FISHERY BULLETIN: VOL. 78. NO. 3 



prey type i available. Like a, the ASV is unaffec- 

 ted by negative or positive preference for other 

 types of prey. As pointed out by Chesson (1978), 

 most other indices of preference, including that of 

 Ivlev (1961), are so affected and their biological 

 meaning is not clear. 



Preference could be affected by many charac- 

 teristics of the prey, only one of which could 

 be considered in this study. Other things being 

 equal, large or more visible prey types could be 

 detected at greater distances (Zaret and Kerfoot 

 1975; O'Brien et al. 1976) and thus have higher 

 ASV's than small or translucent types. Con- 

 sequently, in addition to measuring size of prey, I 

 examined several samples of living zooplankton 

 from the study area and noted, for as many prey 

 types as possible, whether they were opaque or 

 translucent in life and the presence of any pig- 

 ment. 



Ability to escape once detected and attacked 

 would decrease ASV Prey with bioluminescent 

 organs could either be more readily detected than 

 those without or conceivably use them to decrease 

 probability of detection or capture. Aggregation or 

 patchiness of prey could also affect ASV either 

 way depending upon patch size, predator capacity, 

 and the search behavior of the predator. Unfortu- 

 nately, none of these behavioral aspects of preda- 

 tion could be investigated. 



For each of the fish species considered here, I 

 examined four morphological features which 

 could affect preference. Relevant measurements 

 were made to the nearest 0.1 mm with either an 

 ocular micrometer or vernier calipers on at least 

 five specimens spanning the size range of each 

 species considered. The length of the premaxillary 

 was taken as a measure of gape; the diameter of 

 the lens, as a measure of visual ability; and the 

 average space between gill rakers on the lower 

 branch of the first arch, as a measure of minimum 

 particle size that could be retained. These were 

 expressed as linear functions of standard length 

 determined by least squares regression. The filter- 

 ing area of the gill rakers, which could not be 

 directly calculated without knowledge of the angle 

 at which the arch is held during feeding, was as- 

 sumed proportional to the product of the length of 

 the raker-bearing segments of the first arch and 

 the length of the gill raker at the joint between the 

 upper and lower branches. This product or "area" 

 was expressed as a power function of standard 

 length determined by linear squares regression on 

 the logarithms. 



Aside from being affected by characteristics of 

 the fishes and their prey, ASV's could have been 

 biased by problems in the methodology. Any feed- 

 ing in the net (considered above) would tend to 

 increase ASV for large prey retained there and 

 also blur any differences in visibility or escape 

 behavior. Differential rates of digestion and disin- 

 tegration of prey would bias stomach content data 

 toward more resistant and more easily recogniza- 

 ble prey (Gannon 1976). Counting only intact and 

 measurable prey eliminated bias due to differ- 

 ential ease of identification. For example, if all 

 identifiable parts had been counted, the data 

 would have been heavily biased toward 

 Pleuromamma spp. whose spots or "buttons" can 

 be recognized even after the items have completely 

 disintegrated and passed into the intestine, while 

 certain other prey which cannot be identified posi- 

 tively if only one or two features are missing would 

 have been underrepresented. Even among the 

 crustaceans, the rate at which the prey disinte- 

 grates probably varies; Corel ova (1975) indicated 

 that some small cyclopoid copepods remain intact 

 even in the intestine of myctophids. Other types of 

 prey are probably digested much faster than crus- 

 taceans. To at least qualitatively correct for the 

 latter bias, I counted all recognizable remains of 

 chaetognaths, heteropods, other gastropods, 

 siphonophores, and tunicates as "intact" for calcu- 

 lation of ASV's. 



The densities of small zooplankton were un- 

 derestimated due to escapement through the 505 

 ^im mesh of the plankton nets used. Counts of 

 ostracods and certain copepods from an available 

 plankton tow from the study area with 333 /xm 

 mesh on one frame and 505 /um on the other indi- 

 cated that — assuming that the 333 /itm sampled 

 the small prey accurately — prey >1 mm long were 

 adequately retained by the 505 ^tm net. These 

 included most of the prey eaten by the fishes. Two 

 types of frequently eaten prey, large (0.6-0.8 mm) 

 Oncaea spp. and ostracods <1.0 mm were un- 

 derestimated by factors of roughly 4 and 5, respec- 

 tively, in the 505 /u,m sample, and their ASV's are 

 overestimated by the same factors. There were 

 insufficient numbers of other small prey types in 

 the 333/505 sample to provide even roughly reli- 

 able estimates of error. 



Any avoidance of the bongo nets by prey would 

 result in erroneously high estimates of ASV No 

 studies have documented the extent of error due to 

 avoidance by different prey types, but it can prob- 

 ably be assumed to be negligible for the great 



622 



