CLARKE: DIETS OF FOURTEEN SPECIES OF MESOPELAGIC FISHES 



ments of digested prey among the remainder of the 

 stomach contents were also recorded. 



Prey items in the mouth were discarded, but 

 items in the esophagus were included with the 

 stomach contents. The bodies of items in the 

 esophagus were compressed and the appendages 

 were flattened against the body. Such items could 

 conceivably have been eaten in the trawl, but sev- 

 eral lines of evidence indicate that this is an un- 

 important source of error. Hopkins and Baird 

 (1975) reported no evidence of net feeding even 

 when a fine mesh cod end (which would presuma- 

 bly accumulate more zooplankton and restrict 

 water flow) was used. Only a few of the species 

 considered here had items in the esophagus at all 

 frequently, and in all cases such items were the 

 same or very similar to items frequently found 

 among digested or partially digested matter in the 

 stomach. Thus if there was significant net feeding, 

 only some species did so and apparently selected 

 prey from that in the cod end similar to their 

 normal habits. 



The species-size groups for which data are pre- 

 sented here are those from which a reasonable 

 number of intact prey were recorded. If sufficient 

 numbers of specimens were available, I examined 

 specimens until about 100 intact items were re- 

 corded. For other groups, I examined all the fish 

 collected, but eliminated from consideration those 

 for which too few prey items were recorded either 

 because of low numbers of specimens or low inci- 

 dence of prey in the stomach. 



Zooplankton from the bongo net samples were 

 identified and counted from aliquots taken with a 

 plankton splitter. Euphausiids and most adult 

 copepods were identified to species — the former 

 from between all and one-eighth of the sample and 

 the latter from one-sixteenth to one-thirtysecond. 

 Most immature copepods were identified to genus. 

 Ostracods and amphipods from one-sixteenth to 

 one-thirtysecond of the sample were counted and 

 measured to the nearest 0.1 mm. Other taxa were 

 counted from all to one-eighth of the sample. 

 Flowmeters on the plankton nets gave suspect 

 readings; consequently, volumes sampled by each 

 tow were calculated from the duration of the open 

 part of the tow and estimated speed (1 m/s). The 

 densities (per cubic meter) of the different prey 

 types were calculated from the volumes and ad- 

 justed counts. 



The apparent search volume per fish ( ASV) was 

 used as an index of relative preference for the 

 different prey types. For each type of prey noted 



from the stomachs of each category of fish, the 

 ratio of the total number of intact items to the 

 density of that type was divided by the number of 

 fish with intact items in the stomach. Fish 

 examined but with no intact prey items in the 

 stomach were eliminated because they provided 

 no information on preference, they included fish 

 that had not fed at all as well as those with vari- 

 able amounts of digested material in the stomach, 

 and finally their proportion of the total fish 

 examined varied between categories. Thus the 

 ASV's as calculated here apply to fish that had fed 

 recently before capture and take no account of 

 between-category differences in feeding success. 



The ASV is the minimum volume the average 

 fish of each category had to search to capture the 

 observed number of a given prey type. The actual 

 volume searched is larger to the extent that the 

 fish are not 100% effective in detecting, capturing, 

 and ingesting prey. If the fish were equally effec- 

 tive in detecting, capturing, and ingesting all 

 types of prey, the ASV's would be equal. For a given 

 category offish, differences in ASV's between prey 

 types indicate the degree to which the fish were 

 "biased samplers" of the available prey and thus 

 measure relative preference in the broadest sense, 

 i.e., without specifying which aspects of predation 

 were biased. 



The ASV is similar to the index of preference 

 recently derived by Chesson (1978); the relation- 

 ship between the two indices is: 



V, 



OCi = 





where for type i out of m prey types, V", is the ASV 

 and a^ is Chesson's index. Unlike ASV « has no 

 dimensions and is normalized. Assuming that 

 predation does not substantially alter prey densi- 

 ties, i.e., that the number of prey eaten is low 

 relative to the total available, both indices are 

 equivalently related to the probability of a given 

 type of prey being eaten: 



where P, is the probability of prey type i being 

 eaten, andp^ and n, are the density and number of 



621 



