A petrographical study has been undertaken on three groups of pottery from the Cusichaca Valley, Peru, dating to the pre-Spanish Late Horizon Period (ca. 1450-1532 A.D.). Examination of comparative modern pottery from the region and laboratory control pieces indicates that a range of technologies were employed. Inca Fine Ware pastes were prepared by adding crushed basalt-andesite to a refined clay or to a clay naturally low in non-plastics. They were made outside the area, possibly in the adjacent Cuzco Valley, a distance of 50-100 km. Killke Wares are also non-local and have crushed feldspathic rocks added to clays with variable non-plastic contents. Inca Utilitarian Wares and Cooking Pots were made by adding heterogeneous sands or crushed sands/gravels to clays with variable non-plastic contents and are consistent with the local geology, but equally could originate in the Cuzco area.

Pottery, the most common archaeological artifact, is usually recovered from the sites where it was used, not made. However, pottery fabrics contain information which may allow the reconstruction of the earlier stages of a pot's history; its manufacture and the trade network which distributed the wares to the consumers. If the fabrics are studied in thin section, the non-plastic components can be identified; in favourable circumstances, this leads to the location of the pottery workshops, an essential first step in the reconstruction of the trade networks. If comparisons are made with fabrics resulting from known preparation methods, then the ancient manufacturing processes, such as cleaning the clay and adding temper, can be recognised; this in turn clarifies the mode, organization and scale of pottery production. Comparing this technological information with pottery shape and design, indicates the extent to which specific functional or aesthetic requirements in the market were satisfied by the pottery workshop.

The Inca Empire was the last of the pre-Spanish empires of South America. It originated in the South-Central Andes of Peru in the area around Cuzco (Fig 1) during the early part of the second millennium A.D. At the time of the Spanish conquest in 1532, the empire extended southwards from Ecuador to present day Santiago in Chile and eastwards from the Pacific coast, across the Andes, to the tropical lowlands of the Amazon Basin.

Most information about the Inca Empire derives from written records compiled after the Spanish conquest. Although there are a number of studies of Inca architecture (Kendall 1985), other archaeological information for the late pre-Spanish cultures is sparse, particularly for the Department of Cuzco, the heartland of the Inca Empire.

This sparcity of information particularly applies to pottery studies. These have concentrated on typology based on decorative designs, with little work on archaeological assemblages, on petrographical studies of fabrics or on the techniques of pottery manufacture. No pre-Spanish pottery workshops or kiln sites have been identified in the South-Central Andes, although there are a few, modern, non-mechanized potteries where pots are coil-built, painted in traditional designs and open-fired.

It is, however, generally thought that the Inca pottery industry was based in the Cuzco area and that it was created out of the preceding Killke industry during the formative years of the Inca Empire. Rowe (1944) referred to Inca potters mass-producing high quality standardized wares and Morris (1978) described state-controlled distribution of this pottery along 'politically important lines. There is a consensus of opinion that free trade in any commodity was virtually non-existent within the core area of the Empire.

Figure 1. South America and the area of the Inca Empire at the time of the conquest (insert)

In the present study, representatives of three main types of pre-Spanish pottery, namely Inca Fine Wares, Plain Wares (Cooking Pots and Inca Utilitarian Wares) and Killke Wares were chosen for petrographical fabric analysis. Fabric, vessel shape and decorative style were used to discriminate between types of pottery (Lunt 1987). Each type was called a "ware" and these were numbered from Ware 1 to Ware 346. Thin-section petrography including grain-size analysis of the non-plastics (temper and/or clasts) was undertaken for each sample. For comparative purposes, examples of modern coil-built coarse (plain) and fine wares from the Cuzco area plus fired test-pieces made from alluvial clays from the Cusichaca area were also studied.

The pottery was excavated from five neighbouring sites in the Cusichaca Valley, 80 km north of Cuzco. The sites are Huillca Raccay, Quishuarpata, Patallacta, Pulpituyoc and Huayna Quente. All of these sites have standing structures built in the Inca architectural style, which suggests that the sites were built in the late fifteenth century at the beginning of the Inca occupation of the valley (Kendall 1985). One of the sites, Huillca Raccay, shows continuous occupation from at least the beginning of the Late-Intermediate Period (ca. 900 to ca. 1450 A.D.) up to the end of the Late Horizon Period (ca. 1450-1532; chronology of Rowe 1962 and Rowe and Menzel 1967).

The sites represent a new agricultural use for the area by the Incas, a greater population density, an increase in population size and in ethnic composition, and a complex social hierarchy. Similar changes occurred wherever the Incas extended their domination.

The South-Central Highlands of Peru are noted for their geological complexity. The area contains extensive volcano-sedimentary and sedimentary sequences intruded by large plutons of acid to intermediate rocks. Many of the rocks have passed through major orogenic cycles and hence have suffered deformation and metamorphism. Therefore, the bedrock geology and its overlying superficial deposits can contain an enormous range of lithologies.

The local geology of the Cusichaca area is, by contrast, simple. It comprises a Permian red bed sequence of sandstones, conglomerates and breccias unconformably overlying highly deformed Lower Palaeozoic metasediments. More than 90% of these metasediments are quartz-biotite-muscovite-chlorite slates and metashales together with approximately 10% quartzites, grits and conglomerate. A granite-granodiorite complex has intruded these Palaeozoic rocks and locally includes greisenized and tourmalinized facies together with a little sulphide mineralization. Volcanic rocks are very minor in extent but include ashy pyroclastics and rare olivine dolerite dykes (Deruelle 1977).

Close to Cusichaca, the Urubamba river cuts the granite-granodiorite complex but farther upstream it drains the metamorphosed Lower Palaeozioc sequence. Alluvial clays are present along the banks of the Urubamba and their clasts closely reflect the local bedrock geology.

Andesite and basalt lavas form relatively small outcrops, up to 10 sq. km, in the Cuzco Valley to the southeast of Cusichaca. They range is composition from hypersthene-bearing andesites to biotite-bearing basalts and dolerites. They carry plagioclase ± augite ± hypersthene ± hornblende ± biotite ± apatite phenocrysts in a fine-grained to glassy groundmass. Intrusive rocks, diorite and syenite also crop out but only very locally. They are extensively altered and have a primary mineralogy of orthoclase, plagioclase, sphene ± augite ± hornblende ± apatite together with many secondary minerals including abundant epidote. All of the igneous rocks were quarried by the Incas (Gregory 1916).

Over 50,000 sherds were excavated from the Late Horizon Period levels. All of the diagnostic pottery was described in hand specimen in the field and each type was defined as a restricted range of shapes, decorative designs and fabrics.

Over 85% of the Late Horizon Period sherds belong to three types of pottery, all of which were coil-built. Two of these types are decorated (Inca Fine Wares and Killke Wares) and were painted with mineral pigment and polished before firing. The Plain Wares are undecorated. The remaining 15% comprise a variety of decorated and plain wares, each represented by a small number of sherds and are not included in the present study. All of the pottery was fired in open fires under oxidising conditions, at temperatures well below those needed for vitirification (Lunt 1987).

About 20% of the pottery sherds are Inca Fine Wares. These wares were highly standardized, with a limited range of shapes and decorative designs, and were well made. The two most common shapes found in Cusichaca are the aryballus (Fig 2.) and the plate (Fig.3). Inca Fine Wares are found on almost all Inca sites within the Inca Empire, but have their greatest concentration in and around Cuzco.

Approximately half of all the pottery sherds are thin-walled Plain Wares. Of these, 60% contain carbonised food remains and have smoke discoloured surfaces (Fig 4.). These and similar pots appear in many Late Horizon sites and very similar-looking pots are manufactured today for cooking food on open fires. The other 40% of the Plain Wares comprise the Inca Utilitarian Wares (Fig 5.). These pots were made to a high standard using a piecemeal method of assembly similar to that of the Inca Fine Wares (Lunt 1987). It has been suggested (Menzel, 1976) that these pots also were used for cooking, although they rarely have smoked surfaces or carbonised contents. The Inca Utilitarian Wares are found on Late Horizon sites throughout the Inca Empire and, like the Inca Fine Wares, were often copied by provincial potters.

Killke Wares (Dwyer. 1971; Rowe 1944, 1949) make up about 15% of the sherds. The wares have a limited variety of unstandardized shapes and decorative designs: the most common versions found in Cusichaca are shown in Figure 6. Killke Wares are one of a number of similar wares found throughout the Southern Andes from the middle of the first millennium onwards. Their centre of distribution appears to be in the Cuzco Valley, where they predate Inca pottery. They are found as far north as Cusichaca, appearing first in the Late Intermediate Period levels, but continuing into and through the Late Horizon Period.

Figure 2. Inca Fine Ware aryballus. The cross-section shows the major segments
which were coiled separately and assembled when leather hard.
The pigments are black, purple-red and white, applied to a self-slipped, highly polished surface.

Figure 3.Inca Fine Ware plates. Representational head-handles are common.
The plates are painted in black and red (top) and white (bottom) pigments, on a self-slipped, polished surface.
The bottom example shows the most widely distributed form.

Figure 4. Plain Ware: cooking pot. Red-purple paint inside the rim lip is common.
The pots have a wet-hand finish.

Figure 5. Plain Ware: Inca Utilitarian Ware.
The cross section shows the same piecemeal assembly technique as in Figure 4.
Red-purple paint inside the rim is common. The pots were covered in a fine self-slip.

Coarse (plain) wares are made today in a few localities to the south of Cuzco, but because of their high reputation for serviceability, are traded widely across the South-Central Andes. The wares are thin-walled, bag-shaped cooked pots, coil-built but often kiln-fired. Local river clays are picked over by hand to remove the largest clasts and crushed 'slate' is them added as temper 'to make the pots strong'.

These wares are made in a number of villages to the south of Cuzco and are quite widely traded. One of the villages, Raqchi, is identified in oral tradition as the site of an Inca pottery. The raw materials are more carefully prepared than those of the Modern Coarse Wares. Local river clays have their largest clasts removed by hand and are then wet-sieved and washed to produce a very clean clay. Field blocks of basalt-andesite are collected, heated and crushed to a powder until it 'feels like sand' and this is added as temper. The pots are coil-built and open-fired.

Figure 6. Killke Ware. Both pots are painted with black (now brown-grey) and red
pigments on a pale, self-slipped, compacted surface.
The jar form (top) with added rim fillet, is often provided with vertical strap handles.
springing from the top of the rim.

Four samples of alluvial clays from the Urubamba river (2, 4A, 6, 8) and one from the Cusichaca river (10) were collected. Samples 2 and 4A were taken from deposits that are currently exploited as raw materials for the manufacture of stove bricks and other structural items.

The dry clays were gently crushed and visible clasts and vegetable matter were removed by hand. Some of the clay was set aside, made into test-tiles, and fired. The remaining clays were levigated and the finer fraction made into test-tiles and simple thumb-pots (Lunt 1987). All of the test-tiles and thumb-pots were successfully open-fired to temperatures of around 850º C.

The lower size limit for the non-plastic fraction was taken to be 15µm. The lithologies of the >15µm fraction were identified together with a qualitative assessment of their relative abundances, internal grain size of polycrystalline rocks (coarse-, medium- or fine-grained), roundness/angularity, degree of alteration/weathering and the nature of their bounding surfaces (smooth and planar or irregular and indented). The main results are summarized in Tables 1 and 2.

The maximum length of the non-plastic fraction was measured and grain-size analysis of this fraction, using standard sedimentological techniques, (Folk 1974) was performed; counting the first 150 to 180 grains per thin section. These data are plotted on the Ø scale using a class width of ½ Ø and were subjected to statistical analysis. The results are graphically represented in Figures 7a-h. Since the conditions of manufacture (technology) of the modern pottery and experimental test-pieces are known, their thin-section petrography will be described first.

Four examples of Modern Coarse Wares (Ware 8): namely 8-1, 8-2, 8-2 and S144 were sectioned. Although the pottery is essentially unilithic (one lithology comprises more than 90% of the non-plastic fraction), the nature of the temper, its grain-size distribution and ratio of temper to clay are very variable and the wares do not form a coherent group.

The Modern Coarse Wares comprise a crushed temper added to roughly cleaned alluvial clays. The size distribution of their non-plastic fraction (temper and sedimentary clasts) might therefore be expected to be bimodal with a mode for the temper and one for the clasts. Because the sample size is small and the variation of fabrics between them is great, the size distribution data for the four pots have not been combined. They are, therefore, described individually.

S144 closely conforms to the description of a 'slate'-tempered pot. Its temper is unilithic comprising lath-shaped, limonitically-stained, fine-grained metapelite (metamorphosed shale) up to 3mm in length. The temper has planar bounding surfaces that lie parallel to the metamorphic cleavage planes of the pelite, namely along the mica bands. The temper to clay ratio is the highest of any material examined although some of the Inca Utilitarian Wares have ratios that approach it. The size distribution of the temper appears to be close to unimodal with an approximate mean of Ø = 4.5, but is negatively skewed, showing an excess of coarse material, compared to the normal gaussian distribution.

The other examples (W8-1, W8-2 and W8-3) have a loosely-packed temper of unaltered/unweathered amphibole-bearing, granite-granodiorite with minor quantities (<10%) of medium-grained sandstone and trace amounts of altered lava and trachyte. The ratio of temper to clay is low. Temper is subrounded to subangular with a maximum length of 1.7mm and has flat bounding surfaces controlled by feldspar cleavage planes and quartz crystals faces.

W8-1 has a temper set in a clay matrix containing abundant quartz, feldspar and mica laths (up to 25µm in diameter ). The size distribution of the non-plastics is weakly bimodal with a large mode at approximately Ø = 6.5 and a smaller, poorly defined mode at about Ø = 2.75. These are interpreted as being the clay clast fraction (Ø = 6.5) and temper fraction (Ø = 2.75) and would be consistent with a poorly cleaned clay with a small amount of temper.

W8-2 has a temper with a limited size range within clay. The size distribution of the non-plastic component can be interpreted as being bimodal with a main peak at approximately Ø = 2.75 (temper) and small peak at approximately Ø = 5.5 (clay clast fraction). The sizes of the peaks suggest that significant amounts of temper were added to poorly cleaned clays.

Three examples (R1, R2 and S348) were sectioned. All three have a unilithic temper and form a coherent group. The angular temper comprises clinopyroxene and biotite-bearing vesicular basalt-andesite up to 1 mm in length that is loosely-packed (a low temper to clay ratio) within clean clay carrying rare quartz grains (up to 50µm in diameter) and very few mica flakes. No other lithologies are present. The size distribution of the non-plastics (essentially just the temper) is close to unimodal with a statistical mean of Ø = 3.59, (Fig 7a). The petrography is consistent with the description of the manufacturing process, namely of highly prepared and cleaned clays tempered with basalt-andesite.

Figure 7a,b.

The petrography of the fired raw clays shows them to be variable. They range from clean clay (6) with few clasts up to coarse silt in size, to clay with significant amounts of coarse sand size clasts (2, 4A, 8, 10). In all of the test-tiles the ratio of non-plastics to plastics was lower than for any of the pottery fabrics. The clasts reflect the local bedrock geology drained by the rivers, are angular to subrounded and their internal grain size depends upon their lithology. The majority of the rock clasts are unaltered to weathered, epidote-bearing granite-granodiorite with subordinate amounts of fine-grained quartz-muscovite pelite and minor amounts of medium-grained sandstone and arkose. Mud pellets and trace amounts of carbonate, altered lava and trachyte are present, as are discrete grains of quartz, feldspar, biotite and muscovite.

All of the fired test-pieces contain areas with a different firing colour from that of the main clay. The majority of these areas are rounded and small, up to 50µm in diameter, but can reach up to 0.5mm in the unwashed clays. They comprise darker fired clays without any non-plastic component and are interpreted as being natural mud clasts. Other areas are lighter or darker in colour, are lensoidal to subangular in shape and carry discrete quartz and feldspar grains, with a clast to clay ratio equal to, or greater than, the main clay. Some of these areas grade into quartz-feldspar-rich lenses and are interpreted as being inhomogeneities produced in the clay during manufacture of the test-tiles. The more angular examples look very similar to grog.

The majority of the clasts in the raw clays (after removal of the largest clasts and vegetation) are less than 1.0mm in length, but can reach 1.5mm. The size distribution of the clasts appears to be bimodal, but, as the histogram in Figure 7b show, only the coarser half of the distribution has been counted and the distribution is really unimodal with a mean that has the minimum value of about Ø = 5 but is certainly larger than this. This mean is larger (the mean grain size is smaller) than that for any of the pre-Spanish pottery. The distribution shows the clays to be fairly well sorted but the mineralogy and shape of the clasts shows the clays to be sedimentologically immature.

Figure 7c,d.

Levigation of the clays reduced the maximum clast size to 0.5mm and reduced the mean size (i.e. increased the mean value of Ø) of the non-plastic component (Fig 7c). The distribution has the smallest standard deviation (1.14) of any of the material studied showing there to be a tight distribution of clast size. Levigation further reduced the non-plastic to plastic ratio and lowered the abundance of discrete biotite and muscovite flakes; the latter effect is inferred to be the result of the floating off of the micas during washing.

Comparison between the raw and washed alluvial clays and Modern Fine Wares show them to be petrographically very different in terms of the lithology and size distribution of their non-plastic fractions. However, comparison of the clays and Modern Coarse Wares, especially examples W8.1 to W8.3 show similar non-plastic lithologies, but very different non-plastic to plastic ratios.

Four ware types (Wares 13, 26, 27 and 29) were chosen to reflect the range of the Inca Fine Wares. A total of 23 thin section were made:
Ware 13 (F12, W13, W13-1, W13-2, S127, S129A, S205, S229, S306); Ware 26 (W26, W26-1, S223-1); Ware 27 (BDW, S143, S200, S201, S707.45) and Ware 29 (S29-1, S29-2, S29-3, S29-4, S106, S140).

All of the examples have a unilithic temper in clays that are almost totally devoid of a fine-grained (<15µm) non-plastic fraction and all individuals show a similar temper size distribution to that obtained by combining then. Petrographically there are two groups, a minor group of four sections (S29-1 to S29-4) and the rest.

Except for Ware 29, all the Inca Fine Wares are remarkably uniform in thin section and form a very coherent group having an almost totally unilithic temper. The temper consists of angular to subangular basalt-andesite comprising plagioclase ± pyroxene ± biotite ± hornblende (amphibole) ± apatite phenocrysts; trace amounts (<5%) of quartzite are also present. The surfaces of the temper are irregular and embayed and are controlled by the distribution and shape of the phenocrysts. They are well-bonded with the clay that carries rare, small clastic quartz grains up to 50µm in diameter. Biotite laths, up to 200µm in length are present in the clays of the pots tempered with biotite-bearing basalt-andesite (F12, S106, S127, S201 and S223-1) but is absent from those tempered by biotite-poor basalt-andesite. Grog-like material is uncommon and no unequivocal grog was recognised. Only in W26-1 are there minor amounts of grog-like areas containing an exotic fine-grained lava component. The maximum grain size of the temper is 1.25mm and the size distribution of the temper (and clasts?) is close to unimodal with a statistical mean of Ø = 3.81. Their distribution shows a marked negative skewness (Fig. 7d).

Ware 29 is different. Four of the examples (S29-1, to S29-4) have a unilithic temper comprising abundant angular to subangular fragments of a plagioclase ± alkali feldspar ± quartz ± epidote rock similar in appearance to syenite (S29-1, S29-3, S29-4), or to arkose (S29-2), together with trace amounts of trachyte and fine-grained lava. Their maximum length is 0.5mm. Single angular crystals of feldspar and quartz constitute a significant proportion of the temper. The surfaces of the monomineralic temper are flat but the polycrystalline grains have irregular surfaces. The plastic component consists of rare, rounded mud pellets within clays that carry minor amounts of small (<20µm) quartz and feldspar crystals that equally may be finely-crushed temper or sedimentary clasts belonging to the clays.

The size distribution of the temper (and clasts?) is close to unimodal (Fig. 7e) with a mean of Ø = 3.96 which is slightly greater than the average mean for all the Inca Fine Wares, (3.85). A small peak at Ø = >6 may represent a clastic component from the clays.

Comparison between the Inca Fine Wares (other than Ware 29) and Modern Fine Wares shows them to be very similar in terms of the lithology, shape and size distribution of their temper and the very clean (sedimentary clast-free) nature of the clays. These similarities are sufficient to suggest that the Inca Fine Wares were manufactured using the same methods as the Modern Fine Wares. Comparison with the alluvial clays shows that the clays in the Inca Fine Wares fabrics are cleaner and contain no white mica. This suggests either that naturally very clean clays were used by the Inca potters or, that if alluvial clays were used, then sieving as well as, or instead of, levigation, was used to clean the clays.

Figure 7e,f.

These are divided into two main groups; Cooking Pots and Inca Utilitarian Wares. There are a number of ware types within each major group.

Twelve sections were made from the Cooking Pots:
Ware 1 (1-1, 1-1-1, 1-2, 1-2-1); Ware 2 (2-1, 2-1-1, 2-2); Ware 5 (W5, W5-1); Ware 17 (F17, F17-1, F17-2) and nine from Inca Utilitarian Wares: Ware 4 (F4, F4-1, F4-1-3, F4-3-1); Ware 32 (F32, F32-1, F32-2); Ware 120 (S120) and Ware 204 (S204).

Petrographically the Plain Wares are heterogeneous and do not form a coherent group. Some examples have a unilithic temper (W1-1, F32 and F17) others are dilithic (W2-1). The majority, however, are polylithic and these vary from having a granite-granodiorite-dominated to metasediments-dominated temper. These variations in temper lithologies are as great within a ware-type as those found between the Utilitarian Wares and Cooking Pots. Petrographically therefore, it is not possible to distinguish between these two main groups other than by grain size analysis. Hence they can be described together on the basis of their common characteristics which allow them to be distinguished from Inca Fine Wares and Killke Wares.

The majority of the Plain Wares are polylithic with a subrounded to subangular granite-granodiorite > metapelite > sandstone, arkose and quartzite > porphyritic lava > 'grog', temper. Many examples contain feldspathic rocks that are too altered to identify and the percentage of altered/weathered rocks is generally high.

The granite-granodiorite (and rare gneiss) comprise quartz-plagioclase-alkali feldspar with minor biotite and amphibole. They are often highly altered with abundant epidote and alteration of the feldspars (especially plagioclase) to fine clay minerals. The metapelites (metashales and slate) are quartz-muscovite-sericite± biotite-rich and extensively limonitically stained. Epidote grains are common within the sandstone and arkose. The internal grain size of the rock fragments depends upon their lithologies and hence varies from coarse in the granite-granodiorite, to medium in the sandstone, arkose and quartzite, to fine in the metapelites. In none of the samples was the temper of a uniform grain size, unlike the Inca Fine Wares and Killke Wares where the uniformity of grain size within the temper lithologies is striking.

The surfaces of the temper are generally smooth. In rounded to subrounded altered temper the surfaces follow secondary clay minerals, in the subangular granite-granodiorite temper they are controlled by crystals faces or lie along the cleavages planes of feldspar and in the metapelites they are aligned along the muscovite-sericite-biotite bands. Much of the temper is poorly bonded with the clay. The clay carries discrete crystals of quartz, feldspar, muscovite, biotite and epidote up to 50µm in length but as all are constituents of the temper it is not possible to tell if these discrete crystals are temper or an inherent part of the plastic component.

Rounded to subrounded grog-like material is present in all of the Inca Utilitarian Wares but only in W5 and W1-2 of the Cooking Pots. The grog-like areas are distinguished by their different firing colour and the density and identity of their enclosed non-plastic component. In the majority of these areas the non-plastic components are quartz and feldspar and are identical to those found in the main body of the pot. In F4-1-3, however, all of the grog-like areas carry basalt-andesite which is not present elsewhere in the pot and these are interpreted as fragments of Inca Fine Ware. The percentage of this grog is so low (<1%) that it seems unlikely that it was added intentionally.

A minority of the examples are different, W1-1, F17 and F32 are unilithic with an angular temper of unaltered rocks with the same grain size; within clean clays. In F32 the temper is composed of single feldspar crystals, in W1-1 it is granite and in F17 it is porphyritic lava. Other than in F17, porphyritic lava temper is highly altered and very rare or absent. W2-1 is dilithic with unaltered granite plus a coarse-grained quartz-feldspar metamorphic rock (gneiss). With the exception of the basalt-andesite-bearing grog and porphyritic lava, all of the lithologies that make up the temper of the Plain Wares can be found in the local bedrock geology and overlying alluvial clays in the Cusichaca area.

Both Inca Utilitarian Wares (Fig. 7f) and Cooking Pots (Fig. 7g) have close to a unimodal distribution (especially the Inca Utilitarian Wares) and have similar statistical means, namely Ø = 3.53 for the Cooking Pots and 3.38 for the Inca Utilitarian Wares. The mean, in Ø units, for the Inca Utilitarian Wares is the smallest of all the wares (i.e. the mean size of the temper is the largest). The Cooking Pots, however, have a maximum temper size of 1.5mm compared with 1mm for the Utilitarian Wares. In both wares there is a loose correlation between the size of the temper and percentage of altered/weathered rocks and their degree of roundness. Examples with rounded to subrounded altered and weathered rocks as temper have a wider size distribution than those with subrounded to angular temper.

Although the Plain Wares can be distinguished from both Inca Fine Wares and Killke Wares, they do not form a coherent group. Their temper is unilithic to polylithic and the degree of alteration of these lithologies, their internal grain size and the size distribution of the temper are all variable. The Plain Wares share many fabric characteristics with the Modern Coarse Wares and with the experimentally fired alluvial clays.

Figure 7g,h.

Twenty thin sections were prepared from representative ware types of Killke Wares, namely:
F4; F7AG; S162; S162A; Ware 45 (F14/S99, S99, S209); Ware 46 (S312); Ware 47 (47-1, 47-2, 47-3, 47-4); Ware 48 (S48-1, 48-1-1, 48-2, 48-2-1, 48-2-2, 48-3) and Ware 49 (S49, 49-1).

Although multiple examples within a ware type showed petrographical variations (Ware 47, Ware 48) all of the Killke Wares share enough characteristics to be considered a fairly coherent group that is clearly distinguishable from Inca Fine Wares or Plain Wares. Although the temper of the wares can be unilithic (15 pots), dilithic (3 pots) or polylithic (2 pots) they all carry medium-grained feldspathic rocks as part of the temper. Six examples have a unilithic temper of syenite, four of arkose, three of trachyte-andesite and two of fine-grained granite. Three have a dilithic temper of arkose and syenite or arkose and trachyte-andesite and one comprises syenite, arkose and trachyte-andesite. The final example (F4) is very polylithic but carries altered syenite as part of its temper. The majority of the vessels (18) have a subangular temper that is composed of plagioclase ± alkali feldspar ± epidote, (syenite); quartz - altered feldspar - epidote (arkose); plagioclase ± pyroxene ± amphibole porphyritic lava (trachyte-andesite) and quartz-alkali feldspar-plagioclase (fine-grained granite). In 48-2 the amphibole-bearing porphyritic lava is extensively kaolinized. Other lithologies are uncommon, S162A carries approximately 10% of limonitically-stained quartz-mica metapelites (metashales), 47-3 carries approximately 20% amphibole- and pyroxene-bearing andesites and F4, which is unlike all other Killke Wares, is very polylithic and carries glassy and amygdaloidal basic lava, serpentinized basic rocks and vein quartz in addition to altered syenite. The internal grain size of all the lithologies is uniformly medium-grained and the grain boundaries control the surfaces of the temper which are irregular and serrated and form good bonds with the clay. The clay is clean and has small quartz, feldspar and amphibole grains (<50µm in length) which may be finely crushed temper or sedimentary clasts belonging to the clay.

The majority of the Killke Wares contain minor amounts (<5%) of grog-like material. They are poorly bonded with the main clays, have lighter or darker firing colours and are rounded in shape or lensoidal with their long axis lying parallel to the main coil lines of the pot (47-2). The grog-like areas carry small (<15µm in diameter) monocrystalline grains of quartz and lesser amounts of feldspar and have a clast to clay ratio that is similar to, or slightly greater than, the main clays. Their similarity in distribution and appearance to the grog-like areas found in the experimentally fired clays is sufficient to infer a similar origin, namely that they are mud pellets or dried clays that have been incorporated adventitiously into the pottery clay during manufacture.

The size distribution of the temper is very uniform and it has a maximum length of 1.25mm. It has a bimodal distribution with a mode of Ø = 1.75 (Fig. 7h). F14/S99 is an exception, although it is also bimodal, the temper has a mode of Ø = 5.5 and a subsidiary mode of Ø = 4.5 and a maximum temper size of 0.5mm. F14/S99 is tempered by single, angular grains of feldspar plus minor amounts of syenite.

The Killke Wares form a petrographically coherent group, which easily can be distinguished from Modern Coarse Wares, from Modern Fine Wares and from the majority of the Inca Fine Wares. There are similarities between Killke Wares and a subgroup of Ware 29 of the Inca Fine Wares.

Characteristically, the Killke Wares have a temper which consists of a limited number of similar looking feldspathic rocks, all of which are medium-grained and set within clean clays. Comparison of the Killke fabrics with the experimentally fired alluvial clays from the Cusichaca area show that their non-plastic component (temper and clasts respectively) are lithologically different.

Potting clays require a non-plastic component in order to improve their workability and to counteract the stresses that occur during the drying-out and firing of the pots. This is especially true for open-fired pots, where a coarse-textured fabric is essential for successful firing. The non-plastic component can be either a natural part of the clay or deliberately added as a temper.

Although the size distribution, shape and ratio of the non-plastics to clay have all been shown to be important parameters that affect the serviceability of the potting clay, few studies have suggested that specific lithologies are required. If the local clay is unsuitable for pottery making, then traditional potters will temper with local materials. Hence, on this basis, to identify the origin of the temper rocks is to locate the workshop.

The experimental pieces (untempered) and modern pottery (tempered) show that successful pots can be fired from many of the raw materials of the South-Central Andes by employing a variety of manufacturing processes. In order of increasing degrees of preparation these are, untempered raw clay with only the largest clasts and vegetation removed; untempered levigated clays (washed clays); roughly washed clays with crushed temper (Modern Coarse Wares) and washed, clean clays with crushed temper (Modern Fine Wares). This list shows that increased preparation involves an increase in the selection and refinement of the raw materials and this in turn leads to an increase in the standardization of the pottery (the coherence of a ware type) so that individual pots conform more closely to the average pot.

Table 1 shows the petrographical characteristics of the experimental and modern wares together with their known methods of production. It shows that pots with a high degree of preparation have very clean (a low abundance of fine-grained sedimentary clasts) clays, plus a unilithic, angular temper comprising unaltered rocks with a uniform internal grain size and irregular bounding surfaces. Furthermore, the non-plastic size distribution data suggest that a bimodal distribution is found only in pots where temper is added to poorly cleaned clays (Modern Coarse Wares). In untempered, or tempered wares made using very clean clays, the size distribution is unimodal but, as seem in the present examples, with very different mean sizes. Although the size distribution of the non-plastics in the Modern Fine Wares is unimodal, it is possible to demonstrate that the pots are not made from untempered clays, since the fine non-plastic fraction of the clays (<15µm) is quartz-rich unlike the temper, which is quartz-free.

Characteristics shown in Table 1 have been taken together to provide a set of empirical criteria to measure the degree of preparation for the raw materials in pre-Spanish wares. It is recognised that the modern wares have been tempered with crushed rock and that if natural sands were used for tempering then many of these criteria would be inappropriate. Table 2 lists the petrographical characteristics of the pre-Spanish wares, together with an interpretation of their preparation.

The Inca Fine Wares form a very coherent and standardized group. They have very clean clays with an added angular unilithic temper comprising a medium-grained rock (basalt-andesite) which has an irregular surface topography and a unimodal size distribution. All of these features have been shown empirically to be part of the fabric of pots made from highly prepared materials. Furthermore, the similarities between the Inca Fine Wares and Modern Fine Wares are sufficiently close to suggest a common mode of manufacture, namely crushed rock added to very clean clay. As the Cusichaca alluvial clays carry no basalt-andesite clasts, and basalt-andesite is absent from the local geology, the Inca Fine Wares cannot be of local manufacture unless basalt-andesite were imported.

The basalt-andesite temper has very strong petrographical similarities with the andesite and basalt lavas that crop out in the Cuzco Valley, described in Gregory (1916); most notably the presence of biotite and apatite phenocrysts. This suggests that the origin of the temper is from the Cuzco area, which in turn suggests that the pottery workshops were there.

The Killke Wares also form a coherent group, although to a lesser degree than the Inca Fine Wares. They too have clean clay and the majority have a unilithic to dilithic, angular to subrounded temper comprising a restricted number of medium-grained feldspathic rocks with irregular surface topographies. The size distribution is essentially unimodal with a mean size close to that of the Inca Fine Wares. The fabric is interpreted as being the result of cleaned clays with the addition of crushed rock. It is perhaps significant that arkose, trachyte and basalt-andesite look and feel very similar and are interchangeable in their tempering characteristics. As with the Inca Fine Wares, there is little correlation between the temper of the Killke Wares and the bedrock geology of the Cusichaca area and the wares were probably manufactured elsewhere, perhaps using the syenite and diorite rocks found in the Cuzco Valley.

Petrographically, the Cooking Pots form a poorly coherent group. Their non-plastic component is polylithic, subrounded to angular, and has a wide range of internal grain sizes and many smooth/flat boundary surfaces. However, the clays are often clean and the size distribution is markedly unimodal with a mean close to that of the Killke Wares and Inca Fine Wares and very different from the washed Cusichaca clays. Although the fabrics are consistent with untempered clays, the petrographical diversity of the pots would require a number of such clays. The preferred interpretation is that petrographically heterogeneous sand (or crushed sand/gravel) has been added to cleaned clays. There is a high correlation between the Cusichaca area bedrock geology, the lithology of the clasts within the alluvial clays of the Urubamba River and the Cooking Pots, therefore these wares may have a local origin.

The Inca Utilitarian Wares do not form a petrographically coherent group. Their non-plastic component is similar to that of the Cooking Pots, being polylithic, subrounded to subangular, with a wide variation in internal grain size and with a flat/smooth surface topography. The clays are clean to dirty with abundant small (<15µm) sedimentary clasts. The fabrics do not conform with crushed rock tempered pots.

The size distribution of the non-plastics is markedly unimodal and has the smallest mean of any of the material. The size distribution and its mean suggest that these pots are also tempered by sand (or crushed sand/gravel); but the diversity of the wares, especially the variations in the cleanness of the clays, suggests that some may be untempered. The lithologies of the non-plastic component of the Inca Utilitarian Wares are very similar to those found in the alluvial Cusichaca clays and local bedrock geology and these pots may be of local manufacture.

The greatest concentration of these wares is in Cuzco and on immediately adjacent Inca sites. The wares (except for some of the Ware 29) are tempered with basalt-andesite, Basalt-andesite crops out in the Cuzco area (Deruelle 1977; Gregory 1916) and its properties were well known to the Incas who exploited it on a grand scale for building and for making weapons and utensils (Proetzen 1983; Gregory 1916). The use of basalt-andesite as temper may only reflect the proximity of workshops and outcrops, but basalt-andesite is a particularly suitable tempering material. When crushed, its sharp-edged, angular grains with their irregular surface topography form strong bonds with clay (Shepard 1976) and this may have counteracted the inherent structural weakness of coil-built piecemeal-assembled pots which increases with the size of the pot. Many of the Inca Fine Wares aryballi are extremely large (up to 1. 5m) and have been found up to 1,500 km from Cuzco. If Cuzco were the place of manufacture then long journeys over rough terrain would require that the pots be strong.

The total output of Inca Fine Wares was very large. Standardization in all aspects is one of the most striking characteristics and this can be viewed as the natural bi-product of a cost-effective mass-production system. However, standardization and uniformity themselves may have been a goal of production. This would be consistent with known Inca political strategy, namely signalling their presence, interest, or approval by their introduction of recognisable, high-quality products.

Ware 29 of the Inca Fine Wares is in outward respects identical to the rest of the Inca Fine Wares, except that its fired clay is lighter in colour. In many examples, however, the temper is different and shares similarities with Killke Wares. As Ware 29 has not been distinguished previously from other Inca Fine Wares (to the authors' knowledge) there are no specific distributional data for it. Stratigraphically, it is found between earlier Killke Wares and later Inca Fine Wares.

All of these pots are fine wares with no traces of cooking use. In hand specimen the fabrics look very similar. They are amongst the closest stylistic precursors to Inca Fine Wares and share a similar distribution that is centred on Cuzco.

Although the wares are fairly coherent, their differences may represent a series of slightly different technological practices albeit working towards a common product. In the absence of recognised workshop sites it is not possible to tell whether the variations are the result of a number of workshops, each utilising its own adjacent raw materials, or inconsistent practices within a single workshop using a variety of nearby resources. Little of the tempering material is represented in the Cusichaca bedrocks and hence it seems unlikely that the Cusichaca area is the site of the workshop(s).

Although these are often referred to as "cooking pots" and their shape is suitable, these wares are rarely found containing carbonised food remains or with smoke-discoloured surfaces. Rather, like the Inca Fine Wares, they are found in high-prestige burials in the provinces outside of Cuzco and therefore were held in some regard. Also like the Inca Fine Wares the Utilitarian Wares are standardized in appearance.

Although the wares lack a petrographical coherence the grain size distribution of the temper and the clean clay suggests that a similar fabric preparation process was employed to that of the Inca Fine Wares but perhaps using fine sand as temper. The non-coherence of the pots suggests that they may have been made in a number of localities, including perhaps the Cusichaca area, but with some state control to ensure the uniformity of appearance.

There is evidence that these pots were traded over long distances, but as they are small light pots, the requirement that the temper have an irregular surface topography to ensure tight bonding with the clay is less important than for the larger Inca Fine Ware aryballi.

These usually are found with charred contents and discoloured surfaces. Although the pots resemble each other in outward appearance, their petrographical fabrics show the lowest degree of coherence of any pre-Spanish wares and it may be that a spectrum of technologies was used, including the use of untempered clays. It is difficult to envisage these simple, low-grade wares being widely traded and therefore it is more probable that a number of potteries using the local materials were involved. Some of the potteries could have been sites in the Cusichaca area. The temper is consistently coarse: this being the most important characteristic for the serviceability of cooking pots.

This petrographical study is based on the pottery from the latest pre-Spanish deposits at Cusichaca: those which date from the Late Horizon Period. However, of great interest in Andean studies are the transitional periods, where one cultural pattern is disrupted by, or subsumed in, another.

Three topics have been considered in the light of the petrographical evidence: the pottery workshops, the trade networks and the requirements of the consumers. A relatively rapid, major social upheaval such as the spread of the Inca Empire, would be certain to disrupt these systems. Only be placing the pottery samples in their developmental context is it possible to understand many of the ceramic characteristics and account for their coexistence in the latest levels.

The stratigraphy of Huillca Raccay, Cusichaca comprises a long sequence of deposits, the last of which (Late Horizon Period) contains the pottery described in this study. The stratigraphy spans the beginning of the Inca occupation of the Cusichaca Valley up to, and beyond the end of the Inca Empire.

In the underlying deposits below the Late Horizon Period, Killke Wares are the first to appear followed by Ware 29 of the Inca Fine Wares. Although sherds from these earlier horizons were not included in the present investigation, the hand-specimen study indicates strongly that each ware was substantially the same from its first appearance to the top of the sequence. Towards the top of the sequence Inca Utilitarian Wares, Cooking Pots and Inca Fine Wares appear together. These wares form the main components of the assemblages, are present in large quantities and hence represent major trends in consumption for the inhabitants of the Cusichaca area rather than isolated gifts or fortuitous exchange. Modern Fine Wares, rare examples of which lie on the surface of the sites, are separated from the pre-Spanish wares by 450 years during which time Spanish and Spanish-influenced wares dominated the ceramic repertoire of the Southern Andes.

It is accepted that most of the archaeological pottery was traded into the Cusichaca area, hence removed from its developmental context and that is chronology does not necessarily represent the evolutionary chronology of late Andean ceramics. The Inca Fine Wares and Killke Wares are not of local manufacture and bedrock geology and petrography of their temper suggest they were made in the Cuzco area. Here, there is sparse stratigraphical evidence to show that Killke Wares predate Inca Fine Wares and that Inca Fine Wares and Inca Utilitarian Wares are contemporaneous in age - both results are in agreement with the Cusichaca chronology. As there is no independent evidence for the relative chronology of Ware 29, its stratigraphical position at Cusichaca is taken to truly reflect its evolutionary position namely post-dating Killke and pre-dating basalt-andesite-tempered Inca Fine Wares.

Production of Killke Wares spans approximately 200 years into the Late Horizon Period and stylistically shows little variation of discernible technological trends throughout this time. This conservatism suggests that the market for the wares remained unchanged and that the potters had sufficient supplies of similar raw material. The presence of Killke Wares in the Late Horizon Period at Cusichaca suggests that some substratum of the pre-Inca trade network survived to supply an indigenous section of the population with their traditional fine wares.

Petrographically, much of Ware 29 (S29-1 to S29-4) had greater similarities with Killke Wares than with basalt-andesite-tempered Inca Fine Wares although stylistically it is an Inca Fine Ware. It is tempting to interpret this as representing the beginning of the reorganisation of some Killke manufacturing sites into state-controlled mass production at the beginning of the Inca period in the Cuzco area.

The technology of the basalt-andesite-tempered Inca Fine Wares represents a refinement of, and greater standardization in, the earlier processes, most notably in cleaner clays and the use of unilithic temper. Elsewhere, Lunt (1987) has argued that the fabric differences between Killke and Inca Fine Wares were sufficiently marked to require the introduction of technical expertise from the coast to form the basis of the Inca pottery industry. With the present recognition that the Killke fabrics are tempered (contrast Lunt 1987: Appendix 1) and the transitional nature of Ware 29 it can be argued that the Inca technology was the result of a long development in situ. In its most sophisticated form, namely basalt-andesite-tempered Inca Fine Wares, this technology is consistent with the long-held view of state-controlled mass-production. As the Cusichaca evidence shows, this system, once established, had the potential to supply a steadily expanding market throughout the rest of the Inca period.

There is no good technological precursor for the Inca Utilitarian Wares in the pre-Inca levels, but clearly these wares are related technologically to the Fine Wares. Their shapes are quite new, but their manufacturing process, like that of Killke and Inca Fine Wares, used clean clays and temper.

Similarly the Cooking Pots are the first specialized cooking utensils to appear at the Cusichaca sites. Previously a variety of wares, some of which were decorated but which never included Killke Wares, can be seem to have been used in an ad hoc fashion for cooking. In the Late Horizon Period the number of pottery shapes proliferated, suggesting increased specialization perhaps associated with new activities.

This greater variety of wares at Cusichaca may have been the result of new pottery centres, the stimulation of existing ones due to the expansion of traditional markets or a 'pax Incaica' was permitting the Cusichaca inhabitants access to a wider range of wares.

Dr. J.R. Ashworth is thanked for the statistical analysis of the size distribution data and Professor D.D. Hawkes for laboratory facilities. C.R. Ixer is also thanked for retyping this paper and converting it to an on-line publication.

Deruelle, R. 1977, Carte géologique de la déflexion d'Abancay entre San Miguel at Cuzco. Office de la Récherche Scientifique at Technique Outre-mer 1.500,000.

Dwyer, E. B. 1971. The early Inca occupation of the valley of Cuzco, Peru. Unpublished PhD thesis, University of California, Berkeley.

Folk, R. L. 1974, Petrology of sedimentary rocks. Hemphill Publishing Co., Austin, Texas.

Gregory, H. 1916, "A geological reconnaissance of the Cuzco valley", American Journal of Science (4th Series) 41, 1-100.

Kendall, A. 1985, Aspects of Inca architecture. British Archaeological Reports, International Series, 242.

Lunt, S. W. 1987, Inca and pre-Inca pottery. Unpublished PhD thesis, University of London.

Menzel, D. 1976, Pottery style and society in ancient Peru. University of California Press, Berkeley.

Morris, C. 1978, "The archaeological study of Andean exchange systems", in Redman, C.L. (ed.), Social archaeology: beyond subsistence and dating. Academic Press, New York, 315-328.

Proetzen, J.P. 1983, "Inca quarrying and stone-cutting", Nawpa Pache 21, 183-214.

Rowe, J.H. 1944, An introduction to the archaeology of Cuzco. Expeditions to Southern Peru, Report No 2. Papers of the Peabody Museum of American Archaeology and Ethnology, Harvard University, Volume XXVII.

Rowe, J.H. 1962, "Stages and periods in archaeological interpretation", Southwestern Journal of Anthropology 18, 40-45.

Rowe, J.H. and Menzel, D. 1967, Introduction to Peruvian archaeology: selected headings. Peak Publications, Palo Alto, California.

Shepard, A.O. 1976, Ceramics for the archaeologist. Carnegie Institution of Washington, Washington D.C.