The concept of patient-centredness is complex, but is generally seen as an approach that emphasises, on the part of the health professional, attention to patients' psychosocial (as well as physical) needs, the use of psychotherapeutic behaviours to convey a sense of partnership and positive regard, and active facilitation of patients' involvement in decision-making about their care. To date, there is little consistent evidence that doctors' use of a 'patient-centred' consulting style leads to better patient outcomes. However, previous studies have been limited by a lack of conceptual clarity and methodological consensus, and by the absence of a clear theoretical framework linking patient-centredness to outcomes. In this study, three specific, conceptually distinct dimensions of a patient-centred consulting style were operationalised: the 'biopsychosocial perspective', 'sharing power and responsibility' and the 'therapeutic alliance'. These dimensions were measured in terms of three 'socio-emotional' and two 'task-relevant' general practitioner (GP) behaviours using in-depth observational techniques applied to 173 videotaped GP consultations. Theoretically-derived hypotheses were tested concerning relationships between these patient-centred behaviours and two different consultation outcomes: patient satisfaction and enablement. Multivariate regression showed that GPs' patient-centred behaviours did not predict either outcome. The robustness of these findings is considered within the context of study strengths and weaknesses, and implications for future research are discussed.