This research concerns alternative response choice (distractor choice) in analogical reasoning items. Its aim is to take part in the generalisation of Rumelhart and Abrahamson's model (1973) � initially validated with items consisting of mammal's names - by applying it to items consisting of Kohs' configurations. This model is a theory about responses : distractors are selected according their position in a multidimensional space, which accounts for similarity relation between items, and subjects responses are predicted in accordance with proposed distractors.
The similarity space between Kohs' configurations is achieved by processing subjective classifying data with multidimensional scaling. From this space, 22 analogical problems of the type A :B : :C : (Rl, R2, R3, R4) are worked out with reference to subjects. The results of the global analysis point out a good correlation between observed and predicted data. Differential analysis of analogies sharpens this conclusion : when subjects point out an alternative as the best solution to an analogical problem, they base their reasoning on some characteristics of the configurations. If these characteristics are the same that explain the placement of the items in the space, the selected alternative is the predicted one ; but if subjects focus on items' characteristics that occur less significatively in the space, then the alternative pointed out as the best solution is not the one predicted by the model.
If the representation of similarity relations is sufficiently stable as regards the items used, then Rumelhart and Abrahamson's model gives a theoretical framework for distractor selection in multiple choice items.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados