108 



Fishei-y Bulletin 100(1) 



performed a cluster analysis (Ward's minimum variance 

 method) on the volumetric contribution of food items per 

 lobster size class, after standardization. A sudden increase 

 in the joining distance of the clustering sequence repre- 

 sented by the dendrogram represents a natural cutting 

 point for the determination of meaningful clusters (SAS 

 Institute. 1995). In addition, a factor analysis (VARIMAX 

 rotation of the first three principal components) was per- 

 formed on the correlation matiix of the volumetric contri- 

 bution of food items for each 5-mm-CL size class of lob- 

 sters. Cluster and factor analyses were done with JMP 

 statistical software (SAS Institute, 1995). 



Relationships between volumetric contribution and lob- 

 ster CL were described by least squares linear regression 

 for bivalves, rock crab, and flesh. Relationships between 

 percent occurrence of bivalves and rock crab were described 

 by locally weighted (lowess) regi"ession with a SOf smootii- 

 ing factor, and by least-squares regression for flesh. 



Results 



Sample composition, stomach fullness, and 

 types of food items 



The 506 lobsters retained for analyses varied in size from 

 4.3 to 112.4 mm CL (median=35.6 mm CL). Most size 

 classes contained more than 25 lobsters (Table 1). The 

 smallest size class (2.5 to <7.5 mm CL) contained only 

 16 lobsters with a median of 7.0 mm CL; therefore we 

 refer to this group of lobsters as the 7-mm-CL size class. 

 The 21 lobsters >67.5 mm CL were pooled together into 

 a single size class, which we refer to as the 77-mni-CL 

 size class in reflection of their median CL. Females and 

 males accounted respectively for 43.2'^fi and 44.1''r of all 

 lobsters examined; the remainder were too small to deter- 

 mine sex. Lobsters were pooled for analyses irrespective of 

 sex because Weiss ( 1970) and Ennis ( 1973) concluded that 

 diet was the same for both sexes. 



Only two lobsters had empty stomachs and they be- 

 longed to the 10-mm size class. With these two empty 

 stomachs excluded, there was a highly significant relation- 

 ship between the dry mass of stomach contents and lob- 

 ster CL (Fig. 1). Identifiable food items included macroal- 

 gae or benthos that were grouped into broad taxonomic or 

 ecological categories (Table 2). No planktonic organisms 

 were identified from the stomachs, even of the smallest 

 lobsters. However, the crustacean meiofauna group includ- 

 ed the remains of very small crustaceans, some like the 

 harpacticoids and ostracods, known to be bottom-dwell- 

 ing, whereas unidentified minute crustacean lemains may 

 have originated from holo- or mero-planktonic forms or 

 from juvenile amphipods, isopods. or carideans. Sand, silt, 

 and infrequently bits of nylon rope were also found in the 

 stomachs. "Flesh" refers to tissue bolus composed of an- 

 imal soft parts that could not be attributed to a taxon, 

 generally because no distinctive part was found in the 

 stomach along with the tissue or less commonly because 

 distinctive parts from several prey types were present in 

 the stomach but none was attached to the tissue. 



Ontogenetic shifts in diet 



A cluster analysis on the volumetric contribution of food 

 items to lobsters by size class yielded four groups: 7 

 mm, 10-20 mm, 25-60 mm, and 65-77 mm CL lobsters 

 (Fig. 2). These same groups could be seen on a plot of 

 the fii'st three factors of a factor analysis of the correla- 

 tion matrix of the volumetric contribution of food items 

 (Fig. 3). The three factors explained 68. 87^ of the vari- 

 ance (39.9%, 18.2%, and 10.7% for factors 1, 2, and 3). The 

 first factor had strong loadings for crustacean meiofauna 

 (0.96), foraminiferans (0.96), bivalves (0.84), macroalgae 

 (0.82), amphipods (0.78). and rock crab (-0.71). Because 

 lobsters in the 7-mm-CL size class had little rock crab 

 in their stomachs, but relatively high proportions of the 

 other food items, they stood out with a very large score 

 (3.1) on this factor. The next two size classes, 10 and 15 

 mm CL, scored 0.8 and 0.6. respectively. All other size 

 classes scored between and -0.6 on the first factor. The 

 second factor had strong loadings for flesh (0.73), lobster 

 (-0.82), and barnacles (-0.73). Lobsters of the two largest 

 size classes (65 and 77 mm CL) had strong negative scores 

 on this factor (-2.5 and -1.5, respectively), whereas lob- 

 sters of the 10-35 mm size classes scored between 0.5 and 

 1.1. The smallest size class (7 mm CL) and size classes 

 of 40-60 mm CL had scores close to 0. Finally, the third 

 factor had a high loading for carideans (0.74) and some- 

 what smaller loadings for isopods (0.67), coralline algae 

 (-0.57), and pagurids (-0.54). This third factor separated 



