Pilot Study Report: Research on the Harkleroad Ceramic Vessel Collection

Pilot Study Report: Research on the Harkleroad Native American Ceramic Vessel Collection,

San Diego Museum of Man

By Susan Hector, Michael Sampson, Deborah Moss, and Kassandra Nearn

This study was funded by a 2016 Begole Archaeological Research Grant (BARG), Anza Borrego Foundation

August 2017

Introduction

Museum collections of ceramic vessels provide an opportunity to analyze attributes that provide information on manufacturing, use, distribution, and disposal of pottery. The Harkleroad collection at the San Diego Museum of Man consists of vessels from the Anza-Borrego Desert State Park and adjoining region (into Baja California) that have not been analyzed.   Our study uses data from this collection to inform sherd analysis. The goal of the research is to associate vessel attributes (for example: form, decoration, composition) with locations and cultural groups. These associations can then be used to interpret fragmentary pottery sherds found at archaeological sites, and analyze regional and cultural variability.

This research was undertaken with the participation and support of Carmen Lucas, Kwaaymii. The research team is committed to including Native stakeholders in the analysis, both present and future. Many of the vessels in the Harkleroad collection are from Ms. Lucas’s ancestral homeland and tribal territory, and her participation was critical to any success for the program.

Theoretical Framework

Archaeologists attempt to use fragments of material culture to reconstruct, describe, and understand past human behavior and society.   Pottery is an abundant and important component of material culture that provides “insights into the lifeways of people” as well as information about cultural change, trade, ideology, and behavior (Rice 1987: 24-25). The collections, notes, and research conducted by the late Malcolm Rogers of the San Diego Museum of Man provide ample evidence of the importance of ceramic vessels to the native people of the Anza-Borrego Desert and environs (Sampson 2016). In the region, potsherds are one of the most frequent types of artifact found at sites. The sherds are generally so fragmentary that a description of the original vessel is not attempted, and this abundant source of cultural information is frequently ignored or described in general terms. If archaeologists had a large sample of complete prehistoric ceramic vessels to study, it might be possible to use information from these objects to shed more light on the meaning of potsherds found during archaeological investigations. It is the goal of this research to provide that information.

Ceramic vessels, including, ollas, bowls, spoons, figurines, scoops, ornaments, and rattles, were important implements in the everyday life of the indigenous people who traditionally resided in the Colorado Desert, upper Baja California, and adjoining areas during the Late Prehistoric Period (Bean 1978:579; Cline 1979:39, 44, 1984:32-39; Griset 1986:91-97,1996; Heye 1919; Rogers 1936; Van Camp 1979:54-61). Ceramic vessels were employed in a variety of tasks: food storage, water storage and transport, cooking of food, ceremonial, funerary container, children’s toys, transport of live coals, tools, and others (Cline 1979:39, 59; Griset 1996:54-68; Hohenthal 2001:168, 172-173; Rogers 1936:18-20, Plate 9). Ceramic objects used in this region were manufactured by means of the paddle-and-anvil technique (Bean 1978:579; Gifford 1931:42; Hohenthal 2001:170-171; Rogers 1936:5-15; Van Camp 1979:51-53); the paddle typically was a rectangular shaped piece of wood with a handle (Hohenthal 2001: Fig. 7.3; Rogers 1936:8, Plate 4; Van Camp 1979:51), while various kinds of anvils, such as a smooth stone or a special ceramic tool, were employed during the manufacturing process (Hohenthal 2001:Fig. 7.3; Rogers 1936:10-11; Van Camp 1979:51-52). The paddle-and-anvil technique is described and illustrated in Rogers (1936:4-15, Plate 3) and Cline (1979:39-48). Ethnographic accounts indicate that only women manufactured ceramic objects in the region (Cline 1979:43; Gifford 1931:42; Rogers 1936:4-5), but exceptions among certain groups may have existed.

Methodology

The current work, discussed in this preliminary report, represents a pilot study for a larger effort to accomplish this goal. Dr. Hector and Mr. Sampson are engaged in a multi-year project to analyze complete vessels using a systematic, documented methodology. The Harkleroad collection consists of over 500 complete or nearly complete vessels collected over a forty-year period. It is the largest group of complete pots and bowls from the region.

The pilot study initially proposed for the Begole grant consisted of 25% of vessels in the collection (estimated at 125), but additional vessels were included to total 135 in the study. These constitute the first 135 vessels examined at the Museum of Man. The vessels are not stored in numerical order by catalog number, but rather by size to fit the available shelving. Therefore, this sample is close to random.

Many of the vessels were gathered from the area that is now ABDSP, and the remainder are within the Begole grant project area. Out of the 135 vessels in the pilot study, 68 of the ceramic objects were obtained in northern Baja California; approximately half.

The Museum of Man supports this research and gave us permission to complete our study. The pilot study will provide a case study that will guide analysis of the entire Harkleroad collection and other museum collections.

Additional attributes were added to the study following submittal and approval of the grant proposal. The revised recording form is attached as an appendix to this report.

Research conducted by the authors consisted of the following:

  1. The individual bowls, jars, and other ceramic objects were photographed with a scale.
  2. Each object was measured: maximum circumference, opening circumference, and height.
  3. Other attributes were recorded from the object: residual or sedimentary clays, condition and repairs, surface decoration, percentage of fire clouding and charring, vessel shape, neck and rim form, and contents when collected (if any).
  4. It was possible to identify site, location, and cultural group (Kumeyaay [Ipai and Tipai], Cahuilla, Paipai, Kiliwa, etc.) through collection information and/or supplemental research. These data are plotted on a map using GIS technology as either actual locations or buffered points.
  5. The data were initially documented on a paper form in the lab, then entered into an Excel spreadsheet. A database was developed for statistical analysis and as an attribute table for the GIS study.
  6. The Begole Archaeological Research Center will receive copies of a summary version of the data entry form with photographs of each vessel examined in addition to a report describing the results of the research. The Begole Center will receive the detailed Excel spreadsheet for vessels included in the pilot study.

Research Questions

Question 1. Does manufacturing and firing affect sherd preservation? What are we missing in the field? Observations from each vessel can provide important cultural information about vessel manufacture and firing methods (Hohenthal 2001; Rice 1987; Rogers 1936; Schiffer 1987; Shepard 1980). Of particular interest is fire clouding, which may have been an intentional decorative feature (Van Camp 1979). The study will also consider whether fire clouding could have an impact on sherd preservation, an issue that has not been studied (Michael B. Schiffer, personal communication, 1/7/2016).

Question 2. What attributes from vessels can contribute to our understanding of vessel form, function, and disposal during analysis of potsherds collected from the field? Several schemes for typing pottery in the region have been developed based on surface appearance, temper material, and clay (paste) content (e.g., Rogers 1936; Waters 1982b). However, it is clear that, as Griset (1996: 145) has stated, that patterns of ceramic artifact attributes in the context of the cultural or archaeological setting will support an understanding of vessel manufacture, use, and disposal. As background for the research proposed in this document, Dr. Hector reviewed, photographed, and annotated Rogers’ ceramic types for the region as curated at the San Diego Museum of Man (Hector 2014). She identified many of the issues described in Waters (1982a). The goal of the research is to associate vessel attributes (representing form, decoration, composition) with locations and cultural groups. These associations can then be used to interpret fragmentary pottery sherds found at archaeological sites, and analyze regional and cultural variability.

Attribute analysis is a strong analytical tool for categorizing and describing cultural objects (Hector 1984). The attributes for each ceramic vessel are entered into a spreadsheet that can then be used to sort, graph, and analyze the results.

The research conducted is non-intrusive; individual vessels are not damaged or modified in any way. The researchers believe that this study will prove that valuable anthropological insights can be obtained from ceramics without destructive analysis methods (Hector 2007, 2011). In addition, Native American consultation has demonstrated that many native people do not approve of or support the breakage of potsherds in order to conduct analyses.

Data Constraints

As the data collected from the pilot study vessels were entered into an Excel spreadsheet for analysis, observations were made about the collection:

  • Harkleroad usually noted the general location where the vessel was collected, but often referenced his field camp or a local property. Sometimes he did not note the location of collection. Some of the vessels were purchased. In fact, we learned that Mr. Harkleroad was continually buying, trading, and selling vessels. The number that passed through his hands is probably three times the size of the collection at the Museum. The current locations of these other vessels are unknown.
  • A related location problem arose when a geographic name was provided but there are multiple potential locations with that name (e.g., Lost Valley, Indian Creek, etc.). In these cases, we went with the location where he seemed to have made other collections. So in the case of Lost Valley, he collected other pots near the Boy Scout camp so we mapped the “Lost Valley” in the northern part of study area. This may not be correct, and locations will be perfected at the conclusion of the study.
  • Hector and Mr. Sampson discussed the Harkleroad collection with Ken Hedges, retired Museum of Man curator, on October 5, 2016. Mr. Hedges worked closely with Mr. Harkleroad in accessioning the collection, and was in the process of field checking vessel locations with Mr. Harkleroad when the collector passed away. Mr. Hedges was able to map the locations of a number of the vessels on USGS topographic quadrangles. On November 22, 2016, Dr. Hector was able to locate these notes at the Museum of Man. More precision locations are therefore available for some of the vessels collected in the U.S. Many of the Baja California vessels, particularly those taken from the Laguna Salada area, have poor provenience. It is not clear yet how this will affect the final analysis.
  • Malcolm Rogers’ typology is of general use in describing vessel form, but many of the pots did not conform to any of his vessel types. Rather than continually invent new types for unique items, the researchers chose to select a best fit. We still feel that there is no purpose to creating a multitude of vessel shape types. Rather than continue to split, we advocate creating more general types based on function. This position will continue to evolve as the research matures.
  • While analyzing the vessels for the pilot study, the researchers have continued to work at the Museum of Man on the remainder of the collection; we are approaching completion on analysis of half of the collection. An immediate constraint noted in the Harkleroad collection is that most of the pots are vessels taken from rock caches. It will be important to broaden the range of vessels included in future analyses (e.g., we have started to analyze vessels from other museums and private collections to augment the Museum of Man materials). To begin to address this issue, we analyzed two vessels housed at Cleveland National Forest (with site provenience), and a collection of whole vessels housed at the San Diego Archaeological Center. Although these will not be included in the pilot study, their data will be taken into account for the larger project. It may be that most extant complete pots are cached vessels. This will skew the results.

Results of Analysis

Two special studies supported by the grant were conducted for the pilot project. One was a GIS analysis to determine if the limited collection location information is useful. The spreadsheet that was used for the GIS analysis contained more vessels than the final pilot study data set. Several were eliminated from the pilot study for various data-related reasons after the GIS analysis was completed. We decided to keep the larger number in the GIS study.

The other was a series of statistical queries based on identifying sherds in the field using the results of the Harkleroad study. The queries were focused on how to improve the description of potsherds in the field without collection.

GIS Analysis Results

By Kassandra Nearn, NWB Environmental Services, LLC

 Individual points for pottery vessel locations were plotted in ArcMap using estimated locational information prepared by Dr. Hector and provided in an excel spreadsheet. General directional information provided by Mr. Harkleroad’s notes was georeferenced using landmarks and landscape descriptions located on USGS topographic maps and Mexico topographic maps. The Mexico maps were sourced from the Mexico National Institute of Statistics and Geography web archives: http://www.inegi.org.mx/default.aspx.

It should be noted that while a point often communicates a precise location, the points in this analysis are estimated and meant to represent locations within a larger geographic region and cultural area. In no cases was a specific plotted location provided by the collector. While many locations were described in relation to major landmarks such as mountain peaks, others were described in relation to rivers which span a large distance, or to temporary Harkleroad camping places of which the locations are unknown. Once the individual points were plotted, a series of queries were run to change the symbology of the data and create visualizations of overall patterns.

Map 1 Residual and Sedimentary polygon areas—Type

This map provides a visualization of the distribution of the “type” attributes of residual and sedimentary. The queries for this map were:

Type = ‘Residual’

and

Type = ‘Sedimentary’

A new layer was created for each result. A polygon for the Residual and Sedimentary datasets was then created, representing the regions in which each type of pottery was located for this dataset. To create the polygons, a 1000 m (arbitrary) buffer was created around each point dataset. The buffers were dissolved to create a merged polygon. In some areas, there is a large amount of space between points. I initially considered un-dissolving the buffer; however, this allows for some variability regarding the exact point locations—because many of the locations are estimated, they may have been located somewhere else nearby that falls within the general polygon.

Map 1: Type

Map 2 Culture

The Symbology Categories feature in ArcMap was used to auto-generate queries for the ‘Culture’ attribute. The background queries this feature uses would be:

Culture = ‘Ipai’; Culture = ‘Tipai’; etc.

This provided a visualization of where pottery sherds associated with different Native American groups are located for this dataset. The distribution appears consistent with traditionally occupied areas of cultural groups in southern California and Baja.

Map 2 shows the Culture data points overlaid on the Type polygons. It appears Sedimentary is limited to Tipai and Paipai for this dataset, located primarily on the California/Mexico border and in Northern Baja.

Map 2: Culture

Map 3 Culture and Type

This map provides similar information as Map 2, with the addition of recorded cultural areas based on the Handbook of North American Indians, Vol. 8, 1978. Rather than using the polygons, the individual data points are categorized by unique value combinations of ‘Type’ and ‘Culture’.

Map 3: Culture and Type

Overall, the GIS analysis successfully plotted locations of pottery vessels. However, there are some inconsistencies due to vague locations and probably mis-assignment of location due to duplicate names. For example, was the vessel collected from Vallecito in the Anza-Borrego area or Vallecito in Baja California?

After completing the GIS analysis for the pilot study, additional location information was discovered in the files at the Museum of Man. This will greatly improve vessel collection point identification for the study. It is also the intent of the researchers to compile a complete list of Harkleroad collection locations so that we can, hopefully, differentiate between places with the same name. Both of these efforts will be accomplished at the completion of the larger study of approximately 500 vessels.

Statistical Analysis Results

By Deborah Moss, Mathematics Department, Mesa College, San Diego Community College District

The statistical analysis focused on whether certain identifiable attributes on the vessels could be associated with one another to support classification of pottery sherds in the field. The following attributes were compared:

Type (residual/sedimentary) and Object Name (Rogers typology)

Type and Neck Form

Type and Rim Shape

Repairs and Object Name

Object Name and Fire Clouding

Object Name and Wipe Mark presence and location on the vessel

Object Name and Charring presence and location on the vessel

Object Name and Paddle Mark presence and location on the vessel

There are a total of 134 pottery vessels in the data set (there were 135 vessels in the study, and 1976-46-88 was excluded because it was in pieces).

The first challenge was to reduce the 18 types (Object Name) into fewer groups to enable statistical analysis. This led to the following graph, which is a bar graph of Opening Circumference divided by Maximum Circumference (shown on the vertical axis) sorted from lowest to highest. This did not provide any insight with respect to natural breaks in the data set as was initially hoped (i.e. bowls versus jars). For example, a series of similar values followed by a sharp break and then another series of similar values would have indicated two distinct sub-populations with uniform distributions.

The second line of inquiry led to the following histogram, in which six roughly normal and overlapping sub-populations can be seen upon visual inspection. Three are in the large circle on the left and one each are in the other three circles (The apparent peak between the second and third circle is likely an artifact of the overlap of the tails of the distributions to the left and right).

The means can be visually approximated to be near 0.225, 0.325, 0.425, 0.625, 0.825, and 1.075. The purpose of visually approximating the means is to confirm that the data has been properly parsed by comparing the means of the six sub-sets to these approximated values.

Because group 4 can be “lumped” with either group 3 or with groups 5 and 6, it makes more sense to “lump” group 4 with group 3 since that creates nearly a clean break into two categories:

Category 1 – all vessels with an Opening Circumference to Maximum Circumference ratio less than 0.72. These are “jar-like vessels” and roughly correspond to Object Names 1, 2, 3, 4, 5, 6, 7, 9, 13, 14, 15, 18 and 19 with the exception of one Type 16 (1976-46-87, ratio 0.32).

Category 2 – all vessels with an Opening Circumference to Maximum Circumference ratio greater than 0.72. These are “bowl-like vessels” and roughly correspond to Object Names 7, 12, 16, 20, 26, and 27 with the exception of one Type 6 (1976-46-233, ratio 0.902) and one Type 13 (1976-46-23, ratio 0.849).

I recommend these three vessels be rechecked for accuracy of measurements and Type and also be considered for removal from future studies as outliers.

Is there an association between Type (residual, sedimentary) and Object Name?

The “find and replace” feature was used to change “residual” to 0 and “sedimentary” to 1. As described above, Object Names were not used (based on previous analysis which showed sample sizes were too small). Instead, the entire data set was subdivided into vessels with an opening to maximum ratio less than 0.72 (114 out of 134, assigned a value of 0) and vessels with an opening to maximum ratio greater than 0.72 (20 out of 134, assigned a value of 1).

Sample size, n:     134

Degrees of freedom: 132

Correlation Results:

Correlation coeff, r: 0.109238

Critical r:           ±0.1696748

Analysis does not suggest correlation between Type and Object Name when Object Name is parsed into “Jars” and “Bowls”.

Is there an association between Type (residual, sedimentary) and Neck Form?

The “find and replace” feature was used to change “residual” to 0 and “sedimentary” to 1, and to change “straight” to 0 and “flared” to 1.

Six vessels were excluded due to no information in the Neck Form column:

1976-46-101

1976-46-154

1976-46-172

1976-46-193

1976-46-207

1976-46-342

Sample size, n:     128

Degrees of freedom: 126

Correlation Results:

Correlation coeff, r: 0.0857728

Critical r:           ±0.1736229 (alpha = 0.05)

NOTE: Changing the significance level (i.e. alpha = 0.10 or 0.20) does not change the result.

Analysis does not suggest correlation between Type and Neck Form. The takeaway is that Neck Form is primarily a function of the purpose of the vessel and is less informed by the material out of which the vessel is made than by how the vessel is intended to be used.

Is there an association between Type (residual, sedimentary) and Rim Type?

Although the Rim Types (RimShape) are numbers, they are nominal (not ordinal, interval or ratio), so the numerical values cannot be used for regression. Instead, five individual tests were performed, and each time the Rim Type being tested was changed to 1 (using find and replace) and all other Rim Types were changed to 0.

One vessel was excluded due to no information in the Rim Type column:

1976-46-342

For all tests,    Sample size, n:     133

Degrees of freedom: 131

Critical r:           ±0.1703141 (alpha = 0.05)

NOTE: Changing the significance level (i.e. alpha = 0.10 or 0.20) does not change the results.

Rim Type 1:   Sub-sample size: 26

Correlation coeff, r: 0.0330157

Analysis does not suggest correlation between Type and Rim Type 1.

Rim Type 2: Sub-sample size: 11

Correlation coeff, r: -0.1051555

Analysis does not suggest correlation between Type and Rim Type 2.

Rim Type 3:   Sub-sample size: 42

Correlation coeff, r: 0.0717001

Analysis does not suggest correlation between Type and Rim Type 3.

Rim Type 4:   Sub-sample size: 2

Could not be completed due to small sample size.

Rim Type 5:   Sub-sample size: 52

Correlation coeff, r: -0.0544774

Analysis does not suggest correlation between Type and Rim Type 5.

The signs of the correlation coefficients above mean that Residual pottery is very slightly more likely to be correlated with Rim Types 2 and 5 (and vice versa) while Sedimentary pottery is very slightly more likely to be correlated with Rim Types 1 and 3 (and vice versa). These relationships, however, are NOT statistically significant and therefore should NOT be used for prediction.

Based on the above observation and small sample size, Rim Type 4 was excluded from the analysis, Rim Types 2 and 5 were set to 0, and Rim Types 1 and 3 were set to 1. This resulted in the following:

Sample size, n:     131

Degrees of freedom: 129

Correlation Results:

Correlation coeff, r: 0.1045489

Critical r:           ±0.1716148 (alpha = 0.05)

NOTE: Changing the significance level (i.e. alpha = 0.10 or 0.20) does not change the result.

Even when Rim Types are “lumped” based on previous observations, the correlation is still not significant.

As with Neck Form, it should be considered that Rim Type is primarily a function of the purpose of the vessel and is less informed by the material out of which the vessel is made than by how the vessel is intended to be used.

Is there an association between Repairs (patching, drilling) and Object Name?

As described the first question, Object Names were not used (based on previous analysis which showed sample sizes were too small). Instead, the entire data set was subdivided into vessels with an opening to maximum ratio less than 0.72 (114 out of 134) and vessels with an opening to maximum ratio greater than 0.72 (20 out of 134).

An adjustment was made to put the three “misclassified” vessels back into the correct groups in order to meet the minimum requirements for successes/failures. After this adjustment, there were 22 of 115 “jar-like” vessels with “original” (non-modern) repairs. The other 93 either had no repairs or only modern repairs. This is a repair proportion of p1 = 0.191.

There were 5 of 19 “bowl-like” vessels with “original” (non-modern) repairs. The other 14 either had no repairs or only modern repairs. This is a repair proportion of p2 = 0.263.

Results:

Alternative Hypothesis:   p1 not = p2

Pooled proportion: 0.2014925

Test Statistic, z: -0.7234

Critical z:       ±1.9600

95% Confidence interval:

-0.2825002 < p1-p2 < 0.1387931

(The interpretation here for the confidence interval of the difference between the two proportions is that if the confidence interval contains zero, then the proportions are not statistically significantly different from each other.)

Based on the above analysis, there is no statistically significant difference between the proportion of repairs in the two groups.

How many vessels have identifiable damage from fire clouded areas? Where?

% Count Rim Neck Body Base
0 1 0.00% 0.00% 0.00% 0.00%
5 0 N/A N/A N/A N/A
10 5 0.00% 0.00% 100.00% 0.00%
15 8 12.50% 75.00% 100.00% 50.00%
20 21 14.29% 42.86% 95.24% 52.38%
25 5 0.00% 20.00% 100.00% 20.00%
30 22 45.45% 77.27% 100.00% 81.82%
35 8 37.50% 75.00% 87.50% 100.00%
40 15 53.33% 66.67% 100.00% 86.67%
45 2 50.00% 100.00% 100.00% 100.00%
50 10 40.00% 80.00% 100.00% 100.00%
55 2 50.00% 100.00% 100.00% 100.00%
60 7 57.14 100.00% 100.00% 85.71%
65 1 100.00% 100.00% 100.00% 100.00%
70 9 55.56% 100.00% 100.00% 100.00%
75 4 75.00% 100.00% 100.00% 100.00%
80 7 71.43% 71.43% 100.00% 85.71%
85 0 N/A N/A N/A N/A
90 1 100.00% 100.00% 100.00% 100.00%

In general, it appears that fire clouding is most common on the body as it was found in 129 out of 131 cases where fire clouding was present and annotations were made as to the location.

As more of the vessel becomes fire clouded (30% or more), the base is almost always involved, as seen here with 80 out of 88 vessels showing fire clouding on the base when 30% or more of the surface showed fire clouding.

Fire clouding on both the neck and rim show a steady increase with respect to percent of overall fire clouding, with the neck more likely to be involved than the rim.

Is there a correlation between Object Name and Percent Fire Clouding?

As described in the first question, Object Names were not used (based on previous analysis which showed sample sizes were too small). Instead, the entire data set was subdivided into vessels with an opening to maximum ratio less than 0.72 (114 out of 134, assigned a value of 0) and vessels with an opening to maximum ratio greater than 0.72 (20 out of 134, assigned a value of 1).

Five vessels were excluded due to Fire Clouding being reported as “Present” rather than as a percentage:

1976-46-16

1976-46-105

1976-46-110

1976-46-130

1976-46-188

Raw (subjective) values of Percent Fire Clouding (0 – 90%) were used.

Sample size, n:     129

Degrees of freedom: 127

Critical r:           ±0.1729457 (alpha = 0.05)

Critical r:           ±0.1454656 (alpha = 0.10)

Critical r:           ±0.1135744 (alpha = 0.20)

Correlation Results:

Correlation coeff, r: 0.1212778

This result indicates that the Object Name proxy is statistically significantly correlated to percent fire clouding, but only at the 80% significance level. In other words, an object that was placed into the “bowl-like” category of having an opening to maximum ratio greater than 0.72 is positively correlated with having a higher percentage of fire clouding.

Additional investigation showed that raw opening to maximum ratio values are also statistically significantly correlated to percent fire clouding at the 80% significance level (Correlation coeff, r: 0.137889). In other words, parsing the data into “jar” versus “bowl” had no detrimental effect on the relationship between these two variables and serves as a good proxy.

One last test was performed that showed height is statistically significantly correlated to percent fire clouding, but at the 90% significance level (Correlation coeff, r:   -0.153694). In other words, the taller a vessel, the lower the percentage of overall fire clouding. This effect, however, is due to a single outlier with height 81 cm and disappears once this outlier is removed. There is still a statistically significant correlation, but it is at the 80% significance level (like the other measures of size) rather than at the 90% significance level. In summary, vessel size does seem to be statistically significantly correlated to percent fire clouding in that the percentage fire clouding of sherds presumed to have come from the same vessel could be used to make inferences about whether the vessel they came from was “jar-like” or “bowl-like.” This relationship needs to be investigated further with a more robust data set. Of note is the fact that the pilot study did not include the largest size vessels in the Harkleroad collection, because these are stored in the back of the museum lab. These vessels were not analyzed until the summer of 2017, and their data may affect the results of the fire clouding analysis.

Is there a relationship between Object Name and Wipe Mark presence/location?

As described in the first question, Object Names were not used (based on previous analysis which showed sample sizes were too small). Instead, the entire data set was subdivided into vessels with an opening to maximum ratio less than 0.72 (114 out of 134, assigned a value of 0) and vessels with an opening to maximum ratio greater than 0.72 (20 out of 134, assigned a value of 1).

Wipe marks present (overall and in each of the locations indicated) were assigned a value of 1. Absence of wipe marks were assigned a value of zero.

Sample size, n:       134

Degrees of freedom:   132

Critical r:           ±0.1696748 (alpha = 0.05)

Critical r:           ±0.1427023 (alpha = 0.10)

Critical r:           ±0.1114081 (alpha = 0.20)

Correlation Coefficient with Object Name
Wipe Marks Present 0.1703493 – significant at 95%
Wipe Marks on Rim 0.0623358
Wipe Marks on Neck 0.1391918 – significant at 80%
Wipe Marks on Body 0.1620144 – significant at 90%
Wipe Marks on Base -0.0633852
Wipe Marks on Interior -0.0795808

The above analysis indicates a statistically significant correlation between the presence of wipe marks and “bowl-like” vessels. 47 of 114 vessels in the “jar-like” category had wipe marks (p1 = 0.412). 13 of 20 vessels in the “bowl-like” category had wipe marks (p2 = 0.650).

Alternative Hypothesis:   p1 not = p2

Pooled proportion: 0.4477612

Test Statistic, z: -1.9719

Critical z:       ±1.9600

P-Value:           0.0486

95% Confidence interval:

-0.4654504 < p1-p2 < -0.0099882

Based on the above analysis, there is a statistically significant difference between the proportion of wipe marks in the two groups.

Is there a relationship between Object Name and Charring presence/location?

 As described above, Object Names were not used (based on previous analysis which showed sample sizes were too small). Instead, the entire data set was subdivided into vessels with an opening to maximum ratio less than 0.72 (114 out of 134, assigned a value of 0) and vessels with an opening to maximum ratio greater than 0.72 (20 out of 134, assigned a value of 1).

Charring present (overall and in each of the locations indicated) was assigned a value of 1. Absence of charring was assigned a value of zero.

Sample size, n:       134

Degrees of freedom:   132

Critical r:          ±0.2811366 (alpha = 0.001)

Critical r:          ±0.2411606 (alpha = 0.005)

Critical r:          ±0.2218169 (alpha = 0.01)

Critical r:           ±0.1696748 (alpha = 0.05)

Critical r:           ±0.1427023 (alpha = 0.10)

Critical r:           ±0.1114081 (alpha = 0.20)

Correlation Coefficient with Object Name
Charring Present (Exterior) 0.1976840 – significant at 95%
Charring on Rim 0.2332797 – significant at 99%
Charring on Neck 0.2832396 – significant at 99.9%
Charring on Body 0.1926755 – significant at 95%
Charring on Base 0.0315077
Charring on Interior 0.3901504 – significant at 99.9%

The above analysis indicates a statistically significant correlation between the presence of charring and “bowl-like” vessels. This correlation is strongest for the interior, followed by the neck and rim.

The proportion of “jar-like” vessels with charring on both the interior and exterior is p1 = 15/114 = 0.132. The proportion of “bowl-like” vessels with charring on both the interior and exterior is p2 = 11/20 = 0.550.

Alternative Hypothesis:   p1 not = p2

Pooled proportion: 0.1940299

Test Statistic, z: -4.3645

Critical z:       ±1.9600

P-Value:           0.0000

95% Confidence interval:

-0.6451111 < p1-p2 < -0.191731

The proportion of “jar-like” vessels with charring on both the interior and exterior is statistically significantly less than the proportion of “bowl-like” vessels with charring on both the interior and exterior. A sherd with charring on both sides (inside and outside is more likely to have come from a bowl than a jar). Intuitively, this also makes perfect sense. What is unclear is how well interior charring of organic material survives in open air site locations.

Is there a relationship between Object Name and Paddle Mark presence/location?

 As described in the first question, Object Names were not used (based on previous analysis which showed sample sizes were too small). Instead, the entire data set was subdivided into vessels with an opening to maximum ratio less than 0.72 (114 out of 134, assigned a value of 0) and vessels with an opening to maximum ratio greater than 0.72 (20 out of 134, assigned a value of 1).

Paddle Marks present (overall and in each of the locations indicated) was assigned a value of 1. Absence of Paddle Marks was assigned a value of zero.

Sample size, n:       134

Degrees of freedom:   132

Critical r:           ±0.1696748 (alpha = 0.05)

Critical r:           ±0.1427023 (alpha = 0.10)

Critical r:           ±0.1114081 (alpha = 0.20)

Correlation Coefficient with Object Name
Paddle Marks Present -0.0795808
Paddle Marks on Rim -0.0363192
Paddle Marks on Neck 0.0712404
Paddle Marks on Body -0.1035800
Paddle Marks on Base 0.0000000

The above analysis indicates no statistically significant correlation between the presence of paddle marks and whether a vessel is “jar-like” or “bowl-like” vessel.

Additional Results

The researchers analyzed a variety of data with interpretive importance beyond the special studies described above. Brief summaries are included in this section.

Typology and Object Name. In our study of regional ceramic vessels, a variety of ceramic vessel attributes are described and documented that we considered culturally significant. Vessel shape (or form) is one such attribute identified and recorded during the study. We adapted Malcolm Rogers’ (1936:Plate 9) “Yuman and Shoshonean Pottery Shapes” chart as a means of characterizing vessel form in a standardized manner. Rogers’ chart illustrated 28 different vessel shapes, which were in turn subdivided by Rogers into three parts: Western Groups, Desert Groups, and River Groups. A fourth cultural division, Eastern Groups, is listed in the key for the chart on Plate 9, but for an unspecified reason was not delineated as a separate division on the chart. Rogers (1936:16) stated that the Yuman and Shoshonean groups were being divided into four, because environment and geographic location represented significant influencing factors. Rogers (1936:16) observed that the divisions were created “…for the purpose of bringing together those groups among which certain technicalities and pottery shapes are common.” Rogers’ chart includes indigenous groups living in northern Baja California and the northwest mainland portion of Mexico, however, the chart did not specify the Paipai people. Therefore, any vessels originating from Paipai traditional territory might be anomalous in form relative to those depicted in Rogers’ (1936:Plate 9) pottery shape chart.

A certain number of the ceramic vessels we have examined do not in fact fit precisely into a shape as illustrated by Rogers’ chart. When such a difference was noted, we simply chose the vessel shape closest to that illustrated by Rogers. One can expect such divergences in form given the malleability of moist pottery clay, variances in pottery vessel molds used by aboriginal makers, differences in firing circumstances, and, as noted above, the fact that Rogers did not study Paipai ceramic vessels.

The three divisions in Rogers’ vessel shape chart clearly indicates Rogers felt particular ceramic vessel forms were employed by specific groups, thus leading to his conclusion that geography and natural environment played key roles in the distribution of vessel form. In addition, Rogers (1936:Plate 9) correlated the shapes described in his chart with a specific function. The uses of each vessel shape, according to Rogers, is given under “Directory” in his Plate 9 [the numbers in parentheses pertain to numbers given in his illustrated chart of vessel shapes]: storage ollas (1-4, 15), water ollas (5, 14, 23), cooking pot (6), cooking bowls (7, 16, 24), plate (8), food bowls (9, 12, 20, 26), pottery anvil (10), canteen (11, 18, 21), jar (13), tray (17, 25), paint jar (19), scoop (22, 28), and parcher (27).

Rogers (1936:V) sought to only describe “…Yuman ceramic technique in its aboriginal form…” and not discuss any “…recent intrusive practices…” in his 1936 monograph on pottery among Yuman groups. Therefore, we can assume the ceramic vessel shape chart on Plate 9 only depicts vessel forms Rogers thought to be pre-Contact in practice among indigenous people of the region. It remains unclear if he was correct with this hypothesis. We have no reliable means of specifically dating the ceramic vessels in the Pilot Study, given how they were not collected under controlled, scientific means from their archaeological (cultural) context. While we cannot adequately address the issue of Pre-Contact versus more recent forms for the Pilot Study, certain attributes on ceramic vessels were clearly only used by pottery makers in post-Contact times, e.g., flattened bottoms, handles, and decorations dating to the historic period (symbols of the Catholic religion, horses, metal implements, etc.).

The use of a vessel shape typology is fraught with issues. Does the analyst’s typology relate in any way to the intent of the maker/user? Should there be more detailed types, or fewer general types? Despite the issues with Rogers’ typology as described above, we decided to continue using his shapes for this project. It is certainly not a perfect or ideal typology, but serves to describe the general shape of the vessel. Most of all, using this consistently throughout our study will ensure comparability of results.

Repairs and Adhesives. Many of the pots had been repaired by their user. Ten vessels with pairs of drilled holes to pull cracks together were identified. One of the pots had been repaired with fibers, which we identified as milkweed; the 2-ply milkweed twine was threaded through the pair of drilled holes. Some of the pots that had been repaired with drill holes and also had patches.

Adhesive materials were also used to patch holes and cracks. We were able to identify the adhesives used (insect lac, pine pitch, plant resin, mud/clay) using ultraviolet light (Odegaard 2009). Bitumen or petroleum products were not observed on any of the vessels in the pilot study. The use of natural tars as a sealant is common along the California coast, and vessels from that area might feature asphaltum; however no coastal vessels are present in the pilot study.

Some pots in the pilot study had sherd patches that were shaped into disks using an abrasive method, and then used to cover holes in the body to extend the life of the vessel. They were attached with adhesives. The pot was not re-fired.

Adhesives were also identified around the rim of vessels. This was to attach a lid or cap to preserve the contents. Six vessels had evidence for sealing with material around the rim and/or neck. Euler and Jones (1956) described the use of various types of sealants around the necks and rims of storage vessels in the American Southwest. In the pilot study collection, adhesives were found directly on the rim and also around the necks of narrow-necked jars. Vessel lids could have been sherd disks, bowls, or flat rocks.

The pilot study results support two types of vessel covers. One type is a round form, either a shaped sherd or a rock, that fit inside the rim of the vessel. This internal cover would be fit tightly above the neck constriction and was sealed using adhesive. The second type was a cover, typically a bowl or perhaps a fiber mat or textile, that was sealed over the top and neck of the vessel. The adhesives were placed around the exterior of the rim and neck and the cover attached.

Fire Clouding and Sherd Preservation. All of the vessels except two (99%) had fire clouding, present on from 10% to 90% of the vessel. The percentage of fire clouding was not possible to determine on several of the vessels because of coverage by charring. Fire clouding appears to have had an effect on vessel condition, with the clouded areas more susceptible to cracking. Twelve of the fire clouded vessels had damage directly associated with a dark colored fire cloud.

Fire clouds occur when fuel comes into direct contact with unfired pottery, when smoke from the flame contacts the clay, and from contact with oils in human hands (Shepard 1980: 92). Most of the vessels made prior to European contact are fired under low temperature conditions using leaves, sticks, and bark. After contact, livestock dung was used which resulted in a higher temperature firing condition (Rice 1987: 157). These later period pots are more completely oxidized and tend to be redder and more vitrified.

Hohenthal (2001) and Rogers (1936) described ethnographic methods of firing pottery in the study area. Rogers (1936:14) noted that no attempt was made to prevent the fuel from touching the pottery vessels. These carbonized areas would be eliminated if the firing pit temperature was hot enough to oxidize them. Otherwise they would remain as firing clouds. Rogers also noted that his informant told him about putting old cooking vessels back into the fire pit to burn off the carbon.

One of the research questions is focused on whether fire clouding affects sherd preservation since the clouded areas represent incomplete oxidization. Schiffer (1987: 159) noted that porous materials such as low fired pottery weather differently in the environment based on the composition of the clay. Observations in the pilot study support a need for additional research on this topic, since erosion and cracking was observed coincident with dark fire clouds.

Warped Condition before or after Firing. Some of the vessels were extremely warped and others have walls of varying thickness. Occasionally the potter added clay before firing to thicken areas particularly around the neck (none of the patched pots were re-fired). Despite this, they were used successfully for cooking or storage. Warping could occur during drying or firing of the vessel due to lack of care or variation in firing conditions. Rapid temperature changes and shifting fuel can result in irregular vessels (Rice 1987: 153-158). Warped or misshaped sherds from these deformed yet well used vessels may not be correctly identified in the field.

Clay Types. The vessels located in certain areas appear to be neither residual or sedimentary clays, but what we are tentatively calling residuo-sedimentary. These are clays that eroded from granitic sources and were deposited as sediments in the desert. The vessels are most notably from the eastern side of the Laguna Mountains (e.g., Mountain Palm Springs area).

Decoration. Only two of the vessels in the pilot study are decorated in any way. These had scratch marks that appeared to be by design. One of the pots, 1976-46-106, had linear scratches; this vessel had pine pitch on the neck and rim as a sealant. The other scratched vessel, 1976-46-166, had a bisected equilateral triangle (with one longer leg) on the upper body; this image may have been made on the pot after firing. Overall, the sample size in the pilot study was too small for any meaningful conclusions about decoration. None of the vessels in the pilot study have incised rims or paint.

One pot had clay attached to the body of the vessel before firing; this was categorized as applique. The clay was attached as a blob on opposite sides of the vessel, suggesting handles that have fallen off rather than a decoration.

Charring. Bowl interiors are charred, a result of a “dutch oven” effect from cooking. Charring was generally not only on the bottom of a bowl, but inside and around the upper portion of the vessel. Some vessels were only charred around the upper portion and interior.

Wipe Marks. Wiping with a fiber brush or the hand prior to firing can leave obvious marks on the surface of a pot. Sixty of the vessels analyzed had visible wipe marks (44%). Of these, 14 had wipe marks on the rim, 52 had wipe marks on the neck, 12 had wipe marks on the body, and 3 had wipe marks on the base. Since 87% of the visible wipe marks are on the neck of a vessel, it is most likely that a sherd with wipe marks will be from that location.

Seven vessels had wipe marks on the interior of the pot. In order for these to be visible to us, the opening had to be large enough to see into. Therefore, it is unknown how many vessels may have had wiping on the interior but not visible to us.

Paddle Marks. All of the vessels in the collection were made using a paddle and anvil technology. Occasionally the anvil marks are very distinct on the interior of the vessel. Visible paddle marks are more unusual, although sometimes they are so distinct that the grain of the wood in the paddle can be seen. More usually the paddle mark is a curved indentation on the surface of the pot.

Twenty-two of the vessels had visible paddle marks (16%). Of these, 1 had paddle marks on the rim, 8 had the marks on the neck, 18 had paddle marks on the body, and none had paddle marks on the base. Since 82% of the paddle marks are on the body of a vessel, a sherd with paddle marks is most likely from that part of the pot.

Surface. Surface treatment included slipping, burnishing, stucco application, and plain. Four vessels had surface treatment other than a plain finish. Two of the vessels in the pilot study had self slipping, creating a smooth finish with slight texture where the wet clay was manipulated. Two other vessels had stucco finish on the body, creating a rough, cracked texture. Both of the pots with stucco finish were sedimentary clay from the Laguna Salada collection area (Paipai). They are similar to Rogers’ Colorado Stucco sherds in the Museum of Man type collection.

Summary of Results

Question 1. Does manufacturing and firing affect sherd preservation? What are we missing in the field?

Our observations indicate that firing does affect sherd preservation. Although not all fire clouding resulted in vessel surface erosion and deterioration, poor firing was found directly associated with damage to a vessel. Over time, this may result in sherds not being preserved due to low firing. The sample size in the pilot study was too small, so further research will be needed when the entire collection has been examined.

Question 2. What attributes from vessels can contribute to our understanding of vessel form, function, and disposal during analysis of potsherds collected from the field?

The first challenge that we faced was addressing the assignment of vessels into Rogers’ typology, which we referred to as Object Name. Ms. Moss conducted statistical analyses to group the vessels into classifications of Jars and Bowls. This facilitated comparisons and also reflects original usage of the objects.

Sedimentary or Residual Clays and Vessel Form. Can we assign a sherd to the category of jar or bowl based on whether it is sedimentary or residual? Analysis does not suggest any correlation between Type (residual or sedimentary) and Object Name when Object Name is parsed into “Jars” and “Bowls”.

Neck Form. Can we assign a sherd to a Neck Form (e.g., flared, straight, or none) based on whether it is sedimentary or residual? Analysis does not suggest correlation between Type and Neck Form. Neck Form is primarily a function of the purpose of the vessel and is less informed by the material out of which the vessel is made than by how the vessel is intended to be used. A flared neck jar is just as likely to be made of sedimentary as residual clays.

Rim Type. Can we assign a sherd to a Rim Type based on whether it is residual or sedimentary? The correlation is still not significant. As with Neck Form, Rim Type is primarily a function of the purpose of the vessel and is less informed by the material out of which the vessel is made than by how the vessel is intended to be used.

Repairs. If a sherd shows a drilled hole or other repair, can we state whether it is bowl or a jar? Based on the analysis, there is no statistically significant difference between the proportion of repairs in the two groups (bowls and jars).

Fire Clouding. If a sherd has fire clouding, can we say whether it is bowl or a jar? Does the location and percent of fire clouding have meaning? In general, it appears that fire clouding is most common on the body of the vessel. As more of the vessel becomes fire clouded (30% or more), the base is almost always involved, as seen with 80 out of 88 vessels showing fire clouding on the base when 30% or more of the surface showed fire clouding. Fire clouding on both the neck and rim show a steady increase with respect to percent of overall fire clouding, with the neck more likely to be involved than the rim. So, if a sherd has fire clouding, it is most likely from the body of the vessel.

Was that vessel a jar or a bowl? A vessel that was placed into the bowl category of having an opening to maximum ratio greater than 0.72 is positively correlated with having a higher percentage of fire clouding. Height is statistically significantly correlated to percent fire clouding, at the 90% significance level. The taller the vessel, the lower the percentage of overall fire clouding; this conclusion has the most meaning for jars.

Wipe Marks. If the sherd shows wipe marks, can it be identified as a jar or a bowl? Based on the analysis, there is a statistically significant difference between the proportion of wipe marks in the two groups. Bowls are more likely to have wipe marks. For jars, the most common location for wipe marks is on the neck; these sherds would be identifiable due to the smaller curvature.

Charring. If charring is observed on a sherd, can it be identified as a bowl or a jar? The analysis indicates a statistically significant correlation between the presence of charring and bowls. This correlation is strongest for the interior, followed by the neck and rim.

Paddle Marks. Can we assign a sherd with paddle marks as a bowl or a jar? The analysis indicates no statistically significant correlation between the presence of paddle marks and whether a vessel is a jar or a bowl. Paddle marks are most frequently found on the body of a vessel.

Future Directions

This study has proven to be exciting and informative, and has validated the importance of museum collections – even poorly provenienced ones. Obviously well documented collections with complete location data are ideal, but we feel that even without precise locational information the pottery analysis has resulted in important findings. As the study proceeds on the remaining Harkleroad vessels, we feel that more precise locational information will be developed based on increased knowledge of his most frequent collecting areas. For example, it is our opinion that some of the locations identified in the pilot study that produced anomalous outliers are errors that will be corrected when the study is complete by preparing a list of locations frequently visited by Harkleroad. As the study proceeds, a more accurate and complete list of collection locations will be compiled.

It would be very interesting and important to locate a collection of sherds from an archaeological site and apply the results of this study. Even if only a few could be analyzed using the attributes measure for our results, it would improve interpretation.

We would like to acknowledge the assistance from Karen Lacy and San Diego Museum of Man staff, and also the participation of Debbie Moss and Nick Case who provided essential assistance moving pots, measuring, and asking thought-provoking questions. Ms. Carmen Lucas, Kwaaymii Laguna, participated in the analysis and provided important ethnographic and technical ceramic vessel information. Michael Taylor, President, NWB Environmental Services, provided staff support and facilities for the GIS analysis, which turned out to be much more extensive than we planned.

References Cited

Bean, Lowell John

1978   Cahuilla. In California, edited by R. F. Heizer, pp. 575-587. Handbook of North American Indians, Volume 8, W. C. Sturtevant, general editor, Smithsonian Institution, Washington, D.C.

Cline, Lora

1979   The Kwaaymii: Reflections on a Lost Culture. IVC Museum Society, Occasional Paper No. 5, El Centro.

Euler, Robert C. and Volney H. Jones

1956   Hermetic Sealing as a Technique of Food Preservation Among the Indians of  the American Southwest. Proceedings of the American Philosophical Society 100(1): 87 – 99.

Gifford, Edward W.

1931   The Kamia of Imperial Valley. Bureau of American Ethnology Bulletin 97. Washington.

Griset, Suzanne

1986   Ceramic Artifacts. In Excavations at Indian Hill Rockshelter, Anza-Borrego Desert State Park, California, 1984-1985, edited by Philip J. Wilke, Meg McDonald, and L. A. Payen, pp. 80-100. Report on file, California Department of Parks and Recreation, Sacramento.

1996   Southern California Brown Ware. Doctoral dissertation, Department of Anthropology, University of California at Davis.

Hector, Susan M.

1984   Late Prehistoric Hunter-Gatherer Activities in Southern San Diego CountyCalifornia. Doctoral dissertation, Department of Anthropology, University of California at Los Angeles.

2007   Eastern San Diego County Site Evaluations: CA-SDI-4010 and CA-SDI-17817.  ASM Affiliates. Prepared for the Bureau of Land Management.

2011   Evidence for Intensification and Diversification of Activities During the Contact Period in McCain Valley, San Diego County, California. Planning and Research Collaborative (PARC). Prepared for the Bureau of Land Management.

2014   San Diego Museum of Man – Malcolm Rogers Pottery Type Collection.  Manuscript on file with the author.

Heye George G.

1919   Certain Aboriginal Pottery from Southern California. Indian Notes and Monographs, Vol. VII, No. 1, Museum of the American Indian, Heye Foundation, New York.

Hohenthal, William D.

2001   Tipai Ethnographic Notes: A Baja California Indian Community at Mid-Century. Ballena Press Anthropological Papers Number 48. Ballena Press, Novato.

Odegaard, Nancy

2009   Evaluation of Conservation and Preservation Practices in a Southwest  Pottery Collection. University of Arizona, Arizona State Museum

Rice, Prudence M.

1987   Pottery Analysis. The University of Chicago Press, Chicago.

Rogers, Malcolm J.

1936   Yuman Pottery Making. San Diego Museum Papers Number 2. San Diego,  California.

Sampson, Michael P.

2016   The Archaeology of Anza-Borrego Desert State Park and Surrounding Lands:  Seeking Spirituality, Power, and Sustenance in the Colorado Desert. In Press.

Schiffer, Michael B.

1987   Formation Processes of the Archaeological Record. University of Utah Press, Salt Lake City.

Shepard, Anna O.

1980   Ceramics for the Archaeologist. Publication 609. Carnegie Institution of  Washington, Washington DC. Reprinted by Braun-Brumfield, Inc., Ann Arbor.

Van Camp, Gena R.

1979   Kumeyaay Pottery. Ballena Press Anthropological Papers No. 15. Ballena Press.

Waters, Michael R.

1982a  The Lowland Patayan Ceramic Tradition. In: Hohokam and Patayan, edited by Randall H. McGuire and Michael B. Schiffer, pages 275 – 297. Academic Press.

1982b  Appendix G – The Lowland Patayan Ceramic Typology. In: Hohokam and Patayan, edited by Randall H. McGuire and Michael B. Schiffer, pages 537 – 570. Academic Press.