Intelligence can be best understood through the study of nonentrenched, i.e., novel, kinds of tasks. Such tasks require subjects to use concepts or form strategies that differ in kind from those to which they are accustomed. The only partial success of the cognitive-correlates and cognitive-components approaches to intelligence that are in contemporary favor might be due in part to the use of tasks that are more entrenched (familiar in kind) than would be optimal for the study of intelligence. Two nonentrenched tasks are described, one requiring projection into the future of states of objects, the other requiring complex analogical reasoning where multiple terms of analogies can be replaced by alternative answer options. Research into the first task focused upon performance components of task solution (i.e., components used in the execution of strategy); research into the second task focused upon metacomponents of task solution (i.e., components used in the planning of strategy). Correlations of task and component latencies were generally higher than those obtained in most contemporary information-processing research on the nature of intelligence. Some speculations are made about the implications of these results for educational theory and practice. (Author/RL)