The use of behavioural tests aiming to assess the psychological components of stress in animals has led to divergent and sometimes arbitrary interpretations of animal behaviour. This paper presents a critical evaluation of behavioural methods currently used to investigate stress and emotionality. One of its main goals is to demonstrate, through experimental evidence, that emotionality may no longer be seen as a unidimensional construct. Accordingly, following a discussion about concepts, we propose a multiple-testing approach, paralleled by factor analyses, as a tool to dissociate and study the different dimensions of emotionality. Within this multidimensional context, genetic studies (illustrated here by different rat models) are shown to be particularly useful to investigate the neurobiology of stress/emotionality. A genetic approach can be used (i) to broaden and dissect the variability of responses within and between populations and (ii) to search for the molecular bases (i.e. genes and gene products) which underlie such a variability.