Analyses of the faunal materials from the TU4 and TU7 allowed me to address several questions concerning the animal exploitation habits of the Ellikarrmiut people. Through the study of spatial distributions of faunal material abundances in selected samples, I could approach the following issues:
Methods of Faunal Analysis
Before I could begin to analyze the faunal materials from Nash Harbor, I needed to separate animal remains from other archaeological materials, and to identify potentially informative areas for analysis. The Nash Harbor Project employed several University of Oregon students to process recovered materials. Laboratory workers cleaned screen samples with dry natural fiber brushes and damp natural sponges, and sorted the samples into seven broad materials classes: stone tool fragments/debitage, pottery, charcoal/botanical material, historic debris, shell, fish bone, and other bone (mammal and bird). Laboratory workers weighed the material in each class and (where possible) counted individual specimens. I then entered these weights and counts into an Excel spreadsheet to track the relative abundance of faunal material classes within and between proveniences. I used these data to target proveniences with large and variable amounts of faunal material for further study.
I followed four general directives in selecting samples for analysis:
- The faunal material from all samples should represent a broad range of activities related to animal capture and processing.
- The samples should be widely distributed spatially. Originally, I selected samples from several levels in Test Unit 4, with the intention of identfying change in faunal material distribution through time. Since all the samples are roughly contemporaneous, this point is largely moot.
- The samples should be stratigraphically discontinuous, derived from units with different depositional histories, in order to simplify the construction of minimum number of individuals (MNI) data. This will allow me to make the assumption that the remains of a single animal are not distributed among more than one sample.
Following the guidelines above, I selected samples from seven excavation levels for analysis: Levels 6, 14, 18, 20, 24, and 28 from Test Unit 4, and Level 7 from Test Unit 7. Selected sample levels are indicated in the stratigraphic profile maps, Figures 2.2 and 2.4. These samples also bear shorthand laboratory designations (Test Unit/Level): 4/6, 4/14, 4/18, 4/20, 4/28, and 7/7, respectively.
In the second phase of analysis, I attempted to identify to anatomical element and taxonomic category every animal bone or shell fragment from the discretionary sample. The University of Oregon's zooarchaeological comparitive collections lacked certain key species useful for my analyses, including walrus (Odobenus rosmarus), reindeer (Rangifer tarandus) and arctic caribou (Rangifer arcticus), and most of the 89 bird species found on Nunivak Island. To supplement the comparative collection, I employed a variety of literature sources (Cannon 1987; Foster 1991; Gilbert 1973; Smith 1979; Wheeler and Jones 1989) to aid in my identification. In almost all instances, I could utilize these resources to make positive identifications to family, and in many cases to genus. By comparing these results with the range of animals reported in field guides (Burt and Grossenheider 1976; Lamb and Edgell 1986; McConnaughey and McConnaughey 1985; Peterson 1990), I could narrow my identification to the species level in a few instances.
Several anatomical elements proved easier to identify and more information-rich than others. Following Wheeler and Jones (1989:150), I identified certain elements of fish skeletons that preserve well, are easily identified to species, and either occur singly or in pairs. Especially informative fish elements include heavy cranial bones along the medial line (and thus occurring once in a single fish). Some paired elements also proved informative, especially premaxillae, maxillae, and dentaries. Because of extraordinary preservation conditions at Nash Harbor, I could also distinguish elements, such as lightweight bones in the occipital and otic regions, that in other contexts are usually lost to taphonomic factors. Davis (1987:35) noted elements of mammal skeletons providing the same utility, including mandibles, bones of the shoulder and pelvic girdles, and limb and foot bones. The small size of the analyzed mammalian sample limited the number of these indicator bones, and the extent of the comparative collection meant that many bones, especially carpals and tarsals, remain unidentified. Bird remains proved the easiest to identify and often the best preserved. Bird leg, foot, and wing bones, and bones of the sternal and shoulder regions, were especially informative. Analyses at this stage produced number of individual specimens (NISP) data, which reflect the number of bone fragments identified to a particular taxon. In a few instances, I could refit broken bone fragments, but while these refits proved informative for generation of minimum number of indivual data (discussed below), I treated all fragments found at the time of analysis as individual specimens. In other words, if a single bone were found in two fragments, and I refitted those fragments, they still represented NISP = 2.
The next phase of analysis involved the determination of minimum number of individuals (MNI) for each taxonomic group identified above. Minimum numbers provide an estimate of the number of individuals represented in a sample. Using the presence of elements diagnostic of particular taxa—and making some consideration for size and age of the individuals at death—I could derive MNI for all identified taxonomic groups. Except in two instances (the TU4, Level 18 and Level 20 samples, and the Level 24 and Level 28 samples), I regarded each sample unit as discontinuous with other samples, so the MNI for any one species present in a given sample was figured separately from MNI for the same species in other samples.
After identifying remains from excavation samples and computing MNI from those remains, I identified elements from the bulk samples associated with each analyzed unit. Since the volumes of excavated levels and bulk samples varied, and since the bulk samples represented only a tiny fraction of the total soil in any one excavation unit, I could not with any confidence derive meaningful quantitative data from bulk samples. However, the identification of the bones of very small animals in some bulk samples suggested that Nash Harbor residents exploited animals whose remains were not recovered in the 1/4" screen samples.
Since the excavation units providing faunal materials for my analytic sample varied in volume, I studied field notes and maps to estimate the total volume of soil actually excavated in each level. Using these numbers, I computed the Volume Control Constant (KV), to control for variation in excavation volumes. KV represents an excavation level's inverse relationship to the volume of an ideal 10 cm deep excavation level in a 1 m X 1 m excavation unit (.1 m3), and is expressed as:
KV = 0.1/V,
where V is the Excavated Volume of each unit. By multiplying quantitative measures of faunal abundance (e.g. MNI) in a sample with KV, I could directly compare the Minimum Numbers of Individuals (or other measure) for two units, even if they represented samples from units of different excavated volumes. In other words, KV allowed me to pretend that the analyzed faunal samples all derive from ideal 10 cm deep, 1 m X 1 m excavation units.
Faunal Analysis: Problems & Cautions
With small, nonrandom samples recovered using 1 4" screen, from a region largely unexplored archaeologically, there are certain limitations on zooarchaeological analyses discussed in my study.
The analyzed materials are not a random or unbiased sample of the Nash Harbor assemblage. In fact, I chose samples in a decidedly nonrandom fashion, selecting sample units based on sample size and apparent diversity, not for overall representativeness. The analyzed samples represented a tiny fraction of the entire Nash Harbor assemblage, most of which derived from only two midden deposits. The 1/4" mesh used for screening excavated soils led to a bias towards larger animals, or animals with larger remains. I expected that this would lead to an underrepresentation of small fish (e.g. Clupea sp.) and finely pulverized shell fragments, for example. Analysis of faunal material retrieved from bulk soil samples helped correct this bias, at least by indicating that certain species may have been missed. I did not attempt quantitative analyses from bulk samples however, as they are far too small (usually about 2 liters) to represent an entire excavation unit and not uniform in size.
I have already addressed some of the limitations of the University of Oregon comparative collection to my research. While it is a wonderful resource for Northwest Coast archaeologists, it had more limited utility for my studies. I am confident of my identifications at the family level, and in most cases at the genus level. As a result, the broad taxonomic variety of animals (especially birds) found on Nunivak Island is generally underrepresented in my analysis.
Using the arithmetic manipulations described above, I have corrected measures of abundance (bone weight, NISP, and MNI) for variations in the volumes of excavated units. However, these manipulations do not accurately represent a predictable change, since the number of taxa present in a sample varies as a function of NISP or MNI. That is, as the actual number of specimens or individuals rises, the taxonomic richness of a sample should also rise (Davis 1987;Grayson 1981). I will address this difficulty in greater detail in Section 5. Arithmetic manipulations also produced non-integral numbers for measures that are actually counts of individuals (specimens or animals). This leads, for example, to nonsensical measures like "1.8 Microgadus individuals" or "0.6 individual Odobenus bones." These fractions do not represent the observation of fractions of individuals, but rather the number of individual bones or animals per 0.1 m3 of excavated soil. I have avoided rounding corrected measures to the nearest whole number, since this would add a decimal place of uncertainty to my analyses.
Many of my analyses depend on the distribution of faunal remains represented by minimum numbers of individuals. Davis (1987:35–36) questioned the utility of MNI as a measure of abundance between mammalian taxa, although his apprehensiveness concerned comparisons of MNI in large assemblages, subjected to more statistical manipulation than the small Nash Harbor assemblage. Wheeler and Jones (1989: 149) also expressed reservations about using MNI to rank fish remains at a site. Grayson (1981) noted that quantitative faunal data are largely dependent on sample size, an imporant caveat given the small size of the Nash Harbor sample for all taxonomic groups other than fish. Hesse and Wapnish (1985: 114) addressed the spatial distribution of elements in determining MNI; they wondered how far apart two elements must be—horizontally or vertically—before they can be considered as deriving from two individuals. Ultimately, I sided with Grayson (1973) in acknowledging the shortcomings of MNI and other derived data, and utilized them to construct indices of ordinal, rather than absolute, abundance. That is, I could say that certain species were more abundant than others, but could not address how much more important.
Finally, my research pioneers new ground zooarchaeologically. Archaeological data for the entire Yukon-Kuskokwim delta region are generally sparse; except for Chatters’ (1972) brief report, zooarchaeological data were nonexistent. Without baseline data, I can not advance or test hypotheses based on the work of previous archaeologists. Conversely, I expect that many of my results will serve as targets for attack by future, more savvy researchers.