AIM This paper reports the development of a short version of the Caring Nurse-Patient Interaction Scale. BACKGROUND Since the 1980s several instruments have been developed to assess external aspects of caring. They involve using an inductive process of knowledge development to investigate the underlying structure of caring, and few reflect an explicit underlying caring theory. We developed the Caring Nurse-Patient Interactions Scale (CNPI-Long Scale) based on both inductive and deductive processes to assess attitudes and behaviours associated with Watson's 10 carative factors. Two issues led us to abridge our original 70-item scale into a more concise Short Scale (CNPI-Short Scale). First, many of our subscales were moderately to highly correlated, which is an empirical reflection of the theoretical non-independence of the carative factors. Secondly, a 70-item questionnaire was difficult to be deal with in the clinical research setting with severely ill patients because of its length. METHOD Items selected were determined by factor analysis, with specific theoretical and empirical requirements. Data were collected in September 2003 from 377 nursing students beginning their first, second or third year of a nursing programme. RESULTS The Short Scale comprises 23 items, reflecting four caring domains: Humanistic Care (four items), Relational Care (seven), Clinical Care (nine) and Comforting Care (three). All items are related to their theoretical domain alone (i.e. factor loading >or=0.40). Alpha coefficients for the four domains were adequate (0.63-0.74, 0.90-0.92, 0.80-0.94 and 0.61-0.76 respectively). CONCLUSIONS The CNPI-Short Scale, has potential for use in clinical research settings, particularly when questionnaire length is an issue. It is a useful tool for research aimed at demonstrating that caring is indeed fundamental to nursing.