Forty-eight children and forty-eight adults of contrasting degrees of expertise made a series of corrections in order to improve a text (narrative or description) in which three within-statement errors and three between-statement errors had been inserted. Subjects used a simplified word processor (SCRIPREV) which recorded all movements of linguistic units. The purpose of this research was to study revising strategies by examining the correction-sequencing procedures implemented by these subjects. The procedures, which were coded in the form of time series, were compared to the time series of model revising procedures (i.e. effective ones) representing three strategies based on certain predefined functional principles (linguistic level, execution order). The adults used two of these strategies: the Simultaneous Strategy for the narrative, and the Local-thenGlobal Strategy for the description. The children used the Local-then-Global Strategy for the narrative, but did not use any identifiable procedure to revise the description, which they did not manage to totally improve in the expected manner.

Las experiencias presentadas en este artículo han sido concebidas con vistas a precisar una problemática concerniente a las características de la elaboración de una representación cognitiva del universo de la evaluación entre expertos, en una tarea concreta. Los expertos son aquí docentes, y su tarea consiste en evaluar el nivel de dominio de la lengua materna de alumnos de último curso del Ciclo Medio. Los soportes de evaluación son redacciones de francés, de las cuales se presentan varias versiones, haciendo variar la localization de la información concerniente al uso correcto o no de ciertas reglas de escritura. Del mismo modo, se hace variar la información dada a priori al experto en lo que se refiere al nivel del alumno. El seguimiento, en tiempo real, del proceso evaluativo durante la lectura se ve asegurado gracias a un paradigma experimental de ventana móvil. Los resultados permiten, en primer lugar, evidenciar los aspectos secuenciales de la presa de información, forzosamente limitada por el aspecto lineal del objeto, y validar parcialmente un modelo de organización de las representaciones entre expertos. En segundo lugar, encontramos que las heurísticas de juicios (y los sesgos asociados a ellas), puestas de manifiesto en situaciones de laboratorio, están igualmente presentes en una situación más próxima de la realidad cotidiana de una tarea de evaluación.

The current study examined children's ability to analyse pictures of a risky situation, both in relation to the characteristics of the pictures and in relation to the centering/decentering process of cognitive development. Sixty children aged 6, 9 or 11 years were given an objective or subjective version of a story about a risky situation involving road crossing and were asked to reconstruct it by putting six pictures in chronological order. The type of picture series, objective or subjective, had a different effect on the children's understanding and performance, according to the age. The older children were better at ordering the pictures, but on the subjective version only. The picture-version effect on planning time decreased with age; only the younger children took more time to start touching the pictures. On one hand, it is concluded that for the youngest children, objective representations are essential to analysing pictures showing a risk, whereas the oldest children will profit more from a subjective view. On the other hand, subjective representations, which give a more realistic view, provide an excellent tool for testing children's abilities. Subjective representations can be used to detect potentially risky behaviour in virtual situations (static pictures, or multimedia tools), since it permits one to predict at-risk behaviour in the street and to assess the effectiveness of remedial measures.