ebook img

Ceramic Micro-Seriation: Types or Attributes? PDF

13 Pages·0.276 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Ceramic Micro-Seriation: Types or Attributes?

REPORTS CERAMIC MICRO-SERIATION: TYPES OR ATTRIBUTES? Andrew I. Duff Micro-seriation using attributes of decorated ceramics has been shown to accurately refine intrasite and intersite relative dat ing. Using data from Pueblo de Ios Muertos, a nucleated town in west-central New Mexico, this presentation demonstrates that micro-seriation ol type frequencies produces equally accurate results. Typological analysis also provides substantial time savings when compared to attribute recording. Additionally, using types with established temporal ranges permits link age of relative seriation with absolute dates. A combination ol correspondence analysis and k-means cluster analysis was .fimnd to provide sound and easily interpretable results. The results of a typological seriation ol Pueblo de Ios Muertos deposits are examined, and the utility of typological seriation for other contexts is considered. La micro-seriacion de atributos de decoracion cenimica ha sido considerada como la tecnica acertada para refinar !as .fixh as relativas dentro de un sitio y entre varios sitios. Este articulo utiliza datos de Pueblo de Ios Muertos, un pueblo nude ado localizado en el centro-oeste de Nuevo Mexico, para demostrar que la micro-seriacion defrecuencia de tipos ceramicos produce resultados igualmente acertados. El ana/is is tipologico tambien representa un ahorro de tiempo substancial en com paracion con el tiempo que se necesita para analizar Ios atributos decorativos. Ademas, el uso de tipos con rangos tempo rates establecidos permite en/azar la seriacion relativa con fechamientos absolutos. Una combinacion entre Analisis de Correspondencia y el metodo de agrupamiento con k-medias produjo resultados confiables y.fiiciles de interpretar. Aqui se examinan Ios resultados de una seriacion tipologica de Ios depositos de Pueblo de Ios Muertos y se cons ide ra la uti/ida d de la seriacion tipologica en otros contextos. T his analysis has three objectives: to present on micro-seriation. This provides a basis for com techniques valuable in the seriation of parative evaluation of techniques and procedures archaeological data; to evaluate the utility useful for micro-seriation, and reveals that type of type frequency seriation relative to attribute frequencies produce equally accurate results. The based seriation (Le Blanc 197 5); and to examine value of intrasite seriation is considered relative the nature of the occupation and abandonment of to its ability to provide insights into the occupa Pueblo de Ios Muertos, a nucleated town site in tion of Pueblo de Ios Muertos (LA 1585). west-central New Mexico investigated by the Seriation Techniques Cibola Archaeological Research Project. This presentation begins with an overview of Seriation is the process of ordering a number of se1iation techniques, and presents an approach observations relative to some dimension of vari that couples correspondence analysis with k ability (Spaulding 1978:27). In archaeology, the means cluster analysis. A brief introduction to the relevant dimension is often time, but any number Cibola project is followed by description and of dimensions can be seriated. A detailed consid examination of data from Pueblo de Los Muertos, eration of the theoretical underpinnings of seri and comparison of typological seriation results to ation is not warranted here (see Cowgill 1972; attribute seriation performed on many of the same Dunnell 1970; Marquardt 1978; Rouse 196 7), but units by LeB!anc (1975) in his pioneering article a number of critical aspects require discussion. Andrew I. Duff• Department of Anthropology, Arizona State University, Tempe, AZ 85287-2402 American Antiquity, 61 (I), 1996, pp. 89-10 I. Copyright ©by the Society for American Archaeology 89 90 AMERICAN ANTIQUITY [Vol. 61, No. 1, 1996] Central to any chronological seriation is selection K-means cluster analysis (Doran and Hodson of appropriate units of observation. It is also nec 1975: 180-184; Kintigh and Ammerman 1982) is essary to recognize that interpretation is indepen a nonhierarchical clustering procedure for group dent of the ordering procedure (Dunnell 1970; ing cases. One distinct advantage of this tech Johnson 1972; Marquardt 1978; Rouse 1967). nique is that it provides information about clusters A variety of techniques for the ordering of in terms of the input data. If the input data consist archaeological data have been employed. of percentages of different ceramic types, k Occurrence data (presence/absence) and matrix means clusters are described by the mean and reordering often produce satisfactory chronologi standard deviation (in percentages) of the pottery cal ordering given small assemblages and a lim types composing each cluster (this information ited number of artifact categories. More popular can be calculated for clusters created by any clus techniques for seriating numerous archaeological tering algorithm). This differs from other multi cases (sites, features, or deposits) include cluster variate techniques such as MDS, CA, PCA, and analysis approaches (Matson and True 1974 ), and factor analysis where one gets a series of"scores" a variety of multivariate scaling techniques, prin or "loadings" that cannot be so directly related to cipally variants of multidimensional scaling input data. Cluster characterization data presents (MDS) (Cowgill 1968, 1972; Matson and True the opportunity of calculating mean ceramic dates 1974; True and Matson 1970; Washburn and (e.g., South 1978) for each cluster. This can pro Matson 1985). The benefit of combining scaling vide approximate absolute dates to portions of a and clustering techniques has also been noted multivariate display. (Djindjian 1985, 1988; Johnson 1972 ). The Database The approach taken here is to combine corre spondence analysis (CA), a multivariate scaling The data used in this analysis derive from the procedure, and k-means cluster analysis. using the Cibola Archaeological Research Project, a two patteming derived from each to evaluate consistent year survey and excavation project conducted in patteming in the original data (Cowgill 1968). the El Morro Valley of west-central New Mexico Because of its advantages, I have opted to use a (LeBlanc 1978; Watson et al. 1980).2 The Zuni data analysis and display technique (CA) not well (Cibola) region, and the El Morro Valley in par established in Americanist literature, but which is ticular, contains dense occupation between A.D. widely used in Europe for seriation and other appli 1250 and A.D. 1450 in two different settlement cations. I anticipate that this paper will contribute types. The pre-A.D. 1275 pattern consists of spa to a wider appreciation of the possibilities of CA. tially dispersed multiple-room block settlements CA is a technique for data reduction and display that appear to interact as a community (LeBlanc similar, in many respects, to principal components 1978). The post-A.D. 1275 sites are planned, analysis (PCA) and MDS (Hill 1974; Ringrose rapidly constructed nucleated towns with 200 to I 992:615). However, it works directly on counts l ,400 rooms that were occupied for only a few and its strength as a display technique is the simul generations (Kintigh 1985a). taneous projection of both cases and variables in The Zuni region has a well-developed deco the same spatial dimensions. 1 In a sense, this helps rated ceramic chronology that permits accurate one interpret the relative contribution of each vari assignment of occupation ranges, usually to within able to the dimensional representation (80lviken et 50 years. based on ceramic groups (Cotton al. 1982:43). CA has been widely used in Europe 1953:65; Kintigh 1985a). Fine-grained seriation, and has proven useful for seriation (80lviken et al. or micro-seriation (LeBlanc 1975; Marquardt 1982; Djindjian 1985, 1989; Hodson and Tyers 1974) using ceramic attributes, has been used to 1988; Ihm 1987; Laxton and Restorick 1989; show relative temporal intersite and intrasite rela Leredde 1985; Madsen 1988, 1989; Madsen, ed., tionships. In order to make temporal discrimina 1988; Slachmuylder 1985). Awareness of CA is tions within periods defined by ceramic groups, increasing as it becomes incorporated into widely Marquardt (1974) and LeBlanc (1975) focused on used computer statistical packages. ceramic attributes thought to be temporally sensi- REPORTS 91 tive. They employed factor analysis and MDS, isolable, some of which vary with time, while respectively, in their seriations. LeBianc used others may not (e.g., Plog 1983:132-133). physical attributes of ceramics (e.g., Munsell Recording attributes individually may isolate color, slip color) and other variables, such as which attributes are time sensitive. If we grant jar:bowl ratios in his study (1975:26-27, Table 3). this, then, at an intuitive level, it is apparent that His units derived from Pueblo de Los Muertos and types created for time-space systematics should Scribe S, two sites that were the focus of Cibola represent constellations of potentially trend-sensi project excavation efforts (LeBianc 1978). tive attributes. The question then becomes: Does Marquardt employed attribute ratios (1974:97, the homogenizing effect of typological analysis 100, Table 3-8), stressing the need for logically diminish trend-sensitive information to the extent independent variables (1978:304). Marquardt that it is incapable of informing upon fine-grained ( 1974) was more concerned with intersite chronol temporal variation? ogy, and demonstrated the contemporaneity of Given a situation where a number of reason deposits from a number of the nucleated pueblos ably well-dated ceramic types occur in varying (ea. A.D. 1275-1400) in the El Morro Valley. Both frequencies within excavated assemblages from a sought to avoid pitfalls associated with typological single site, the potential for capturing temporal analyses by employing attributes (e.g., Le Blanc variation of attributes via typological frequency 1975:24-26; Marquardt 1978:284, 296-297; analysis appears likely. Cowgill ( 1972) has noted Redman 1977:46, 1978: 187; Watson 1977:387). that loss of some information may actually be The argument against using typological data is desirable for seriation. I sought to determine if in need of reevaluation. The primary reasons cited analysis of type frequencies alone could repro for rejecting typological data are that small pieces duce the results of the more detailed seriation frequently cannot be assigned to type, resulting in work successfully conducted by Marquardt loss of information (LeBianc 1975 ), that type iden ( 1974) and Le Blanc ( 1975). If so, I hoped to iden tification is too subjective and variable from per tify the advantages and limitations of a type-based son-to-person (Plog 1983: 136; Washburn 1983; approach. The data used in this analysis come Washburn and Matson 1985:75-76), that types from Pueblo de Los Muertos, a nucleated town in were originally created to systematize (normative) the El Morro Valley that was one focus of Cibola time-space relationships, and that types obscure or project excavations. The site provides an ideal test homogenize variability (Redman 1977:46). case because the attribute seriation is well known The lack of an adequate number of sherds and readily accessible (Le Blanc 1975 ). assignable to type is generally not a problem. In Decorated ceramics from the Cibola project fact, many of the ceramic attributes recorded by were originally tabulated using a detailed attribute the Cibola project (Redman 1978) seem to require coding system (Redman 1977). The data dis a sherd that is probably large enough to assign to cussed and analyzed here consist of decorated type. Typological assignment is much faster than ceramic type counts from Pueblo de Ios Muertos. recording a number of attributes for each sherd, as The sherds were tabulated by Watson and Kintigh. is creating the resultant database. There are sub Although other types are present, only counts of jective elements involved for attribute or type the following decorated ceramic types were uti recording. Types represent associations of "recur lized: Tularosa Black-on-White, St. Johns rent attributes." Whether these are statistically Polychrome, Springerville Polychrome, Pinedale significant or not is less relevant (Cowgill Polychrome, Kwakina Polychrome, and 1972:386; Rouse 1967:166-167; Spaulding Heshotauthla Polychrome. Other ceramic types 1978:29-30). The fact that types were specifically are comparatively rare, providing little additional designed to help order observations along a temporal information. Approximate date ranges chronological dimension (Dunnell 1970:308; for these ceramics and the relative order of their Watson 1977) often makes them appropriate for appearance are presented in Figure 1. Carlson temporal seriation. Types are constellations of ( 1970) presents detailed descriptions and pho attributes, many of which are individually tographs of these types except forTularosa Black- 92 AMERICAN ANTIQUITY [Vol. 61, No. 1, 1996] I Tularosa Black-on-white St. Johns Polychrome Springerville Polychrome CaD. f- ~ Pinedale Polychrome Kwakina Polychrome '-- Heshotauthla Polychrome - Figure 1. Approximate date ranges I for Zuni-area decorated ceramic 1200 1250 1300 1350 1400 types used in this analysis. Years A.D. on-White. Tularosa Black-on-White occurs pri marily in jar form, and can be considered an Statistical Package written by Kovach.3 Each of approximate design equivalent of St. Johns these procedures can be found in more widely Polychrome. available statistical packages. Typological information was not recorded dur Analysis of Cibola Type Frequency Data ing the original coding, so I have only used col lections that have since been typologically Several analyses were conducted to ascertain the analyzed. All ceramics were shaken through /< degree to which relative seriation based on type inch mesh, with typological assignment attempted frequencies (and/or percentages) could aid in for all pieces larger than Y, inch (x;-inch mesh was archaeological interpretation. Initially, units were used for excavation). Sherds not identifiable to simply aggregated into upper and lower levels type were tabulated as such. The most problem from each provenience. This provides information atic typological distinctions are between St. Johns about general relationships among units and can Polychrome and Heshotauthla Polychrome, bowls produce a well-ordered seriation. However, analy characterized by a gradient from thick-to-thin sis of individual levels or small aggregates of lev exterior white lines and from matte-to-subglaze els was thought to more accurately reflect the to-glaze interior paint dating A.D. 1275-1300. complex relationships inherent in these archaeo Because only two persons, often working logical deposits. together, typed all of the sherds, consistent biases Attribute and Type Seriation related to individual differences in typological assignment should be reduced (e.g., Plog Correspondence analysis and k-means both isolate 1978:159, 1983:136). the same axes of variability within Pueblo de Ios The data consist of counts from multiple prove Muertos assemblages. A plot of the first two CA niences and several excavated levels. Occasionally, axes (Figure 2) shows the relative contributions of adjacent levels or entire units with small sample the various types. Individual cases (levels) are sizes or only excavated for a few levels were com plotted with symbols representing the k-means six bined. Table I presents ceramic counts by type for cluster solution. K-means groups together exca all of the proveniences used in this analysis, with a vated levels containing similar percentages of dec column specifying comparable levels used by orated types. This demonstrates that both LeBlanc (1975). For the k-means analyses pre techniques arrange cases in similar manner. Axis I sented here, ceramic type counts have been con accounts for 70 percent of the variance, and can be verted to (row) percentages to minimize the effect interpreted as time. Axis I places St. Johns of sample size (Cowgill 1968:372). I have Polychrome and Tularosa Black-on-White, the employed a k-means program written by Kintigh earlier types, and earlier units to the left, with later (Kintigh and Ammerman 1982). The CA analysis types and units to the right. The second axis, was performed on counts using the Multi-Variate accounting for 16 percent of the variance, captures REPORTS 93 Table l. Analysis of Type Counts from Pueblo de Ios Muertos Proveniences. Tularosa St. Johns Springerville Pinedale Kwakina Heshotauthla Unit Level(s) Case Black-on-White Polychrome Polychrome Polychrome Polychrome Polychrome LeBlanc (1975) 19 I & 2 I 23 44 7 0 12 18 19 5 2 9 22 7 3 15 23 19-5 19 6 & 7 3 84 147 11 5 26 21 19-6 20 I & 2 4 22 59 3 2 45 50 20 3 & 4 5 13 33 2 0 13 25 20 5 & 6 6 28 42 3 4 29 87 20-5 20 7 & 8 7 23 64 5 2 5 18 20 9 8 10 42 0 0 5 2 20-9 20 10 & 11 9 47 62 0 0 0 104 4 10 20 144 9 4 105 2 11 47 31 4 0 16 16 105-2 105 4 & 5 12 43 228 6 11 (105-6) 105 7 13 23 38 0 5 ( 105-8) 108 I 14 32 57 18 11 64 56 108 2 15 19 39 8 9 53 112 108-2 108 3 16 37 29 9 15 51 88 108 4 17 21 17 3 26 43 108 5 18 15 25 0 0 13 12 108-5 108 7 19 29 53 0 0 8 0 108-7 225 2 & 3 20 8 26 5 I 8 17 225 6 21 10 40 2 2 4 4 225 7 22 10 96 I 0 4 0 300 all 23 12 38 3 I 4 15 303 5 24 18 29 3 3 4 9 (303-2 and 303-4) 303 11 & 12 25 19 38 2 0 9 6 (303-8) 410 all 26 11 6 3 14 15 303-11 (303-13) 418 all 27 14 23 7 9 25 81 419 I & 2 28 6 15 2 5 31 419-2 419 3 29 20 55 4 19 12 44 419 4&5 30 28 78 2 2 13 26 419-4 419 6 31 9 72 4 0 16 3 419-6 419 7 32 4 40 0 0 6 0 500 I 33 57 130 9 38 38 500 2 34 20 95 0 0 13 5 500-2 (500-3) 502 2 35 39 47 0 0 14 3 502 3 36 28 49 0 0 13 4612 I 37 2 43 4 2 15 38 4612 2 38 0 37 I 18 4733 3 & 4 39 18 126 2 2 11 17 4915 2 40 5 57 0 4 2 Note: The last column presents units used by Le Blanc ( 1975:Table 4), with those in parentheses adjacent to levels used here. a combination of functional and possibly temporal Examining the data, then comparing these differences. The primary distinction on Axis 2 is results with LeBlanc's MDS (1975:Figure 7) on a between Tularosa Black-on-White and all other case-by-case basis, requires the plotting of CA types, and Tularosa Black-on-White is the only cases in a single dimension using Axis 1. There is type that occurs primarily in jar form. Tularosa a mild arch, or "horseshoe," detectable in the CA Black-on-White appears to persist in the record plot.4 The arch begins at the lower left, rises slightly later than St. Johns Polychrome (its tem toward the center, and tapers off to the lower right. poral equivalent), probably because decorated jars Plotting points that are curvilinear in a single have longer use-lives than decorated bowls (see dimension alters, somewhat, the relationship LeBlanc 1975:28). among points. However, that the "horseshoe" is 94 AMERICAN ANTIQUITY [Vol. 61, No. 1, 1996] KEY: • ... Ceramic type *1 ...... Case number cluster1 *.& •• C luster2 Cluster 3 *+ •Ciuster4 -ClusterS • -ClusterS ~... .'. N ~ Heshotauthla • Pinedale Poly • Figure 2. Correspondence analysis plot of Pueblo de Ios Muertos ceramic assemblages with case numbers from Table 1. Axis 1-70% relatively weak in this data set (a function of the rank-order measures of association were calcu large proportion of variance captured by the first lated, and both (Spearman's r5 = .956, Kendall's r axis relative to the second) reduces this distortion. = 885) indicate a very strong positive association. Comparing the order of common units indi Both measures vary from -1.0 to 1.0, with 1.0 cates agreement between LeBlanc's MDS and indicating a perfect positive relationship. The sub CA. Table 2 presents a rank-ordering of common stantial agreement between these analyses sug units. The CA units are ordered using the first gests that approximately equivalent results can be axis, and LeBlanc 's are ordered by MDS obtained through the use of typological data. (l975:Figure 7). There are 13 units common to Returning to interpretation of the CA plot, it is both analyses (refer to Table 1). Many are simi difficult to objectively interpret points along a larly ordered, especially the earlier units. Two nonlinear continuum, such as those produced by CA or MDS. However, the graphical presentation Table 2. Rank-Order Association. is extremely helpful for visualizing relationships Unit Level(s) Order of cases to variables (types). K-means can more CA MDS formally characterize these relationships by pro Early viding information directly referable to input 20 9 I I 108 7 2 2 data. Table 3 presents the cluster characterization 500 2 3 3 values for the six-cluster solution seen plotted in 419 6 4 4 Figure 2, including the mean and standard devia 19 6&7 5 5 tion of type frequencies and cases assigned to 303 11 & 12 6 7 each cluster. In addition to illustrating that pat 419 4&5 7 6 105 2 8 8 terning is not a spurious result of the CA tech 108 5 9 9 nique, the k-means cluster characterization 19 5 10 13 information can be used to assign probable 20 5&6 11 11 absolute dates to portions of the CA plot using the 419 1&2 12 10 mean ceramic date (MCD) formula (South 1978). 108 2 13 12 Late Multiplying the mean percentage of each CA ~ correspondence analysis, MDS ~ multidimensional ceramic type by the midpoint of their temporal scaling, Spearman's r, ~ .956, Kendall's 't ~ .885 range (Figure 1) , and then dividing this by the Note: MDS data from LeBlanc 1975: Figure 7. REPORTS 95 Table 3. K-means Solution Information. Cluster Unit/Level(s) MCD Tularosa St. Johns SJ2ringerville Pinedale Kwakina Heshotauthla I 2019. I 04/4, I 05/4 & 5 A.D. 1283 11.5/3.3 74.5/7.0 1.3/1.3 .51 1.0 8.4/3.5 3.8/2.9 225/6, 225/7, 419/6 419/7. 50012. 4 733/3 & 4 4915/2 2 19/6 & 7. 20/10 & !I A.D. 1283 32.8/5.3 54.1/5.2 1.1/1.4 .2/.6 9.1/4.6 3.1/3.0 105/7, 108/7,303/11 & 12 50212, 502/3 3 19/1 & 2, I 05/2. I 08/5 A.D. 1295 26.9/7.5 39.9/7.0 3.6/2.2 1.011.8 13.1/4.5 15.5/2.0 30315, 50011 4 2017 & 8, 300/all, 4!9/4 & 5, A.D. 1294 13.7/8.0 55.7/4.8 2.9/1.3 1.5/.2 5.1/2.5 21.1/6.0 4612/2 5 19/5, 20/1 & 2, 20/3 & 4, A.D. 1308 11.3/5.6 34.3/6.0 4.9/2.8 3.6/3.9 17.2/6.3 28.7/3.7 108/1,225/2 & 3, 4!9/3, 4612/1 6 20/5 & 6, I 08/2, I 08/3, I 08/4. A.D. 1321 13.0/5.6 16.8/4.5 3.1/1.7 3. 7/1.6 19.3/6.1 43.1/7.2 410/all, 418/all, 410/1 & 2 Note: Ceramic types represented are Tularosa Black-on-White, St. Johns Polychrome. Springerville Polychrome, Pinedale Polychrome, Kwakina Polychrome, and Heshotauthla Polychrome. Mean ceramic date (MCD), type percentage mean/standard deviation, and cases are presented by cluster. sum of the mean percentages, produces an MCD recently demonstrated that MCD can provide for each cluster (South 1978). For the calculations results as accurate as other chronometric tech used here, I have employed a midpoint date of niques, significant concern about the approach A.D. 1275 for Tularosa Black-on-White and St. persists. Refinements in the MCD approach that Johns Polychrome because the cutting dates for produce date ranges rather than mean dates (e.g., the earlier occupation of the site immediately Steponaitis and Kintigh 1992), and in the dating adjacent to Pueblo de los Muertos (ScribeS) sug of ceramics it is applied to, should yield more gest the majority of the earlier construction post reliable absolute date referents. dates A.D. 1250 (Watson et aL 1980:205). The Occupation and Abandonment of Pueblo de purpose of using South's formula here is to !os Muertos demonstrate the value of k-means cluster charac terization in seriation. MCDs for each cluster are Le Blanc ( 1975) did not engage in specific interpre also listed in Table 3. The MCDs for clusters 1 tation of his seriation data; an attempt to do so is and 2, and 3 and 4 are nearly identicaL This sug made here because our ability to "micro-seriate" gests that while these clusters ( 1 and 2, 3 and 4, deposits is meaningless unless it can provide other respectively) may be characterized by different Wise unrealized insights. The construction types, they are temporally equivalent. The cases in sequence ofPueblo de los Muertos is relatively well these clusters, respectively, occupy the same por documented based on extensive wall trenching tions of Axis 1 in the CA plot (Figure 2), support (Marquardt 1974:37, 39; Watson et aL 1980). ing the inference of this axis as primarily Initially, two rows of rooms were constructed in a representing time. large square (Figure 3 ), with another interior tier The MCD formula produces dates that are then added. Watson et aL (1980:207) suggest the remarkably consistent with other information site was not occupied until this was completed. known from Pueblo de los Muertos (see follow Dendrochronological dates suggest initial construc ing). This demonstrates the utility of k-means tion at about A. D. 1285. The MCDs for clusters 1 cluster data. Although Christenson ( 1994) has and 2 agree with this date. Somewhat later, a fourth 96 AMERICAN ANTIQUITY [Vol. 61, No. 1, 1996] --- --- 105 Figure 3. Plan of Pueblo de Ios Muertos (LA 1585, after Watson et al. 1980:Figure 4 and Kintigh 1985b:Figure 4.32). Note that Rooms 225, a southeastern room, and 502, a plaza room, are not labeled. 0 20 Meters 300 bank of rooms was added to the interior, con supported by the similarity of the room 500 and structed atop a light accumulation of plaza trash. 502 deposits to those within rooms of Pueblo de Given the limited number of rooms actually Ios Muertos proper. The potential access corridor, excavated (13) relative to an estimated total of designated 419, appears to have accumulated 500 to 850 rooms (Kintigh 1985a:50-51; trash throughout the occupation. Marquardt 1974:43; Watson et al. 1980:207), it is Several of these rooms appear to continue to difficult to extrapolate anything more than a sug accumulate trash for the remainder of the site's gestive sequence of occupation. We must also occupation, suggesting they were abandoned keep in mind that we are witnessing the abandon early within Pueblo de Ios Muertos's life history. ment of rooms evidenced by their filling with Rooms 19, 300, and 303 appear to begin being trash. Most rooms contain at least one hearth and trash filled toward the middle of the occupation appear to have been used for habitation range, while rooms 108 and 225 both might have (Marquardt 1974:Table 2-2). Rooms 303 and 418 been built atop early trash accumulations and represent kiva locations, although 418 is a con occupied until relatively late. Room 303, an verted habitation room (Marquardt 1974:43, 45; abnormally large room with a flagstone-lined Watson et al. 1980:207). Provenience 419, origi floor and a sipapu (a hole in the floor that repre nally thought to have been a room, turned out to sents the place of emergence from the under be an access corridor through the west wall world) (Marquardt 1974:43 ), apparently fell out (Watson et al. 1980:207). of use rapidly. A schematic presentation of the temporal rela The dominance of units with early trash depo tionships between units and levels is presented as sition corroborates the interpretation of a rela Figure 4. The earliest trash deposition includes tively short, but intensive occupation at Pueblo de the lower levels of nearly all of the rooms exca Ios Muertos. From the limited excavation sample, vated. The lower levels of rooms 20, 104, 105, it appears that many rooms were occupied very 108, 225, 500 and 502, and trench deposits 4733 briefly, then became convenient refuse disposal and 4915 date to the earliest period of deposition. locations. Spatially, the rooms in the northern Marquardt's ( 1974:40) suggestion that Pueblo de portion of the site appear to have declined earli Ios Muertos was constructed around another con est. After some initial trash deposition, room 108 temporaneously occupied pueblo is potentially appears to have been occupied until relatively REPORTS 97 7 6 2&3 225--- KEY: 4&5 7 2 I I ·----Stratigraphic error ~----·---- D 105 4 Anomalous case 104 ___ ~1 2 4915 I 7 6 4&5 3 1 &2 419 ~- I I I I I -----1~0&~1~1- -9 ----7~i_s 3&1~4~1~----~ 20 I I ________ m 7 5 4 3 2 - 108-- I I :! I I I ·c: 2 ;:) 500 I 3 2 502 ___ ----~U 3&4 4733 ________I 6i~7----~.&~---------J 19 11 &12 5 303 ~----------L-J All 300 ~---------------~1 2 4612 ----- ---------~----LI ____ _. All 410 --------- --- ~----·------------ I All 418 ________________________________I Early Late Figure 4. Relative seriation of Pueblo de Ios Muertos excavated units and lev els from correspondence analysis of typological data. late. The filling of adjacent rooms (104 and 10 5) entry corridor and room I 08 both contain late with trash indicates that rooms in this vicinity deposits, but room I 08 deposits accumulated rela continued to be occupied. The early deposits from tively rapidly. The western section of the pueblo these northern rooms have a somewhat greater might have been among the latest abandoned due association with Tularosa Black-on-White domi to its proximity to the entrance/exit and Muerto nated assemblages, possibly suggesting capture of Creek (Watson et al. I980). a functional emphasis related to decorated jars. CA stratigraphically misplaces a few levels However, this association varies through time when projected onto a single dimension (Figure between levels of the same unit. 4). These cases are highlighted and deserve com The remainder of the excavated rooms appear ment. Levels 4 and 5, and 7 in Room I05 are to be rapidly abandoned and filled with trash dur reversed. These levels are assigned to clusters I ing the middle of the occupation. This may corre and 2, which the MCD suggests are essentially spond with an early A.D. 1300s decline posited by contemporaneous. Levels I and 2 in Room I9, LeBlanc ( 1978). This pattern incorporates the bulk levels 5 and 6 in Room 20, and level I in Room of the site, with evidence for continued occupation I 08 appear to be cases anomalous relative to the strongest for units along the west. Rooms 410 and remainder of the levels from the same unit. The 4I8 contain shallow deposits and are among the surface level misplacements may be related to latest deposits at the site. Room 4I8 was converted post-depositional factors such as increased sherd to a kiva from a habitation room (Marquardt breakage due to trampling, altering the relative I974:43, 45; Watson et al. I980:207), and may contribution of some types. One could selectively have been in use until final abandonment. The remove troublesome levels to achieve a perfect 98 AMERICAN ANTIQUITY [Vol. 61, No. 1, 1996) senat10n; however, stratigraphic misplacements poral variation in their frequencies, the only dif may be instructive. To determine if local overrep ference will be in the temporal resolution. resentation (e.g., several sherds from a single ves The combination of k-means and CA tech sel), disturbance, or other factors may be niques has the potential to produce an extremely responsible requires additional scrutiny. useful approach to seriation. CA is quantitatively That the bulk of the site was apparently con robust and visually informative, and provides structed and abandoned rapidly has already been information about which variables account for the suggested (LeBlanc 1978; Watson et al. 1980). greatest amount of variance within a group of The typological seriation results confirm these assemblages (as do factor and principal compo suggestions, but also indicates relatively late use nents analyses). Combining this with k-means and trash-filling of some units at Pueblo de Ios clustering produces case groupings that are not Muertos. It appears that much of the site fell out always visually distinct. The properties of these of use rapidly and perhaps was never completely clusters are directly referable to the original data, utilized. There was probably a greatly reduced and the MCD formula was used to highlight the presence at the site of somewhat greater duration potential of relating dimensional representations than initially thought. The latest MCD is A.D. with absolute dates. Refinements to the MCD 1321 (Cluster 6). A limited number of later deco approach and in the dating of the decorated rated types (e.g., Pinnawa Glaze-on-White) sug ceramic types employed can only increase the gest that occupation may have persisted until precision of this approach. CA has many uses out about A.D. 1350. Additional investigation at side of seriation. It is especially useful as an Pueblo de Ios Muertos should concentrate on the exploratory procedure, allowing one to readily northeastern and west-central portions of the site (and visually) assess major axes of variation in in order to determine the degree of post A.D. almost any data set. 1310-1320 occupation. Typological data may not always be appropri ate for seriation problems. If only one or two Summary and Conclusions ceramic types are present, or types are defined by This presentation has demonstrated that it is possi attributes that do not vary with time (e.g., func ble to achieve accurate, fine-grained temporal tion or rim form), attribute-based analyses are seriation using type frequencies. The type fre clearly warranted. However, it is often unclear quency data from Pueblo de Ios Muertos are con which attributes should be recorded, and what sistent with the relative seriation of units achieved they will inform upon. A frequent approach is to by LeBlanc's ( 1975) intrasite attribute seriation. record information on several attributes expecting Although I have not presented information here, that some will contain temporal information. CA typological seriation of excavated collections from may be particularly helpful in this case. CA can several El Morro Valley towns produces results isolate axes of variability, and the display of the comparable to those achieved by Marquardt attributes along with cases can help isolate which (1974 ), who also employed attribute seriation. are temporally sensitive. Performing this analysis I should like to point out that although Pueblo on a subset of the overall assemblage would per de Ios Muertos was occupied for a relatively short mit substantial time savings, as one could con period (ea. 60 years), and that most of the ceramic centrate on coding only time-sensitive attributes types have relatively short durations, "micro-seri for the remainder of the assemblage. ation" is equally relevant to most archaeological The success of seriating type frequencies is situations. Marquardt ( 1978:284) has stressed that worthy of note. The coding of several attributes micro-seriation is simply seriation in which the involves a substantially greater time investment temporal resolution achieved is fine. I suggest (Christenson 1994:300), and this analysis sug that typological seriation employing the combina gests the results (for seriation) are no more useful tion of techniques outlined above should be than type data. Typological totals (and/or percent equally valuable in regions where ceramic types ages) are the standard reporting technique for span several hundred years. Provided there is tern- most published analyses and contract archaeology

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.