a very low count because it was fouled on the net or 

 jammed against the frame during some portion of 

 the tow, but have been in apparently good working 

 order when retrieved. For consistency during sub- 

 sequent analyses, all ratios >2.000 were consid- 

 ered to be due to a malfunction of one wheel caus- 

 ing a very low reading for the possible distance 

 towed. We feel that this cut off was justified on the 

 basis of the shape of the histogram of ratio fre- 

 quencies (Figure 2). Seventeen percent of all hauls 

 fell into this malfunction category. The remaining 

 33% of ratios, between 1.105 and 2.000, rep- 

 resented a high level of normal operational varia- 

 tion. In all cases of ratios >2.000 it was an un- 

 usually low wheel reading which produced the 

 high ratio, not an unusually high one. Ratios 

 >2.000 were not further considered in the 

 analyses. 



Consistency with Other Estimates of Distance 



Wheel counts were significantly positively cor- 

 related (959( level) with time duration and loran 

 determined estimates of distance (Table 1 ), but the 

 correlations were not high. Much more precise 

 estimates of time duration distances were avail- 

 able on the 17 tows employing a time-depth gage. 

 All the gaged hauls were taken between 2,500 m 

 and 3,000 m where bottom conditions were rela- 

 tively uniform. The correlations between the 

 wheel counts and distance based on gage time on 

 bottom were significant ( -1-0.89 at the 95% level) 

 and much higher than with wire-out determined 

 values. This marked improvement in correlation 

 indicates that errors in determining actual bottom 

 contact time were the major source of difference 

 between wheels and other measures of distance. 



The 17 time-depth gage records were compared 

 with the wire-out time on bottom estimates. On 

 average the trawl was actually on bottom 23 min 



Table l. — Correlation coefficients between measurements of 

 distance towed. All values are significantly greater than 0.0 

 (95% level). Asterisk denotes values significantly higher in 

 selected data. Significance was tested following 

 2-transformation. Data selection consisted of removing ex- 

 ceedingly large position changes and samples with a wheel ratio 

 >2.0 (see text). 



Selected datan = 257 



longer than predicted, varying from 75 min longer 

 to 65 min shorter. 



The time-depth gage readings were also used to 

 determine a rough estimate of the amount of 

 wheel slippage. Using linear regression the gage 

 measurement of bottom time was taken as the 

 independent variable and the wheel counts as the 

 dependent variable. If the average speed over bot- 

 tom was the intended 2.0 kn, and if no slippage 

 occurred, then the slope of regression should have 

 been 30.85 (counts per minute). While sig- 

 nificantly different from zero, the slope of regres- 

 sion (19.2) was also significantly lower than the 

 expected 30.85 at the 95% level. This major dis- 

 crepancy may have been due to a ship's speed over 

 bottom consistently much less than 2.0 kn, consid- 

 erable wheel slippage, or both. If ship's speed over 

 bottom actually did average 2 kn, then an estimate 

 of the worst slippage was 40%. If the ship's speed 

 was consistently low, then the wheels performed 

 better by slipping less. 



Wheel Counts Versus Catch Data 



No consistent positive relationships were found 

 between the catch of the trawl and the wheel count 

 estimates of distance. This lack of the desired posi- 

 tive correlation was difficult to evaluate because 

 catch was controlled by performance of the trawl 

 and the actual distribution of the fauna. Faunal 

 variation among areas sampled may have masked 

 variation in catch due to differences in distance 

 towed. 



In one comparison the echinoderm catch of 22 

 pairs of consecutive tows in approximately the 

 same bottom area at continental shelf depths was 

 examined. According to wire-out determinations 

 of time on and off bottom all 44 tows were on 

 bottom 20 min. Echinoderm catch was considered 

 because these asteroids, echinoids, and 

 holothuroids represent relatively immobile large 

 benthic organisms. The number of species and 

 total specimens were taken separately as mea- 

 sures of catch size. The number of times the longer 

 haul of the pair (as indicated by the magnitude of 

 the low wheel count on each sample) had the 

 greatest catch was tallied for all pairs of samples 

 and compared against random expectations using 

 a nonparametric sign test (Table 2). While having 

 a greater catch in most cases, the longer tows did 

 not take significantly higher numbers of 

 echinoderm species or specimens. 



In addition we examined the catch of the abun- 



794 



