Correlation of aggregation with seven other life-history traits of Tetrahymena thermophila. Those named "PC" are combinations of traits obtained from Principal Component Analyses; we describe their essence here but full details are given in . The degree of cell aggregation under food rich conditions showed a negative relationship with short-distance dispersal including variation within and among strains, but not in more restrictive among strain analysis (A). Elongation strategy under starvation conditions was markedly associated with the tendency to aggregate, with strains where some cells elongated far more than others (up to becoming dispersal morphs) and for a shorter time showing stronger aggregation than strains where all cells elongated similarly for a long time (B). Strains that tended to aggregate strongly were less efficient as single-cell colonizers (C). Strains with small and elongated cells under food rich growth conditions showed a higher tendency to aggregate than strains with big and round cells, this effect was only present when within-strain variation was included (D). More aggregative strains showed reduced survival and average elongation abilities under starvation conditions when within strain variation was included in analyses (E). Strains growing faster and reaching a higher final cell density in the presence of nutrients were less inclined to aggregate (F). Growth strategy (K vs r) showed no relationship with aggregation (G). Because replicates of a given strain were not linked between experiments, we used a randomization procedure  to correlate parameters from different experiments, similar to the one used by : the replicates of a given strain were randomly associated across experiments 1000 times, and a correlation was computed for each random association. n: sample size (limited by the experiment with the smaller sample size); r: mean Spearman's correlation over the 1000 random associations; s: proportion of significant correlations over the 1000 random associations; P: probability of obtaining s if the null hypothesis of no correlation is true. Points on each graph reflect the means of five random associations between the two traits to illustrate both between and within strain variations. The second line of statistics at the top of each graph gives results for Spearman correlations based on means of the 10 strains only, discarding variation between replicates of each strain.