Recent studies (e.g., Anderson, 1965; Jacobs & Vandeventer, 1968; Wittrock, 1967) have demonstrated that young children can learn, retain, and transfer general cognitive problem-solving skills and strategies. In each case the instruction has been carried out in a laboratory-like setting with fairly well-defined procedures. The success of these studies may lead to attempts to teach "intelligence" directly. This paper argues that a taxonomy of intelligence tasks is required to provide the dimensions for measuring transfer following teaching for intelligence. At the present time Guilford's Structure-of-Intellect model is considered too macrocosmic for this purpose. A micro-taxonomy was therefore constructed within the domain Guilford refers to as cognition of figural relations. First, 201 aptitude and intelligence tests were examined as possible sources of relevant tasks. Thirteen hundred thirty-five test items were located and defined as our universe of interest. Twelve logical relations were derived for use in classifying these items. A manual written to explain these 12 relations was successful in communicating how they apply. The relations were adequate for classifying 95% of the items in our universe. It was concluded that the taxonomy based on the 12 logical relations could be meaningfully used for evaluating attempts to teach the cognition of figural relations domain of intelligence.