For gastric cancer patients who have no peritoneal seeding at a macroscopic level but positive results in the peritoneal lavage cytology (PLC), the prognostic benefit expected by surgical resection is still controversial. During the period 1975-1994 as series of 417 consecutive patients without distant organ metastases underwent surgical resection for gastric cancer that had invaded the subserosal or deeper layers of the stomach wall. Immediately after laparotomy, the pouch of Douglas was washed with 100 ml of physiologic saline solution, and the fluid was collected for cytologic examination (four slide glasses) using Giemsa and Papanicolaou staining methods. According to the macroscopic (P) and cytologic (Cyt) results, the 417 patients were classified into three groups: P(+) (n = 97); P(-)/Cyt(+) (n = 25); and P(-)/Cyt(-) (n = 295). Their 3-year survival rates after surgical resection were 4%, 24%, and 48%, respectively (p = 0.0001: P(-)/Cyt(+) vs. P(-)/Cyt(-); p = 0.0018: P(-)/Cyt(+) vs. P(+). Among the 25 P(-)/Cyt(+) patients, postoperative survival was not associated with the T stage, N stage, cellular atypism, or cluster formation but with the number of cancer cells per slide during PLC. The 3-year survival rate was 35% for the subgroup with fewer than 10 cancer cells per slide (17 patients) and 0% for the other subgroup with 10 or more cancer cells per slide (8 patients) (p = 0.017). For P(-)/Cyt(+) patients, who represent a subgroup of gastric cancer patients with an intermediate survival rate between the P(-)/Cyt(-) and P(+) patients, the number of cancer cells observed during PLC offers a potent prognostic indicator for the gastrectomy.