In a randomised trial, general practitioners and nurses in 21 practices were trained in patient-centred consulting and use of materials for people with Type 2 diabetes (GPs 0.5 days; nurses 1.5 days; two optional follow-up half-days). Twenty practices formed the comparison group. Professional beliefs, attitudes and behaviour were measured (pre-trial, close-of-course and end-of-trial), supported by patient reports of nurse behaviour (141 trained: 108 comparison patients, 1 year after diagnosis). A total of 49 practice nurses responded (29 trained; 20 comparison). Trained nurses rated relative importance of patient-centred to professional-centred care as greater than comparison nurses. Trained nurses became less keen on the approach during the trial, and perceived time constraints persisted. Patients diagnosed later in the study were less likely to recognise intervention materials. Trained nurses rated delivery of important aspects of care and satisfaction with style of care as lower than comparison nurses, but patients were more positive about delivery of care from trained than comparison nurses. Although nurses rated patient-centred care as important, whether or not they had been trained as part of the trial, the short, generalizable training programme significantly reduced nurse perceptions of their ability to deliver it. Nonetheless, patients reported that important aspects of diabetes care were delivered more if their nurses had been trained in patient-centred consulting. This raises issues concerning measurement scales completed by trained professionals.