Subjects who held either strong, moderate, or mild opinions on an issue were asked to write arguments in favour of their opinion, and opposed to their opinion. Half did in a pro-con order and half in con-pro order. They then recorded their opinion again. In addition to opinion change the dependent variable was the numbers of pro and con statements written. It was found that most persuasion
... [Show full abstract] occurred among those holding strong or moderate opinions on the issue with a suggestion that subjects with strong opinions were affected by argument order, con-pro being more effective than pro-con. A significant proportion of the subjects with mild opinions showed boomerang effects. The numbers of pro statements consistently exceeded con statements under the pro-con order, and con statements consistently exceeded pro statements under the con-pro order. The absolute numbers of both pro and con items were directly related to initial strength of opinion. Implications for dissonance, incentive, and learning models of opinion change are noted.