Sources of behavioral variance: Implications for sample size decisions

Results from an intensive study of the sources of variance in a variety of chimpanzee behaviors are used to elucidate research design problems, and criteria for sample size selection are outlined. These results indicate that: (1) groups should be stratified (subdivided) only when the stratification factor is known to be strongly correlated with the behaviors of interest; (2) given a behavioral observation system with high reliability and well-trained observers, little will be gained from assigning more than one observer to each observation; (3) taking multiple observations per subject per time period substantially strengthens a research design, so that even very small sample sizes can yield generalizable results; and (4) observations should be spread as uniformly over the time period as possible, particularly for behaviors which change over time.