388 
Fishery Bulletin 106(4) 
spatial differences in procurement of samples by includ- 
ing only sites located at the eastern end of the island 
that were associated with rocky intertidal coastlines. 
We further controlled for possible seasonal differences 
in procurement by including only midden deposits that 
were stratigraphically associated with winter villages 
(as denoted by the presence of large semisubterranean 
house ruins). Sample sizes for Pacific cod in each of 
these faunal assemblages ranged between 48 and 3219 
specimens (individual bones). Actual sample sizes for 
measured Pacific cod skeletal elements ranged from a 
low of 10 to a high of 507 elements. By archaeological 
standards, the number of fish bones recovered from these 
middens was large, and thus the data set was relatively 
robust. Nevertheless, we recognize the possibility of 
potential sample-size effects, and present the entire data 
set as a case study. 
The majority of Pacific cod bones recovered contained 
significant biometric information because they were not 
fragmented and came from well-preserved deposits. 
Because length figures prominently in modern fisheries 
management research (Shin et al., 2005; for the Gulf 
of Alaska see Thompson et al., 2006), allometric rela- 
tionships developed by Orchard (2001, 2003) were used 
to establish the live length of an individual from the 
skeletal elements of Pacific cod. As an exercise to pres- 
ent the types of data that can be reconstructed from 
ancient Pacific cod bones, the fecundities of Pacific 
cod were estimated from their length distributions, by 
means of regressions developed by Karp (1982). More 
accurate reconstructions could be based on average 
fecundity relationships specifically recorded for Pacific 
cod in the Gulf of Alaska, although these data were 
not available at the time of publication of this article. 
Thus, the assemblages of skeletal remains recovered 
from middens in our study contained information 
about Pacific cod populations that could be compared 
to information on modern population structures for 
a large-scale view of changes in size and abundance 
over time. 
Measurements were taken from the ascending process 
of premaxillas and the centra of trunk vertebrae from 
Pacific cod derived from eight archaeological sites span- 
ning the period ca. 2550 BC to 1540 AD (radiocarbon 
years were calibrated to calendar years by using the 
methods of Stuiver and Reimer, 1993). Although these 
data were time averaged, the samples were from single 
stratigraphic units dated by multiple radiocarbon mea- 
surements and likely represent a single depositional 
episode that took place over one or a few seasons. From 
these data, the estimated length of the individuals that 
contributed each element was computed from Orchard’s 
(2001, 2003) regression equations and the combined 
measures were used to produce length-frequency dis- 
tributions for each time slice. 
The modern length distribution of Pacific cod was 
obtained from the longline survey done by the Alaska 
Fisheries Science Center in the Gulf of Alaska in 2005 
(n = 3308; Thompson et al., 2006). Longline surveys 
are a fishing method that may be most comparable 
with prehistoric jigging practices. However, it should be 
noted that modern longline gear may subtly select for 
smaller size fish, as discussed by Halliday (2002). The 
mean longline survey data (Thompson et al., 2006) are 
presented as frequencies in specific size bins, and mean 
length was calculated by using procedures in Gedamke 
and Hoenig (2006). 
We also measured the relative abundance of Pa- 
cific cod in the prehistoric middens using a measure 
known as an abundance index (Al). This measure can 
be used to track shifts in the relative abundance of 
taxa in relation to other taxa in a faunal assemblage 
and represents a normed ratio of a highly ranked (in 
terms of human foraging efficiency) taxa A to a lower 
ranked taxa B, measured as AI=A/\A+B ] (Bayham, 
1979; Ugan and Bright, 2001). Values close to 0 indi- 
cate a complete absence of taxa A, and values close to 
1 indicate a dominating presence. We used an abun- 
dance index because they are generally robust and are 
resistant to taphonomic (conditions affecting preser- 
vation) and collection biases, as long as these biases 
are systemic to all the assemblages being compared 
(Ugan and Bright, 2001) — a situation applicable to 
the Sanak Island data. Although simple to calculate, 
the measures can be powerful. For example, shifts 
in the index can reflect changes in human foraging 
efficiency because they incorporate body-size-based 
caloric relationships. In the model, it is assumed that 
larger bodied prey are often more highly ranked than 
smaller bodied prey (because they contain more calo- 
ries per unit of effort); therefore when a highly ranked 
or large-bodied prey is compared to a lower-ranked, 
or smaller-bodied prey, the values can be used to de- 
termine the occurrence of resource depressions, or 
declines in foraging efficiency. Abundance indices 
are a well-established and peer-reviewed method of 
measuring changes in the frequencies of taxa through 
time from archaeological data (Broughton, 1994, 1997; 
Butler, 2000; Ugan and Bright, 2001; Nagaoka, 2002; 
Betts and Friesen, 2006). 
Problems can arise in the interpretation of Al mea- 
sures when inappropriate B taxa, or lower-ranked taxa, 
are chosen for comparison with higher ranked A taxa. 
Betts and Friesen (2006) demonstrated that these is- 
sues can be overcome by 1) carefully selecting B taxa 
that occur in moderate, but relatively stable frequen- 
cies in the assemblages, and 2) comparing multiple 
low-ranked B taxa and assessing overall trends. Here 
we compared Pacific cod to smaller-bodied species of 
fish — Cottidae, Hexagrammidae, Pleuronectidae, and 
Oncorhynchus. The former three taxa were found in the 
same resource patch (i.e., in the nearshore jig fishery), 
and likely entered the procurement system as bycatch. 
The last taxon, salmon, was likely captured at stream 
mouths and head waters with weirs and nets, and was 
included in our analyses for comparative purposes. 
Mean lengths of Pacific cod recovered from different 
eras were compared by using t-tests and analysis of 
variance, and linear regressions were used to deter- 
mine the significance of trends over time. 
