Réponse à une question ouverte: incidence du mode de questionnement

Two methods of data collection (mailing and telephone interviews) were applied to the same open-ended question : “What is an ideal breakfast ?” on equivalent samples (N=1000). Lexicometric analysis and descending classification demonstrate important lexical, syntactic and semantic differences in the corpuses. Telephone answers are shorter, lexically and syntactically poorer ; they focus on objective product descriptions (breakfast ingredients) with a prototypic approach. In contrast, written answers, longer and richer, are more global, conceptual and idiosyncratic, they also include contextual aspects (time, place, attendees...) and personal judgments. This paper was presented as part of the conference secondes journées internationales d'analyse statistique des données textuelles. Montpellier, 21-22 octobre 1993.