112 Timothy S. McCay, Joshua Laerm, M. Alex Menzel 



and William M. Ford 



cherry {Prunus serotina), and birch (Betula spp.) in the canopy, sparse woody 

 vegetation below the canopy, and lush herbaceous vegetation. Northern hard- 

 wood forests were dominated by black oak (Quercus velutina), northern red oak 

 (Q. rubra), yellow birch (B. luted), and black cherry in the canopy. Rhododen- 

 dron {Rhododendron maximum) and mountain laurel (Kalmia latifolia) were 

 common shrubs, and the composition and density of herbaceous vegetation was 

 variable. 



COMPARISON OF TRAPPING METHODS 



At each of 12 plots we established one selective transect and one linear 

 transect of pitfall traps in July 1994. Both transects consisted of 20 traps placed 

 at 5-m intervals, were approximately parallel, and were separated by 50 m. Pit- 

 falls in selective transects were placed along logs, rocks, and stumps where our 

 previous experience had indicated that chances for shrew capture might be good. 

 Traps in linear transects were placed without regard to microhabitat conditions. 

 Pitfalls were tapered plastic cups (11-cm lip diameter and 14-cm depth) partial- 

 ly filled with preservative and set flush with the ground surface. 



In August 1995, we constructed a series of five Y-shaped drift-fence 

 arrays at each of four plots randomly chosen from among the 12 original plots. 

 Each array consisted of three, 3-m "arms" of 36-cm-wide aluminum flashing 

 radiating from the center of the array. Arms were set at 120° angles, and flash- 

 ing was buried to 3 cm to prevent mammals from burrowing under the fences 

 (Handley and Varn 1994, Kirkland and Sheppard 1994). Nine pitfall traps were 

 set in association with each array, such that three were placed in the middle, and 

 two at the ends of each of the three arms. The five arrays were set in a line 

 approximately parallel to, and 50 m from, the previously established transects at 

 these plots. Individual arrays were spaced 25 m apart, so that the length of the 

 array series was equal to the length of the transects (100 m). 



We operated the two types of transects at 12 plots from 9 to 23 July 

 1994 for a total of 3,360 trapnights (TN) per method. We operated all three 

 methods at four plots from 4 to 1 1 August, and again from 1 8 November to 2 

 December 1995. Trapping effort was equal at the two types of transects (2,240 

 TN), but greater at the arrays (4,040 TN). Because pitfalls associated with an 

 array are interdependent, it is not meaningful to compare sampling effort 

 between transects and arrays. Thus, we used methods of analysis that were not 

 influenced by differences in sampling effort. All specimens were identified to 

 species and accessioned into the collections of the University of Georgia Muse- 

 um of Natural History. 



The distributions of capture frequencies using each survey method were 

 compared using likelihood-ratio tests of independence (Agresti 1990). Rejection 

 of the null hypothesis of independence indicated that the methods produced dif- 

 ferent distributions of capture frequencies, and thus different perceptions of 



