Cet article vise à améliorer la compréhension du développement du traitement des feed-back dans le raisonnement hypothético-déductif, indépendamment des connaissances ou croyances de l’individu concernant la thématique spécifique de l’énoncé du problème. Pour cela, 108 enfants de 7 à 14ans ont résolu une version simplifiée du Mastermind pour laquelle un programme de sélection des feed-back a été utilisé afin de générer des situations structurellement comparables. Quatre indices ont permis d’attester de l’accroissement des performances à cette tâche de raisonnement hypothético-déductif au cours du développement. À partir de l’analyse des protocoles individuels, cinq stratégies de résolution ont été identifiées : trois peu performantes, insensibles aux feed-back et deux prenant en compte le feed-back à un niveau local ou à un niveau global (stratégie experte), plus performantes et dont la mise en œuvre s’accroît au cours du développement. Ces stratégies reflètent des interprétations plus ou moins adaptées du problème.

In hypothetico-deductive reasoning, it has been shown that young children validate the tested hypothesis when these are consistent with their previous knowledge about the thematic used in the problem statement and not necessarily in function of their observations. In the studies, little attention has been paid at the processing of feedback received after the hypotheses are tested. This paper aims to improve the understanding of this particular stage via a developmental analysis of the strategies used by 7 to 14years-old children (n=108) without experience of the task, that solve a simplified version of Mastermind game for which no previous knowledge interferes with the feedback processing. The experiment is characterized by the use of a feedback selection program that ensures obtaining structurally comparable situations and an individual protocol analysis using an original method of strategies identification. The first hypothesis is that during childhood and adolescent years, there is an evolution of hypothetico-deductive reasoning, tested by means of four performance indicators: the failure rate, the number of trials used to solve the task, the number of repetitions of a prior incorrect response and the rate of incorrect processing of feedback. The second hypothesis is that the development of hypothetico-deductive reasoning consists in more elaborate feedback processing, leading to more efficient strategies, tested by means of the analysis of individual protocols and three performance indicators. Results show that the four performance indicators of older children are better than younger, without experience of the task, and that the way the feedback is processed develops toward the use of more efficient strategies. The analysis of individual protocols proves the existence of several feedback processing strategies. Five strategies were identified: three inefficient strategies did not process the feedback (random strategy, perseverance strategy and permutation strategy), one strategy processes the feedback at a local level (local feedback strategy) and one strategy processes the feedback at an overall level (optimal feedback strategy, i.e. the expert strategy). The most efficient strategy, for which feedback processing is the most complex, is most frequent in the course of development with a shift away from strategies that ignore the feedback. Strategies are matched with incorrect interpretations of the task and incorrect interpretations of the role of feedback.