Hl'NTKR and N'lCHOLL: NORTIIKRN ANCHdXV SCIIOOI.INC TIIKKSIIOLD 



for each lig-ht treatment. These measurements are 

 indices of only one characteristic of a fish school, the 

 tendency of individuals to maintain contact with each 

 other and thereby remain in a social group. The 

 polarization of individuals in a school is frequently in- 

 cluded in definitions of schooling as the cohesive 

 movements of a school require that fish maintain 

 relatively constant headings and individual distances 

 which gives the polarized appearance of moving 

 schools. This characteristic of fish schools was not 

 measured in our study, thus the visual threshold we 

 estimated was one for the maintenance of schooling 

 in the broadest sense, that is the existence of a group 

 maintained by visual attraction. 



Radiometric Procedures, Calibrations, 

 and Computations 



Radiometric equipment used in this study con- 

 sisted of an Optronics spectroradiometer (model 

 714-V) (calibrated against a radiometric standard) 

 and a photometer (Gamma Model 700). The spectro- 

 radiometer was used to measure the spectral irra- 

 diance produced by the sources at full lamp output 

 but the treatment levels of irradiance were below the 

 sensitivity of the spectroradiometer. Absolute 

 measurements of light intensity were made 25 cm 

 above the water surface (the difference between this 

 position and within the water would be < 10%). The 

 light treatments were varied by placing neutral den- 

 sity filters in each of the light sources; the neutral 

 density filters were calibrated on an optical bench. 

 Test levels we used were computed from the filter 

 factors for the neutral density filters. The photo- 

 meter was used to check irradiance levels prior to a 

 test, but we believe the computed values to be more 

 accurate. Irradiance distribution in the tank was 

 mapped using the photometer and the treatment 

 values weighted by tank area so that they represent- 

 ed the average irradiance 25 cm above water 

 surface. 



Our laboratory estimates of the visual threshold 

 for schooling were used to calculate the maximum 

 possible depth of schooling in the sea for various 

 levels of incident irradiation and water types. 

 Threshold values in W cm"^ were converted to an- 

 chovy effective units (W cm-'^^nch. eff.) by weighting 

 the spectrum in the apparatus by the relative sensi- 

 tivity of the scotopic curve of the anchovy Engraulis 

 encrasicholus from an electroretinogram by Pro- 

 tasov (1964) (Table 1). Two levels of night illumina- 

 tion were used, full moon at 3 m depth (2.78 x 10 ~^ 

 W cm -2), and starlight at the earth's surface (1.08 x 

 10" 10 W cm"^) (both measurements from Munz and 



McFarland 1977). The depth (meters) in the sea (Z) 

 at which a given threshold (E,) value occurred was 

 calculated using the equation of Baker and Smith 

 (1982): 



In 



where £■„ is the incident radiation (full moon or star- 

 light) in anchovy effective units. K, is the wave 

 length specific attenuation coefficient and is the sum 

 of coefficients for pure water (K,,), dissolved organic 

 matter iK,j), and chlorophyll a (K,). Tables of coeffi- 

 cients, and equations for calculating these attenua- 

 tion coefficients, are given by Baker and Smith 

 (1982). In our calculations we assumed that the 

 dissolved organic matter was constant at 0.7 mg 1 ~ ' 

 which is typical of the anchovy habitat. We calcu- 

 lated K^ for a range of chlorophyll (Chi) a concentra- 

 tions ranging from 0.1 to 10 mg Chi a m -^ and at 25 

 nm intervals from 425 to 600 nm for each Chi a con- 

 centration. Each K, value for 25 nm increments was 

 weighted by the appropriate anchovy scotopic 

 sensitivity, and the average anchovy weighted 

 ^t anch. eff. ^as used in the final calculation of Z. 



Many uncertainties and possible biases exist in 

 such an extrapolation from laboratory- to sea condi- 

 tions: Cloud cover was not considered nor were 

 possible effects of bioluminescence; spectral irra- 

 diance values for full moon and starlight of Munz and 

 McFarland (1977) may not be representative of con- 

 ditions in the anchovy habitat although they are 

 relatively close to those given in photometric units 

 by Brown (1952); variation in dissolved organic mat- 

 ter is not considered; the radiance distribution over 

 360° in the tank probably does not resemble that in 

 the sea (only downwelling irradiance was considered 

 here); use of the action spectrum based on an electro- 

 retinogram of a dark adapted E. encrasicholus eye 

 instead of one for schooling of E. mordax; and of 

 course, the usual statistical uncertainties. Despite 

 these uncertainties and biases we believe our esti- 

 mates of schooling depth are the most accurate to 

 date thanks to the models developed by Baker and 

 Smith (1982). 



RESULTS AND DISCUSSION 



The schooling threshold based on the index of 

 dispersion was between 4.8 x 10"^^ and 7.8 x 10" '^ 

 W cm-2 (Fig. 1; Table 2). At the lower value and in 

 darkness the index of dispersion (s^/x fish per 



237 



