RÉSUMÉ Cette étude porte sur l’utilisation de la ponctuation par des rédacteurs de descriptions d’objets dans des textes techniques. L’étude repose sur le modèle qui résulte des travaux de Fayol et collaborateurs (Fayol, 1997). Ces travaux ont montré que la force de séparation d’un signe de ponctuation est en forte corrélation avec le degré de liaison entre les états ou faits qui se succèdent dans le texte. Plus ceux-ci sont liés, plus faible sera la force de séparation du signe de ponctuation mis entre eux. Ces travaux ont testé cette hypothèse par calcul de cette corrélation et ceci, dans le cadre de tâches de rédaction rigoureusement contrôlées et contraintes. La présente étude teste cette hypothèse dans le cas d’une tâche de rédaction moins contrainte, mais en rapport direct avec l’activité professionnelle des adultes concernés et donc plus plausible pour eux. De plus, une démonstration plus forte est recherchée, au-delà de corrélations, par un relevé systématique des contre-exemples au modèle. Les résultats montrent une très grande validité du modèle. Or ce modèle est très simple. Trois applications pratiques sont suggérées : une utilisation méthodologique du modèle pour l’étude des représentations ; l’explication du modèle dans la formation des rédacteurs techniques ; l’introduction du modèle dans les systèmes de génération de textes par ordinateur.
The use of punctuation in technical descriptions : marking the hierarchical decomposition of the object described This study deals with the way writers use punctuation marks when they describe complex objects in technical texts. The first experimental studies on the use of punctuation (and connectors) by non-professional writers were carried out by Fayol and his colleagues and resulted in the model proposed by Fayol (1997 b). These studies showed first, a large between-subjects agreement regarding the hierarchy in punctuation marks (from the paragraph to the comma) according to the intensity of linkage (or separation) that they induce between propositions. Secondly, they demonstrated that the frequency and the nature of punctuation marks are highly correlated to the degree of linkage between the states and/or events successively mentioned. In a text, the stronger the linkage between two successive propositions, the lower the separation value of the punctuation mark between them. The results of these studies were obtained by requesting subjects to write texts, starting from very common scripts (e.g., going to the dentist). In that way, subjects did not encounter interpretation difficulty and researchers had a highly predictable hierarchical structure among the propositions. Moreover, the Fayol model rests on correlations computed between the values of between-elements linkage on the one hand, and the frequency and values attached to punctuation marks on the other. The objective of the present experiment was twofold : 1 / to assess the model in a professional context, using a writing task that was indeed less strictly controllable but, at the same time, less constrained and more plausible, because of its direct relation with the subjects’ professional field ; 2 / to test the model very strictly, by tracking down all occurrences of examples to the contrary in the whole set of elementary patterns of punctuation. Adults enrolled on a business documentation course, received three maps of a documentation centre which varied according to the number of pieces of information and the level of detail given. They were requested to write three texts describing these maps so that they could be given to users to guide them in the room. Results strongly confirm the Fayol (1997) model. Out of 353 elementary patterns under analysis, 303 patterns were consistent with the model, the proportion being [f = 85,8 % ; 95 % confidence interval : 81,7 % < f < 89,3 %]. So, when they transcribe a representation structured as a decomposition of a complex object into units described by features, writers process punctuation marks ordered according to their linkage/separation strength, and use them to mark in text surface, item clustering and hierarchical levels of the complex object decomposition tree. The model is fairly simple. From a practical perspective, it is proposed that punctuation protocol be used as one of the methodological tools for studies of experts’ mental representations. Also, it is claimed that explicitation of the model could benefit technical writers’training. Finally, it is suggested that the implementation of the model in text generators should be considered.
