BARKLEY: SELECTIVITY OF TOWED-NET SAMPLERS 



is an increase in the apparent reaction distance. 

 Instead of the true value, Xo, we would expect 

 to obtain an apparent value which approaches 

 (xo + radius of the school) for large schools. 

 If Stolephorus schools at all sizes represented 

 on Figure 4, the different values of Xo/R, 6 versus 

 8, may reflect the presence of smaller and larger 

 schools, differing in radius by about 2R or 1 m. 



Some support for this apparently fanciful 

 argument can be found in the Stolephorus catch- 

 es obtained at night by Murphy and Clutter 

 (1972). Night purse-seine catches showed 

 marked peaks in abundance of fish in the 7-8 mm 

 and 11-12 mm size classes and relatively low 

 abundances in the 4- , 9- , and 14-mm class in- 

 tervals. Simultaneous tows with the 1-m net, 

 on the other hand, yielded precisely opposite re- 

 sults. The 1-m net apparently undersampled 

 abundant size, classes and was more effective 

 with the less abundant size classes. This is pre- 

 cisely what is expected if the purse seine catches 

 schools of various sizes equally well, whereas the 

 1-m net underestimates the abundance of 

 schooled fish, more or less in proportion to school 

 diameter, provided only that the fish school by 

 size and that changes in abundance are associ- 

 ated with changes in school size, rather than 

 changes in the number of schools. 



This detailed treatment of a single set of 

 length-frequency data is intended only as a dem- 

 onstration of the amount of information which 

 can be extracted from such observations in the 

 light of theory, when and if measurements of 

 typical reaction distances and escape velocities 

 have been made. Results of this as-yet-tentative 

 analysis may be summarized by saying that 

 Stolephorus in the 1-m net sample apparently 

 consist of: 



A. Animals in the 1.5- and 2.5-mm class in- 

 tervals, which are partially lost through the 

 meshes and probably are unable to avoid 

 the net. 



B. Animals 3.5 mm long, which seem to be 

 adequately retained by the meshes and ap- 

 parently are too small to effectively avoid 

 the net. 



C. Animals 4.5 to 11.5 mm long, which react 

 at about 3-4 m distance (proportionately 

 more, or less, if their swimming speed is 



less, or more, than 10 body lengths per sec- 

 ond) . Maximum Pc for this group is about 

 0.35 for the 4.5 mm fish, dropping to 0.002 

 for the 11.5 mm fish. 

 D. Animals 12.5 to 14.5 mm long, which have 

 anomalously large values of Pc when com- 

 pared to group C, above. If these differ- 

 ences are significant, this implies that their 

 reaction distances or swimming speeds, or 

 both, may be abnormally low. 

 In the following examples there are no mea- 

 surements of absolute abundance to compare 

 with towed-net catch data, so assumptions will 

 have to be made as to the relative contributions 

 of population structure and avoidance to the out- 

 come of sampling. Before making such assump- 

 tions it will be useful to consider the relative 

 effects of these two factors in the case of Sto- 

 lephorus. 



On Figure 4 it can be seen that population 

 abundance of Stolephorus, Nl, changed by a 

 factor of 10 over the length interval sampled by 

 the 1-m net. Over this same size interval the 

 catch per class interval {Cl) changed by a 

 factor of 1,000 or more. Thus Nl accounts for 

 no more than 1% or 2% of the observed changes 

 in catch length frequency, the remaining 98% 

 to 99% being attributable to avoidance, at least 

 for animals large enough to be fully retained by 

 the mesh. In this strictly relative sense a 10- 

 fold change in population density with size is in 

 fact negligible, however important it may be in 

 another context. 



The problem of relative significance can be 

 placed in perspective by considering whether 

 data such as the 1-m net Stolephorus catches 

 might be used to estimate population structure. 

 Daylight 1-m net catches appear to be so over- 

 whelmingly influenced by avoidance that even 

 a 100-fold change in Nl might not be measurable 

 since it could be obscured by a mere 10% error 

 in estimating the effects of avoidance. In this 

 sense it would be fair to assume that a popula- 

 tion's structure is uniform, constant for all class 

 intervals sampled by a given net if changes in 

 Nl could not have been measured by that net. 

 This can reasonably be taken to be true when- 

 ever the right-hand slope of the catch curves 

 {Cl) exceeds the slope of the population curve 



807 



