Animal temperament describes behavioural differences between individuals that are consistent across time and contexts. Variation in animal temperament is rapidly gaining interest and attention within behavioural and evolutionary ecology. If we are to understand the causes and consequences of temperament variation within and between populations we need to determine the selection pressures that affect temperament in natural environments. To date, however, the vast majority of temperament studies have been carried out on captive-bred individuals. This review highlights potential problems that arise from using captive animals to elucidate the ecological and evolutionary functions of temperament in wild populations. For example, development, learning and environmental variability can all affect behaviour. Thus, both environment and gene-by environment interactions can affect the fitness functions of different temperaments, and hence selection. We stress the need for measurements of repeatability and heritability, and the importance of biological and ecological validation of temperament tests in wild animals. We describe the limited evidence from wild populations of the fitness consequences of temperament variation, and the use of intra- and inter-specific comparisons to prove adaptation. To identify multiple axes of behavioural variation, and how these interact with environments that vary spatially and temporally, we need long-term studies on wild populations – yet few studies of this nature currently exist. Finally, and perhaps counter-intuitively, we suggest that there is much to be gained from incorporating some of the approaches and statistics employed in the much longer established field of human personality.