In Denmark, as in many other countries, there are extensive efforts to revise texts that emanate from public authorities to make them easier to understand and more appropriate for an audience of lay citizens (a development akin to the campaign for plain English). Such efforts focus mainly on practice and not on evidence derived from research. Furthermore, insofar as the effects of revisions of texts from public authorities and private businesses have been tested by research, the revisions under examination have been conducted by researchers or at least for research purposes. This study takes a different approach by focusing on textual revisions made ‘at-source’ by employees at the Danish Tax Authority and independently of any research project. Taking as its point of departure the communicative goals laid down by the Authority, the main question addressed in this article is whether three ‘at-source’ textual versions vary in respect of their recipients’ levels of comprehension and perceptions. Three different versions of a letter from the Danish Tax Authority (an unrevised version and two versions that had been revised in different ways) were distributed at random. Each respondent read one version and then answered a questionnaire about his/her level of comprehension and about his/her perceptions of the letter. A total of 714 questionnaires were collected. The analysis shows that the revised letter versions are, in fact, easier to understand, that the perceptions of these versions do vary, and that the variations are in accordance with the communicative goals of the Authority. However, the results also indicate that there is room for further improvement and they may suggest that the textual revision practices that are applied – or at least held up as ideal – by Danish authorities are not as effective as sometimes assumed.