Figure 8 Relationship between mean fitness and average cell size. (A) Data from all 12 populations and 9 time points for which both fitness and size were measured. (B) Hypothesis I: There exists a rigid functional relationship between average cell size and mean fitness, such that the latter can be improved only by increasing the former. Longitudinal (one population over all time points) and cross-sectional (all populations at one time point) regressions would yield the same slope. (C) Hypothesis II: The functional relationship between size and fitness is malleable, so that replicate populations may diverge in the relationship between size and fitness. Although longitudinal regressions (dashed lines) may be strong, cross-sectional regressions (solid lines) need not show any systematic coupling between size and fitness. (D) Dashed and solid lines show actual longitudinal and cross-sectional regressions, respectively, using the data shown in A. All regressions were performed according to model II procedures that are applicable when both variables are measured with error, but the corresponding error variances are known (Mandel, 1964). From analyses of variance performed on repeated measures, the ratio of the error variances for average cell size and mean fitness was 0.138 (adjusted for sample sizes of two and three, respectively, and averaged over all populations and generations).