JAHN ET AL : FOOD-SEEKING LARVAL WHITE CROAKER 



was first seived through 330 p.m mesh, took about 

 10 minutes to obtain; the portion retained on the 

 330 |jim mesh was added to the contents of the 

 large plankton net. All pump samples were pre- 

 served in 5% formalin. 



Laboratory 



All fish larvae and eggs were sorted from the 

 large zooplankton samples and identified. All 

 specimens of white croaker, the only species 

 abundant in all six collections, were measured 

 with an eyepiece micrometer in units of 0.024, 

 0.062, or 0.159 mm, depending on magnification. 

 Length was measured from tip of snout to end of 

 straight (NL) or fiexed (FL) notochord or to the 

 end of the hypural plate when this margin was 

 vertical (SL). A further designation of de- 

 velopmental stage indicated the amount of yolk 

 present: "free embryos" (Balon 1975) had a 

 relatively massive yolk sac and may or may not 

 have had functional eyes and mouths; more 

 advanced individuals with a much-reduced or 

 totally resorbed yolk sac, fully pigmented eyes, 

 and an apparently functional mouth were 

 designated "feeding-stage" larvae, or simply 

 "larvae". 



All larvae, plus a maximum of 20 free embryos 

 with apparently functional mouths from each col- 

 lection, were dissected for gut contents analysis 

 by methods described in Arthur (1976) and 

 Gadomski and Boehlert (1984). Length, rather 

 than width, of prey items was measured, because 

 it was considered a more conservative property of 

 often crushed specimens and because our concern 

 was not so much with what the larvae could eat 

 (Hunter 1981) as with what they did eat. Lengths 

 of prey items (of copepods, cephalothorax length) 

 were recorded in 50 fxm classes up to 200 iJim, by 

 100 ji-rn classes from 200 \x.m to 1 mm, and by 0.5 

 mm classes at larger sizes. In a few cases, these 

 size categories were inconvenient, and more in- 

 clusive ranges were used. 



Water bottle samples of phytoplankton and mi- 

 crozooplankton were prepared following proce- 

 dures in Utermohl (1931). From a thoroughly 

 agitated sample, a 50 mL subsample for net 

 phytoplankton was taken and placed in a settling 

 chamber overnight (about 14—18 hours). Cells 

 were identified and counted in 10 ocular fields, 

 and mean density (cells per liter) calculated as 

 the number counted scaled by the proportion of 

 the area of the 10 fields (20.6 mm^ total) to the 

 area of the slide (510.7 mm^). 



Microzooplankton was filtered from a 500 mL 

 subsample onto a 35 fxm mesh screen, washed 

 from the screen into a 50 mL settling tube and 

 allowed to settle overnight. All organisms >50 

 fxm were counted and identified to taxon and size 

 category, using the same system as for larval fish 

 gut contents. Densities were scaled to number per 

 liter. 



The 100 ^jLm zooplankton samples were concen- 

 trated to 200 mL, then subsampled twice using a 

 10 mL Stempel pipette. Organisms were identi- 

 fied and classified to size categories as described 

 above for larval fish prey. Counts from two sub- 

 samples were averaged and expressed as number 

 per m"^. 



Data Analysis 



The microzooplankton (from water bottles) 

 data set consisted of six vertical profiles of four 

 sampling heights each. Principal components 

 analysis was used to look for vertical layering and 

 time-correlated changes in the makeup of these 

 assemblages. A list of taxa present in three or 

 more samples from at least one sampling height 

 was chosen. Abundances were log-transformed 

 [logio(jc + D], and principal components com- 

 puted from the covariance matrix. Component 

 scores for each of the 24 samples were used to 

 make plots in which two- and three-dimensional 

 groupings were sought that could be clearly re- 

 lated to sampling height or to the sequence in 

 which the samples were taken. The taxa having 

 high loadings on axes (components) identified 

 with time and vertical trends were subsequently 

 scrutinized individually. A similar analysis was 

 done for phytoplankton, but omitted here in the 

 interest of brevity. 



Gut contents were conveniently analyzed by 

 lumping taxa into the 10 categories: dinoflagel- 

 late, tintinnid, rotifer, polychaete larva, lamelli- 

 branch larva, crustacean nauplius, copepodite 

 and adult copepod, amphipod, invertebrate egg, 

 and "other". Unidentifiable matter was ignored 

 in all comparisons. To test for differences in diet 

 between subsets of larvae, we used an adaptation 

 of the "bootstrap" (Efron 1982). The test criterion 

 was the percentage of prey comprised by a major 

 item in one of the two groups of guts. The null 

 hypothesis that two sets were not different was 

 simulated by combining the two data sets and 

 then, through repeated sampling, determining 

 the probability of observing the criterion percent- 

 age from such a mixture. 



253 



