The study had two main aims: (i) to develop a model incorporating research self-efficacy and research disposition as predictors of publication output; and (ii) to assess the fit and robustness of this model to dichotomous groups based on gender, academic qualifications and rank. Data were drawn from a survey of lecturers (n = 331) working in two large Australian universities. The survey response rate was 33.6% and the lecturers were affiliated with a diverse range of research disciplines. Well-fitting measurement models for both research self-efficacy and disposition were produced using AMOS. However, a structural model which included these two measurement models would not satisfactorily fit the full sample. Subsequent testing yielded acceptable results when separate analyses employing the group variables of gender, academic qualifications and rank were undertaken. The model with rank offered the best fit of the data and, across all models, research self-efficacy proved to be the most important predictor of output. The results of the study provide an understanding of the scope and nature of research self-efficacy and, in addition, contribute to knowledge about the factors that underlie research disposition. As a consequence, the results have implications for career development programmes designed to foster research skills and further research. Additionally, the results point to some possible interventions aimed to boost research endeavour and output.