FISHERY BULLETIN: VOL. 79, NO. 4 



Table 2. — Parameters used in the calculation of maximum likelihood values of the population mean dadt (/u) and 

 variance i a'^) for northern anchovy larvae responses to clown fish predators, using the method of Cohen 1 1961). Logarithm- 

 transformed data for da/dt were used in these calculations. A taken from tables in Cohen 1 1961); /? = (N -n)'N,y =S'^'iX 



Xo) 



'M 



X + y )X - Xo), and a'^ 



y (X -Xn|2. 



DISCUSSION 



These experiments were performed to examine 

 responses of northern anchovy larvae to attacks 

 by biting (i.e., nonfi Iter- feeding) fish planktivores 

 such as juvenile fish preying on zooplankters. The 

 nature of the predator is relatively unimportant. 

 Inexperienced larvae are likely to respond to an 

 attack in the same way because in all cases failure 

 to do so is terminal. The other alternative is that 

 prey have specific predator images to which they 

 respond, which seems improbable. 



One objective was to evaluate an experimental 

 technique that could be easily repeated and 

 applied further (e.g., to problems of larval vul- 

 nerability and starvation interactions, effects of 

 alternate prey species, prey density, and other 

 questions affecting larval predation mortality). 

 For this reason it was considered important ini- 

 tially to use a predator that was readily available 

 at a uniform size. Natural fish predators could not 

 be obtained in good condition and in a reasonable 

 size range for the period of the experiment. Clown 

 fish were found to be an excellent substitute. 



However, experiments were performed at 20° C 

 which is at the upper extreme of the temperature 

 range of northern anchovy, but this probably had a 

 small effect on the observations. The energy re- 

 quired for the short periods of high level activity in 

 an escape swimming burst is undoubtedly derived 

 mainly from anaerobic metabolism, which is 

 known to be less dependent on temperature than 

 aerobic metabolism (see, e.g., Bennett 1980). Ex- 

 perimental data on burst swimming performance 

 are only available for rainbow trout and these data 

 confirm the small effect of temperature in the 

 range of 10° to 25° C (Webb 1978). 



Two primary measures of response to attack 

 were obtained, the proportion of larvae showing 



a startle response and the apparent looming thresh- 

 old for that response. The proportion of larvae re- 

 sponding to attack increased linearly with length 

 (Figure 1). This result is surprising. Kimmel 

 ( 1972 ) has shown that several days are required for 

 the Mauthner cell, which initiates the startle re- 

 sponse, to extend caudally. The eyes are not func- 

 tional before first feeding, so they could not 

 initiate a startle response (O'Connell 1981). 

 Neuromasts are present from hatching but if they 

 were involved in stimulating a startle response, 

 very high percentages of early larvae should show 

 responses to attack. These developmental events 

 suggest a rapid increase in percent response to 

 attack would be expected in the yolk-sac stages, 

 approaching maximum response levels at about 

 first feeding, as found for responses to electric 

 shock stimuli (Webb and Corolla 1981). 



However, development rates are likely to vary 

 widely in a population of larvae. It is for this 

 reason that food is first added on the second 

 day after hatching in spite of the fact that the 

 mean time to first feeding is about 4 d (Hunter 

 1976). The percentage of yolk-sac larvae respond- 

 ing to attack was quite small and was probably 

 due to early maturation (most likely of the visual 

 system) of the most rapidly developing individuals. 



The ALT changed with larval size, decreasing 

 rapidly as larvae grew in length. This decline 

 could have been due to improved acuity in the 

 visual system and/or maturation of neural path- 

 ways processing visual information. This later 

 could include shorter response latencies with age. 

 The changes in ALT may be the basis for the in- 

 creasing proportion of avoidance responses in 

 larger larvae. 



The only other measurement of looming 

 thresholds for fish are those of Dill (1974a, b) for 

 postlarval zebra danio, Brachydanio rerio, in re- 



732 



