Languages differ in how they lexicalize and grammaticalize spatial information, thereby constraining how speakers organize spatial information to encode motion in discourse. Such cross-linguistic differences raise new questions concerning the relation between language and thought. In this context we examined how speakers of two typologically different languages, English and French, performed several tasks, all of which involved motion events: a production task (describing visual scenes showing motion events), a non-verbal categorization task (grouping visual stimuli of these events), a verbal categorization task (deciding which visual stimulus best corresponds to a sentence describing a motion event). In addition, all three tasks were coupled with an eye-tracking paradigm measuring on-line how participants allocated their visual attention when exploring these events. Subjects' verbalizations during the production task differed substantially in the two groups as a function of language-specific factors. French speakers focused mostly on Path information (lexicalized in the verb), while English speakers expressed Manner (in the verb) and Path (outside of the verb) equally often, thereby producing denser utterances. Subjects' preferential choices during the categorization tasks (in both of its verbal and non-verbal versions) were guided by different criteria. However, a more important language effect was observed in the verbal version of this task as compared to its non-verbal version. Finally, although speakers in both groups allocated more attention to Path information overall during their visual exploration of the events, their focus ٭ Address correspondance to Efstathia Soroli, 2 of attention also varied across groups, suggesting a greater focus on Path in French as compared to English, as well as a different unfolding of attention during the processing of the visual stimuli. However, the eye-tracking data must be interpreted with care since they also showed variable patterns that seem to be stimulus dependent. In conclusion, we argue that our results support a moderate version of the relativity hypothesis according to which typological language properties have a clear impact on linguistic behaviours, but also on non-linguistic behaviours, although to a lesser extent, in some tasks, and with some stimuli.