Population differences in responses of red-legged frogs to introduced bullfrogs - Rana aurora

Ecology, Sept, 1997 by Joseph M. Kiesecker, Andrew R. Blaustein

We used a 2 x 2 factorial design, replicated three times, with treatments being the presence or absence of adult bullfrogs, interacting with red-legged frog tadpoles that are either syntopic or allotopic with bullfrogs. The experiment was conducted from March 22 to March 30 1995 in the south marsh at E. E. Wilson Wildlife Refuge. We used 12 rectangular, open-bottom 8 x 4 m pens (32 [m.sup.2]) constructed of 1-[mm.sup.2] mesh fiberglass screen that was pressed [approximately equal to]20 cm into the soft mud substrate. Enclosures were placed in a linear array parallel to the water's edge, in a randomized-block design. The depth of water in each enclosure ranged from [approximately equal to]10 cm to 1.5 m. Enclosures were cleared of their macrofauana (e.g., corixids, notonectids, salamanader larvae) by repeated sweeping with a net prior to experimentation. To minimize potential confounding effects of spatial gradients (e.g., temperature, depth), we blocked enclosures for random assignment of treatments.

Rana aurora tadpoles from both populations were matched for developmental stage (Gosner stage 25; Gosner 1960) and similar size (1.57 [ or -] 0.247 g; range 1.077-1.83 g). Rana catesbeiana adults were matched for size across each treatment (103.4 [ or -] 3.733 g; range 88.27-126.13 g). One R. catesbeiana adult was added to each enclosure and was given 24 h to acclimate, after which 50 R. aurora tadpoles were added. Other than the R. aurora larvae, the adult bullfrog was the only vertebrate in the enclosures. After 7 d, enclosures were checked and the percentage of R. aurora surviving in each enclosure was determined. At the end of the experiment, we also captured each bullfrog and flushed its stomach.

Statistical analyses

The number of times tadpoles crossed the center line was summed for experiments 1 and 2 for each test. We also calculated the mean number of individuals under refuge for each 30-s reading. For experiments 1 and 2, we initially tested whether there was a difference in the tadpole initial responses (first 5 min) and final responses (last 5 min), using paired t tests. However, because there was no difference between the initial and final responses for any population tested (P [greater than] 0.13 for all comparisons), the data from initial and final responses were combined for further analyses. This resulted in only one observation of activity and refuge use for each trial. We then used multivariate analysis of variance (MANOVA) to test for stimulus effects on the dependent variables, activity (mean number of line crosses) and shelter use (mean number of tadpoles under refuge) (Tabachnick and Fidell 1989). We evaluated the independent significance of activity and shelter use with the total structure coefficients from a discriminant function analysis (Reznick 1990). Total structure co-efficients are the correlations between an individual's discriminant function score and its value for the dependent variable; a significant correlation indicates that the variable contributed significantly to the discrimination among treatments.