The relationship between the amount of information processing in an Attribute Learning task and autonomic activity, measured by skin conductance response (SCR), was investigated. The amount of information processing was manipulated by type of concept and feedback. Furthermore, the influence of practice and verbalization was studied. Three types of concepts were used: conjunction, exclusion and joint denial. The stimulus population consisted of four three-valued dimensions. The results showed that (1) SCR did not differ among concepts; (2) SCR varied with type of feedback, it being smallest at blanks and largest at infirming feedback; (3) SC'R was related to the number of confirmations preceding infirmation; (4) subjects mainly used the strategy 'reject hypothesis only after information'; (5) subjects usually selected not-yet-tested hypotheses; (6) practice influenced performance; (7) verbalization did not result in more efficient use of information; however, the appropriateness of the experimental set-up to study this factor was questioned; (8) conjunction was easier to learn than the other concepts. The results were discussed in terms of uncertainty reduction, resulting from testing hypotheses in a concept learning task.