Big changes to the way in which research funding is allocated to UK universities were brought about in the Research Excellence Framework (REF), overseen by the Higher Education Funding Council, England. Replacing the earlier Research Assessment Exercise, the purpose of the REF was to assess the quality and reach of research in UK universities-and allocate funding accordingly. For the first time, this included an assessment of research 'impact', accounting for 20% of the funding allocation. In this article we use a text mining technique to investigate the interpretations of impact put forward via impact case studies in the REF process. We find that institutions have developed a diverse interpretation of impact, ranging from commercial applications to public and cultural engagement activities. These interpretations of impact vary from discipline to discipline and between institutions, with more broad-based institutions depicting a greater variety of impacts. Comparing the interpretations with the score given by REF, we found no evidence of one particular interpretation being more highly rewarded than another. Importantly, we also found a positive correlation between impact score and [overall research] quality score, suggesting that impact is not being achieved at the expense of research excellence.