Investigating the relationship of working memory tasks and fluid intelligence tests by means of the fixed-links model in considering the impurity problem Karl Schweizer ⁎ Institute of Psychology, Johann Wolfgang Goethe-University, Mertonstr. 17, 60054 Frankfurt a. M., Germany Received 21 December 2005; received in revised form 29 November 2006; accepted 29 November 2006 Available online 12 January 2007 Abstract The impurity of measures is considered as cause of erroneous interpretations of observed relationships. This paper concentrates on impurity with respect to the relationship between working memory and fluid intelligence. The means for the identification of impurity was the fixed-links model, which enabled the decomposition of variance into experimental and non-experimental parts. A substantial non-experimental part could be expected to signify impurity. In a sample of 345 participants error scores and reaction times, which were obtained by the Exchange Test, represented working memory, and Advanced Progressive Matrices served as measure of fluid intelligence. The four independent latent variables of the model associated with error scores and reaction times led to a multiple correlation .67 with the latent variable of fluid intelligence. However, there was impurity since the decomposition by means of the fixed-links model showed that only 45% of the common variance was due to working memory. © 2006 Elsevier Inc. All rights reserved. Keywords: Cognitive process; Fluid intelligence; Structural equation modeling; Latent curve model The question of whether working memory contri- butes to intelligence has stimulated a large number of studies. As consequence, many correlational results suggesting the existence of a substantial relationship are available. Ackerman, Beier and Boyle (2005) report a metaanalytic investigation of 57 studies and suggest a correlation of .48. The inspection of the individual results reveals that this field of research shows a high degree of heterogeneity. There are rather low besides very high correlations. The results obtained by means of structural equation modeling are most impressive. Some studies even suggest near identity of working memory and intelligence with respect to individual differences. Typically, the relationship is investigated at the latent level in considering a number of (slightly differing) measures (e.g., Buehner, Krumm, & Pick, 2005; Colom, Abad, Rebollo, & Shih, 2005; Colom, Rebollo, Palacios, Juan-Espinosa, & Kyllonen, 2004; Colom & Shih, 2004; Engle, Tuholski, Laughlin, & Conway, 1999; Kyllonen & Christal, 1990). The heterogeneity of results demands for an explanation. Actually, there is a number of potential explanations. For example, the difference between correlations observed at the manifest level on one hand and at the latent level on the other hand provides an Available online at www.sciencedirect.com Intelligence 35 (2007) 591 – 604 ⁎ Tel.: +49 69 798 23350; fax: +49 69 798 23847. E-mail address: K.Schweizer@psych.uni-frankfurt.de. 0160-2896/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.intell.2006.11.004