PICKERING, Trevor A.; ADOLPH, Stephen C. ; Harvey Mudd College ; Harvey Mudd College : Intra-individual variation and the estimation of maximal sprint performance

Researchers often quantify sprint speed of animals in the laboratory as a measure of whole-animal performance. Maximum speed of each individual is usually estimated by taking the fastest of a small number of trials (typically 2-5) per individual. We investigated the statistical consequences of estimating maximal speed as a function of the number of trials per individual using the western fence lizard (Sceloporus occidentalis). We ran each lizard (N = 21) 20 times at each of two temperatures, 20°C and 35°C. Sprint speed did not increase or decrease over time, indicating no training or fatigue effects. About 50% of the overall variation in speed at each temperature was due to within-individual (among-trial) variation. We used these data to perform statistical sampling experiments in which we randomly drew a sample of a given size from each individual�s set of 20 runs, and took the highest value from each individual�s sample as its maximal speed. The estimated maximum speed increased as a function of the number of trials per individual; for example, 5 trials per individual yielded an estimate of maximal speed that was 90% of the true maximum, on average. Sprint speeds at 20°C and 35°C (both maximum and mean) were positively correlated among individuals. The number of trials influenced the estimated magnitude of this correlation: for example, 5 trials per individual yielded a correlation coefficient that was 90% of the true value, on average. Therefore, intra-individual variation in performance leads to an underestimation of both maximal speed and the correlation between maximal speeds at different temperatures. The magnitude of this bias decreases as the number of trials per individual increases. It is likely that the magnitude of this bias also decreases as the amount of intra-individual variation (relative to inter-individual variation) decreases.