21 



Two programs for handling these data were then written by programmers 

 in the Computation Facility of the University of California, San Diego, 

 using Fortran as the machine language. The first program was written 

 as a test of whether this system of data storage could be used satisfactorily 

 for grouping together species into natural classifications as related to the 

 distribution of physical factors. Earl Ferguson of the Computation 

 Facility wrote most of the program, with some assistance from Anna 

 Devore. The computor was programmed to list all species found at various 

 depths, temperatures, oxygen limits, sediment types and geographic 

 localities. Unfortunately, the limits of these variables were chosen arbi- 

 trarily, based on previous inspection of the data. The computor first listed 

 all stations falling within the various ranges of each variable, using station 

 cards only. The next step entailed a search through all of the species-data 

 cards, in which the machine stored the numbers of all species occurring 

 at all stations previously listed for each category. Finally, all species 

 numbers were matched with the species-name cards, and complete print- 

 outs were made of all the species names occurring under each increment of 

 depth, temperature, oxygen, sediment type and geographical region 

 (example on fig. 5a). The entire operation, taking 28 minutes, would have 

 taken several months by hand. The next step in determining the influence 

 of various physical factors on the distribution of invertebrates would be 

 to have the machine choose the limits of the variables by a multi-correla- 

 tion analysis. This has not been done as yet, since a somewhat simpler 

 method was devised for a preliminary analysis of the assemblages. 



The second program was modified from one originally written for 

 Edward Facer of Department of Oceanography at Scripps Institution, 

 and was written by Robert and Eileen Mitchell. Facer's program 

 was devised for data catalogued in a somewhat different form, and so had 

 to be rewritten to utilize the present data. Facer (1957, 1963) demonstrated 

 that through a contingency matrix of joint occurrences of common species, 

 using the geometric mean of the proportion of co-occurrences, that cer- 

 tain natural groupings of animals would result from a sampling of a 

 region, which would correspond to environmental boundaries. Using only 

 the presence or absence of species at every sample location, minus a 

 term that corrects for sample size, the computor was asked to produce 

 statistically valid groupings from 350 species taken more than five times 

 in this sampling program. The other 800 species taken in this study were 

 considered too infrequently taken to be considered in this analysis. The 

 computor program, which did not take into account type of sampling de- 

 vice, first compared the distribution within the 270 stations of every spe- 

 cies with every other species. Pairs of species, with pre-established lower 



