KOHMATSU— DIET OF LARVAL SALAMANDERS 



65 



linear regression of HW on SVL were 

 examined. The HW of larvae was signifi- 

 cantly correlated with SVL (r,=0.899, 

 p<0.001: for log,,, transformed data). Each 

 larva was then classified as either large- or 

 small-headed according to the residuals of 

 HW on SVL (plus or minus value, respec- 

 tively), although there was no significant 

 difference in SVL or HW between large- 

 and small-headed larvae (tested a posteri- 

 ori: SVL, t42=0.149, p>0.882; HW, 

 t42=l-567, p>0.125, both for log,o trans- 

 formed data). Wet mass of total diets of 

 larvae was compared between large- and 

 small-headed groups using Student's t- 

 test. The proportional contribution of 

 each prey category to the total stomach 

 contents was determined for each head 

 shape group. For this analysis, 

 terrestrial invertebrates were pooled 

 together irrespective of their taxonomic 

 allocations due to the small sample size for 

 each invertebrate order. Overall diet 

 composition by wet mass was compared 

 between the two groups using a MANOVA 

 with individuals as replicates. Subse- 

 quently, the mean proportion of each 

 dietary category was separately analyzed by 

 ANOVA. In order to standardize vari- 

 ances for all statistical tests, all absolute 

 values and proportions were log^ trans- 

 formed and arcsine-squarc-root trans- 

 formed, respectively. 



Results 



Total wet mass of large-headed larvae 

 was significantly larger than that of small- 

 headed larvae (t=3.77, p<0.001). One of 

 the small-headed larvae had an empty 

 stomach. The overall dietary composition 

 of large- and small-headed larvae differed 

 significantly (Wilks' lambda=0.51, 

 F=5.05, p<0.001 by MANOVA; Table I). 

 A large proportion of the diet of the large- 

 headed larvae comprised terrestrial inverte- 

 brates, whereas for the small-headed 

 larvae, detritus (fine litter remains) and 

 Diptera larvae were the main stomach 

 contents, with terrestrial invertebrates com- 

 posing no more than 2% of the diet (Fig. 

 2). Separate ANOVAs showed significant 

 differences in the percentage of terrestrial 

 invertebrates and Diptera larvae between 

 the two groups of larvae, but not in the 

 remaining dietary categories (Tabic 1 ). 



Discussion 



The results of the stomach cimucih 

 analysis showed thai ilic prc\ iicius of 

 larval H. retcirdalus vary in accordancc- 

 vvith variation in head morphology (Fig. 2). 

 The large-headed larvae ate larger prey 

 items than the small-headed lar\ae (.Appen- 

 dix). In larxal H. reiardatus, possession of 

 a wide head and nuniili is advantageous in 

 that it permits ilic larva to feed on larger 

 prey (OlKlaclu, 1994). Nevertheless, the 



Mil t 



Results t)!' ANOV.As {o\ diet projUM t ion ol llviiohiiis relardaliis. 



So, II i. c 





\1N 



1 



r* 



Tcncstrial unci Ichralcs 



1. 42 



1.2.S65 



17.0.^6 



0.(XX)2 



DipU'i a hii \ ac 



1, 42 



(1.932 



0.9077 



0.0044 



1 plk'UK-i iiplci a 



1, 42 



0.1)057 



0.9112 



0.3453 



I 1 iihupi CI a 



1. 42 



0.01 



1.40195 



0.2(X>4 



( opcpoJa ( ( \ I lopoul.i ) 



1, 42 



0.0(X)5 



0.(X)7I 



0.9334 



SalaiiuiikU-i 



1. 42 



0.026 



0.5307 



0.0925 



Deti nils 



1. 42 



0.964S 



6.58 



0.014 



♦ Raiiiccil alpha lcvcl=0.007 (i.e., 0.005/7) 



