AIMS--To assess the reliability of the detection of erythrophagocytic amoebic trophozoites in stool samples in the diagnosis of dysentery associated with invasive Entamoeba histolytica. METHODS--Amoebic culture was carried out on single stool samples collected from patients from Mexico, Colombia, and Bangladesh. The stools had been examined by light microscopy. Amoebic dysentery was diagnosed when erythrophagocytic E histolytica trophozoites were observed in a case of bloody diarrhoea. E histolytica isolates were characterised by isoenzyme electrophoresis and results correlated with microscopical findings in stools. Statistical analysis was performed using the chi 2 test. RESULTS--Where erythrophagocytic amoebae had been observed in dysenteric stool specimens the E histolytica phenotype was invariably invasive (p < 0.0001). Observation of erythrophagocytic amoebae in dysentery is 100% specific and predictive of infection with invasive E histolytica. When amoebic culture-positive cases only are considered it is 96% sensitive. In this study E histolytica of zymodeme XIV was more commonly associated with amoebic dysentery than zymodeme II. There was no significant difference between the carriage rate of invasive and non-invasive E histolytica in non-dysenteric diarrhoea. Asymptomatic subjects carried non-invasive E histolytica more frequently than invasive E histolytica. Patients with non-amoebic dysentery, when shown to be infected with E histolytica, carried non-invasive strains (12%). CONCLUSIONS--Sensitivity and specificity of microscopical examination of a single stool specimen for diagnosing amoebic dysentery is very high; intestinal carriage of invasive E histolytica detected by culture is not necessarily an indication of active disease as patients with diarrhoea and asymptomatic subjects shed invasive and non-invasive E histolytica. There are possibly two subpopulations of invasive E histolytica with different pathogenic potential which can be differentiated by zymodeme analysis.