To assess how pressures from patients on doctors in the consultation contribute to referral and investigation. Observational study nested within a randomised controlled trial. Five general practices in three settings in the United Kingdom. 847 consecutive patients, aged 16-80 years. Patient preferences and doctors' perception of patient pressure and medical need. Perceived medical need was the strongest independent predictor of all behaviours and confounded all other predictors. The doctors thought, however, there was no or only a slight indication for medical need among a significant minority of those who were examined (89/580, 15%), received a prescription (74/394, 19%), or were referred (27/125, 22%) and almost half of those investigated (99/216, 46%). After controlling for patient preference, medical need, and clustering by doctor, doctors' perceptions of patient pressure were strongly associated with prescribing (adjusted odds ratio 2.87, 95% confidence interval 1.16 to 7.08) and even more strongly associated with examination (4.38, 1.24 to 15.5), referral (10.72, 2.08 to 55.3), and investigation (3.18, 1.31 to 7.70). In all cases, doctors' perception of patient pressure was a stronger predictor than patients' preferences. Controlling for randomisation group, mean consultation time, or patient variables did not alter estimates or inferences. Doctors' behaviour in the consultation is most strongly associated with perceived medical need of the patient, which strongly confounds other predictors. However, a significant minority of examining, prescribing, and referral, and almost half of investigations, are still thought by the doctor to be slightly needed or not needed at all, and perceived patient pressure is a strong independent predictor of all doctor behaviours. To limit unnecessary resource use and iatrogenesis, when management decisions are not thought to be medically needed, doctors need to directly ask patients about their expectations.