Six experiments concerned people's ability to estimate the degree and sign of covariation represented in a bivariate distribution of stimuli with which they had just been presented as a series of pairs of stimuli. The stimuli were pairs of numbers, pairs of lines of variable lengths, or word-line pairs. In the latter case, subjects were asked to think of the words in terms of either pleasantness or familiarity; hence, the covariation relationship was between the normative pleasantness or familiarity value of the word and a line of variable length. In the sixth experiment, subjects were presented with two word-line pairs and were asked to access the covariation of both simultaneously. In most cases, the estimates reflected the sign and degree of covariation of the stimuli quite well. The estimates did not reflect accurately the stimulus covariation when the stimuli were numerical and the sign of the correlation of the stimuli was negative. A distinction is made between intuitive and strategic processes in the assessment of covariation.The ability to predict and estimate characteristics of people, events, and things would seem to be essential in everyday interactions. Knowing that certain behaviors of other people tend to be associated with other behaviors allows one to prepare for the future and behave appropriately. One way to view predictive ability and the knowledge upon which it is based is to consider that it results in part from assessments of the covariation of stimuli derived from experience. That is, one learns from experience that certain kinds of social signals from another person are associated with aggression, affection, or other tendencies. One could view these signals as being correlated with the tendencies. As with most correlations, the relationship is not perfect and has some uncertainty associated with it.Previous research indicates that people are generally quite poor at judgments about the actual degree of covariation represented in stimuli they inspect (Nisbett & Ross, 1980). Studies have been reported in which subjects were shown pairs of correlated stimuli, such as letters, representing symptoms and diagnoses (Smeldelund, 1963), mood and the weather (Shaklee & Mims, 1982), and so on. Sometimes the stimuli have been presented in tabular form (see Beyth-Marom, 1982, for a summary). Another popular method used to study covariation assessment has been to show subjects a number, ask for a prediction about the value of its pair, and then show the correlated pair as feedback (Birnbaum, 1976;