These experiments deal with the relation between the ability of a person to carry out efficient action in a situation, and that of the same person to answer questions about the situation. Existing knowledge on the topic has various gaps. In particular it is relatively common to show performance changing without change in verbal knowledge, but relatively rare to show changes in verbal knowledge without change in performance. A second major gap is in defining the kinds of task on which discrepancies may be expected to occur, since in everyday life we do quite frequently expect people to know what they are doing. Accordingly, seven groups of subjects were studied controlling two artificial systems, one concerned with transportation and one with a model of the economy. The discrepancy between verbal knowledge and performance was changed by altering the number or salience of the relationships being learned, and this is at least one factor in making tasks show the discrepancy. Several instances were found of change in verbal question answering opposed to or in the absence of any change of decision performance; thus, the discrepancy is not asymmetric affecting only one measure of learning. This implies that all the variables examined affect processes specific to one or other measure, rather than changing a common database. The results challenge a common view of the discrepancy between performance and verbal accounts, and suggest rather that there are alternative modes of processing in human decision making, each mode having its own advantages.