OBJECTIVE: The study of psoriatic arthritis (PsA) is hampered by the absence of a widely accepted, validated case definition. We investigated whether the physician's opinion can be used as a gold standard when developing classification criteria for peripheral joint PsA. METHODS: UK rheumatologists who had published on PsA and attendees at 3 international meetings on PsA held in the UK were polled by questionnaire. There were 3 phases. The first questionnaire asked whether rheumatologists believed in the construct of PsA. The second survey developed a list of features thought to distinguish patients with PsA from other forms of peripheral arthritis. The final phase was development of a series of 61 "paper" patients with various combinations of the features of PsA. The paper patients were assessed by 15 rheumatologists who were asked whether, in their opinion, the patient had PsA. Latent class analysis was used to identify subgroups of patients and cross-tabulations were used to identify which clinical and laboratory features were associated with each subgroup. RESULTS: Rheumatologists agreed on the construct of PsA and that not all patients with psoriasis and an inflammatory polyarthritis have PsA. Latent class analysis identified 3 classes, corresponding to definite PsA; a middle group that was very likely to be given a diagnosis of PsA by some rheumatologists (high diagnosers), but unlikely to be given the diagnosis by others (low diagnosers); and a third group corresponding to "probably not PsA." CONCLUSION: For the group of patients with "definite PsA" the physician's opinion can be taken as the gold standard when developing classification criteria. However, for patients in the "middle group" there will always be disagreement with the gold standard whether the standard is based on the opinion of the high diagnosers or the low diagnosers.