Definitions of the concept of intelligence typically emphasize two aspects, the ability to solve complex problems (the complexity aspect) and the ability to acquire new knowledge (the learning aspect). Complexity has been the most emphasized aspect in the psychometric approach to research on intelligence. The study investigated whether a change of test item sequencing, intended to increase test complexity, would cause increased involvement of general intelligence (G). Three types of problem-solving items (Groups, Series, and Bongard) were administered in a sequence (the homogeneous treatment, n = 363). In the treatment intended to increase complexity, items of different types were presented by turns (the heterogeneous treatment, n = 1,778). Three reference tests measuring verbal, spatial, and reasoning ability were used in the analysis. Contrary to what was expected, the items presented homogeneously showed higher G loadings. The reason for this might be that in the reasoning tests used, processes of learning took place from one item to another, and that the high-G individuals could take better advantage of this opportunity in the homogeneous than in the heterogeneous treatment. The consequences of the results for testing in general and for computer adaptive testing in particular are discussed.