Cores

A very high number of cores were recovered from the 1x1 m workshop test unit (n = 352) and these cores were classified using the strict definition that the cores have not a single positive percussive feature. In the excavated sequence at Q02-2u3, cores show a distinct standardization in production between level 5 and level 4, a change in reduction strategy in level 3, and finally in the upper most levels 1 and 2 the pattern reverts to characteristics similar to that observed in the lowest levels.

Level		Obsidian		Rotations						Total
		Ob1	Ob2	0	1	2	3	4	5
1	No.	57	2	15	21	15	6	2	0	59
	% Lvl	96.6%	3.4%	25.4%	35.6%	25.4%	10.2%	3.4%	0%	100%
2	No.	46	3	18	20	4	6	1	0	49
	% Lvl	93.9%	6.1%	36.7%	40.8%	8.2%	12.2%	2.0%	0%	100%
3	No.	65	5	31	25	11	2	0	0	69
	% Lvl	92.9%	7.1%	44.9%	36.2%	15.9%	2.9%	0%	0%	100%
4	No.	90	4	25	31	24	10	3	0	93
	% Lvl	95.7%	4.3%	26.9%	33.3%	25.8%	10.8%	3.2%	.0%	100%
5	No.	53	1	26	17	7	3	0	0	53
	% Lvl	98.1%	1.9%	49.1%	32.1%	13.2%	5.7%	0%	0%	100%
6	No.	7	1	1	3	0	1	1	0	6
	% Lvl	87.5%	12.5%	16.7%	50.0%	.0%	16.7%	16.7%	0%	100%
7	No.	15	3	4	3	3	3	2	2	17
	% Lvl	83.3%	16.7%	23.5%	17.6%	17.6%	17.6%	11.8%	11.8%	100%
Total	No.	333	19	120	120	64	31	9	2	346
	% Lvl	94.6%	5.4%	34.7%	34.7%	18.5%	9.0%	2.6%	0.6%	100%

Table 7-17. Cores at Q02-2-u3 by excavation level showing counts and proportions of Ob1 and Ob2 material and rotations.

Figure 7-16. Q02-2-u3 Cores by Level showing changes length and weight.

These patterns can be summarized as follows. First, in the early levels, such as in basal level 7, large nodules were available and they were knapped more extensively, and then the cores were discarded while they were still relatively large in size, particularly if they were made of Ob2 obsidian. Rotations:In these earlier levels a high proportion of cores were rotated extensively, showing conservation behavior. Ob2 obsidian:three cores (16%) in level 7 contained heterogeneities. Some of these cores had a relatively high number of rotations, with three rotations on an Ob2 core in level 7 and there were four rotations on a lone Ob2 core from level 6. It appears that the presence of bubbles and other heterogeneities was a less significant deterrent to production in the early levels of workshop activity.

Subsequently characteristics of intensification and decline were observed in middle and upper levels of Q02-2u3. Discarded cores became increasingly more ovate through the intensification period of levels 5 through 3 (see Figure 7-16), where weight increased without a concurrent increase in length. Higher up, in level 2, the mean weight of cores plummeted by 15 grams. These changes in the morphology of discarded cores suggest that more detail on size variability among cores could shed light on the processes at work. These multivariate patterns can be illuminated through cluster analysis.

Cluster Analysis with Cores

Using the cluster analysis of cores from the workshop test unit, a few groupings of cores emerge from clustering on six variables. These more general groupings permit the examination of changes in core attributes in terms of types that reflect the underlying variability in the dataset.

Cluster analysis is a classification method that arranges a set of cases into clusters that are more similar to each other than they are to cases from other clusters (Aldenderfer 1982;Shennan 1997:234-258). The values from six ratio and scale variables of 334 complete obsidian cores from Q02-2-u3 workshop test unit were examined using hierarchical cluster analysis. Hierarchical clustering produces nested clusters with increasing variance within clusters as one ascends the cluster hierarchy.

Cores were clustered by first exploring different clustering methods on the Z scores of following measures: Weight, Length, Medial Width, Medial Thickness, Rotations, and Percent of Cortex. The four common hierarchical clustering methods: Single Linkage, Complete Linkage, Average Linkage and Ward's Method were conducted. This exploration revealed that regardless of the method selected, three cluster solutions were most commonly observed in the clustering agglomeration schedules. Three clusters generated by a Complete Linkage clustering solution provide the tightest clusters with measures on cores. Complete Linkage clustering, also referred to as "Maximum" and "Furthest-Neighbor" Method, is an approach where the dissimilarity between two groups is equal to the greatest dissimilarity between a member of cluster 1 and a member of cluster 2. Complete Linkage clustering tends to produce tight clusters of similar cases and the algorithm delays including new members in a cluster; furthermore, it does not tend to produce chaining.

	Function
	1	2
Weight (g)	.803*	.537
ThickMed (mm)	.723*	.109
% Cortex	.476*	-.366
Rotations	-.304	.654*
WidthMed (mm)	.193	.514*
Length (mm)	.264	.486*
* Largest absolute correlation between each variable and any discriminant function.

Table 7-18. Canonical discriminant function for Q02-2u3 Cores.

Clusters derived from cores are expressed in two discriminant functions shown in Table 7-18. This table shows the structure matrix for cores in order of size of correlation. Function 1 describes core sphericity using the Weight and Thickness measures, while Function 2 describes tabular shaped cores (Width and Length) with high Rotations.

Figure 7-17. Canonical discriminant chart for core clusters from Q02-2u3.

The variables were standardized using Z scores so that they had a mean of 0 and a standard deviation of 1, which has the result of regularizing measures like Percent Cortex (ranging from 0-100%) so that they cluster more effectively against the values of measures like Medial Thickness (ranging from 0.5 to 34cm). A description of the morphology of these three core clusters will be followed by an examination of the relative frequencies across levels.

Cores were clustered into three groups with low eigenvalues and relatively proportional counts in each cluster. An inspection of the clustering measures across the three groups revealed that the three clusters described three types of cores that were apparent during the lithic analysis.

Figure 7-18. Boxplots showing three clusters of Q02-2u3 cores with size measures. The tabular shape of C2 cores is apparent in their thinness relative to their length and width.

Characteristics of Three Core Clusters from Q02-2u3
	C1(n=226)	C2(n=56)	C3(n=52)
Size	Light, short, narrow cores that started on small nodules.	Mid-weight, and long and wide, but as thin as C1 cores.	Heavy, long, wide, and very thick. Ovate in shape. From larger nodules relatively unreduced.
Cortex and Rotations	Moderate amounts of cortex, moderate number of rotations on large number of cores.	Virtually no cortex and high number of rotations on cores.	High amounts of cortex, few rotations.
Ob2 Fraction	Few (n=10), or 4%.	Moderate (n = 8), 14%	None (n=0)
Comments	Smaller initial nodules.	Started on large nodules, and most heavily reduced.	Sampled but discarded early.

Table 7-19. Q02-2-u3 characteristics of three clusters from core attributes.

Did these clusters of cores result from intrinsic characteristics of three types of starting nodules, or are the clusters the result of reduction strategies and the particular reduction history of that core? C2 cores, for example, appear to have started as "tabular cores" but following some reduction they were discarded at the workshop despite being fairly large.

There is little correlation between clusters and attributes of the material such as Ob2 (hetereogeneous), cortex type, or color and banding characteristics. C1 and C2 cores separate easily into two groups: the C1 consisting of small cores and C2 with long, narrow cores that were heavily reduced. However, the cluster 3 (C3) cores are more enigmatic because they were discarded relatively early in the reduction sequence and these cores are evident throughout the sequence the workshop so the reason for their discard is not immediately obvious.

Figure 7-19. An example of a C1 core [L03-162.303], arrows indicate percussion.

Figure 7-20. An example of a C2 core [L03-162.305].

Figure 7-21. An example of a C3 core [L03-162.280] showing a small area of thin cortex.

Raw material availability and the opportunities for reduction in the particular reduction trajectory of a given core appear to determine the cluster in which a discarded core is categorized using this system. Following this logic, if the abundance beginning in Level 5 is due to quarrying work at the quarry pit, 600m up the slope, then additional large nodules would have become available from this quarry and with increased raw material availability, larger cores were discarded at the workshop during this time of abundance.

Figure 7-22. Clustered Cores from Q02-2u3 workshop test unit showing counts by level.

Changing proportions of these clusters of cores between levels, particularly between levels 5 and 2, suggests that there was a changing availability of raw material. A distinctive pattern emerges that centers on level 4 as the probable apex of production of large flake blanks for export from the workshop. In level 4 there are a relatively large number of C2 cores with 3 or 4 rotations and yet they are over 50mm long and appear to have been discarded far earlier than would have occurred in other levels in the workshop. The varying counts and weights of these clusters evident in Figure 7-22 show a peak in large, cortical cores (C3) in level 5 that are replaced by heavily reduced tabular cores (C2) in level 4, and finally both groups give way to an increase in counts of small, moderately retouched cores (C1) in level 3.

The presence of quantities of C2 cores in Level 4 is particularly intriguing because the discard of cores with little cortex (and thus heavily worked), yet still evidently long and tabular, suggests that very large cores were abundant in this phase. By level 3 the C2 cores were replaced by the smaller C1 cores that are moderately reduced, which are taken to mean that large cores were no longer available and surpluses of large cores remaining in the system from earlier times had been consumed. The most parsimonious explanation for this abundance of large, tabular C2 cores in level 4 is that a new source of obsidian were being exposed through labor investment in quarrying in the vicinity of the workshop. A¹⁴C date from level 4 calibrates to between 1500 - 1380 BCE, or the end of the Early Formative, precisely when obsidian is most abundant at consumption sites like Qillqatani (discussed in Section 3.4.2).