The research on emotional reactivity usually implies the use of standardised behavioural tests that provide a quick idea of the effect of a treatment on the reactivity of subjects to potentially dangerous situations. Many validity criteria have been considered to evaluate these tests. This validity concept supports the idea that animals' behaviour in these tests model human anxiety. Generally, those criteria repeatedly labelled as "ethological validation" refer to the analogy between animals and human in the meaning of the test situation. Although the content of the ethological validation concept is heterogeneous, it is steadily related to a fixed interpretation of the behavioural items produced in a given experimental setting. The basic assumption of such reasoning is that the behavioural items would always be expressed in the same behavioural context whatever the subject, its gender, strain or species, thoroughly asserting a predefined subjective state. Using multivariate and textual analysis, we found evidence that the "ethological validation" recourse to an a priori interpretation for a given behavioural variable may be deceptive. We defend the idea that the meaning of a behavioural variable should be restricted to the general context where it arose. Theoretical propositions and methodological options are discussed