23.02.2013 Views

Plains Indian Studies - Smithsonian Institution Libraries

Plains Indian Studies - Smithsonian Institution Libraries

Plains Indian Studies - Smithsonian Institution Libraries

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

166 SMITHSONIAN CONTRIBUTIONS TO ANTHROPOLOGY<br />

ber of individuals per taxon is consistent within<br />

a sample though differing between sites and not<br />

controlled by sample size. What varies between<br />

the sites is the pattern of skeletal recovery. In the<br />

case of the Near Eastern samples, the distribution<br />

of the different bone types within each site's<br />

subsamples are similar. In the case of the bison<br />

samples reported by B. Gilbert (1969:283), this is<br />

clearly not the case. The Woodland sites have a<br />

relative absence of toe bones compared to the<br />

Middle Missouri and Coalescent sites. The result<br />

is widely different effective numbers of individuals.<br />

The point of these comments is to demonstrate<br />

that the curvilinear relationship produced<br />

empirically by a plot of many MNI/E combinations<br />

does not reflect patterned bias in the way<br />

either estimates a consequences population of<br />

whole animals. It is possible to say that any factor<br />

that (1) increases the number of categories of<br />

element types (a trephic factor like the discovery<br />

of new morphological distinctions); (2) increases<br />

the recovery rate of small bone fragments (a<br />

sullegic factor like reducing screen size); or (3)<br />

changes the skeletal recovery pattern (a cultural<br />

factor like a change in the location of animal<br />

processing activities) will vary the relationship<br />

between MNI and E. Whether either effectively<br />

mirrors the finds population is a separate question.<br />

Another method of estimating taxon frequency<br />

in the finds population has been proposed (Gilbert<br />

and Steinfeld, 1977; Hesse and Perkins, 1974;<br />

Holtzman 1979; Perkins, 1973; Wapnish et al.,<br />

1977). It attempts to overcome one weakness of<br />

the MNI approach by using a greater proportion<br />

of the sample, and one weakness of the E approach<br />

by correcting for the variable number of<br />

elements identifiable in different species. Called<br />

the weighted abundance of elements (WAE) by<br />

Holtzman (1979:80), and relative frequency (rf)<br />

by Perkins (1973), it is calculated by dividing E<br />

by the number of bone elements that are potentially<br />

identifiable in a species skeleton. Two<br />

choices exist in the way the division is done. With<br />

one, E is divided by the number of element types<br />

that are actually present in the sample. This is<br />

roughly equivalent to an average MNI. Each<br />

element type is considered an estimator of relative<br />

species frequency. Taking the average minimizes<br />

the risk associated with a single element estimator<br />

(roughly the comparison between a mean and a<br />

mode as a measure of central tendency). With the<br />

other, E is divided by the potential number of<br />

identifiable element types, whether or not they<br />

are actually represented in the sample. With both<br />

choices it is recommended that element types<br />

with extremely low frequency of recovery be omitted<br />

from both values in the calculation. The first<br />

choice reflects the view that only a portion of an<br />

animal's carcass, in many cases, is likely to be<br />

consistently interred in a cultural deposit (for<br />

instance, only foot bones) and that calculations<br />

using the second method would underestimate<br />

taxon frequency. The second method reflects the<br />

view that, in general, whole carcasses are interred<br />

in cultural deposits, and it is the effects of attrition<br />

that reduce the number of element types recovered.<br />

Since the rf estimator is prone to error<br />

from differential preservation, it is critical that<br />

the divisor be chosen with some consideration of<br />

what the choice implies about preservation.<br />

Holtzman (1979) has compared the performance<br />

of rf/WAE and MNI in a series of computer<br />

simulations. He concluded that "in all simulations<br />

the WAE estimates showed generally<br />

smaller mean squared errors than the MNI estimates,<br />

even under most conditions where the<br />

WAE bias was larger than the MNI bias."<br />

Where the distorting effects of natural or cultural<br />

factors are suspected to be large, it is advantageous<br />

to dispense with trying to estimate the<br />

taxon frequency in the finds or consequences<br />

populations. Procedures of this type have been<br />

described by Binford (1978) and Lyman (1977;<br />

1979). These approaches consider faunal remains<br />

as animal parts rather than as whole animals.<br />

The potential resource value (in terms of meat,<br />

fat, sinew, grease) is estimated for each bone<br />

fragment type and multiplied by the bone element<br />

frequency. The output of this kind of analysis<br />

is a description of the resources represented,<br />

not an estimate of the proportion of each species

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!