This thesis empirically investigates the use of impression management in the narrative sections of the annual reports of UK listed companies. Impression management is examined by testing the obfuscation hypothesis which claims that firms with poor performance have a tendency to obfuscate negative organisational outcomes. For this purpose, the thesis provides an assessment of the extent to which reading difficulty and self-presentational dissimulation are associated with the disclosure of favourable or unfavourable results ('good/bad news') in annual financial statements, conditional on a firm's size and sector of operations. Impression management has previously been studied in the context of agency theory explanations of managerial and investor behaviour. This study contributes to the understanding of impression management in a corporate reporting context by first reviewing relevant theoretical work in behavioural finance, social psychology, and linguistics. Social psychology provides additional insights into the managerial motivation to engage in impression management, the circumstances fostering managerial impression management, and preferred managerial strategies. Behavioural finance offers insights into the effectiveness of impression management. Research in linguistics and social psychology provides the basis for developing new methodologies for measuring impression management in corporate narrative documents which overcome the validity problems inherent in conventional measures. Three new methodologies are introduced. The first develops cohesion-based measures of reading difficulty that focus on grammatical devices within and between sentences, including the number and density of cohesive ties and the proportion of new and given information (MMAX2). The second methodology provides multiple cohesionbased measures of readability, as applied in web-based readability scoring (CohMetrix). The third methodology measures impression management in the form of selfpresentational dissimulation (i.e. portraying a public image of firm performance and prospects inconsistent with a managerial view of firm performance and prospects), using linguistic markers which include word count, self-reference, reference to others, the use of emotion words, and cognitive complexity. The empirical analysis that is reported in this thesis is based on a sample that is balanced across industrial sectors and representative of the size distribution of firms. Results show firm size and not 'good/bad news' to be the determining factor in reading difficulty. Although the main effects model shows 'bad news' to be directly related to reading difficulty, this association is no longer significant when 'good/bad news' is interacted with firm size. Results suggest that large firms are more likely to produce corporate narrative documents which are less cohesive (and thus more difficult to read) than small firms. This is not interpreted as impression management, but as an indication that firms might tailor their corporate narrative documents to the reading strategies of their target readership groups. Thus, large firms seem to cater to the needs of high-knowledge readers (professional investors or readers largely familiar with the infom1ation content of the chairman's report), and small firms to the needs of low-knowledge readers (individual investors or readers largely unfamiliar with the information content of the chairman's report). Results regarding impression management in the form of self-presentational dissimulation suggest that the linguistic markers are not indicative of impression management in the form of selfpresentational dissimulation, but of other psychological issues.