FISHERY BULLETIN: VOL. 79, NO. 4 



length which separated the two intervals of statis- 

 tical homogeneity for that prey. (Prey species 

 which were not present over the entire range of 

 fish sampled were not included, e.g., Apisto- 

 branchus ornatus and Sternapis fossor. ) For exam- 

 ple, the value of E for the polychaete Glycera 

 capitata for fish <37 cm was - 0.12. For fish 37 cm 

 or larger, the value of jE rose to 0.55; the difference, 

 0.67, was associated with a division point of 37 cm. 

 Although the resulting relationship (Figure 5) 

 may not be amenable to tests of statistical signifi- 

 cance, a positive trend exists. Prey which showed 

 an increase in frequency in larger fishes also 

 showed a greater increase in electivity by these 

 larger fish. 



Nonparametric ranking statistics were used to 

 test the possible effects of prey mobility, feeding 

 method, and protective structures (polychaete 

 tubes) on size-related increases in prey frequency. 

 The rank for each prey was determined by the fish 

 length at which the prey showed a significant in- 

 crease in frequency (Table 3). For each test, prey 

 were assigned to categories based on characteris- 

 tics as described in published literature (Barnes 

 1968; Smith and Carlton 1975; Jumars and 



20 25 30 35 



FISH LENGTH (cm) 



40 



Fauchald 1977) or as described by workers famil- 

 iar with local fauna (Jones et al.^"). A lack of dif- 

 ference between summed ranks for each category 

 would imply that the distributions of different 

 prey types (motile, discretely motile, or sessile, in 

 this example) were the same over all fish sizes at 

 which any prey frequency increased; e.g., a prey 

 taxon which was found at higher frequency in fish 

 >21 cm was just as likely to be motile or sessile as 

 one which was found at higher frequencies in fish 

 35 cm or longer. 



The results of these nonparametric tests (Table 

 6) show that no one motility type predominated in 

 prey taxa frequent in either small or large fish 

 sizes. Feeding locale, a possible indicator of expo- 

 sure of a taxon to predation at the sediment sur- 

 face, was not significant in explaining size-related 

 variations in prey frequency. Feeding type, which 

 is related to both degree of exposure and motility, 

 also showed no trend when ranked over fish 

 lengths. The only statistically significant rela- 

 tionship appeared when ranks for 11 tubed 

 polychaetes were compared with ranks for 7 un- 

 tubed polychaetes. Tubed taxa generally had a 

 higher rank than untubed taxa. This implied that 

 large fish selected tubed more often than untubed 

 polychaetes. However, three of the four polychaete 

 taxa which decreased in frequency in larger fish 

 (Table 3) have tubes. Thus, the presence or absence 

 of tubes in polychaetes did not always appear to be 

 an important criterion for variation in prey fre- 

 quency with fish size. 



H. Jones, G. Bilyard, and K. Jefferts, School of Oceanog- 

 raphy, Oregon State University, Corvallis, OR 97331, pers. com- 

 mun. March 1978. 



Table 6. — Results of nonparametric tests of effects of prey 

 characteristics on fish size related increases of prey frequency. 

 Prey are ranked by fish size at which prey frequency increased. 



Test and categories 



Value of 

 test statistic 



Critical value 

 for significance' 



Prey motility: 



Motile 



Discretely motile 



Sessile 

 Feeding mechanism: 



Tentaculate 



Burrowing (deposit 

 feeder) 



Carnivorous (raptorial) 

 Protective structures: 



Tubed polychaetes 



Untubed polychaetes 

 Feeding locale: 



Surface 



Subsurface 



2H = 0.029 

 NS 



2H = 3.508 

 NS 



^Us = 58 



3Us = 57 

 NS 



■^ = 5.991 

 df = 2 



X^ = 5.991 

 df = 2 



Us = 58 



ni = 11,n2 = 7 



Us -- 72 



n, = 11.02 = 9 



Figure 5.— Increase in the Ivlev electivity of prey vs. length of 

 fish at which frequency of that prey increased significantly 

 Numbers designate prey taxa (see Table 3). 



'Significant at 95% confidence level. 



^H: result from Kruskal-Wallis test (3 categories) (Sokal and Rohlf 1969). 

 ^Us result from Wilcoxson two sample test (2 categories) (Sokal and 

 Rohlf 1969). 



758 



