Cognitive abilities - aspects of structure, process and measurement

Abstract

The overall purpose of the thesis is to describe the development of the Swedish system of measuring cognitive abilities applied at enlistment of conscripts. The Enlistment Battery has been used for more than fifty years to classify 18-year old men into military positions for their compulsory service. Throughout the years the batteries have been influenced by the changes in the factor analytically derived models of cognitive abilities. Seven earlier versions of the battery are presented, as well as a review of the development of the models of the structure of intelligence from Spearman to the Scandinavian hierarchical model of Undheim and Gustafsson that equalizes general ability and Fluid intelligence. This most recent theoretical development has strongly guided the design and evaluation of the latest version (1994) of the battery. The multivariate charac¬ter of cognitive ability, of tests and of test performance was acknowledged in the way that latent variable esti¬mates of general intelligence, Crystallized intelligence and General visualization (as nested factors) comprised the test results in the first computerized version. This is reported in the second study. The first study examines the construct validity of the preceding version. A confirmatory factor analysis approach was used in which the orthogonal hierarchical structure was confirmed. The two last studies are directed towards a theoretical deepening concerning aspects of test performance and test evaluation. The impact of item sequencing on the construct vali¬di¬ty of complex problem solving tests was investigated. It was found that item sequencing that appeared to increase the opportunity to learn throughout the test produced a better Fluid intelligence test than an item sequencing that implied less opportunity to learn. From earlier empirical studies and from theories of intellectual development it was assumed that cognitive abilities are more differentiated in high ability subjects than in low ability subjects. This was studied empirically for the enlistment population through a multi-group latent variable approach and was supported by the results in that the broad ability factors Crystallized intelligence and General visualization captured more variance on higher levels of general ability than on lower levels. Implications for future advance in cognitive testing are discussed regarding aspects as the impact of item sequencing effects on adaptive testing, and the opportunity to make use of the multi¬dimen¬sionality of tests and even of test items in test construction and evaluation.