Recent studies have suggested that both the subjective judgement of a physician and standardized clinical models can be helpful in the estimation of the probability of the disease in patients with suspected pulmonary embolism (PE). We performed a multi-center study in consecutive in- and outpatients with suspected PE to compare the potential diagnostic utility of these methods. Of the 517 study patients, 160 (31%) were classified as having PE. Of these patients, 14% had a low probability as estimated by the treating physician, while 25 to 36% were categorized as having a low clinical probability with the use of two previously described clinical models. The objectively confirmed prevalence of PE in these three low probability categories was 19%, 28% and 28%, respectively. The three methods yielded comparable predictive values for PE in the other probability categories. We conclude that a physician's clinical judgement alone and two standardized clinical models, although comparable, perform disappointingly in categorizing the pre-test probability in patients with suspected PE.