To verify the hypothesis that chronic atrophic gastritis (CAG) is a precancerous lesion for gastric cancer, we investigated the correlation between the prevalence rates for CAG determined by levels of serum pepsinogens (I and II) of randomly selected, healthy, blood donors and the mortality rates for gastric cancer among four prefectures (Akita, Iwate, Miyagi, and Okinawa) in Japan. The prevalence rates for CAG according to a criterion that the level of pepsinogen I is below 70 ng/ml and the pepsinogen I to pepsinogen II ratio is below 3.0, were the highest in Akita for both sexes (38.8 percent in males, 37.8 percent in females) and the lowest in Okinawa (13.9 percent in males, 12.3 percent in females), where mortality rates for gastric cancer are the highest and the lowest in Japan, respectively. The correlation between the prevalence rates for CAG and the standardized mortality ratios for gastric cancer among these four prefectures was statistically significant in males (r = 0.97, P = 0.03), but less significant in females (r = 0.92, P = 0.08). These data strongly support the hypothesis that CAG is a precancerous lesion. The limitations of our cross-sectional study and the advantages of measuring the levels of serum pepsinogens in epidemiologic studies and in mass-screening programs for gastric cancer are discussed.