Aims: To test the hypothesis that the published morphological criteria permit reliable segregation of small cell carcinoma of the lung (SCLC) and large cell neuroendocrine carcinoma (LCNEC) cases by determining the interobserver variation. Methods and results: One hundred and seventy cases of SCLC, LCNEC and cases diagnosed as neuroendocrine lung carcinoma before LCNEC had been established as a diagnostic category were retrieved from the archives of the assessor's institutes. A representative haematoxylin and eosin section from each case was selected for review. Batches of cases were circulated among nine pathologists with a special interest in pulmonary pathology. Participants were asked to classify the cases histologically according to the 2004 World Health Organization (WHO) criteria. The diagnoses were collected and kappa values calculated. Unanimity of diagnosis was achieved for only 20 cases; a majority diagnosis was reached for 115 cases. In 35 cases no consensus diagnosis could be reached. There was striking variability amongst assessors in diagnosing SCLC and LCNEC. The overall level of agreement for all cases included in this study was fair (kappa = 0.40). Conclusions: Using non-preselected cases, the morphological WHO criteria for diagnosing SCLC and LCNEC leave room for subjective pathological interpretation, which results in imprecise categorization of SCLC and LCNEC cases.