Paste classification

Clearly frequency profile analysis may be applied to very complicated index schemes involving multiple attributes and attribute states. For the analysis of the Taraco Peninsula surface data, however, I designed a very simple index scheme, which is outlined below. Given that I had a limited amount of time to complete the analysis of a substantial ceramic dataset (approx. 94,000 sherds) I decided to use only a single attribute in defining the phase index profiles. The attribute I chose was paste, for two principal reasons; first, Lee Steadman had already demonstrated that there were quite significant shifts in the relative frequencies of pastes between the phases of her Chiripa sequence; second, simple paste classification is a very rapid process and enabled me to complete the analysis within the available time.

My paste classification, designed as it is simply to distinguish the three phases of the Chiripa sequence, is considerably less elaborate than that of Steadman (see Table 3.1 for the relationship between my paste groups and Steadman's pastes). As Cowgill has observed with reference to seriation, ``often one can throw away much of one's information and still obtain a good chronological ordering'' ([Cowgill 1972]). Thanks to Steadman's work, I was able to avoid collecting much ceramic information at all, and focus instead on only a few factors known to have chronological significance. Thus the attribute ``paste'' has in the present analysis only four states: paste groups 1, 2, 3 and other. It should be emphasized that all of these paste groups are subsets of Chiripa ceramics; thus, all are fiber-tempered.

1. Paste Group 1: Any sherd containing one or more large, opaque, angular quartz fragments. This group is similar to Steadman's Paste 21, which as she says accounts for 48% of the Late Chiripa assemblage ([Steadman 1999]: 66). This paste group is easily identified without magnification. Composed principally of Steadman's Pastes 19 and 21. Identified without magnification.
2. Paste Group 2: Any sherd not fitting the description of paste group 1 which has translucent, rounded quartz inclusions. These inclusions are the predominant tempering agent. This paste group is most common in the Middle Chiripa phase. Composed principally of Steadman's Pastes 19 and 26. Identified under 10X magnification.
3. Paste Group 3: Any sherd not fitting the description of groups 1 or 2 which has a large amount of mica visible in the matrix and on the surface. This paste group is most common in the Early Chiripa phase. Composed principally of Steadman's Pastes 17, 18 and 19. Identified without magnification.
4. Paste Group Other: Any Chiripa sherd not belonging to any of the three paste groups defined above. Principally, in my sample, Steadman's Pastes 16, 19 and 23.

The analytical process which I applied in order to generate counts for each of the four paste groups is represented schematically in Figure 3.2. It is important to note that in my scheme no sherd can belong to more than one group, and the determination of group membership must be made in a particular order. Thus, a sherd with both angular opaque quartz inclusions and abundant mica in the paste would, according to my process, be classified as paste group 1, not as paste group 3. This is because the assignment to paste group 1 comes earlier in the analytical sequence. My classification is therefore a taxonomy, in Dunnell's sense of the term ([Dunnell 1971]).

Figure 3.2: Chiripa ceramic analysis process flow chart

I would like to emphasize again that my paste groups differ from Steadman's pastes in several ways. Though I use some of the same attributes she uses to defined pastes, I evidently construe them more broadly than she. For example, the presence of transluscent, rounded quartz inclusions is used by Steadman to define her Paste 26. The same attribute is also used by myself to define my paste group 2. And, indeed, as Table 3.1a shows, virtually all of Steadman's Paste 26 sherds (over 90%) fall into my paste group 2. However, many other sherds from other pastes also fall into paste group 2, especially sherds Steadman had classified as Pastes 19 and 30. This would seem to indicate that I am including a broader range of objects in my definition of ``transluscent, rounded quartz inclusions'' than is Steadman. The same is true of the other attributes I use for the other paste groups. This is not a real problem, of course, but it is yet another indication that anyone wishing to duplicate my methodology will need to obtain a bag of pasted ceramics and generate their own index profiles for the Chiripa phases beforehand. A calibration step is apparently necessary for each analyst.

Table 3.1: Correlation of CAT Paste Groups and Steadman's Pastes.

• a) counts

	Paste Group
Steadman Paste	1	2	3	Other Chiripa	Total
16	0	5	1	10	16
17	0	13	144	7	164
18	6	10	51	2	69
19	21	79	65	40	205
20	0	9	12	4	25
21	72	0	0	0	72
22	0	1	2	0	3
23	1	4	5	8	18
24	0	0	1	3	4
25	0	2	20	2	24
26	1	90	1	0	92
30	0	12	6	0	18
Total	101	225	308	76	710

• b) proportions

	Paste Group
Steadman Paste	1	2	3	Other Chiripa	Total
16	0.0	2.2	0.3	13.2	2.3
17	0.0	5.8	46.8	9.2	23.1
18	5.9	4.4	16.6	2.6	9.7
19	20.8	35.1	21.1	52.6	28.9
20	0.0	4.0	3.9	5.3	3.5
21	71.3	0.0	0.0	0.0	10.1
22	0.0	0.4	0.6	0.0	0.4
23	1.0	1.8	1.6	10.5	2.5
24	0.0	0.0	0.3	3.9	3.6
25	0.0	0.9	6.5	2.6	3.4
26	1.0	40.0	0.3	0.0	13.0
30	0.0	5.3	1.9	0.0	2.5
Total	100	100	100	100	100

For the purposes of my analysis, only sherds of paste groups 1-3 were counted. Therefore, if a is the number of sherds of paste group 1, b the number of sherds of paste group 2, c the number of sherds of paste group 3 and n equals a+b+c, then a frequency profile takes the following form $\left(\frac{100a}{n}\right)$- $\left(\frac{100b}{n}\right)$- $\left(\frac{100c}{n}\right)$ where `-' is used as a delimiter and not as a subtraction operator. Therefore, if a particular group of sherds has 21 of paste group 1, 46 of paste group 2, 14 of paste group 3 and 25 of paste group other, the assemblage profile would be $\left(\frac{100*21}{21+46+14}\right)$- $\left(\frac{100*46}{21+46+14}\right)$- $\left(\frac{100*14}{21+46+14}\right)$ or 26-57-17.