Fig. 4. Magnitude of pairwise correlations between speeds for two different trials i and j as a function of the number of trials separating the two (|i–j|). Correlations declined significantly with separation for both 20°C and 35°C but most of the variation was unexplained (Mantel tests; see Results).