In a prospective study we compared the diagnostic accuracy of endoscopic fine needle aspiration cytology with that of brush cytology and forceps biopsy in relation to gross tumour pattern and site in 265 confirmed consecutive cases of malignancy of the oesophagus, stomach, colon, and rectum. Aspiration cytology gave the highest diagnostic accuracy (94%), which was significantly better than that of brush cytology (84.9%) and biopsy (87.2%) (p less than 0.005). The difference was mainly related to tumour pattern. When compared to brush cytology and biopsy aspiration cytology was significantly better in submucosal tumours (92.9% v 7.1% and 14.3%, p less than 0.001); in infiltrative malignancies (95.8% v 90.1% and 78.9%, p less than 0.01), and in ulceronecrotic malignancies (90.9% v 36.4% and 45.4%, p less than 0.05). In polypoid malignancies there was a significant trend (p less than 0.05) in favour of forceps biopsy, with a diagnostic yield of 100% compared with 95% for aspiration cytology and 93.3% for brush cytology. The accuracy of the different techniques was not significantly related to the site of the tumour. The cumulative accuracy of aspiration cytology and biopsy was significantly better than that of biopsy and brush cytology (98.5% v 90.9%, p less than 0.005). Aspiration cytology was diagnostic in 21 of 24 lesions that were negative with both brush cytology and biopsy. There were no false positive cytology or histology results. We conclude that aspiration cytology is a simple, safe, and reliable technique with a high diagnostic yield and is of particular value in submucosal, infiltrative, and ulceronecrotic tumours.