In order to construct the polygons mentioned above, it v.-as nGcossary 

 to compute the average depth for each haul. I-'or tho 19.39 hauls, when the 

 stray angle ivas measured tv/o or throe times durllis a haul, the angles were 

 simply averaged and the depth computed from the average angle and the length 

 of the towing wire. In 19Ul, when angles yfera measured- each minute during 

 a haul, it v;as possible to construct a plot of the course of each haul, 

 (fig, 1). Average depths were computed from the plots by measuring the 

 area bounded by each one and its baseline, ;OTd dividing by the Length of 

 th3 baseline. 



EFFECT OF TEIffERATUIU! 



Inspection of the vertical profiles for relative numbers of eggs and 

 larvae, togetjier with the cori*esponding temperature profiles (fig. 2) in- 

 dicates a possible relationship betvrcen temperature and vertical distribu- 

 tion. Such a relationship could arise as the result of the seeking of an 

 optimvmi temperature by the larvae, and by the adult fish which lay the eggs. 

 Since pelagic fish eggs tend to remain in v/ater of the same specific gravity 

 as that in which they were fertilized (V/alford, 1939), the vertical distri- 

 butioft of eggs shoiild tend to reflect that of the parent fish at; th^ time 

 of spavrtiing. This relationship is complicated by vertical turbulence above 

 the thermocline, v;hich tends to disperse passive bodies like fish eggs. 



T'^atever the nature of the relationship betv/een temperature and the 

 distribution of pilchard eggs, the larvae may be expected, to respond directly 

 to changes in temperature, since they are capable of locomotion., Again 

 the relationship- is probably complicated by other factors, such as the amount 

 of fodder organisms present. 



To measure the correlation between temperature and concentrations of 

 eggs and larvae, the relative numbers v/ore plotted as regressions on degrees 

 centigrade (fig. 3)» Hauls above a depth of 10 meters were omitted because 

 of the reversals in egg and larvae profiles which often occurred in that 

 layer. Also excluded ;vere zero hauls below the first. Inclusion of either 

 of these two categories of hauls v.'ould obscure the decrease from the maximum 

 concentrations dovm to zero, v;hich is the point of chief interest in the 

 regressions in so far as thoy arc to be used in determining the depth of 

 net hauls for r^jgular surveys. Correlation coefficients v;-erc .590 for eggs, 

 .55? for small larvae, and .5U8 for large larvae, corresponding to proba- 

 bilities of .010, .OUl, and .02U of chance occurrence. Combining these 

 probabilities by the method of Fisher (1936, p. 105). indicates, that the 

 result, as a whole, is highly significent (P = ,0009). 



Of special interest in the foregoing correlations are the intersections 

 of the computed regression lines v;ith their baselines, since those should 

 indicate the temperature belovf T;hich wc vrould not expect to find eggs or 

 larvae. The indicated lainimura temperatures are 10.5 for eggs, lii.l;° for 

 small larvae, and 12.0° for large larvae. Since thxa regressions for large 

 and small larvae v/ere found by Fisher's (1936 p. 1U6) "t" test not to differ 

 significantly in slope, they were combined, giving an indicated minimum of 

 12.0° for both size categories. 



184 



