254 



Fishery Bulletin 94(2). 1996 



Number ol maior prey types 



Figure 3 



Frequency distributions for the major prey types 

 (five, including unidentifiable category) present 

 in bagged (above horizontal axis) and unbagged 

 (below axis) juvenile pink snapper, Pristi- 

 mopoides filamentosus, collected on nine dates 

 during February-August 1994. 



of body size was not expected. It is possible that the 

 date effect not related to size was caused by increases 

 in the abundance or availability of prey between win- 

 ter and summer. 



In comparisons of prey volumes between bagged 

 and unbagged specimens, the effects of body size and 

 date of collection clearly need to be recognized, and 

 adjustments made. Although this would seem intu- 

 itively necessary if sampling design does not control 

 for fish body size and collection date, these factors 

 influenced our results despite the fact that our sam- 

 pling design specifically controlled for them. The 

 possibility that the effects of body size and collection 

 date may be too large to adjust statistically after- 

 the-fact (if left uncontrolled by basic sampling de- 

 sign) should be taken into account in future studies. 



Artifacts of regurgitation 



Some food studies of physoclistous marine fishes (e.g. 

 Bowman, 1986) have indicated or have strongly sug- 

 gested that ingested prey are lost due to stomach 

 eversion resulting from expansion of swimbladders 



nagged 

 Unbagged 







5 10 15 20 25 30 35 40 45 65 70 75 80 85 90 95 

 Longest axis of largest prey Item (mm) 



Figure 4 



Frequency distributions of maximum prey sizes (all types, 

 including unidentifiable category) for bagged and unbagged 

 juvenile pink snapper. Pristimopoides filamentosus, collected 

 on nine dates during February-August 1994. 



when specimens are retrieved from even moderate 

 (50 m) depths. This appears especially true for 

 piscivores and other predators of large prey that have 

 simple, undifferentiated stomachs (Treasurer, 1988). 

 Undetected regurgitation can lower the precision 

 of estimates by adding erroneously empty or low-vol- 

 ume samples to the data set. There are also at least 

 three ways in which undetected loss of stomach con- 

 tents can bias dietary studies. These include 1 ) un- 

 derestimates of the amount of food ingested, 2) un- 

 derestimates of diet composition, including the taxo- 

 nomic or functional variety of prey, and 3) size-selec- 

 tive regurgitation of prey. The latter can either in- 

 flate or reduce the effect of 1) depending on the na- 

 ture of the size bias, which in turn may influence 

 the precision of the estimates. 



Precision of estimates 



Total loss of stomach contents due to regurgitation 

 occurred infrequently (4 out of 92 fish). However, the 

 large reduction in prey volume for unbagged fish in- 

 dicates that partial loss of prey because of regurgi- 

 tation was a frequent occurrence. Partial prey loss 

 did not inflate the variances of prey volume estimates 

 for unbagged snapper; CVs (mean/SD100% ) of the 

 date-specific estimates of prey volume were equiva- 

 lent for bagged and unbagged fish (67% and 62%, 

 respectively). Perhaps this reflects an effect of bias — 

 selective loss of large prey — offsetting a decrease in 



