A total of 611 patients with carcinoma of the oesophagus or gastric cardia were operated on between July 1982 and December 1989. Resection was performed in 491 patients (one-stage, 483; two-stage, eight), bypass operation in 97, and 23 had exploration alone. The anastomoses of 580 patients with one-stage resection and bypass operations were evaluated. Hand-sewn anastomosis using a single layer of continuous absorbable monofilament suture was performed in 304 patients (221 resections and 83 bypasses). A stapled anastomosis was performed on 276 patients (262 resections and 14 bypasses). Following resection, there were 11 (5 per cent) anastomotic leaks in the hand-sewn group and ten (3.8 per cent) in the stapled anastomosis group (P = 0.69). Excluding anastomotic leaks, hospital mortality and anastomotic recurrence, stricture occurred in 18 of 172 hand-sewn anastomoses (10.5 per cent) and in 57 of 195 stapled anastomoses (29.2 per cent) (P less than 0.001). In patients who had bypass operations there were 12 anastomotic leaks, ten in the hand-sewn group (12.0 per cent) and two in the stapled anastomosis group (14.3 per cent). Only two of the discharged patients with bypass developed anastomotic strictures, a low incidence probably because of short survival. In addition, there were 245 subsidiary anastomoses made in the abdomen by the hand-sewn method as part of the reconstructive procedure, and there was one leak. The results of this non-randomized study suggest that hand-sewn anastomosis using a single layer continuous technique for the oesophagus is as safe as the use of circular staplers; hand-sewn anastomosis is less likely to become stenotic.