RESULTS 



General 



Chromis aggregate in schools of 10 - 100 individuals, and usually stay 

 in the vicinity of some fixed landmark. If they are confronted with a 

 potential threat, such as a bar jack, ( Caranx ruber ) , or an unfamiliar 

 plastic model, they turn and swim toward this "home" landmark. They 

 choose the same landmark regardless of the direction of the intruder's 

 approach. If the intruder is approaching rapidly (1 ra/sec or more) an 

 observer notices rapid motion, with all fish in the school responding 

 simultaneously. If the intruder's speed is slow (1/2 m/sec or less) 

 the fish react more gradually and apparently not at the same time. 



To quantify the response, we chose reaction time and reaction distance 

 (see previous section) in preference to a subjective measure of re- 

 action strength, percent of failures to respond, or number of respond- 

 ing fish. Trials were run about 2 min. apart. A series of trials in 

 which 4 min., 2 min., 1 min., and 1/2 min. inter-trial intervals were 

 used showed no significant habituation (decrement in response with 

 decrement in interval) in the escape response. This justifies the 

 arbitrary choice of 2 min. used in all other experiments. 



Effect of Model Speed 



It was remarkable that for a given model, the reaction distance seemed to 

 be independent of the model's speed. Our subjective impression had been 

 that the reaction was considerably stronger at faster speeds. We thus 

 expected that the fish would react at a greater distance for a faster 

 model than for a slower one. Fig. 1 shows graphs of reaction distance 

 as a function of speed for various models. The lumped data for all 

 runs, all days (medium white ellipse) shows a great deal of variability, 

 but fails to show a slope significantly different from at the 90% 

 level. It is not clear from this line whether the mean distance 

 represents a characteristic of the response, or whether there is so much 

 scatter that it makes no sense to consider reaction distance as a 

 measure of the response. If the fish were reacting at a constant 

 distance, one would have the relationships 



(1) D = V X T 



(2) T = D X (1/V) 



where D is the reaction distance, T is reaction time, and V is speed 

 of the model. Therefore, if D is a constant, a plot of T versus 1/V 

 should yield a line with slope = D and intercept = 0, The data of 

 Fig. 1 are replotted in this manner in Fig. 2. Evidently there is a 

 strong tendency for an early reaction at faster speeds, since the 

 slopes of these lines are positive and not greatly different from the 

 predicted values (mean reaction distances from Fig. 1). In most 



VI-203 



