Objective – To assess the reliability and to describe the categories and the procedure to apply the VRMICS/P (Verona-Medical Interview Classification System/Patient). Setting – The interviews used for the reliability study were audio taped. Five general practitioners (GPs) working in two general practices in South-Verona recorded their consultations. Sample – 50 interviews selected randomly from 120, 10 for each GP. The selection criterion for the participating patients was a GHQ-12 score of 3 and the consultation for a new illness episode. Main outcome measures -The VR-MICS/P classifies patients' verbal behaviours into 21 categories, 15 of them are defined by form (cue or statement) and content. Procedure – Two trained raters classified 50 interviews. Before applying the classification system each interview is divided into units which are numbered to define doctor's and patient's sequence of speech. Results – The reliability of VR-MICS/P was satisfactory (Kappa 0.85). Similarity Index (Dice, 1945) for categories varied between 0.71 and 0.94. Reliability for form and content classification was satisfactory too (Similarity Index between 0.81 and 0.89 and between 0.84 and 0.94, respectively). Conclusions – The VR-MICS/P is a reliable measure for describing patients' verbal behaviours during medical interviews. It can be used together with the VR-MICS/D (Verona-Medical Interview Classification System/Doctor; Saltini et al., 1998) to describe the medical interview, the quality of doctor-patient interview and can be used as a measure of patient centredness.