FISHERY BULLETIN: VOL. 79, NO. 4 



< 



o 

 a: 

 u. 



a. 

 < 

 o 



UJ 



UJ 



IT. 



< 



Z 



o 



(- 

 q: 

 o 



Q. 



o 

 tr 



Q- 



r 120 



00 2 04 06 08 1 



PROPORTION OF LARVAE RESPONDING TO ATTACK 



Figure 2. — The relationship between the proportion of northern 

 anchovy larvae responding to an attack and the proportion es- 

 caping an attack by clown fish. 



by the clown fish w^as not correlated with larval 

 size. 



The proportion of larvae attempting to escape 

 was therefore related to the proportion escaping 

 (Figure 2). Using functional regression analysis 

 (see Ricker 1979), the relationship was found to be 

 linear, and gave a value of 30% of larvae attempt- 

 ing to escape being captured. This is within the 

 expected range obtained from observations on 

 each group of larvae. 



The startle response included a turn. The angle 

 of the prey escape path relative to the predator 

 strike path showed no particular relationship with 

 larval length (Figure 3A), but this was primarily 

 due to the large angles measured for the 0.391, 

 0.399, and 0.421 cm larvae. Without these data, 

 the escape angle would have increased with larval 

 length, implying larger larvae were better at 

 avoiding the predator's strike path. However, 

 larvae in the three length classes with large 

 angles did not differ from others in any obvious 

 way, and therefore the data cannot be rejected. 



However, the distance traveled in an escape 

 swimming burst increased with larval total length 

 (Figure 3B). Using regression analysis (type II) 

 the relationship was best described by a power 

 function, but r^ values were relatively low. Mean 

 escape speeds in an escape swimming burst also 

 increased with larval size (Figure 4). These speeds 

 were normally about half the mean burst speeds 



IS 



z 

 < 



I 

 I- 

 < 

 a. 



a. 



< 

 o 



80 



40 



UJ 



> 



< 

 q: 



o 



z 

 < 



H 

 If) 



Q 



UJ 

 Ol 



< 



o 



4 - 



I 



— 

 8 - 



i 



J L 



J L 



B 



Escape distance traveled = 2 96L°^^ 



(r2--a599 ; N=II3) 



_L 



^ 00 02 04 06 08 10 I 2 

 TOTAL LENGTH (cm) 



Figure 3. — The relationship between (A) the angle of the prey 

 escape path to the predator's strike path and (B) escape distance 

 traveled both as functions of total length of northern anchovy 

 larvae attacked by clown fish. Vertical bars show ±2 SE. 



obtained in a forced swimming burst (Webb and 

 Corolla 1981), but escape speeds comparable with 

 the forced maximum were seen during chases. 



Analysis of ALT's for larval escape responses is 

 somewhat complex. These ALT's were assumed to 

 be distributed in the population like any other 

 character. However, not all larvae responded. 

 Since nonrespondants were eaten before respond- 

 ing they were assumed to have higher ALT's than 

 larvae showing escape responses. Therefore, the 

 true distribution of ALT's in the population is not 

 known as values above some observed level are 

 missing. Such a population is said to be censored, 

 and since values are missing from one end of the 

 distribution only, it is said to be singly censored. 

 Cohen (1961) has described methods to calculate 

 the maximum likelihood mean and variance of 

 such censored samples. However, before this can be 

 done, the nature of the ALT distribution must be 

 considered. 



The nature of the distribution of ALT's in the 

 population was evaluated using probit analysis, 

 developed to examine related problems in toxicol- 

 ogy (Sprague 1969). ALT values were ranked and 



730 



