In an experiment to study the effects on attitudes of requiring subjects to use evaluatively biased language, 84 schoolchildren aged 13–14 years completed a questionnaire to measure their attitudes on the issue of adult authority over teenagers, before and after writing an essay on this issue in which they were either required to incorporate words from a list all of which implied a positive evaluation of a pro-authority position or a negative evaluation of an anti-authority position (pro-bias condition), or required to incorporate words where the implied evaluations were reversed (anti-bias condition), or were given no words to incorporate (control condition). Relative to controls, pro-bias subjects showed as a shift towards a more pro position and anti-bias subjects became more anti irrespective of their initial attitudes (pro-bias versus anti-bias comparison, p<.01). However, when tested 6 days later most of this effect had disappeared, particularly in the case of subjects whose initial attitudes were least pro. At this final session, subjects also rated attitude statements on the issue in terms of scales constructed from the pro-bias and anti-bias word lists. In accordance with previous research, the more pro subjects' attitudes, the more they showed greater polarization of judgement on the pro-bias than the anti-bias scales (p<.000l). It is concluded that a person's attitude may be related to the kind of evaluative language he will apply to an issue, and that when a person is induced to use language implying a particular evaluation of an issue, he may change his attitude, at least in the short term, so as to be more congruent with the language he has used.