We extend the prior literature on biased disclosure decisions by examining whether, when and how managers bias the tone of forward-looking narratives. In order to measure tone we employ techniques of manual content analysis and we aggregate positive, neutral and negative statements into an overall measure of tone.We then analyse the frequency of positive and negative statements for firms with large impending year-on-year changes in sales and operating profit margin, and we regress tone cross-sectionally on four managerial incentive variables that are unrelated to the private signal about future trading, namely loss status, sign of earnings change, business risk, and the existence of an analyst earnings forecast. We find that firms with large impending performance declines bias the tone in the outlook section upwards. Also, we find that loss firms, risky firms and firms with an analyst earnings forecast provide a more positive tone, while firms with an earnings decline provide a more negative tone. Finally, we observe that for a majority of our managerial incentive variables the main vehicle of biasing the tone is to change the number of negative statements, not the number of positive statements. Overall, our findings are difficult to reconcile with predictions from signalling models, but they are consistent with the alternative view of impression management. Our results have policy implications. In particular, they suggest that there is a need to reconsider the current largely unregulated nature of forward-looking narratives.