Background: The method used for the pancreaticojejunal anastomosis (PJA) following the Whipple procedure is the most controversial aspect of the operation. For the 6 years 1987-1992, in 88 patients resected for periampullary tumours we used the isolated defunctioned Roux loop duct-to-mucosa PJA method with good results. Methods: To investigate whether the success of this method could be attributed to duct-to-mucosa suturing, we performed 21 consecutive ‘dunk’ anastomoses without changing any other aspect of the reconstruction. Twenty-one resections immediately preceding the 21 study patients were used for retrospective comparison. A pancreolauryl test, measuring serum flourescein at 30-min intervals by photometric analysis, was carried out in 5 patients in each group to assess exocrine function for a minimum of 6 months after operation. Results: Median age (61.9 years), indication for operation, and postoperative stay were similar in the two groups. Four pancreatic fistulas (duration 3-7 days) developed in the ‘dunk’ group and none in the duct-to-mucosa group. There were two major complications in the ‘dunk’ group and four in the duct-to-mucosa group. One patient died in the ‘dunk’ group, representing an overall mortality of 2.4%. The mean peak fluorescein concentration at 30 min was 5.36 ± (SEM) 0.58 µg/ml in the ‘dunk’ group and 5.72 ± 0.71 µg/ml in the duct-to-mucosa group; in the control group (n = 5) it was 13.4 ± 0.63 µg/ml, using a modified method of the pancreolauryl test in serum. Conclusions: We conclude that the anastomotic integrity of the PJA is compromised by using a ‘dunk’ single-layer method. It is recommended that in patients with a duct > 2 mm, a two-layer duct-to-mucosa technique is utilized, with the previously described isolated Roux loop.