Introduction Role modelling is a key component in the training of doctors that influences professional behaviour, identity and career choices. Clinical teachers and residents are often unaware of this, thereby risking transmission of negative behaviour. On the other hand, awareness positively affects role model behaviour. To assess role model behaviour, the Role Model Apperception Tool (RoMAT) was developed and validated in general practice training. The aim of the current study was to validate the RoMAT in the hospital-based training setting. Methods The authors asked first to last year residents, regardless of their specialty, to participate after written approval from their clinical teachers. The tool was completed online in 2017. The authors performed a principal component analysis and investigated internal consistency, construct validity, inter-rater reliability, known-groups comparisons and floor and ceiling effects. Results Of the 473 residents contacted, 187 (40%) completed the questionnaire. As in the primary validation study, the authors extracted two components: ‘Caring Attitude’ and ‘Effectiveness’, explaining 67% of the variation with a Cronbach’s alpha of 0.94 and 0.93 respectively. Evidence for construct validity was found and there were no floor or ceiling effects, but inter-rater reliability was low. Discussion The RoMAT was internally consistent and valid to assess role model behaviour of the clinical teacher towards the resident in the hospital-based training of medical specialists. The poor inter-rater reliability, most likely due to homogeneous RoMAT responses, should be borne in mind when evaluating RoMAT scores on individual clinical teachers. Electronic supplementary material The online version of this article (10.1007/s40037-019-00527-6) contains supplementary material, which is available to authorized users.