FISHERY BULLETIN: VOL. 83. NO. 3 



2.4 - 

 2.3 



1.7 - 

 1.6 



1.5 - 

 1.4 

 13 

 1.2 

 1.1 



1.0 



0.9 •- 



2xSE 



• -Group 1 

 o- Group 2 



ci 



SCHOOLING 



THRESHOLD 



(Geometi 



TIC mean) T 



1 ir 



RANDOM 



"Dark" 10" 



'2 10"" 10''° 10'^ 10-8 "Daylight" 



IRRADIANCE Wcm"^ 



FiciKK I.-\'isual threshold for schooling in northern anchovy. 

 Schooling is indicated by the index of dispersion (s^/x)- A ratio of 1 

 implies no schooling as it indicates a random distribution. Each 

 point is a mean calculated from 8-10 photographs and bars are ± 2 

 X standard error of the mean. No error bars are given for one value 

 (2.34 ± 0.47) because it falls far beyond the rest of the values. Mean 

 dispersion is shown for four test levels of downwelling irradiance 

 (log scale), "dark" (below the sensitivity of a 931 A photomultiplier). 

 and "daylight" (1.496 x 1 -■'' W cm " 2) which was the norma! day- 

 time level of irradiance in the apparatus. 



quadrat) did not differ from unity indicating that the 

 fish were randomly distributed. Hence no schooling 

 existed at the lower irradiance value and in darkness, 

 whereas at the higher value the fish were clearly 

 aggregated. These values delimit a region of about 2 

 log units of irradiance where one cannot be certain if 

 schooling occurs or not. The actual threshold for 

 schooling must fall somewhere in that region, and we 

 have arbitrarily considered the threshold value to be 

 the geometric mean irradiance of the above two irra- 

 diances (6 x 10"^^ W cm"-) thereby reducing the 

 uncertainty in the threshold value to about 1 log unit. 

 In subsequent tables and figures we give the upper 

 and lower bounds of the region as well as the 

 threshold value, however. 



The variance of the dispersion index, a measure of 

 the variation in school dispersion among photo- 

 graphs, increased sharply at irradiances above the 

 threshold indicating a wide variation in the disper- 

 sion of fish among photographs. This can be expected 

 because schooling fish react to fright stimuli, feed- 

 ing, and many other conditions by altering interfish 

 distances, thereby changing the cohesion or degree 

 of dispersion of the school (Blaxter and Hunter 

 1982). At light levels below the visual threshold, fish 

 are unable to respond socially to such stimuli, hence 

 the variation among photographs is low. 



Mean distance to the nearest neighbor followed 

 the same pattern as we have described for the index 

 of dispersion. Values in darkness and at the lower 



Table 2.— Mean and standard deviation of the Index of dispersion and 

 mean distance to the nearest neighbor for various irradiance levels. 



'Unfiltered tungsten lamp. 



'Filtered lamp - spectrum given In Table 1. 



'Below sensitivity of 931A photomultiplier. 



238 



