The presence and likelihood of an alternative diagnosis to pulmonary embolism is an important variable of the Wells' prediction rule for establishing clinical probability. We assessed whether evoking specific alternative diagnoses would reduce the probability of pulmonary embolism enough to forego further testing. We retrospectively studied a cohort of 965 consecutive patients admitted for suspicion of pulmonary embolism at three medical centers in Europe in whom the presence of an alternative diagnosis at least as likely as pulmonary embolism was recorded before diagnostic testing. We divided the patients into 15 categories of alternative diagnoses evoked. We then assessed the prevalence of pulmonary embolism in each diagnostic category and compared it to the prevalence of pulmonary embolism in a reference group (patients with no alternative diagnosis or a diagnosis less likely than pulmonary embolism). The prevalence of pulmonary embolism in the reference group was 48%. The presence of an alternative diagnosis as or more likely strongly reduced the probability of pulmonary embolism (OR 0.15, 95% CI: 0.1-0.2, p < 0.01). In almost every diagnostic category, the prevalence of pulmonary embolism was much lower than in the reference group with an odds ratio below or near 0.2. Bronchopneumonia (OR 0.4, 95% CI 0.2 to 0.7) and cancer (OR 0.6, 95% CI 0.3 to 1.5) reduced the likelihood of pulmonary embolism to a lower extent. Evoking an alternative diagnosis at least as likely as pulmonary embolism reduces the probability of the disease, but this effect is never large enough to allow ruling it out without further testing, especially when bronchopneumonia or cancer are the alternative diagnoses considered.