Bottleneck size^{a} | Step number (P value)^{b} | Fitness gain ± SEM^{c} |
---|---|---|
10 | ≥2 (0.0092) | 0.260 ± 0.066^{***} |
33 | ≥3 (0.0006) | 0.834 ± 0.047^{*} |
100 | ≥2 (≤0.0001) | 0.815 ± 0.064^{*} |
333 | ≥4 (≥0.0001) | 0.756 ± 0.037^{**} |
1,000 | 1 (≤0.0001) | 1.027 ± 0.050 |
2,500 | 2 (0.0132) | 1.000 ± 0.034 |
10,000 | 1 (0.0014) | 0.888 ± 0.083 |
1^{a} | 1 (0.0001) | -1.017 ± 0.063 |
↵^{a} Bottleneck size of 1 corresponds to the decline population (Figure 2) and is included for comparison.
↵^{b} Step number estimated from Figure 3. P value is the significance level associated with the addition of the final step to the model (Kleinbaum and Kupper 1978; Elena et al. 1996). For bottleneck sizes of 10, 33, 100, and 333, step numbers are minimal estimates because the recovery was incomplete (see footnote c).
↵^{c} Gain for each recovery population determined from data in Figure 3 as log_{10}(WF) - log_{10}(W_{0}), where W_{F} is the mean of all fitness measures after the final step, and W_{0} is the mean of all fitness measures before the first step. Thus, fitness gain is determined over all steps, and was positive because W_{F} > W_{0}. Fitness change during the decline was similarly determined, but was negative because W_{F} < W_{0}. A recovery was incomplete if the absolute value of the fitness gain was significantly less than |- 1.017 ± 0.063|, the absolute value of the fitness decrease during the decline (^{*}P < 0.05, ^{**}P < 0.01, and ^{***}P < 0.001 by a t-test; Sokal and Rohlf 1994).