Mann-Whitney U tests with a set at 0.017 (for t12 samples) to account for multiple comparisons.

Analysis of covariance (ANCOVA) could not be used to evaluate midpoint and endpoint data for

organ sizes because the assumption that covariate values have similar distributions and ranges

for all treatment groups (Quinn and Keough 2002) was violated.

 All data were analyzed using SPSS for Windows (Release 11.0.0). Only turtles that

survived, displayed no external signs of illness, and continued to gain mass until the time of

tissue sampling were included in the analyses. Data are expressed as means + standard errors

with alpha set at 0.05 unless otherwise noted.

 Results

 OM, energy, nitrogen, and lipid contents of food samples (n = 7, collected at biweekly

intervals) were consistent throughout the experiment (Table 2-1). Although differences in energy

content of pellets among weeks approached significance, the relative difference between the

highest and lowest energy values measured was only 2.87% (for DM) and 2.73% (for OM).

 R and R-AL turtles had comparable values for all repeated measures data collected during

weeks 0 through 5. Data for these two groups were analyzed and are presented separately for this

time period to demonstrate that there were no differences between groups prior to the switch to

an ad libitum diet for R-AL turtles.

 Intake in week 0 (during which all turtles were feeding ad libitum) was similar for all three

treatment groups (ANOVA, F2,112= 0.946, p = 0.392). Repeated measures ANOVA of intake

during weeks 1 through 5 (n = 37 AL, 39 R-AL, and 39 R) and during weeks 6 through 12 (n =

17 AL, 29 R-AL, and 29 R) revealed significant time and treatment effects as well as significant

interactions between time and treatment group (Table 2-2). The pattern of AL intake was

significantly different from the patterns of R and R-AL intake during weeks 1 through 5

(Tukey's HSD post hoc test, p < 0.0001 for both comparisons), and intake patterns of all three