Raven’s Progressive Matrices, manipulations of complexity and measures of accuracy, speed and confidence

This paper examines the effects of complexity-enhancing manipulations of two cognitive tasks – Swaps and Triplet Numbers tests (Stankov, 2000) – on their relationship with Raven’s Progressive Matrices test representing aspects of fluid intelligence. The complexity manipulations involved four treatment levels, each requiring an increasing number of components and relationships among these components. The accuracy, speed of processing, and confidence measures were decomposed into experimental and non-experimental parts and represented by the latent variables within a structural equation model. In the fitted model, four latent predictor variables had substantial path coefficients to Raven’s Progressive Matrices test. Experimental accuracy scores for both Swaps and Triplet Numbers tests have significant predictive validity. Thus, complexity-enhancing manipulations affect correlations fluid intelligence captured by the Raven’s test. In addition, two non-experimental latent variables (speed from Triplet Numbers and confidence from Swaps) have significant path coefficients.