Only those bones and bone fragments recovered from undisturbed contexts were included in the material analyzed from the Wall and Fredricks sites. In other words, bone from the plowzone was excluded. The vast majority of the analyzed faunal remains from the Wall site was from four 10´10-ft units of undisturbed sheet midden. Although three burial pits were excavated at this site in 1983, the fill from only one of those pits contained more than a few poorly preserved bone fragments. Therefore, the remains from the fill of only one burial pit and four squares of midden make up the sample analyzed from this site. The 1983-1984 faunal assemblage from the Fredricks site was recovered from the fill of fourteen pits. Nine of these were burial pits, one was a fire pit, one was a storage pit, and three pits were of indeterminate function. No sheet midden was found at the Fredricks site.
Identical analytical procedures were used on the assemblages from both sites. All of the bone recovered in the 1/2-inch and 1/4-inch mesh screens was analyzed. There were numerous tiny, unidentifiable fragments of bone retrieved by the 1/16-inch screen. Because it would have been a time-consuming and (probably) pointless task to separate all of these minute fragments from the fine gravel that was also recovered in this size screen, only those bones and bone fragments which appeared to be identifiable were pulled from the 1/16-inch washings. The bones and bone fragments from each excavated unit (10´10-ft square of midden, or feature) and from each level or zone within each excavation unit were kept separate during analysis. Also, bones from different-sized screens were not combined during analysis.
The basic procedures followed in identifying and analyzing the faunal remains from the two sites closely follow those outlined by Smith (1976): (1) each bone fragment was initially sorted into one of three groups (unidentifiable, identifiable only to class, or identifiable as to skeletal element); and (2) each of these fragments (whether it was identifiable or not) was examined for evidence of modification such as burning or cutting.
For those bones that could be identified beyond the level of class, the side of the body (when applicable) and portion of the bone (proximal, distal, or shaft) was noted. After that, a taxonomic identification was made for each of the identifiable bones and bone fragments. Several of the variables that affected whether a fragment could be identified beyond family or order were: "(1) the specific skeletal element in question (i.e., rib versus mandible), (2) the amount of diagnostic surface present, (3) the ability of the person identifying the specimen, (4) the size of the comparative collection being employed, and (5) the degree of morphological similarity of species within the taxonomic group" (Smith 1976:281). To help minimize problems introduced by variables (3) and (4), a group of 205 bones and bone fragments was sent for identification to Elizabeth Reitz at the Zooarchaeological Laboratory, University of Georgia. This sample consisted of bones that appeared to be identifiable but for which the type collection at the Research Laboratories lacked comparative specimens.
In addition to determining the total number of fragments in each taxonomic category, all of these fragments were weighed.
When possible, the age and sex of the animal represented by a particular fragment was assessed. In most cases, these characteristics could be determined only for the remains of white-tailed deer. Deer age was estimated by noting whether or not the epiphyses of the long bones were closed, and by using Severinghaus's (1949) criteria of tooth development and wear. Sex of the deer was determined by using the pelvic girdle criteria set forth by Edwards et al. (1982). Attempts to determine age and sex of several other species, such as rabbits, squirrels, and raccoons, were less successful than for deer. This problem resulted, in large part, from characteristics of the faunal assemblages themselves. Many of the bones, or portions of bones, that display the characteristics used to distinguish between animals of different ages or sexes simply were not present in the remains being studied.
Information obtained from the procedures discussed above constitute primary data or "direct quantification of identified material" (Wing 1979:119). Several factors can influence how accurately these primary data reflect the original faunal sample. All bones, for example, do not stand an equal chance of being represented in an archaeological assemblage. The survival of bone after it has been discarded is affected, primarily, by two factors: its physical condition at the time of disposal, and the nature of the environment in which it was placed. Whether a bone was burned, boiled, or roasted affects its chemical and physical properties, which, in turn, influences preservation (Chaplin 1971:15). Also, the basic structure of the bone must be considered. Teeth and phalanges are stronger than ribs and vertebrae, and thus are less likely to be destroyed (Payne 1972:68).
The manner in which a particular bone was discarded further affects its survival. If the bones were buried in a trash pit, for example, the rate of disintegration would depend on factors such as the "acidity or alkalinity, degree of aeration, movement of water, bacterial population, as well as the structure and seasonal properties of the soil" (Chaplin 1971:16). If it remained on the surface of the ground, it would be more accessible to scavengers, more likely to be damaged by weather, and more susceptible to being stepped on and crushed.
Excavation techniques also affect the number and kinds of bones eventually available for analysis. The portion of the site excavated, sieving techniques utilized, and steps taken to protect the fragile bone after excavation all affect the sample.
For these and other reasons, one can assume that any collection of archaeological bone will represent only a portion of the faunal remains originally associated with the site. Thus, the primary data obtained probably will not provide enough information for reliable interpretations of what the assemblage represents in terms of past behavior. For this reason, secondary data, "which involve interpretation, extrapolation, or estimations based on primary data" (Wing 1979:118) are neccessary. Examples of secondary data include calculations of minimum numbers of individuals, and estimates of usable meat weight.
Chaplin (1971) lists three of the most commonly named methods for quantifying the species represented by a collection of animal bones: (1) the fragments method, (2) the weight method, and (3) the minimum-numbers method. Whereas there are advantages to each method, Chaplin and most others (e.g., Daly 1969; Klein and Cruz-Uribe 1984; Smith 1976; Styles 1981; and White 1953) prefer the minimum-numbers method.
With the fragments method one counts the total number of identifiable bones and fragments of each species and determines the ratio of different bones or different species. The number of identified specimens (bones or bone fragments) per species is sometimes abbreviated as NISP (Grayson 1979; Klein and Cruz-Uribe 1984; Payne 1975). The NISP is little more than a list of bones of different animals present in an assemblage. The number of bones of a particular species represented in an assemblage does not necessarily indicate what percent of the diet of the original inhabitants was made up of the meat from that animal. For example, some species of animals have more bones than others. Also, although hunters may bring back the entire carcass of a smaller animal, they are liable to return with only the more useful parts of a larger one. Thus, only the broadest questions about subsistence can be answered using NISP.
In another approach, used to arrive more directly at conclusions about the relative dietary importance of each species, the analyst weighs the bone from each species and then multiplies that weight by a factor to determine the amount of meat represented by each type of animal. In using this method, however, every scrap of bone must be utilized in order to arrive at an unbiased approximation of amount of meat (Daly 1969:149). Because much of the bone analyzed usually is fragmented, it is nearly impossible to place each scrap into its appropriate species category. Further, it is impossible to account for all of the bone missing from the site or not retrieved during excavation. Also, the weight of the bone is affected by whether or not it was burned or charred and by the thoroughness with which it was cleaned and dried after excavation. Another objection to the weight method is the fact that it begins with the assumption that there is a fairly constant relationship between the weight of an animal and the weight of its bones. Although there is a correlation between these two factors, the relationship is variable (Smith 1975a:100). To counteract this bias it would be necessary to apply a different live weight value for each age and sex category for each species analyzed. Because it is not always possible to identify the species to which a fragment belongs, let alone the age or sex of the animal, the weight method is only appropriate for use with relatively few completely identified fragments.
The minimum numbers of individuals (MNI) method avoids many of the problems that plague the other two methods. Using the simplest form of this procedure, the minimum number of animals of each species is determined by counting the maximum number of any particular bone. When possible, the age, sex, and size of the animal is taken into account to increase the accuracy of this method. This analytical procedure is superior to the other procedures for a number of reasons.
The minimum number of animals that the bones could have come from is an indisputable fact. It is, moreover, a direct measure of a number of animals involved and is an abstraction of the true number of animals involved only within fixed limits. It also involves no assumptions about differential preservation of bone which can not be checked by examination of the specimens or by a site inspection. It is therefore using verifiable facts throughout (Chaplin 1971:70).
Grayson (1973:70) notes that the minimum-numbers method "provides us with units which are necessarily independent of one another, and which may therefore be validly used in further statistical manipulation."
Despite these advantages, the minimum-numbers method also has several shortcomings. First, there is more than one way to derive the minimum number figure from an assemblage. Variation in the way in which faunal material from a site is grouped, for example, affects the results of analysis. If the material is separated into clusters according to the stratum and excavation unit in which it is found, it will yield the largest estimation of MNI. If the excavation unit is ignored, the minimum number decreases, and if neither excavation unit nor stratigraphy is used in grouping the material, the number will be even smaller (Grayson 1973:433). The comparability of the data produced by the minimum-numbers method is still suspect unless the analyst explicitly states how he arrived at his figures.
All three methods were used to quantify the faunal remains from the Wall and Fredricks sites. The NISP method was used because it was calculated automatically as the bone fragments were identified. Also, the weight of the bone identified for each taxonomic category was calculated. Comparison of the relative abundance of each species, as revealed by the number of identified fragments and by the weight of these fragments, provided information useful not only in determining the possible importance of these animals to the original inhabitants, but also information about the conditions (such as fragmentation or preservation) that affected how much of the assemblage could be identified and to what taxonomic level. The weights of the identified bones were not converted to meat weights because of the vast array of biases introduced by the use of the weight method.
The minimum numbers of individuals method was relied on most heavily in interpreting the two faunal assemblages. In comparing the assemblages from the Wall and Fredricks sites, MNI was calculated from each site as a whole, with neither the excavation unit nor site stratigraphy taken into consideration. Although it yielded the smallest number of individuals, this method was necessary because of the different contexts from which the two assemblages were recovered.