Background. Although (general practitioners) GPs have a role in managing patients with advanced cancer, little is known about their referral decisions. Aim. The aim of this study was to explore, using structured vignettes, how GPs might manage patients presenting with advanced cancer. Design. A self-administered survey consisting of structured vignettes was administered to GPs in Australia. Fifty-six vignettes describing patients who may benefit from palliative care and/or treatment were constructed encompassing seven advanced cancer diagnoses (cerebral metastasis, lung metastases, renal cancer, bone metastases, ulcerating skin metastases, spinal metastases and stridor) and three clinical variables (age, prognosis and mobility). Seven vignettes were presented to each respondent. Respondents were asked if they would refer the patient and the benefits of different treatment modalities. Participant responses were compared with responses provided by an expert panel. Logistic regression and parametric tests were used to estimate odds of referral. Setting/Participants. The respondents were GPs, currently registered and practicing in Australia. Participants were selected randomly from a national list of practitioners. Results. Four hundred and seven questionnaires were received. There was wide variation (31%- 97%) in the proportion of respondents who agreed with the expert panel. The odds of referral for radiotherapy varied the most. Significant predictive variables included patient age, mobility and prognosis and respondent demographics. Conclusion. GPs' referral decisions for patients with advanced cancer appear to deviate from expert opinion and can be predicted using respondent and patient characteristics. If these data were reflected in clinical practice some patients may not be offered helpful palliative treatment options.