Three experiments were conducted, in each of which subjects first rated an issue using the evaluation scales of the semantic differential and subsequently wrote sets of arguments concerning the issue They their rated each argument for strength of then-agreement or disagreement with it The results in all studies showed that subjects wrote a greater number of attitude-consistent arguments than attitude-inconsistent arguments (a balancing effect) The preponderance of attitude-consistent arguments increased as attitude became more extreme In all studies strength of agreement with either pro or con arguments was a function of attitude, but strength of disagreement was not Subjects wrote more arguments with which they agreed than arguments with which they disagreed (a positivity effect), but this effect only occurred when subjects were not specifically instructed to partition arguments into those they agreed with and those they disagreed with. Level of dogmatism or intolerance of ambiguity did not affect the number of attitude-consistent versus attitude-inconsistent arguments written or strength of agreement/disagreement with them. Results were discussed in terms of a balance model of information processing and the effects of the social situation on the recall process