270 



Fishery Bulletin 94(2), 1996 



larvae (n=l,667) were measured to the nearest 0.1 

 mm SL. Standard length of mackerel larvae pre- 

 served in ethanol and formalin was corrected for 

 shrinkage by using the relationships between fresh 

 and preserved measurements obtained from the sub- 

 set of larvae videotaped at sea (/ -2 =0.98). The stan- 

 dard length of a random subset of the 3 other domi- 

 nant larval fish species (51 silver hakes, 467 yellow- 

 tail flounders, 200 redfishes, preserved in ethanol) 

 was measured. Mackerel lengths represent fresh 

 measurements except those used for comparison with 

 other species, in which case all lengths are ethanol- 

 preserved measurements. 



Gut content analysis 



Undamaged Atlantic mackerel larvae were selected 

 for gut content analysis. The digestive tract was sepa- 

 rated from the body and dissected under a light mi- 

 croscope. Whenever possible, prey were identified 

 and measured. The position of the prey in the trac- 

 tus (foregut, midgut, and reargut) and the state of 

 digestion of the prey (slightly, partly, and well di- 

 gested) were noted. Because crustacean prey were 

 flattened dorso-ventrally, cephalothorax length, 

 rather than width, was measured. Fish larvae were 

 particularly damaged by digestion and only a small 

 proportion of them (36 out of 710) could be identified 

 and measured directly under the light microscope. 



Indirect identification of some digested fish larvae 

 was based on discriminant analysis of their otoliths. 

 To establish the discriminant function, between 18 

 and 29 ethanol-preserved specimens of each of the 

 four dominant larval fish species sampled in July 

 1991 (Atlantic mackerel, silver hake, yellowtail floun- 

 der, and redfish), ranging in size from hatching length 

 to 10 mm, were selected from the plankton collec- 

 tions and measured. The sagittae and lapilli of each 

 fish were extracted and 8 measurements were taken 

 on each of the 4 otoliths: longest and shortest diam- 

 eter (LD and SD); longest and shortest radius (LR 

 and SR); surface and perimeter (SU and PE); and 

 longest and shortest diameter of nucleus (LN and 

 SN). The discriminant function used 15 measures and 



ratios of measures (LRg/LRj , 



LN 



su s /su L , 



PE S / 



SU S , PEg/LD, , SN S /LN L , SR S /SD, , LD S /LR, , SD S /SD, , 

 UySU L , SDg/SRg, SDg/SN L , SD L , LNg/SN L , SRg/SU L ), 

 in order of decreasing discriminatory power, where 

 the subscripts S and L refer to the sagitta and the 

 lapillus, respectively (from the same side of the fish ). 

 The discriminant function, which was validated by 

 using the known larvae, correctly classified between 

 67' i ;ind 89% of the larvae of each species (Table 1). 



The discriminant function was used to identify an 

 additional 38 (out of 710) digested fish larvae for 



Table 1 



Validation of the linear discriminant function used to iden- 

 tify fish larvae found in the gut of larval Atlantic mack- 

 erel, Scomber scombrus, sampled from 18 to 23 July 1991, 

 southwest of Sable Island (Scotian Shelf). The discrimi- 

 nant function is based on the morphometry of the otoliths 

 of known larvae (see text). Successful identification is the 

 percentage of larvae of a given species correctly identified 

 by the discriminant function. Type-1 error is the propor- 

 tion of fish larvae incorrectly assigned to another species 

 ( H u rejected when it is true). Type-2 error is the propor- 

 tion of fish larvae identified as a given species that were 

 actually of another species iH accepted when it is false). 



Silver Yellowtail Atlantic- 

 hake flounder mackerel 



Redfish 



Successful 

 identification {%) 



78 



Type-1 error (%) 22 



Type-2 error (%) 8 



Sample size 54 



1 1 



29 



:;s 



which one undamaged sagitta and one undamaged 

 lapillus (from the same side of the fish) were recov- 

 ered. The length of identified fish larvae was back- 

 calculated on the basis of the best polynomial regres- 

 sion between one of the otolith dimensions and the 

 ethanol-preserved standard length for the known 

 specimens of the species (LD S , r 2 =0.87 for Atlantic 

 mackerel; LD g , r 2 =0.93 for silver hake; SR g , r 2 =0.80 

 for yellowtail flounder; SD S , r 2 =0.75 for redfish). 



Zooplankton were identified, enumerated, and 

 measured in the RMT collections (64-//m and 333- 

 pm net samples) that contained Atlantic mackerel 

 larvae. The selectivity of Atlantic mackerel larvae 

 for a category of prey was estimated by using 

 Chesson's alChesson, 1978): 



a , 



(dj / p j) / (£di/ Pi ), for i = I,... N, 



where a is the selectivity for prey taxon j, N is the 

 number of prey taxa, d andp are the frequencies of 

 prey taxon^ in the diet and the plankton respectively, 

 and d and p t are the same frequencies for the t'th 

 prey taxon. Copepod eggs in the diet of large fish 

 larvae may have reflected the ingestion of egg-bear- 

 ing females and, accordingly, eggs were not consid- 

 ered in the calculation of selectivity indices. Follow- 

 ing Fortier and Harris ( 1989), a, was first computed 

 for each individual Atlantic mackerel larva captured 

 in the RMT and then averaged by length classes (<5, 

 5 to <6, 6 to <7, 7 to <8, >9 mm). Variations in a reflect 

 changes in the preference of the predator, a is not 



