AIMS: To explore the ability of a physiotherapist, using a standardised musculoskeletal physiotherapy assessment protocol, to accurately identify the structures potentially responsible for shoulder symptoms against a standardised arthroscopic shoulder diagnostic assessment, and to determine the physiotherapists' ability to influence post-test diagnostic accuracy. STUDY DESIGN: Consecutive case-based cohort study. SUBJECT SELECTION: All participants were selected by two orthopaedic surgeons for arthroscopic investigation during a 6-month period. SETTING: Private orthopaedic clinic. METHOD: All consenting participants selected for arthroscopic investigation were examined by the physiotherapist prior to arthroscopy. Presence and priority of impairments/diagnoses were recorded on a standardised form. Inter and intra-rater reliability and diagnostic accuracy were tabulated. STATISTICAL ANALYSIS: Proportional agreement on diagnostic incidence (broad) and priority (strict) using 2×2 contingency tables for sensitivity, specificity, positive and negative predictive value and positive and negative likelihood ratios were calculated. Post-test probabilities were analysed to determine the influence of a positive or a negative finding. RESULTS: 211 participants, aged 14-79 years were included. Overall prevalence of subacromial pathology was (77%) and, disorders of the passive restraints (29%). For both negative and positive findings, post-test probabilities were not notably altered; although positive findings yielded greater value in the decision-making modelling. The physiotherapist's ability to identify individual pathology (eg, tendon rupture vs tendinopathy, capsular vs labral) was lower than recognition of pathology within the broader diagnostic category. CONCLUSIONS: The physiotherapist's ability to diagnose individual pathologies was inconsistent. Indirectly, this raises the issue of whether signs and symptoms identified under arthroscopic surgery are reflective of a lesion/pathology reflective of a specific tissue.