Survey of Shrews 113 



shrew assemblage structure. We also partitioned the data table involving all 

 three methods into several independent, four-fold (2-by-2) tables (Lancaster 

 1949) to better determine patterns of dependence. For example, capture rates of 

 Sorex spp. (both species combined) and Blarina brevicauda were compared 

 between transects (both types combined) and arrays. For each four-fold table, 

 we calculated the corresponding odds ratio and tested the hypothesis that the 

 odds ratio was equal to unity (Agresti 1990). 



MICROHABITAT ANALYSES 



In July 1994, we measured several microhabitat variables surrounding 

 each of the 240 pitfall traps of the linear-transects. Because these traps were 

 placed without regard to microhabitat conditions, surveys provided an unbiased 

 sample of conditions at the forest floor and could be compared to capture fre- 

 quencies of each species at those locations. Only traps associated with linear 

 transects were considered in this analysis. 



At each trap station, we established a circular plot with a 2.5-m radius. 

 Within each plot we measured the diameter and length of all coarse woody debris 

 greater than 4 cm in diameter. We also measured the greatest length and width 

 of all rocky debris, and the diameter at the forest floor of all stumps within each 

 plot. These measurements yielded an index to the abundance of fallen logs, 

 rocks, and stumps surrounding each pitfall trap. We also measured the distance 

 from the pitfall trap to the nearest fallen log, rock, or stump. 



Microhabitat measurements were compared to shrew capture frequen- 

 cies using Pearson product-moment correlations. We regressed capture frequen- 

 cy of each species against distance to nearest structure (Neter et al. 1990). 



RESULTS 

 METHOD COMPARISON 



In 3,360 trapnights (TN) at selective transects in 1994 we collected 358 

 individuals representing four species (Table 1). In 3,360 TN at linear transects 

 we collected 126 individuals from the same four species. Sorex cinereus was the 

 most commonly captured shrew, followed by S. fumeus, Blarina brevicauda, and 

 S. hoyi. Sorex hoyi was uncommon at our sites and, therefore, was omitted from 

 all statistical analyses. We captured 2.8 times as many individuals in selective as 

 in linear traps, and this ratio was relatively constant among species (Table 1). 

 Consequently, the distribution of shrew captures (relative abundance of each 

 species) did not differ between these two methods (G 2 = 0.722; P = 0.697; df = 

 2). 



