A panel review of histologic specimens was carried out as part of a multi-centre case-control study of gastric cancer (GC) and diet. Comparisons of diagnoses of 100 GCs by six pathologists revealed agreement in histologic classification for about 70-80% of the cancers. Concordance was somewhat higher when using the Lauren rather than the Ming or World Health Organization classification systems. Histologic types from reading biopsy tissue agreed with those derived from surgical specimens for 65-75% of the 100 tumours. Intra-observer agreement in histologic classification, assessed by repeat readings up to 3 years apart by one pathologist, was 95%. The findings indicate that, although overall concordance was good, it is important to standardise diagnoses in multi-centre epidemiologic studies of GC by histologic type.