Forty-nine dogs were made diabetic by total pancreatectomy. Fifteen untreated pancreatectomized animals survived a mean (+/-S.E.) of 7.0 +/- 1.1 days with a mean (+/-S.E.) plasma glucose level of 402 +/- 26 mg/100 ml before death. The pancreata of 32 dogs were distended with cold (4 degrees ) Hanks' solution, minced, digested with collagenase (600 U/ml tissue) for 15-25 minutes, and autotransplanted either into the splenic artery (three dogs), directly into the splenic pulp (21 dogs), or into the portal vein (ten dogs). Tissue infusion into the splenic artery resulted in infarction and persistent hyperglycemia. Direct implantation into the splenic pulp of tissue digested for 15, 20 and 25 minutes resulted in permanent normoglycemia (fasting plasma glucose < 150 mg/100 ml) in 7 of 8, 7 of 7, and 6 of 6 dogs respectively. Glucose tolerance test mean (+/-S.E.) K values (% decline of plasma glucose concentration/minute) in these groups two weeks after transplantation were 1.20 +/- 0.20%, 1.60 +/- 0.25 and 0.70 0.08% respectively, indicating that 20 minutes digestion was best for intrasplenic transplantation. Tissue prepared in the optimal manner (20 minutes digestion) and embolized into the liver resulted in normoglycemia in three of eight dogs, and a mean (+/-S.E.) K value of 0.77 +/- 0.10%. Both dogs receiving tissue dispersed for 25 minutes into the portal vein remained hyperglycemic. In the dogs subjected to intraportal transplantation, portal pressure rose from a mean (+/-S.E.) of 6.5 +/- 0.6 cm H(2)O before to 21.9 +/- 2.2 cm H(2)O immediately after tissue embolization, but declined to 6.5 +/- 1.0 cm H(2)O by ten weeks in animals becoming normoglycemic. We conclude that in dogs direct implantation of pancreatic tissue into the splenic pulp is superior to embolization into the portal vein or splenic artery because the splenic circulation is not compromized, portal hypertension is obviated, and glucose metabolism is best controlled as judged by glucose tolerance test K values.