Abstract
Effect of instruction on dimensionality of tests. The concern for this topic spreads for its relevance for the most popular models of Item Response Theory. Nevertheless, the doubts on the fulfillment of the assumption of unidimensionality in the case of achievement tests are generalized. Doubts that have led to the substitution of the original concept for weaker versions, for instance, the "essential unidimensionality" (Stout, 1987). This paper presents the partial result of an exploratory study, developed to examine the effect on the response dimensionality to an achievement test of the differences in instructional experiences. DIMTEST procedure (Stout, 1987) identifies the lack of essential unidimensionality in the set of 9 ítems (<0.001) designed to assess achievement in content subject to different instructional strategies for two grupos of examines (N=650). Lastly, this paper analyses the consequences of results for a possible differential functioning of the set of ítems that not met the requirement of essential unidimensionality.
