Purpose -The purpose of the study is to examine whether public servants as users of accounting information are responsive to changes in the content of an accounting report and, if so, how responsive they are. Design/methodology/approach -The study applies a theoretical framework, linking accounting information and accounting use by focusing on users' "mental models". The empirical context for the study is a municipality located in Northwest Russia. Three public servants were followed over several days for the purpose of studying their use of a formal accounting report. Later, those users were also presented with a different version of the same accounting report that was re-constructed by the researcher based on normative accounting theory. The study documents the responses of public servants and presents those in a narrative form. Findings -The study reveals that though the original formal financial accounting report is extensively used, the public servants mostly reject the usefulness of the modified version of the report. The modified accounting report puts public servants in a "discomfort" zone, where the users' mental models come in conflict with the information in the modified report. Following that, the study uncovers that the use of the traditional accounting report is based on three different mental models developed by users over time and guiding the use of accounting. Those mental models are described by three metaphors: "balanced matrix", "water tank" and "fair rules". Originality/value -There is an unreasonable expectation that change in the public sector accounting system from cash toward accrual information will improve the quality of decisionmaking by public servants. The claim needs evidence that change in the accounting information supply will actually lead to change in information use. The paper demonstrates that change in use requires more substantial change in the users' mental models. This change is difficult, timedemanding and requires the development of tailor-made training programs. The paper is also a response to van Helden's (2016) call for more observational studies that can give a more accurate picture of use.