were more aesthetically appealing. power differentials. Moreover, it is likely that the quarrying of
Cut stone blocks were also used for low relief and three- andesitic blocks required specific knowledge, indicative of
dimensional sculptures (Fig. 3a). While the production of higher investment in human resources, compared to quarrying
sculptures, instead of mural paintings, was probably related boulders.
to aesthetic preferences and the nature of activities (e.g., rit-
uals) and messages conveyed by the iconography (e.g., Sarro Procurement techniques of cut stone blocks
1991), the use of cut stone blocks in sculptures, as opposed to
modeled clay or stucco, was probably due to their functional Few quarries used to procure for cut stone blocks have been
(or structural) superiority. It is highly difficult, and likely im- identified in Mesoamerica, with the exception of Mitla in
practical, to make large sculptures, such as those of the serpent Oaxaca (Robles García 1992; see also Gillespie 1994 for an
heads adorning the Feathered Serpent Pyramid (FSP), out of Olmec monument carving site). This has resulted in a poor
modeled clay or stucco. We suggest that cut stone blocks were understanding of the techniques used for rock extraction.
preferred for certain architectural elements due to their func- Quezada et al. (Quezada Ramírez et al. 2018) and Quezada
tional, structural, and aesthetic superiority over alternative ma- Ramírez (2016, pp. 221–225) conducted an
terials and thus serve to highlight qualitative differences ethnoarchaeological study of techniques for the procurement
among buildings. Also, we emphasize that the use of cut stone of cut stone blocks in Central Mexico. Cut stone blocks are
Archaeol Anthropol Sci

Fig. 3 Cut stone blocks used at Teotihuacan: a sculptures at the Feathered courtyard at the West Plaza Complex, a sub-complex of the Street of the
Serpent Pyramid; b cut stone blocks dismantled from the Feathered Dead Complex; e stairs of a temple at the West Plaza Complex; and f an
Serpent Pyramid and discarded at the rear of the pyramid; c a step for entrance to the La Ventilla II apartment compound
rooms at the La Ventilla III apartment compound; d steps delimiting a

still used in construction (e.g., stone pavements and houses) in Teotihuacan, and systematic quarrying techniques such as
Mexico and come from several quarrying locations. those described by Quezada Ramírez et al. (2018) may have
Quezada Ramírez et al. (2018) observed the techniques been employed. If so, this may suggest that some specialized
used to extract stone blocks of andesite at a quarrying location group or groups of stone workers existed, although it is be-
in the Guadalupe range on the west coast of Lake Texcoco, yond the scope of this paper. Unlike quarrying unshaped
where the Aztecs also exploited cut stone blocks for monu- rocks, extracting nearly square or rectangular blocks
mental construction (López Luján et al. 2003). The extraction (preforms) required knowledge of specific rocks (such as the
process consisted of the following steps. The first step was to comprehension of veins and fracture dynamics). Thus, the use
choose a specific location based on the desired size of the final of cut stone blocks, specifically andesitic rocks, was expen-
products and the rock’s propensity for fracturing. Holes were sive not only in terms of preparation and transportation costs
made using iron sticks along a selected fracture line (which but also in terms of human resources in knowledge. It should
took several days) and filled with dynamite in order to detach be noted, however, that not all the cut stone blocks were quar-
a large block or blocks from the bedrock. The second step was ried through the systematic techniques described above. In
to reduce these large blocks to smaller blocks or preforms. particular, those of tezontle were probably quarried as un-
Workers did this by locating veins on the blocks and shaped blocks and then shaped through percussion and abra-
fracturing them along the veins using a wedge and a sion since tezontle is vesicular and may lack the veins that
hammer. The preforms were then carried to specific facilitate straight fractures. It is possible that some types of
construction locations. Quezada Ramírez et al. (2018) note andesitic rocks were also quarried as unshaped blocks and
that fracture lines in the preforms are largely straight. processed later.
Although iron tools and dynamite did not exist in
Prehispanic times, this suggests that similar steps were taken
to quarry stone blocks of andesite, which probably required Geological sampling of possible sources
more labor expenditure than modern quarrying (Abrams
(1994, p. 48) estimates the ratio of labor costs with steel tools The provenance study is based on the comparison of geolog-
to those with stone tools at 1:1.5). ical samples from possible source areas and archaeological
As mentioned above, andesitic rocks are one of two major samples of cut stone blocks collected from architectural com-
raw materials used for fashioning cut stone blocks at plexes in the city. No sources were known at the beginning of
Archaeol Anthropol Sci

this study and, due to constraints of time and resources, not all the northeastern portion of the Basin of Mexico, which per-
potential source areas were surveyed. Thus, the results of this tains to stepped plains forming part of the Trans-American
study remain preliminary and seek to provide baseline data for Volcanic Arch (Mooser 1968, pp. 32–22; Vázquez-Sánchez
future research. Below we present procedures of geological and Jaimes-Palomera 1989, p. 136). Consequently, most vol-
and archaeological sampling and methods employed to char- canic events in the two regions are associated with the Trans-
acterize rock samples. Then, we discuss the results with re- American Volcanic Arch and produced different types of an-
spect to the procurement organization of cut stone blocks. desitic, dacitic, and basaltic rocks from the Miocene to the
As mentioned above, andesitic rocks are widely available Quaternary (Vázquez-Sánchez and Jaimes-Palomera 1989).
in the Teotihuacan Valley, and the proximate sources outside In each possible source area, the lead author collected one
the valley are located in the Texcoco region, south of the to 11 samples, maximizing different colors and textures. It
Teotihuacan Valley (Fig. 4). The lead author (Murakami) should be noted that sample size at each area was affected
chose these two regions for geological sampling of andesitic by accessibility (some lands were private and access was
rocks, and more focus was given to sampling within a 10–15- prohibited). In most cases, samples were taken from outcrops
km radius of the city of Teotihuacan (i.e., southern and west- and in some cases from boulders. A total of 80 samples were
ern portions of the Teotihuacan Valley and the northern por- collected from 16 possible sources (Table 1 and Fig. 4). The
tion of the Texcoco region) due to time constraints and the sample location was recorded with a handheld GPS. Below
hope of distinguishing local and non-local materials for this we describe each sampling location and the type of rock re-
study. The Teotihuacan Valley and the Texcoco region occupy ported in the literature.

Fig. 4 Location of geological


samples in the Teotihuacan Valley
and the Texcoco region
Archaeol Anthropol Sci

Table 1 Sample size and characteristics of geological samples of andesitic rocks

Source area Location Context n XRF Thin section Identified rock types Color variations
(abbreviation) (n) (n)

Cerro Colorado W Teo Valley Boulder 4 3 2 Andesite Light pink/brown to light gray
(COL)
Cerro Malinalco W Teo Valley Boulder 11 10 3 Andesite Light brown, brown, light
(MAL) Gray, gray, dark gray, mottled
pink/gray
Zacualuca (ZACL) W Teo Valley Boulder 5 3 1 Dacite, andesite Purple, pink, dark gray
Cerro Nixcuyo W Teo Valley Boulder 1 1 0 Andesite Purple
(NIX)
Cerro Clacaiteipa W Teo Valley Boulder 2 1 0 Andesite Dark pink, mottled dark gray
(CLA)
Belem (BEL) S Teo Valley Quarry 9 8 2 Dacite, andesite, rhyodacite, Mottled white/brown, brown
rhyolite
Cerro Xoxoqui S Teo Valley Boulder 6 6 3 Andesite Gray, purple, dark pink
(XOX)
Santiago La Cruz S Teo Valley Boulder 5 5 2 Andesite Light to dark gray
(CRZ)
Cerro Patlachique S Teo Valley Boulder 4 4 2 Andesite Light brown to brown
(PAT)
Cerro La Cruz S Teo Valley Quarry 8 2 1 Andesite-dacite Purple, gray, dark pink/brown
(CRUZ)
Cerro Huixtoyo S Teo Valley Quarry 6 1 0 Andesite-dacite Purple-dark gray
(HUI)
Tepetitlan (TPT) Texcoco Boulder 2 2 1 Andesite-dacite Dark gray/pink
Region
Jolalpan (JOL) Texcoco Quarry 7 7 2 Andesite-dacite Pink to gray/pink to dark gray
Region
La Flor (FLOR) Texcoco Secondary 4 4 3 Andesite, dacite, rhyolite Gray/pink
Region deposit
Santo Tomas Texcoco Boulder 1 1 1 Andesite Gray
(STOM) Region
Coatlinchan Texcoco Quarry and 5 3 2 Andesite, dacite Light to dark gray, light brown
(COAT) Region boulder
Total 80 61 25

The Teotihuacan Valley In the northwestern portion of the valley, samples were
collected from a mountain adjacent to the village of Santiago
Andesitic rock samples were collected from 11 cerros or Zacualuca (hereafter referred to as Zacualuca), Cerro Nixcuyo
mountains in the Teotihuacan Valley (Figs. 4 and 5). (just east of Zacualuca), Cerro Clacaiteipa (north of
Andesitic rocks constitute the oldest formation in the Zacualuca), Cerro Colorado (north of the modern town of
Teotihuacan Valley (Díaz Lozano 1979, p. 32; Hernández San Juan Teotihuacan), and Cerro Malinalco (or Cerro
Javier 2007, pp. 30–39; Mooser 1968; Sotomayor Maninal in the geological map) (Fig. 4). Díaz Lozano (1979,
Castañeda 1968). Díaz Lozano (1979) reports the results pp. 38–40) reports the presence of gray quartzitic andesite
of petrographic thin section analysis of rocks from multiple (western side of Cerro Nixcuyo), dark red hypersthene andes-
mountains within the valley, and the sampling locations for ite (western side of Cerro Nixcuyo), reddish altered andesite
this study were selected based on his descriptions of rocks (southeastern side of Cerro Nixcuyo, Cerro Colorado, south-
(see also Biskowski (1997) who conducted INAA on ba- eastern side of Cerro Malinalco), and dark red pyroxene an-
saltic and some andesitic rocks within the valley). desite (southeastern side of Zacualuca). It seems that all these
Sampling locations can be grouped into two major areas, andesitic rocks are from Miocene volcanoes (Mooser 1968;
the northwestern and southern portions of the valley. Díaz Sotomayor Castañeda 1968). Mooser (1968) assigns these
Lozano (1979, p. 32) also mentions that andesitic rocks are rocks to andesite and rhyodacite, whereas Sotomayor
present in an area east of Cerro Gordo, but the lead author Castañeda (1968, p. 42) assigns Zacualuca rocks to augite
was unable to visit this area due to time constraints in the andesite (pyroxene andesite) and those from Malinalco to
field. hornblende basalt. The geological map published by INEGI
Archaeol Anthropol Sci

(1979) reports andesite from Quaternary volcanoes in pyroxene andesite (Cerro Huixtoyo). All of these andesitic
Zacualuca and Cerro Clacaiteipa and gray dacite from Upper rocks are assigned to rhyolite in the geological map published
Tertiary volcanoes in Cerro Malinalco. by INEGI (1979). Mooser (1968) states that the majority of
At the southern limit of the Teotihuacan Valley stands the the mountains in the Patlachique range (Fig. 5e) are Miocene
Patlachique range, which divides the Teotihuacan Valley and except for Cerro Patlachique, which is assigned to a Pliocene
the Texcoco region (Fig. 1). It consists of multiple mountains volcano. He assigns rocks from Miocene volcanoes to
that belong to the same formation that originated in the andesite and rhyodacite, whereas those from the Pliocene are
Miocene (Vázquez-Sánchez and Jaimes-Palomera 1989, p. assigned to andesite and latite. Sotomayor Castañeda (1968, p.
148). Rock samples were collected from a hill adjacent to 42) assigns the andesitic rocks in the Patlachique range to
the modern village of Belem, Cerro La Cruz (adjacent to the biotite latite and hornblende andesite.
modern village of Santiago Tepetitlan; hereafter referred to as To the east of Cerro Gordo, Díaz Lozano (1979, p. 41)
Santiago La Cruz), Cerro Xoxoqui (whose western slope is cut reports pink compact andesite near Hueyapan, northwest of
by a road to Texcoco), Cerro Patlachique (or Cerro Patlachico Otumba. This type of rock is found between basaltic flows
in the geological map) on the northeastern portion of the that covered andesitic rocks. He mentions that stone blocks
range, and Cerro Huixtoyo and Cerro La Cruz on the western were quarried extensively at this location, although it had been
side. Of these mountains, Belem is a disused modern quarry abandoned at the time of his survey in the early 1900s.
site (Fig. 5a), and Huixtoyo and La Cruz are still functioning
as rock quarry sites. The Texcoco region
Díaz Lozano (1979, pp. 33–37) reports pink dacite
(Belem), gray to pink dacite (eastern side of Cerro Major andesite outcrops extend in the mountain ranges that
Patlachique), gray andesite (Santiago La Cruz and the eastern form the eastern limit of the Basin of Mexico (e.g., Heizer and
portion of Cerro Patlachique), pinkish light gray altered an- Williams 1963). The geology of the Texcoco region is not well
desite (the northeastern portion of Cerro Patlachique), banded studied, and there is little information on the variation of an-
blue gray/pink compact andesite (La Cruz), and brown desitic rocks available in the region. The mountain ranges that

Fig. 5 Mountains sampled in this study: a a historical quarry at Belem; b a modern quarry at Jolalpan; c a modern quarry at Coatlinchan; d Cerro
Malinalco viewed from the Street of the Dead; and e Patlachique range (the Sun and Moon Pyramids can be seen)
Archaeol Anthropol Sci

form the eastern limit of the Texcoco region (and thus the that sedimented into sand layers. The boulders likely originat-
Basin of Mexico) were formed from two major units, volcanic ed from multiple locations and include different types of an-
deposits of the Early Pliocene and the Tlaloc Formation of the desitic rocks. Of these source areas, Coatlinchan is the south-
Quaternary, which covered most of the earlier deposits. The ernmost location, about 25 km directly south of Teotihuacan.
composition of rocks from the Early Pliocene varies from It was included because a large Teotihuacan-style statue
andesitic to dacitic (Vázquez-Sánchez and Jaimes-Palomera (called the Idolo de Coatlinchán) was found there (Heizer
1989, p. 149), and that of the Tlaloc Formation includes an- and Williams 1963) and was identical in mineralogical com-
desitic, latitic, and dacitic (Vázquez-Sánchez and Jaimes- position to the statue of the so-called Water Goddess found
Palomera 1989, p. 157). Vázquez-Sánchez and Jaimes- near the Moon Pyramid.
Palomera (1989, p. 158) mention that the majority of rocks
are andesite and dacite of amphibole, lamprobolite
(oxyhornblende), and pyroxene. The geological map of
INEGI (1979) shows the presence of hornblende andesite Selection of archaeological samples
and lamprobolite andesite.
Samples were collected from five possible source areas Thirteen architectural complexes were included in the analysis
(Figs. 4 and 5): a mountain adjacent to the modern village of of cut stone blocks (Table 2): the Moon Pyramid, the
San Antonio Tepetitlan (hereafter referred to as Tepetitlan); Quetzalpapalotl Palace Complex, the Sun Pyramid, the
Cerro Teponaxtle (adjacent to the modern village of Jolalpan Street of the Dead Complex, the Ciudadela, La Ventilla I, La
or Jolalpa; hereafter referred to as Jolalpan) on the southern Ventilla II, Atetelco, Tetitla, Zacuala Palace, Yayahuala,
side of the Patlachique range; La Flor near the city of Texcoco; Tepantitla, and La Ventilla III (no samples were collected from
Santo Tomás, east of the city of Texcoco; and mountains near the Moon Pyramid). All cut stone blocks came from structures
the modern town of Coatlinchan (hereafter referred to as built during the Tlamimilolpa through Metepec phases (ca.
Coatlinchan A and Coatlinchan B). Among these, Jolalpan A.D. 250–650), except for some blocks from the first Sun
and Coatlinchan A are modern quarry sites (Fig. 5). La Flor Pyramid, which was built during the Miccaotli phase (ca.
is also a modern quarry site, formed by a deposit of boulders A.D. 150–250).

Table 2 Sample size and summary data of archaeological samples of andesitic cut stone blocks

Location (abbreviation) Recorded blocks (n) Sample (n) XRF (n) Thin section (n) Mean size (m3) Range (m3) Estimated total
volumea (m3)

Central precinct
Moon Pyramid (MP) 107 0 0 0 0.031 0.012–0.061 151.97
Quetzalpapalotl Palace Complex (QP) 149 28 11 3 0.021 0.002–0.057 29.68
Sun Pyramid (SP) 41 41 16 4 0.147 0.046–0.3 1038.72
Street of the Dead Complex
Viking Group (VG) 31 28 6 1 0.028 0.004–0.125 1.77
West Plaza Compound (WPC) 344 69 33 10 0.016 0.002–0.112 50.39
Compound of the Superimposed 11 10 2 0 0.014 0.009–0.02 0.15
Structures (CSS)
Ciudadela (CD) 46 46 27 15 0.175 0.003–0.497 2070.34
Intermediate elite compounds
La Ventilla I (LVI) 52 46 19 9 0.045 0.012–0.228 11.25
La Ventilla II (LVII) 22 23 11 3 0.015 0.004–0.056 0.36
Atetelco (Atet) 175 23 10 1 0.016 0.005–0.047 4.93
Tetitla (Tet) 19 19 10 3 0.006 0.002–0.011 0.11
Zacuala Palace (Zac) 39 32 12 1 0.007 0.002–0.017 0.99
Yayahuala (Yay) 34 21 10 3 0.022 0.002–0.231 0.73
Tepantitla (TP) 12 7 1 0 0.037 0.006–0.066 2.78
Non-elite compounds
La Ventilla III (LVIII) 16 15 8 1 0.012 0.005–0.035 0.34
Total 1098 408 176 54 3364.52
a
Estimated total volume includes those cut stone blocks that were not recorded in the field
Archaeol Anthropol Sci

In total, the size, color, and texture of 1098 cut stone blocks color groups in this study. These color groups are not meant
were recorded and 408 samples collected from 12 architectur- to be equivalent to source groups, for we think same color
al complexes (Table 2). The selection of samples was judg- groups may have come from several different sources and a
mental and multiple criteria were combined differently for single source area may also contain rocks of different color
each architectural complex. When cut stone blocks were not groups.
numerous, nearly all were recorded and sampled with the ex-
ception of some blocks of identical material. This was the case Petrographic thin section analysis
at the Viking Group and the Complex of the Superimposed
Structures within the Street of the Dead Complex, as well as Based on the results of microscopic analysis, 54 archaeologi-
the surrounding apartment compounds (with the exception of cal samples and 25 geological samples were selected for pet-
Atetelco). At the Quetzalpapalotl Palace Complex, the West rographic thin section analysis (Tables 1 and 2). For archaeo-
Plaza Complex, and Atetelco, the color and texture of all ob- logical samples, at least one specimen was included from each
served cut stone blocks were recorded (except for those coated microscopic color group. Petrographic analysis was employed
with lime plaster and heavily weathered blocks) and samples mainly to evaluate microscopic color groups and corroborate
were selected to maximize different colors and textures. The the classification of archaeological samples. Thus, the color,
state of preservation and the public visibility of the location texture, and structure of groundmass and the size and type of
were also taken into account in choosing the sampling loca- phenocrysts were recorded. Subsequently, archaeological
tion. At the Feathered Serpent Pyramid and the Sun Pyramid, samples were assigned to possible source areas, which were
around 40 samples were arbitrarily selected from dismantled further examined based on geochemical analysis.
cut stone blocks scattered around the pyramids (Fig. 3b).
From each selected cut stone block, a small piece (ca. 5 mm Geochemical analysis
to 5 cm) was detached using a hammer and a chisel.
XRF was used within archaeological provenance research to
identify chemical signatures of each sample. XRF is more
Analytical methods commonly used for obsidian analysis but has successfully
been applied to extrusive igneous rocks, including basalt
The provenance study of the cut stone blocks was carried out (Greenough et al. 2001; Latham et al. 1992) and andesite
combining three different techniques: reflected light micro- (Jones et al. 1997; Ogburn 2004; Williams-Thorpe and
scopic observations on the freshly broken edge of samples, Thorpe 1993; see also Schaaf et al. 2005). Based on the results
petrographic analysis of thin sections, and geochemical anal- of microscopic analysis, 61 geological samples and 176 ar-
ysis through XRF. There are advantages and disadvantages to chaeological samples were selected for XRF analysis
each technique, so the three methods used together comple- (Tables 1 and 2). These samples were analyzed at the
ment one another. Petrologically different rocks may form University of Missouri Research Reactor (MURR)
from the same magma, depending on the cooling rate and Archaeometry Laboratory.
other factors. Conversely, petrologically similar rocks may XRF analysis was conducted by placing the entire artifact
form from chemically different magmas. Thus, it is important within the chamber of an ElvaX energy dispersive X-ray fluo-
to combine geochemical and petrographic (including micro- rescence spectrometer. Measurements were made using a
scopic) analyses. tungsten anode. Quantitative analysis of the resulting spectra
was calibrated with data from the analyses of 19 different
Microscopic analysis geological standard reference materials1 (SRMs). Reference
samples were made by combining powdered SRMs with a
Microscopic observation of freshly broken edges with a wax binder in a ratio of approximately 4:1 and pressing the
reflected light microscope is a useful and expedient technique
1
to identify color, texture, and some minerals at no cost. For Standard reference materials from the following institutions were used to
calibrate the ElvaX spectrometer: National Institute for Standards and
certain types of rock, it can be used as a substitute for thin
Technology (NIST): SRM278 (obsidian rock), 688 (basalt rock), and 1633b
section analysis. A limitation, however, is the difficulty in (coal fly ash); United States Geological Survey (USGS): AGV-1 (andesite),
identifying small and underdeveloped crystals. All archaeo- BCR-1 (basalt), BHVO-1 (Hawaiian basalt), BIR-1 (Icelandic basalt), G-2
(granite), GSP-1 (granodiorite), RGM-1 (rhyolite), SDC-1 (mica schist),
logical samples (n = 408) and the 61 geological samples were
STM-1 (nepheline syenite), and W-2 (diabase); and Geological Survey of
analyzed through microscopic observation (Tables 1 and 2). Japan (GSJ): JA-1 (andesite), JB-2 (basalt), JG-2 (granite), JGB-2 (gabbro),
The color and texture of the groundmass and the size (e.g., JR-1 (rhyolite), and JSy-1 (nepheline syenite). Reference values for NIST and
small, medium, large) and type of phenocrysts were recorded USGS standards are given in the sixth edition of Tables for Neutron Activation
Analysis (Glascock 2006). Reference values for GSJ standards are from un-
for each sample. Based on these variables, the archaeological published neutron activation data produced at MURR and are available upon
samples were classified into different groups referred to as request.
Archaeol Anthropol Sci

combined powder/wax matrix into a solid pellet. Analyses of order to make data interpretation easier. Of the
an additional 30 whole-rock obsidian samples were conducted abovementioned pattern-recognition techniques, this study
to validate the resulting calibration. Following this protocol, employed PCA to transform the data from the original corre-
analysis with the calibration routine produced elemental con- lated variables into uncorrelated variables (Baxter 1992;
centration values for 13 elements (Na, K, Ca, Ti, Mn, Fe, Zn, Baxter and Buck 2000; Neff 1994, 2002).
Ga, Rb, Sr, Y, Zr, and Nb) with varying degrees of estimated Based on PCA results, archaeological samples were divid-
uncertainty (Table 3). ed into several provisional groups. The group membership of
For this study, a non-destructive approach was chosen for each archaeological sample was then assessed through the
its time and cost-effectiveness, but it entailed disadvantages Mahalanobis distance (or generalized distance), a metric that
that should be kept in mind. First, the irregular surfaces (i.e., makes it possible to describe the separation between groups or
freshly broken edges) analyzed here may have interfered with between individual samples and groups on multiple dimen-
transmitted and incident X-rays, producing erratic results; a sions. The Mahalanobis distance of a specimen from a group
flat and polished surface is preferable. Second, the samples centroid (Bieber et al. 1976; Bishop and Neff 1989) is defined
may not have exhibited a homogeneous distribution of ele- by:
ments throughout the analyzed surface. This was particularly h it h i
true for some of the volcanic rocks, which showed large phe- D2y;x ¼ y−X I x y−X
nocrysts of hornblende, feldspar, biotite, pyroxene, and am-
phibole. Typically, this problem is addressed by pulverizing where y is the 1 × m array of logged elemental concentrations
samples into a homogeneous mixture of relatively uniform for the specimen of interest; X is the n × m data matrix of
crystal size, thus allowing a determination of the Baverage^ logged concentrations for the group to which the point is being
chemical content of the entire rock (i.e., the combined com- compared, with X being 1 × m centroid; and Ix is the inverse of
position of groundmass and phenocrysts). As such, elemental the m × m variance-covariance matrix of group X.
determinations for the samples may be more accurately and Mahalanobis distance is analogous to expressing distance
precisely made using more sensitive bulk-sample techniques from a univariate mean in standard deviation units because it
such as neutron activation analysis, in which the sample is takes into account variances and covariances in the multivar-
homogenized into powdered form and pressed into a pellet, iate group. Like standard deviation units, Mahalanobis dis-
or selective analysis of highly polished surfaces. However, all tances can be converted into probabilities of group mem-
of these options require substantial commitments of time and bership for individual specimens. For relatively small
resources. In this study, microscopic and petrographic analy- sample sizes, it is appropriate to base probabilities on
ses were employed for the samples that showed large and Hotelling’s T2, which is the multivariate extension of the
irregular phenocrysts to complement XRF analysis. univariate Student’s t.
The interpretation of compositional data obtained from the In addition to PCA and Mahalanobis distance, we conduct-
analysis of archaeological materials is discussed in detail else- ed statistical analysis on subsets of the samples using bivariate
where (e.g., Baxter and Buck 2000; Bieber et al. 1976; Bishop plots of the color groups identified through microscopic and
and Neff 1989; Glascock 1992; Harbottle 1976; Neff 2000) petrographic analyses. Similarity chemical composition de-
and is only summarized here. The main goal of data analysis is pends largely on the similarity of the original magma and
to identify distinct homogeneous groups within the analytical not necessarily on petrological similarity. Thus, it is critical
database. Initial hypotheses about source-related subgroups in to combine petrological and geochemical data to reliably iden-
the compositional data can be derived from non- tify source areas.
compositional information (e.g., archaeological context, visu-
al description, etc.) or by applying various pattern-recognition
techniques to the multivariate chemical data. Some of the
pattern-recognition techniques that have been used to investi- Results
gate archaeological data sets include cluster analysis (CA),
principal components analysis (PCA), and discriminant func- Microscopic and petrographic analyses
tion analysis (DFA). Each technique has advantages and dis-
advantages that may depend upon the types and quantity of Archaeological samples were grouped into 24 major groups
data available for interpretation. through initial microscopic analysis: eight gray, four light
The variables (measured elements) in archaeological and brown, seven brown, two white, two purple, and one mottled
geological data sets are often correlated and frequently large in brown/white. Some of these groups were further divided into
number. This makes handling and interpreting patterns within subgroups for a total of 35 identified groups (including sub-
the data difficult. Therefore, it is often useful to transform the groups) (Table 4). Following petrographic analysis, some of
original variables into a smaller set of uncorrelated variables in these groups were combined (Brown 1b, 2, 5, and 6 were
Table 3 Summary data of elemental concentrations (ppm) for geological samples of andesitic rocks

Source area Na K Ca Ti Mn Fe Zn Ga Rb Sr Y Zr Nb

Colorado (n = 3) Mean 27,674.86 11,094.22 28,575.03 3810.74 306.51 22,277.51 63.25 16.43 16.77 683.66 36.09 162.19 0.87
Std. Dev. 726.01 2777.02 7148.18 432.16 71.25 914.28 11.33 0.33 6.07 11.65 15.62 19.59 0.29
Archaeol Anthropol Sci

CV (%) 2.62 25.03 25.02 11.34 23.25 4.10 17.92 2.00 36.21 1.70 43.27 12.08 33.39
Malinalco (n = 10) Mean 24,235.56 11,930.21 30,799.53 4780.81 298.90 25,961.79 58.67 16.21 28.14 705.59 44.04 168.16 4.51
Std. Dev. 2871.83 3154.88 12,714.78 1262.42 116.61 2928.49 11.54 0.29 8.06 115.55 21.38 17.75 4.22
CV (%) 11.85 26.44 41.28 26.41 39.01 11.28 19.67 1.77 28.64 16.38 48.55 10.56 93.52
Zacualuca (n = 3) Mean 21,469.73 17,897.55 30,373.46 3243.37 279.40 26,145.84 53.06 15.91 58.25 335.70 26.04 168.31 3.92
Std. Dev. 4059.11 1425.20 7970.98 767.13 75.76 5137.27 3.83 0.08 15.38 35.93 1.89 15.00 3.27
CV (%) 18.91 7.96 26.24 23.65 27.12 19.65 7.21 0.50 26.40 10.70 7.27 8.91 83.48
Nixcuyo (n = 1) Mean 16,837.36 12,309.78 29,932.83 3906.19 243.43 34,018.89 73.03 16.41 56.25 365.20 31.67 167.48 3.30
Clacaiteipa (n = 1) Mean 23,009.57 8527.06 33,920.53 4153.79 318.77 28,400.23 62.69 16.03 34.57 512.08 31.49 140.00 1.88
Belem (n = 8) Mean 29,265.47 18,529.20 13,690.27 2115.66 341.48 16,278.81 46.80 15.69 85.15 488.71 53.50 147.21 4.74
Std. Dev. 2685.16 1927.87 3357.87 889.70 163.50 2505.10 9.95 0.27 12.79 47.80 46.06 24.74 1.51
CV (%) 9.18 10.40 24.53 42.05 47.88 15.39 21.26 1.69 15.02 9.78 86.09 16.80 31.75
Xoxoqui (n = 6) Mean 25,763.81 16,740.86 14,770.18 4160.94 505.73 25,454.32 50.04 15.99 77.76 877.95 46.60 165.10 4.87
Std. Dev. 6797.46 1498.71 2483.26 780.97 313.09 7132.64 8.61 0.28 3.82 85.66 12.20 4.77 2.53
CV (%) 26.38 8.95 16.81 18.77 61.91 28.02 17.21 1.76 4.91 9.76 26.18 2.89 51.98
Santiago La Cruz (n = 5) Mean 20,583.42 12,185.18 22,847.12 5533.19 221.92 28,553.58 49.50 16.10 43.13 1086.88 79.52 183.01 3.93
Std. Dev. 4873.01 2545.30 5438.83 945.74 104.68 4968.62 14.20 0.41 8.30 118.39 24.55 11.13 3.11
CV (%) 23.67 20.89 23.81 17.09 47.17 17.40 28.69 2.53 19.24 10.89 30.88 6.08 79.28
Patlachique (n = 4) Mean 16,682.44 18,722.08 35,150.81 6269.63 306.41 35,059.03 75.94 16.78 14.80 885.45 58.06 172.27 1.13
Std. Dev. 1350.85 967.37 1038.24 337.18 24.41 1658.06 4.63 0.08 2.71 20.87 5.63 10.43 0.32
CV (%) 8.10 5.17 2.95 5.38 7.97 4.73 6.10 0.50 18.33 2.36 9.69 6.05 28.26
Cerro La Cruz (n = 2) Mean 27,099.79 14,559.28 19,952.43 3535.83 175.39 18,892.40 43.97 15.71 77.64 594.11 42.45 168.62 7.93
Std. Dev. 135.14 192.16 774.04 525.11 49.36 302.87 0.81 0.11 0.38 20.42 0.78 14.35 1.25
CV (%) 0.50 1.32 3.88 14.85 28.14 1.60 1.84 0.68 0.49 3.44 1.83 8.51 15.73
Huixtoyo (n = 1) Mean 27,753.45 18,594.12 24,922.27 3052.64 196.33 17,864.89 49.02 15.80 83.85 519.47 31.45 163.22 6.03
Tepetitlan (n = 2) Mean 27,796.70 16,077.95 19,945.84 3281.11 280.98 19,723.45 50.78 15.92 61.67 574.27 33.59 156.95 5.06
Std. Dev. 796.52 585.94 3094.02 368.92 32.93 1507.02 9.95 0.27 12.62 43.89 9.15 17.67 4.25
CV (%) 2.87 3.64 15.51 11.24 11.72 7.64 19.59 1.68 20.47 7.64 27.23 11.26 84.03
Jolalpan (n = 7) Mean 33,360.60 21,719.07 14,078.22 1788.49 378.69 11,995.95 36.94 15.22 104.16 319.96 26.74 120.23 5.88
Std. Dev. 3453.27 3448.17 1942.11 408.00 70.52 3313.34 4.94 0.18 13.90 43.03 6.07 20.67 3.07
CV (%) 10.35 15.88 13.80 22.81 18.62 27.62 13.37 1.19 13.34 13.45 22.69 17.19 52.33
La Flor (n = 4) Mean 26,590.04 23,666.00 18,314.46 2592.15 447.31 21,750.28 45.53 15.76 49.67 446.95 22.73 143.68 5.22
Std. Dev. 1213.63 5550.80 3030.93 1610.90 70.31 2272.06 2.98 0.10 27.69 149.40 7.08 17.45 4.16
Archaeol Anthropol Sci

79.73

68.58
4.15
2.85
0.39
grouped together based on the presence of oxyhornblende),
Nb resulting in 32 groups. Moreover, it appeared that four out of
the 408 archaeological samples were basalt and therefore ex-

151.44
12.15

20.70
13.67
98.63
cluded from subsequent analysis. Among the color groups,
Zr

Gray 1 and Gray 2 comprised the majority of samples (about


35 and 39%, respectively), followed by Gray 3 (about 4%),
31.14

33.00
24.08
17.07

7.95 Mottled Brown/White (3%), Light Brown 1 (3%), White 1


Y

(2.2%), and Brown 1 (2%). Although the samples are not


statistically representative, it is unlikely that the overall pattern
354.44
741.78

103.11
33.43

29.09

changed drastically. During the recording process in the field,


Sr

it was noted that fine-grained gray andesite comprised the


majority of non-sampled cut stone blocks.
55.75

34.43
51.85
66.40
11.35

Through the comparison of color groups with geological


Rb

samples (Table 4; see Figs. 6, 7, and 8), some groups were


provisionally assigned to specific sources. Samples in the
15.31
15.78
0.64

0.30
1.92

Light Brown 1 group are nearly identical to geological sam-


Ga

ples from Cerro Patlachique, and those in the Mottled Brown/


White group are very similar to geological samples from
35.28
48.06

19.38
6.54

9.31

Belem (see below). Other color groups were also assigned to


Zn

possible sources but less confidently than our assignments for


the two groups mentioned above.
18,341.26
20,914.93
6795.85

In addition to the groups similar to the geological sources


10.45

32.49

sampled, there are other groups most likely originating from


Fe

sources that were not sampled. The Gray 1 and 2 groups,


which comprise the majority of samples, are totally different
316.99
226.04
15.72

11.49
3.63
Mn

from the geological samples, being only somewhat similar to


Coatlinchan B. The White 1 group is also completely different
from sampled geological sources; it is probably trachyte. The
2982.65
3362.85

858.97
62.15

28.80

White 2 group was assigned to oxyhornblende (lamprobolite)


CV coefficient of variation (standard deviation divided by the mean, then multiplied by 100)
Ti

andesite, which was not observed in our geological samples


(but Colorado and Cerro Patlachique do contain
22,969.83
20,403.57

oxyhornblende). A major source of oxyhornblende andesite


8435.00
16.55

41.34

is found in the Guadalupe range on the west coast of Lake


Ca

Texcoco (López Luján et al. 2003), and so it is possible that


the White 2 group came from the Guadalupe range.
22,146.19
6465.19
6834.67

Sotomayor Castañeda (1968) suggests that some grinding


23.45

29.19

tools at Teotihuacan were made of rocks from the


K

Guadalupe range.
29,887.19
26,011.29
4655.08
17.90
4.56

Geochemical analysis
Na

Preliminary analysis of the compositional data verifies that


Std. Dev.
CV (%)

CV (%)

most samples analyzed here can correctly be classified as ei-


Mean
Mean

ther andesite or dacite according to the discriminant diagram


of Winchester and Floyd (1977). Source samples analyzed in
this study fall predominantly within the andesite and dacite
range of volcanic rocks, with some outliers from La Flor and
Table 3 (continued)

Santo Tomas (n = 1)
Coatlinchan (n = 3)

Belem falling within the compositional range of rhyolite and


rhyodacite (Fig. 9a). This is not unexpected given that samples
Source area

from La Flor derived from secondary (i.e., boulder) deposits


that may represent varying lithologies, and samples from
Belem were classified as dacite through petrographic analysis.
Table 4 Summary data of color groups identified through microscopic analysis cross-tabulated with macrogroups created through XRF analysis

Color group n (%) Rock type Description Result of XRF (n) Possible source

Teotihuacan High Rb
Valley Macrogroup Outliers
Macrogroup
Archaeol Anthropol Sci

Gray 1a 114 (28%) Dacite Fine to glassy groundmass with small mafics 1 28 2 Non-local (Tepeapulco/Pachuca?)
(biotite and hornblende) scattered and
some small to medium plagioclase phenocrysts
Gray 1b 30 (7.4%) Dacite/andesite Fine to glassy groundmass with abundant small 6 9 1 Non-local (Tepeapulco/Pachuca? Texcoco
mafics (biotite and hornblende) and small region?)
to medium plagioclase phenocrysts
Gray 2a 79 (20%) Andesite Glassy to fine groundmass with few but 26 1 Texcoco region
developed plagioclase phenocrysts and
small hornblende scattered. Gray 2
group is characterized by the presence
of well-developed large
(> 2 mm) phenocysts of plagioclase
Gray 2b 26 (6.4%) Andesite Glassy to granular groundmass with small mafics 11 1 3 Texcoco region
(hornblende and biotite) scattered and
abundant small to large plagioclase
phenocrysts
Gray 2c 16 (4%) Andesite Glassy to fine groundmass with small-medium 7 Texcoco region
mafics (hornblende and biotite) scattered
and large plagioclase phenocrysts
Gray 2d 24 (5.9%) Andesite Glassy to fine groundmass with small to large 10 Teotihuacan Valley
mafics (hornblende and biotite) and Texcoco region
developed plagioclase phenocrysts
Gray 2e 10 (2.5%) Andesite Fine to medium groundmass with scattered/ 4 1 Teotihuacan Valley
crowded medium mafics (hornblende and
biotite) and developed feldspar phenocrysts
Gray 2f 4 (1%) Trachyte/dacite Glassy groundmass with zoned oxidized 1 1 Non-local (Tepeapulco/Pachuca?)
voids, plagioclase phenocrysts (some zoned Texcoco region
growth), and scattered small
hornblende and biotite
Gray 3a 9 (2.2%) Andesite Fine to glassy groundmass with abundant 8 Texcoco region
medium-large plagioclase phenocrysts
and few mafics
Gray 3b 5 (1.2%) Andesite Fine to glassy groundmass with abundant 3 Texcoco region
plagioclase phenocrysts of varying sizes and
few mafics. Gray 3 group is characterized by
the reduced amount of mafics
Gray 3c 2 (0.5%) Andesite Fine to glassy gray groundmass with medium 1 1 Texcoco region
to large plagioclase phenocrysts and few
mafics. Some olivine phenocrysts
as accessories
Table 4 (continued)

Color group n (%) Rock type Description Result of XRF (n) Possible source

Teotihuacan High Rb
Valley Macrogroup Outliers
Macrogroup

Gray 4 5 (1.2%) Trachyte/rhyolite Fine to granular groundmass with abundant 2 2 W Teotihuacan Valley
and crowded plagioclase phenocrysts Non-local
with some small hornblende.
Some samples are very glassy with
quartz phenocrysts, similar to Gray 8
Gray 5 4 (1%) Dacite-andesite Fine purplish glassy groundmass with 3 Teotihuacan Valley
few phenocrysts (some plagioclase
and hornblende)
Gray 6 4 (1%) Dacite-andesite Fine to glassy groundmass with xenomorphic 3 Texcoco region
phenocrysts. Some plagioclase, biotite,
and olivine phenocrysts
Gray 7 1 (0.2%) Andesite Very fine to glassy groundmass with some 1 Teotihuacan Valley
developed elongated plagioclase and biotite
phenocrysts and small rounded mafics
scattered. Some olivine phenocrysts
Gray 8 1 (0.2%) Rhyolite-rhyodacite Glassy mottled dark gray and light brown 1 Non-local (Tepeapulco/Pachuca?)
groundmass with some quartz and plagioclase
phenocrysts and few mafics
Light Brown 1 12 (3%) Oxyhornblende dacite Fine light brown groundmass fluidal structure 2 Patlachique
with oxyhornblende phenocrysts scattered
Light Brown 2 4 (1%) Trachyte/rhyolite Very fine to glassy groundmass with few 2 Non-local (Otumba? Tepeapulco/Pachuca?)
phenocrysts and mafics
Light Brown 3 2 (0.5%) Rhyolite Mottled fine (microcrystalline; light brown) and 1 Non-local (Otumba? Tepeapulco/Pachuca?)
glassy (white) groundmass with quartz and
plagioclase phenocrysts. Small mafics
scattered
Light Brown 4 5 (1.2%) Andesite Fine light brown groundmass with abundant 3 Patlachique range
small-medium plagioclase and some alkaline
feldspar phenocrysts. Few mafics.
Brown 1a 6 (1.5%) Andesite Fine to glassy gray groundmass with some well- 4 Teotihuacan Valley
developed plagioclase and abundant
hornblende and biotite phenocrysts
Brown 1c 2 (0.5%) Andesite Glassy to granular pink groundmass with 2 Teotihuacan Valley
abundant well-developed plagioclase
and some hornblende and biotite phenocrysts.
Some oxyhornblende
Brown 1b, 2, 5, and 6 6 (1.4%) Oxyhornblende Fine to grassy pink-brown groundmass 4 1 Colorado
andesite with plagioclase phenocrysts of varying
sizes and scattered oxyhornblende phenocrysts
Archaeol Anthropol Sci
Table 4 (continued)

Color group n (%) Rock type Description Result of XRF (n) Possible source

Teotihuacan High Rb
Valley Macrogroup Outliers
Macrogroup
Archaeol Anthropol Sci

Brown 3 1 (0.2%) Andesite Glassy mottled gray and pink groundmass with 1 Patlachique range
some plagioclase and hornblende phenocrysts.
Some quartz and alkali feldspar
Brown 4 1 (0.2%) Andesite Very fine to glassy groundmass with some 1 Patlachique range
plagioclase phenocrysts of varying sizes and
small hornblende. Some olivine phenocrysts
Brown 7 1 (0.2%) Andesite Fine to glassy groundmass with few 1 Cerro La Cruz
plagioclase phenocrysts and abundant and
scattered hornblende
White 1 9 (2.2%) Rhyolite/trachyte Very fine to glassy groundmass with some 5 Non-local (Otumba? Tepeapulco/Pachuca?)
plagioclase phenocysts and small
hornblende scattered. Some
specimens are almost devoid of phenocrysts
White 2 6 (1.5%) Oxyhornblende Very fine to glassy groundmass with some 3 Texcoco region
andesite medium to large plagioclase phenocryst and
oxyhornblende or corroded hornblende
scattered
Purple 1 1 (0.2%) Dacite-andesite Glassy mottled dark gray and pink groundmass 1 Texcoco region
with abundant well-developed and
elongated plagioclase phenocrysts and some
hornblende and olivine
Purple 2 1 (0.2%) Rhyolite-dacite Fine purple groundmass with few plagioclase 1 La Cruz?
phenocrysts and some small mafics scattered
Mottled 13 (3.2%) Dacite Fine to glassy mottled white and brown 4 2 Belem
Brown/White groundmass with some plagioclase non-local
phenocrysts and abundant small
oxyhornblende phenocrysts scattered
Total 404 (100%) 111 39 26 Total = 176
Archaeol Anthropol Sci

Fig. 6 Cross-polarized micrographs of geological samples (see Table 1 for the abbreviation of source areas): Plg = plagioclase, Hbl = hornblende, Bio =
biotite, Qtz = quartz, Ol = olivine, Oxy = oxyhornblende, Afs = alkali feldspar, Opx = orthopyroxene

In terms of lithology, archaeological samples from PC-transformed data fail to reveal source-specific sub-
Teotihuacan show markedly greater variation relative to the group structure.
source samples (Fig. 9b), ranging from transitional andesite/ A second PCA conducted on only the samples comprising
basalt to trachyte and peralkaline rhyolite (i.e., comendite and the source database (i.e., excluding artifacts) reveals that, un-
pantellerite). Of the architectural complexes sampled here, the like with the entire dataset, alkali metals and Nb account for
West Plaza Complex shows the greatest variation in terms of most of the variation in the dataset (Table 5). Biplots of these
volcanic lithologies. scores suggest that at least some of the better-represented
PCA of the XRF data reveal that greater than 90% of the sources can be successfully discriminated in multivariate
cumulative variation in the 237-member dataset can be space (Fig. 10a, b). This is particularly true if the boulder-
described by six principal components (PCs). The first derived samples from La Flor are removed from consider-
six PCs are most heavily loaded by immobile elements ation, something entirely warranted considering that these
(e.g., Nb, Y, Zr, Ti, Sr), suggesting that some petrogenic samples may have derived from multiple geological forma-
differences may be detectable within the sample. The first tions and locations.
PC is strongly loaded on Sr and Ti in the negative direc- Bivariate plots of logged elemental concentrations also re-
tion. The second PC is positively loaded on Nb. Biplots of veal slight subgroup structure among individual sources
Archaeol Anthropol Sci

Fig. 7 Cross-polarized micrographs of gray groups: Plg = plagioclase, 29. i CDC-8. j CDC-19. k TetC-19. i QPC-15. m YayC-2. n WPCC-52. o
Hbl = hornblende, Bio = biotite, Qtz = quartz, Ol = olivine. a CDC-13. b SPC-3. p LVIIIC-12 (see Table 2 for the abbreviation of architectural
LVIC-2. c SPC-14. d AtetC-3. e SPC-16. f CDC-15. g CDC-17. h CDC- complex)

(Fig. 10c). Unfortunately, there is no straightforward correla- the artifact samples suggest that the tightly clustered group of
tion between elemental abundance and geological source. For samples originates from a single source. Their distinctiveness
example, the samples from Cerro Colorado and Cerro from other samples is attributed to either (1) a fundamentally
Malinalco tend to plot with each other in bivariate and multi- different petrology, (2) a chemically distinct source, or (3) a
variate space, as might be expected since these sources are combination of both. Cross-validation of these samples with
relatively close together in geographic space, but the single the rock classification diagrams presented earlier suggests that
sample from Santo Tomas is also compositionally similar, they fall within the dacite/rhyodacite range of compositions;
despite the fact that this source is located over 40 km to the thus, the chemical data may reflect petrology. Yet, it should be
south. noted that these artifacts are compositionally distinct from the
Examination of the artifact samples analyzed in this study dacites and rhyolites present in the source database, and there-
reveals slight clustering, suggestive of significant chemical fore, the data appear to reflect as well. We prefer a conserva-
similarities among samples. While the majority of samples tive approach for handling these results and thus discern at
reflect the broad compositional variability represented by the least two compositional macrogroups: one reflecting the broad
various sources in the Basin of Mexico, a tightly clustered compositional variability observed within the Teotihuacan
group of artifact samples falls well outside of this range. Valley and the Texcoco Region (designated as the
Specifically, these samples have consistently higher concen- Teotihuacan Valley Macrogroup) and the other representing
trations of Rb and K (Fig. 10d) and lower concentrations of Sr, a different chemistry that is suspected—but not proven—to be
Zr, Zn, Fe, and Ca. The chemical differences observed within non-local (called the High Rb Macrogroup). Using
Archaeol Anthropol Sci

Fig. 8 Cross-polarized micrographs of brown groups: Plg = plagioclase, 42. d WPCC-25. e CDC-28. f WPCC-16. g WPCC-65. h LVIC-39 (see
Hbl = hornblende, Bio = biotite, Qtz = quartz, Ol = olivine, Oxy = Table 2 for the abbreviation of architectural complex)
oxyhornblende, Afs = alkali feldspar. a CDC-35. b LVIC-32. c CDC-

Mahalanobis distance-based probabilities, these two groups petrographic analyses (Table 4). This accords well with the
can be shown to be statistically distinct from one another, microscopic observation results of Gray 1 samples, which
and artifact samples unassignable to either group can easily suggest they are from non-sampled geological sources. All
be identified (called outliers) (Fig. 11). of the White 1 samples were assigned to outliers, which also
accords well with microscopic observation. These samples
were identified as rhyolite or trachyte through microscopic
Discussion of the combined results and petrographic analyses. Compositional data also indi-
cate that some are rhyolite and others trachyte. These
The comparison of color group and PCA macrogroup assign- types of rock were not collected during the sampling.
ments (Table 4) indicates that some color groups can be se- Some Gray 4 (rhyolite and trachyte), Gray 8 (rhyolite-
curely assigned to sources not sampled in this study. The ma- dacite), and Light Brown 3 (andesite) samples were also
jority of the samples composing the High Rb Macrogroup are assigned to outliers; microscopic analysis indicated no
those assigned to Gray 1a or 1b through microscopic and similar geological samples.

Fig. 9 Classification of volcanic rocks based on immobile elements (after Winchester and Floyd (1977)) showing source samples from the Basin of
Mexico (a) and artifact samples from Teotihuacan (b). Samples not classified as andesite or dacite (rhyodacite) are labeled with MURR ANIDs
Archaeol Anthropol Sci

Table 5 Principal component


analysis of the 61-member dataset PC1 PC2 PC3 PC4 PC5 PC6
of volcanic-stone sources from
the Basin of Mexico % Variance 50.863 22.92 6.569 5.397 4.12 3.817
Cum. % Variance 50.863 73.783 80.352 85.749 89.87 93.686
Total Structure Coefficients
Ti −0.307 0.4583 0.0868 −0.2055 0.5661 0.5231
Ca −0.2268 0.1262 0.4671 −0.1053 0.0348 −0.3139
Sr −0.2182 0.2812 −0.3265 0.1178 −0.0905 0.1563
Fe −0.2021 0.205 0.023 −0.0619 0.1805 −0.4232
Y −0.1121 0.4492 −0.5738 0.2126 −0.2428 −0.2022
Zn −0.1091 0.1477 0.0287 −0.145 0.031 −0.2997
Zr −0.0346 0.1408 −0.0103 0.0114 0.0444 −0.2468
Ga −0.0153 0.0168 −0.0025 −0.0107 0.0016 −0.0339
Mn 0.0704 −0.0974 −0.3986 −0.8265 −0.0872 0.0132
Na 0.1103 −0.1259 −0.0148 −0.0723 −0.1772 0.3364
K 0.1753 −0.1151 −0.124 −0.2482 0.3849 −0.3399
Rb 0.5587 0.0171 −0.2495 0.2812 0.5546 −0.042
Nb 0.6164 0.6085 0.3166 −0.1832 −0.2846 0.0301

The first six principal components are shown, accounting for greater than 90% of the cumulative variation in the
dataset. The first two components are most heavily influenced by alkali metals, niobium, and manganese (see Fig.
10a). Numbers in bold show moderate-high correlation with respective component

PCA assignment of macrogroups also provides some prob- or broader regions (i.e., the Teotihuacan Valley or the Texcoco
lematic results. While Light Brown 1 samples were assigned region). The results are summarized in Table 4.
to Cerro Patlachique through microscopic analysis, these sam- Based on the results of PCA mentioned above, four ele-
ples were designated outliers by PCA. It is possible that the ments, Rb, Sr, K, and Fe, were selected to show the variability
irregular surfaces and/or non-homogeneous distribution of el- among geological and archaeological samples. Comparison of
ements produced erratic results. Brown 6 (dacite-andesite) is Rb and Sr reveals some correlation between compositional
somewhat similar to Malinalco samples but was classified an data and geographical areas (Fig. 12). Specifically, geological
outlier. It is not clear whether this is due to analytical uncer- samples from each of the northeastern (Santiago La Cruz and
tainty or whether it is from another source. Although we sug- Xoxoqui) and western (La Cruz, Huixtoyo, and Tepetitlan)
gested earlier that White 2 samples (oxyhornblende andesite) portions of the Patlachique range show clear clustering and
possibly came from the Guadalupe range on the west coast of can be isolated from the rest of the samples. Samples from the
Lake Texcoco, these samples were assigned to the southern portion of the Patlachique range (Jolalpan) and those
Teotihuacan Valley Macrogroup by PCA. It is possible that from the Texcoco region (La Flor and Coatlinchan A and B)
there are oxyhornblende andesite sources in the Texcoco re- also exhibit clustering, although a sample from the western
gion or the Guadalupe and Texcoco sources share similar Teotihuacan Valley (Zacualuca) does fall within the range of
compositions. At any rate, oxyhornblende andesite is not pres- variability of the samples from the south. Conversely, a sam-
ently documented within the Teotihuacan Valley, suggesting ple from the Texcoco region (Santo Tomas) falls near the
that the White 2 group likely came from outside the valley. range of the Malinalco samples. Since our sampling in the
Geochemical compositional results were analyzed further Texcoco region was not extensive, it is possible that signifi-
by comparing bivariate plots against color groups. This allows cant overlap exists between sources in the western
us to illustrate the compositional variability among petrologi- Teotihuacan Valley and the Texcoco region. We suggest the
cally similar rocks. Geological samples were divided into gray possibility that this overlap is due to the similar geological age
andesite-dacite (including purple), brown andesite-dacite (in- of some sources, although there are controversies with respect
cluding light brown), and mottled brown/white dacite, which to the dating of geological events in these regions. Thus, it is
were then compared against archaeological samples of the likely that the variability among the western Teotihuacan
same color groups (i.e., gray, brown, and mottled brown/ Valley and the Texcoco region can be accounted for by both
white groups). Samples of the White 1 and 2 groups were petrology and the geological age of sources.
excluded from this analysis. Based on this comparison, each The bivariate plot of K against Fe also shows a gen-
color group was assigned to specific sources, possible areas, eral division between the Teotihuacan Valley and the
Archaeol Anthropol Sci

Fig. 10 Bivariate plots based on geochemical data. a Biplot of the first membership). d Bivariate plot of rubidium and potassium
two principal components scores for geological samples (samples from concentrations for both geological and archaeological samples (ellipses
La Flor are removed). b Biplot of the first and third component scores. c represent the 1σ confidence interval for group membership in geological
Bivariate plot of rubidium and strontium concentrations for geological sources. Extreme outliers removed). Note clustering of artifacts at the
samples (ellipses represent the 1σ confidence interval for group upper left of chart (high Rb-high K)

Texcoco region. Samples from the western Teotihuacan Gray andesite-dacite


Valley and the northeast portion of the Patlachique range
are clustered together, and both of which can be isolated Based on the tentative divisions among geographically dis-
from the Texcoco samples. Samples from the west por- tinct source areas for gray andesite-dacite, archaeological sam-
tion of the Patlachique range fall between these two re- ples were compared against geological samples using the
gions. As with the plot of Rb and Sr, there is an overlap same bivariate plots. As indicated by PCA, the majority of
between the two regions; samples from Coatlinchan A Gray 1 samples are outside of the range of both Teotihuacan
are located within the range of the Teotihuacan Valley and Texcoco sources (Fig. 12b).
samples. A sample from Santo Tomas in the Texcoco The source area for these samples is unknown, but we
region is well outside the range of the Texcoco sources suggest a location somewhere in Central Mexico. The biplots
but also apart from the Teotihuacan sources as well. clearly show that Gray 1 samples form part of a continuum in
Given the uncertainty of non-destructive XRF analysis the chemical composition of geological samples. The nearest
as employed here, it is possible that the chemical com- source areas for dacite/rhyodacite include the Otumba region
position of the Santo Tomas sample may be in error, or and an area between the northeastern limit of the Teotihuacan
perhaps, this particular piece of stone is not representa- Valley and Pachuca, both of which contain obsidian sources.
tive of the Santo Tomas source locality. Geochemical composition of obsidian has been well studied
Archaeol Anthropol Sci

Fig. 11 Bivariate plots showing the two macrogroups of artifacts defined plot of potassium and iron concentrations. b Bivariate plot of rubidium
within the assemblage from Teotihuacan (compositional groups of and zirconium concentrations (outliers are labeled with MURR ANIDs)
geological sources are shown by 1σ confidence ellipse). a Bivariate

and those studies indicate that Pachuca obsidian has a higher other types of rocks with different cooling processes, specifi-
concentration of Rb (e.g., Carballo et al. 2007). Since obsidian cally rhyolite and probably rhyodacite and dacite. In addition
forms from rhyolitic lava, the geochemical variations between to Pachuca obsidian, two other obsidian sources, the
these two sources of obsidian should also reflect variations of Zacualtipan and Paredon (northwest and southeast of

Fig. 12 Bivariate plots of Rb and Sr concentrations for geological samples (a), Gray 1 group (b), Gray 2 group (c), Gray 3 group (d), and the rest of gray
and purple groups (e)
Archaeol Anthropol Sci

Pachuca, respectively) obsidian sources, also have high Rb based on Rb and Fe contents. The Sr and K concentra-
concentrations (Carballo et al. 2007, Table 3). This suggests tions allow for further division of the Patlachique range
that an area along the Sierra Madre Oriental, which runs from sources into northwestern, western, and southern portions.
the northwest to southeast in Mexico, is formed from Rb-rich Samples from Cerro Patlachique cluster with sources from
rocks. Based on these observations, we suggest that the geo- the northeastern portion of the Patlachique range accord-
logical source of the High Rb Macrogroup is located not with- ing to Sr and K contents but depart from them according
in the Basin of Mexico, but to the north of the Teotihuacan to Fe and Rb contents. This may attest to the difference in
Valley, probably near Pachuca. geological age between Cerro Patlachique and adjacent
Within the Gray 1 group, some Gray 1b rocks were mountains as suggested by Mooser (1968).
assigned to the Teotihuacan Valley Macrogroup (Fig. 12b). The comparison of archaeological samples of Light
It is possible that they came from the Texcoco region where Brown and Brown groups against geological samples in-
samples contain higher concentrations of Rb and K and lower dicates that all the archaeological samples fall within the
concentrations of Sr and Fe compared to sources within the range of Teotihuacan sources with the exception of the
Teotihuacan Valley. Microscopic and petrographic analyses few outliers (Light Brown 2 and 3) defined by PCA men-
suggest that some Gray 1b samples are somewhat akin to tioned above. The plots (Fig. 13) show that Light Brown
samples from La Flor. 1 samples are clustered with Patlachique samples even
Gray 2 samples fall within the range of Teotihuacan and though they were assigned as outliers by PCA.
Texcoco sources, with the majority likely coming from the Microscopic and petrographic analyses show that Light
Texcoco region (Fig. 12c). Most Gray 2a and 2b samples are Brown 1 samples are strikingly similar to those of Cerro
within the range of Texcoco sources, and Gray 2e samples Patlachique samples (Fig. 13) and can thus be assigned to
cluster with Teotihuacan sources, although not necessarily Cerro Patlachique. As previously noted, the non-
within the range of sampled sources. Gray 2c and 2d are likely destructive methodology employed here may have intro-
from a variety of sources in the Teotihuacan Valley and duced analytical uncertainty in the XRF data, so we are
Texcoco region. It should be noted, however, that geological hesitant to rely exclusively on this source of data alone as
samples similar to Gray 2 samples were not identified, so a means of discriminating sources.
future research should attempt to locate their sources. Light Brown 2 and 3 samples are also considered out-
Gray 3, 5, 6, and 7 and Purple 1 samples fall within the liers within the PC-transformed data, and the bivariate
range of either or both Teotihuacan and Texcoco sources plots show that they are not clustered with any
(Fig. 12d). Microscopic and petrographic analyses reveal a Teotihuacan sources. Petrographic analysis suggests that
similarity between some archaeological and geological speci- they are trachyte and/or rhyolite-trachyte, not present in
mens (see Figs. 6, 7, and 8), which is consistent with geo- our geological dataset. While Light Brown 4 samples fall
chemical data. However, it was not possible to assign archae- geochemically within the range of the western and south-
ological samples to specific geological sources. Geological ern portions of the Patlachique range (Fig. 13), we do not
samples from the Texcoco region are underrepresented in this find any similar petrology in our source materials. It is
study and it is likely that varying lithologies with similar geo- possible that they came from sources in the Patlachique
chemical composition exist in this region. range not sampled in this study.
Gray 4, Gray 8, and Purple 2 samples were identified as Brown 1b, 2, 5, and 6 samples were assigned to Cerro
other types of rocks (i.e., rhyolite and trachyte) and designated Colorado through petrographic analysis, which is consistent
outliers by PCA. We discuss the possible sources of these with geochemical analysis, with the exception of Brown 6
outliers at the end of this section. (Fig. 13). These samples contain varying amounts of
oxyhornblende and developed plagioclase phenocrysts.
Brown andesite-dacite Within PC space, Brown 6 is an outlier, but the bivariate plots
clearly show that it is closely clustered with Colorado/
Brown variants of andesitic rocks are nearly limited to the Malinalco samples.
Teotihuacan Valley and the southern side of the Brown 1a, 1c, 3, and 4 samples were not assigned to
Patlachique range, although this does not preclude the any source though petrographic analysis, but the bivariate
possibility that there are unknown sources of brown an- plots show that they are within the range of Teotihuacan
desitic rocks in the Texcoco region. As with gray types, sources (probably the Patlachique range). Of these sam-
bivariate plots of four elements (Rb, Sr, K, and Fe) indi- ples, the Brown 4 sample exhibits similar petrological
cate that compositional variability is more or less corre- composition in some respects to those from La Flor in
lated with geographical areas (Fig. 13). Several sources in the northern Texcoco region. Since this source is a sec-
the Patlachique range (except for Cerro Patlachique) are ondary deposit, it is possible that some of the original
distinguishable from West Teotihuacan Valley sources rocks derived from the Patlachique range. All these
Archaeol Anthropol Sci

Fig. 13 Bivariate plots of Rb and Sr concentrations (left) and K and Fe concentrations (right) for geological and Brown group samples. Micrographs of
geological and archaeological samples of similar composition are also shown

samples are characterized by well-developed phenocrysts petrologic similarities, the archaeological samples of the
of plagioclase, hornblende, and/or biotite. Mottled Brown/White group were assigned to Belem through
Petrographic observation showed that one Brown 7 sample microscopic and petrographic analyses. The comparison of
is very similar to the Cerro La Cruz samples from the western archaeological samples against geological samples from
Patlachique range (Fig. 13). However, the bivariate plots sug- Belem and Zacualuca based on geochemical data generated
gest that Brown 7 came from the western Teotihuacan Valley, by XRF analysis (Fig. 14) shows that most archaeological
probably near Nixcuyo. samples cluster with the geological samples from Belem.
Two of the archaeological samples were designated as outliers
Mottled brown and white by PCA as mentioned above, and in the bivariate plots, they
do not cluster with any source material. However, given their
Mottled brown and white rocks were identified as dacite and striking similarity in petrology to other samples, it is likely
collected from only two sources within the Teotihuacan that some factors, such as non-homogeneous distribution of
Valley: Belem in the northeastern portion of the Patlachique elements over the surface, resulted in poor characterization by
range and Zacualuca in the western Teotihuacan Valley. It is our XRF method. We suggest that all the Mottled Brown/
possible that more sources of these rocks exist, but based on White samples are from Belem.
Archaeol Anthropol Sci

Fig. 14 Bivariate plot of Rb and


Sr concentrations for the Mottled
Brown/White group and micro-
graphs of geological and archae-
ological samples

Geochemical outliers and trachyte is located in the Soltepec range at the eastern
limit of the Teotihuacan Valley, also where the Otumba
Out of 26 outliers identified through PCA, two samples of obsidian source is located (Fig. 15; Hernández Javier
Light Brown 1 show strikingly similar composition to geo- 2007). Since Otumba obsidian was heavily exploited dur-
logical samples from Cerro Patlachique as discussed ing the Teotihuacan period, it is highly likely that other
above. Six of the outliers were identified as rhyolite (two types of rock were also quarried in this area.
of which are peralkaline rhyolite), which pertain to Gray, The rest of the outliers (n = 16) are andesite-dacite,
Light Brown, and White color groups. Additionally, two whose colors include gray, light brown, brown, purple,
samples (Gray and White groups) were assigned to tra- white, and mottled brown/white. Some of these outliers
chyte. Most trachyte and rhyolite specimens contain high have high Rb concentrations. It is possible that both the
concentrations of Rb, so we suggest the possibility that the Tepeapulco/Pachuca area and the Otumba area provided
source areas for these types of rocks and the High Rb those rocks. Lastly, we suggest the possibility that other
Macrogroup are located near each other, probably north sources within the Basin of Mexico, such as the
of Tepeapulco and/or around Sierra de las Navajas Guadalupe range on the west coast of Lake Texcoco
(Pachuca) as discussed above. Aside from the (López Luján et al. 2003), were also exploited. There
Tepeapulco/Pachuca area, the nearest source of rhyolite are several Teotihuacan period settlements on the eastern
Archaeol Anthropol Sci

Fig. 15 Map showing possible


non-local (beyond 10 km radius)
sources (dashed lines) of andesitic
rocks in Central Mexico. Pink-
purple tones represent extrusive
igneous rocks (based on the
INEGI digital map)

and western sides of Lake Texcoco that reflect interaction emphasized that source assignments are still preliminary, and
of varying intensity with Teotihuacan (Clayton 2013). further analysis coupled with extensive sampling of source
Both Azcapotzalco and Axotlan, located on the west coast materials is needed to determine the source of each sample.
of Lake Texcoco, show a close connection with With this caveat in mind, overall patterns show that non-local
Teotihuacan, so it is possible that their residents quarried sources (beyond 10–15 km) were exploited to a greater degree
rocks in the nearby Guadalupe range (Fig. 15) and pro- than local sources in the Teotihuacan Valley. Around 80% of
vided them for urban construction at Teotihuacan. the analyzed samples were assigned to the source of the High
Provenance studies of ceramics in Central Mexico have Rb Macrogroup (possibly the Tepeapulco/Pachuca area) and
successfully discriminated different source areas, such as the Texcoco sources. We suppose that the actual proportion of
the Texcoco region, the Teotihuacan Valley, the Otumba cut stone blocks from non-local sources is higher because the
region, and the Cuauhtitlan region (adjacent to the majority of blocks recorded (but not sampled) in the field are
Guadalupe range) (e.g., Garraty 2013; Neff et al. 2000), fine-grained gray andesite-dacite, which is probably the same
as well as three subregions of the Texcoco region (Minc as the Gray 1 group.
2009). These studies are encouraging as the parent rock of In addition to non-local sources, some mountains in the
local clay and temper is supposedly from the adjacent Patlachique range (southern Teotihuacan Valley) were
mountains. In fact, the geochemical composition of obsid- exploited. Specifically, this study identified rocks from Cerro
ian and ceramics at Otumba shows a substantial overlap Patlachique, Belem, and possibly Cerro La Cruz. Compared to
(though obsidian is much more homogeneous) (Neff et al. the southern valley, rock sources in the western portions of the
2000). Thus, further research is warranted for determining Teotihuacan Valley were underexploited. These overall pat-
the provenance of outliers and the High Rb Macrogroup. terns apply to most architectural complexes.
The use of cut stone blocks was highly restricted to some
buildings in the central precinct during the Tlamimilolpa and
Implications for the procurement Early Xolalpan phases, whereas during the Late Xolalpan to
organization of andesitic cut stone blocks Metepec phases, it was restricted to some intermediate elite
compounds (Murakami 2010). In the following discussion,
Petrographic and XRF analyses strongly suggest that the ma- we focus on the Tlamimilolpa and Early Xolalpan phases
jor sources of andesitic cut stone blocks are not located in and examine the changing procurement organization of cut
close proximity to the city but outside a 10–15-km radius from stone blocks based on diachronic changes in the spatial distri-
the city center. The Tepeapulco/Pachuca area and the Texcoco bution of different source materials among architectural com-
region are the likely source areas (Fig. 15). It should be plexes in the central precinct.
Archaeol Anthropol Sci

The FSP and the Quetzalpapalotl Palace Complex (QPC) blocks from outcrops. Alejandro Pastrana’s research at
contain samples of cut stone blocks from Early Tlamimilolpa Pachuca (Pastrana and Domínguez 2009) has demonstrated
contexts. At both structures, the majority of cut stone blocks that mining groups resided at the quarry, probably for part of
were likely brought from the source of the High Rb the year, to exploit obsidian. Moreover, Teotihuacan period
Macrogroup and, to a minor extent, from the Texcoco region. settlements are found near the quarry, whose residents likely
Gray types predominate among these rocks. In addition, participated in the quarrying of obsidian. It is possible that the
brown type rocks from the southern (e.g., Patlachique and procurement of andesitic cut stone blocks was organized in a
Belem) and western (e.g., Colorado) Teotihuacan Valley are similar fashion.
present but constitute a minor proportion. With respect to sources within the Teotihuacan Valley, it is
The variety of rocks appears to have increased in the central possible that procurement was organized by the state to some
precinct during the Early Xolalpan phase. The major sources extent, guild-like institutions, and/or through community-
of the High Rb Macrogroup and Texcoco region continued to level labor obligations (e.g., Aztec tequitl; see Carballo
be exploited, but the number of different brown types from the 2013) at Belem in the Patlachique range and Colorado/
Teotihuacan Valley increased. Specifically, cut stone blocks of Malinalco in the western valley. These two source areas pro-
Light Brown 4 and Brown 1 through 6 types appeared during vided some large-sized blocks to the FSP. While our sampling
this phase. White 1 and 2 and Purple 1 types were newly was not random, if we assume that our samples are represen-
exploited. All of these types likely came from the southern tative, a simple calculation shows that about 400 blocks
or western valley, with the exception of White types. This (0.2 m3 on average) or 80 m3 of rocks were quarried at each
indicates that the mountains in the southern and western valley of these two source areas (based on the calculation that around
were more extensively exploited than during the Early 10,000 blocks were used for the FSP). The procurement at
Tlamimilolpa phase. The White 1 type probably came from each source was probably not as extensive as this, but even
the eastern valley and the White 2 type from either the to extract 50 or 100 large blocks would require some system-
Texcoco region or the western coast of Lake Texcoco. atic exploitation.
Sampled complexes rebuilt in this phase include the QPC, Cut stone blocks from other sources within the valley were
the Sun Pyramid, three complexes within the Street of the probably obtained by small-scale work groups either indepen-
Dead Complex (the Viking Group, the West Plaza Complex, dently or through tribute or patron-client relations. As pointed
and the Complex of the Superimposed Structures), and the out above, the number of different types of rock increased dur-
Ciudadela. Of these complexes, the West Plaza Complex con- ing the Early Xolalpan phase. The majority of cut stone blocks
tains the most variety of rocks. Including subtypes from our from sources within the valley were used for staircases and
color groups, a total of 20 different types of rock were used at steps delimiting courtyards and rooms. The sizes of these indi-
the West Plaza Complex, followed in frequency by the Viking vidual blocks are considerably smaller than those used for the
Group (11 types), the QPC (10 types), and the Sun Pyramid (9 FSP and the Sun Pyramid. This suggests that procurement did
types). The number of types at the West Plaza Complex is not require extensive specialized knowledge and/or groups,
clearly related to the larger sample and amount of cut stone meaning average persons could have quarried these blocks.
blocks used at the complex, but it might also attest to an Furthermore, procurement activities at a greater number of
importance attached to the complex. It is important to note sources and/or locations within a single source, as opposed to
that the number of rock types does not necessarily equate to an intensive quarrying of a single source, imply that procure-
the number of rock sources, as different types of rock may be ment was less systematically organized than the quarrying at
present in a single mountain. Further provenance studies are the Tepeapulco/Pachuca and Texcoco sources. Most likely in-
necessary to assign each color group identified in this study to dependent work groups quarried limited numbers of blocks at
a specific source location, and much is dependent on the geo- certain locations because they were required to provide stone
logical scale of what we may treat as a source location. blocks to the state and/or because of their value as commodity.
The predominance of a few sources and the near exclusive
use of cut stone blocks for state buildings suggests that their
procurement was organized directly by the state (e.g., labor Conclusion
tax akin to Aztec coatequitl; see Carballo 2013) or indirectly
through other guild-like institutions (e.g., patron-client rela- This study indicates that a limited group of sources of andes-
tions) during the Early Tlamimilolpa and Early Xolalpan itic rocks, most likely non-local, were heavily exploited for the
phases. Considering the large size of cut stone blocks used construction of the central precinct and other buildings. This
for the FSP and the Sun Pyramid, it is likely that the observation, coupled with the unequal distribution of cut stone
Tepeapulco/Pachuca and Texcoco sources were systematical- blocks (Murakami 2010) and the skill required to quarry large
ly and extensively exploited by a group (or groups) of special- blocks, suggests that procurement was systematically and cen-
ists, who possessed the knowledge to extract large-sized trally organized by a specialized group or groups of masons.
Archaeol Anthropol Sci

Given the concentrated distribution of cut stone blocks within Acknowledgements The lead author (Murakami) would like to thank the
Consejo de Arqueologia for their permission to conduct research, as well
the central precinct, it seems reasonable to assume that these
as Alejandro Sarabia, the director of the Zona Arqueológica de
specialized groups were closely associated with the state gov- Teotihuacan, for providing logistical support for archaeological sampling
ernment. While it is difficult to determine the nature of the at Teotihuacan. The study presented in this paper was made possible by
relationship between the specialized groups and the state with financial support from the National Science Foundation Dissertation
Improvement Grant (BCS-0836716) awarded to Murakami and a
the currently available evidence, we suggest two possibilities:
National Science Foundation grant (0504015) awarded to MURR.
patron-client relations and/or relations embedded in a tribute Finally, we thank Chris Oswald for his assistance in the analysis of the
system. In either case, we argue that the procurement of an- samples by XRF and Alex Jurado for editing help for an earlier draft of
desitic cut stone blocks was administered by the state govern- this paper. Comments and suggestions by two anonymous reviewers
greatly helped clarifying our argument and the scope of the paper.
ment. If we are correct that the Tepeapulco/Pachuca area was
the major source of the High Rb Macrogroup, then it is likely
that elite groups at Tepeapulco administered the procurement
and transportation of rocks. The presence of possible second- References
ary centers in resource-rich areas can also be observed in the
Tula region, where elites of Chingu (Fig. 15) probably admin- Abrams EM (1994) How the Maya built their world: energetics and
istered the production and transportation of lime (Barba ancient architecture. University of Texas Press, Austin
Pingarrón et al. 2009; Murakami 2016). Along with this cen- Barba Pingarrón LA (2005) Materiales, técnicas y energía en la
construcción de Teotihuacan. In: Gallut MER, Peralta JT (eds)
tral procurement and distribution system, we think that other
Arquitectura y urbanismo: pasado y presente de los espacios en
systems at intermediate and small scales coexisted for the Teotihuacan. Memoria de la Tercera Mesa Redonda de
procurement and distribution of cut stone blocks. All of these Teotihuacan. Instituto Nacional de Antropología e Historia,
discussions illustrate the importance of provenance studies of Mexico City, pp 211–229
key resources for understanding the nature of intra-regional Barba Pingarrón LA, Córdova Frunz JL (2010) Materiales y energía en la
arquitectura de Teotihuacan. Instituto de Investigaciones
and inter-regional relations and the expansion of the Antropológicas, Universidad Nacional Autónoma de México,
Teotihuacan state. Mexico City
While the resolution of the geochemical data is insuffi- Barba Pingarrón LA, Blancas J, Manzanilla LR, Ortiz A, Barca D, Crisci
cient to rigorously define source-specific compositional GM, Miriello D, Pecci A (2009) Provenance of the limestone used in
Teotihuacan (Mexico): a methodological approach. Archaeometry
groups, the combined analysis by microscopic observation,
51:525–545
petrography, and XRF served to identify specific source Baxter MJ (1992) Archaeological uses of the biplot—a neglected tech-
areas. We note that, while the initial classification based on nique? In: Lock G, Moffett J (eds) Computer applications and quan-
microscopic observation of freshly broken edges allowed us titative methods in archaeology, 1991, vol BAR International Series.
to analyze all collected samples, it was slightly modified Tempvs Reparatvm, Oxford, pp 141–148
Baxter MJ, Buck CE (2000) Data handling and statistical analysis. In:
following the results of the petrographic analysis. Ciliberto E, Spoto G (eds) Modern analytical methods in art and
Petrography is a powerful tool for determining source areas, archaeology. John Wiley and Sons, New York, pp 681–746
but is limited by the fact that the spatial extent of possible Bieber AMJ, Brooks DW, Harbottle G, Sayre EV (1976) Application of
source areas is immense, and more importantly, varying li- multivariate techniques to analytical data on Aegean ceramics.
thologies exist in a single source area. These factors make it Archaeometry 18:59–74
Bishop RL, Neff H (1989) Compositional data analysis in archaeology.
difficult to obtain representative samples from possible In: Allen RO (ed) Archaeological chemistry IV, vol 220. Advances
source areas. In this respect, this study has demonstrated that in chemistry. American Chemical Society, Washington, D.C., pp
rocks with different lithologies but, from the same area, 576–586
share a similar geochemical composition. Thus, geochemical Biskowski MF (1997) The adaptive origins of Prehispanic markets in
central mexico: the role of maize-grinding tools and related staple
analysis can help to identify source areas irrespective of
products in early state economies. Ph.D. diss. University of
having detailed knowledge of lithologies. That said, similar California, Los Angeles
geochemistry does not necessarily indicate geographic ori- Carballo DM (2013) Labor collectives and group cooperation in Pre-
gin, as shown by the geochemical overlap between the west- Hispanic Central Mexico. In: Carballo DM (ed) Cooperation and
ern Teotihuacan Valley and the Texcoco region (but they do collective action: archaeological perspectives. University Press of
Colorado, Boulder, pp 243–274
have different lithologies). Clearly, the three techniques
Carballo DM, Carballo J, Neff H (2007) Formative and Classic period
employed in this study have their respective advantages obsidian procurement in Central Mexico: a compositional study
and disadvantages, and so it is critical to combine multiple using laser ablation-inductively coupled plasma-mass spectrometry.
methods. This study provides baseline data for future re- Lat Am Antiq 18:27–43. https://doi.org/10.2307/25063084
search, which should include more extensive geological Clayton SC (2013) Measuring the long arm of the state: Teotihuacan's
relations in the basin of Mexico. Anc Mesoam 24:87–105
sampling, petrographic analysis, and more sophisticated geo- Cowgill GL (2000) The central Mexican highlands from the rise of
chemical analyses (e.g., INAA, ICP-MS) to determine spe- Teotihuacan to the decline of Tula. In: Adams REW, MacLeod MJ
cific source areas of andesitic cut stone blocks. (eds) The Cambridge history of the native peoples of the Americas.
Archaeol Anthropol Sci

Volume II: Mesoamerica, Part 1. Cambridge University Press, Murakami T (2015) Replicative construction experiments at Teotihuacan,
Cambridge, pp 250–317 Mexico: assessing the duration and timing of monumental construc-
Díaz Lozano E (1979) Rocas y minerales del valle. In: Gamio M (ed) La tion. J Field Archaeol 40:263–282
población del Valle de Teotihuacan, Tomo II. Instituto Nacional Murakami T (2016) Materiality, regimes of value, and the politics of craft
Indigenista, Mexico City, pp 27–66 production, exchange, and consumption: a case of lime plaster at
Garraty CP (2013) Market development and pottery exchange under Teotihuacan. J Anthropol Archaeol 42:56–78
Aztec and Spanish rule in Cerro Portezuelo. Anc Mesoam 24:151– Neff H (1994) RQ-mode principal component analysis of ceramic com-
176 positional data. Archaeometry 36:115–130
Gillespie SD (1994) Llano de Jícaro: an Olmec monument workshop. Neff H (2000) Neutron activation analysis for provenance determination
Anc Mesoam 5:231–242 in archaeology. In: Ciliberto E, Spoto G (eds) Modern analytical
Glascock MD (1992) Characterization of archaeological ceramics at methods in art and archaeology. John Wiley and Sons, New York,
MURR by neutron activation analysis and multivariate statistics. pp 81–134
In: Neff H (ed) Chemical characterization of ceramic pastes in ar- Neff H (2002) Quantitative techniques for analyzing ceramic composi-
chaeology. Prehistory Press, Madison, pp 11–26 tional data. In: Glowacki DM, Neff H (eds) Ceramic source deter-
mination in the greater Southwest. Monograph 44. Cotsen Institute
Glascock MD (2006) Tables for neutron activation analysis. The
of Archaeology, Los Angeles, pp 15–36
University of Missouri Research Reactor Center, Columbia
Neff H, Glascock MD, Charlton TH, Charlton CO, Nichols DL (2000)
Greenough JD, Gorton MP, Mallory-Greenough LM (2001) The major- Provenience investigation of ceramics and obsidian from Otumba.
and trace-element whole rock fingerprints of Egyptian basalts and Anc Mesoam 11:307–321
the provenance of Egyptian artifacts. Geoarchaeology 16:763–784 Ogburn DE (2004) Evidence for long-distance transportation of building
Harbottle G (1976) Activation analysis in archaeology. Radiochemistry 3: stones in the Inka empire, from Cuzco, Peru to Saraguro. Ecuador
33–72 Lat Am Antiq 15:419–439
Heizer RF, Williams H (1963) Geologic notes on the Idolo de Coatlichán. Pastrana A, Domínguez S (2009) Cambios en la estrategia de la
Am Antiq 29:95–98 explotación de la obsidiana de Pachuca: Teotihuacan, Tula y la
Hernández Javier I (2007) Geología y geomorfología volcánica de la Triple Alianza. Anc Mesoam 20:129–148. https://doi.org/10.1017/
región de los yacimientos de obsidiana de Otumba en el sector norte S0956536109000133
de la Sierra Nevada de México. Tesis de licenciatura, Universidad Quezada Ramírez ON (2016) El Templo Mayor de Tenochtitlan:
Nacional Autónoma de México sistemas, materiales y técnicas constructivas. Tesis de licenciatura.
INEGI (1979) Carta geológica: Texcoco E14B21, Mexico y Tlaxcala, Escuela Nacional de Antropología e Historia, Mexico City
Escala 1:50,000. Instituto Nacional de Estadística, Geografía e Quezada Ramírez O, Pascal C, López AG (2018) De la cantera a
Informática, Mexico City Tenochtitlan: Presencia de rocas de origen volcánico en la
Jones GT, Bailey DG, Beck C (1997) Source provenance of andesite construcción y la escultura del Templo Mayor. La explotación de
artifacts using non-destructive analysis. J Archaeol Sci 24:929–943 andesita en la cantera de San Bartola Tenayuca, una aproximación
Latham T, Sutton PA, Verosub KL (1992) Non-destructive XRF charac- etnoarqueológica. Mexico City
terization of basaltic artifacts from Truckee, California. Robles García NM (1992) La extracción y talla de cantera en Mitla,
Geoarchaeology 7:81–101 Oaxaca. Tecnología para la arquitectura monumental. Arqueología
López Luján L, Torres J, Montúfar A (2003) Los materiales constructivos 2(7):85–112
del Templo Mayor de Tenochtitlan. Estudios de Cultura Náhuatl 34: Sarro PJ (1991) The role of architectural sculpture in ritual space at
137–166 Teotihuacan, Mexico. Anc Mesoam 2:249–262
Margáin CR (1967) Sobre sistemas y materiales de construcción en Schaaf P, Stimac J, Siebe C, Macías JL (2005) Geochemical evidence for
Teotihuacan. In: XI Mesa Redonda. Sociedad Mexicana de mantle origin and crustal processes in volcanic rocks from
Antropología, Mexico City, pp 157–211 Popocatepetl and surrounding monogenetic volcanoes, Central
Mexico. J Petrol 46:1243–1282
Millon R (1973) The Teotihuacan map. Part 1: text. Urbanization at
Sotomayor Castañeda A (1968) Estudio petrográfico del area de San Juan
Teotihuacan, Mexico, vol 1. University of Texas Press, Austin
Teotihuacan, Edo. de México. In: Lorenzo JL (ed) Materiales Para la
Millon R (1981) Teotihuacan: city, state, and civilization. In: Bricker V, Arqueología de Teotihuacan, Serie Investigaciones XVII. Instituto
Sabloff JA (eds) Supplement to the handbook of middle American Nacional de Antropología e Historia, Mexico City, pp 39–49
Indians, vol.1: Archaeology. University of Texas Press, Austin, pp Sugiyama S (2005) Human sacrifice, militarism, and rulership: material-
198–243 ization of state ideology at the feathered serpent pyramid,
Minc LD (2009) Style and substance: evidence for regionalism within the Teotihuacan. Cambridge University Press, Cambridge
Aztec market system. Lat Am Antiq 20:343–374 Vázquez-Sánchez E, Jaimes-Palomera R (1989) Geología de la Cuenca
Mooser F (1968) Geología, naturaleza y desarrollo del Valle de de México. Geofis Int 28:133–190
Teotihuacan. In: Lorenzo JL (ed) Materiales para la arqueología de Williams-Thorpe O, Thorpe RS (1993) Geochemistry and trade of eastern
Teotihuacan, Serie Investigaciones XVII. Instituto Nacional de Mediterranean millstones from the Neolithic to Roman periods. J
Antropología e Historia, Mexico City, pp 29–37 Archaeol Sci 20:263–320
Murakami T (2010) Power relations and urban landscape formation: a Winchester JA, Floyd PA (1977) Geochemical discrimination of different
study of construction labor and resources at Teotihuacan. Ph.D. diss. magma series and their differentiation products using immobile el-
Arizona State University ements. Chem Geol 20:235–343

You might also like