The purpose of this paper is to re-examine the spatial diffusion of Spanish influenza in Japan, which swept over the world three tunes from 1918 to 1921. According to the com-mon opinion they have said that the epidemic of Spanish influenza had not a clear diffusion process in Japan. But minutely examining the trend of the first outbreaks reported in “Influenza”, a report of the Spanish influenza, and “Statistics of the Causes of Death of the Empire of Japan”, the author believed to be doubtful to conclude so. For example, Fig. 2 and Fig. 3, showing the inter-prefectural diffusion process estimated from the peak month of the number of influenza deaths, suggest that the epidemic tended to diffuse from the western part to the eastern part in Japan. Then, in order to investigate its diffusion process in detail, the author analyzed the monthly influenza mortality of each prefecture from July 1916 to June 1926, applying S-mode factor analysis. The 120 (the number of months)±46 (the number of prefectures) data matrix was subjected to factor analysis, and three factors with eigenvalues of greater than 1.0 were extracted (Table 1). Judging from Fig. 4 and Table 2 showing the correlations between some variables and factor loadings, the following interpretation seems to be possible. Factor I, accounting for 78.8% of the total variance, is related to the contagious opportunity prescribed by the distance from the major ports (Kobe, Moji, Nagasaki, Shimonoseki and Osaka etc.) in the western part in Japan (Fig. 6), and distinguishes the Western Japan from the rest of the country. Factor II, accounting for 8.5% of the total variance, is related to the contagious opportunity corresponding to the hierarchical urban system, and distinguishes the urban areas from the other areas. Factor III, accounting for 7.9% of the total variance, is related to the factors promoting the prevalence of epidemic within the household (temperature, average number of persons per household, % of people under 15 years and over 60 years, personal income and number of doctors per 1, 000 persons), and distinguishes the Eastern Japan from the rest of the country. Further cross correlations for lag values from -3 to +3 months between factor scores were calculated to clarify the time lead and lag relationship between three factors (Fig. 7). Consequently it turned out that there was not a marked time lag between Factor I area and Factor I area, but the epidemic could spread from Factor II area to Factor III area two months behind. Therefore, on the basis of these cross correlations, the diffusion route could be presumed roughly as follows; The Spanish influenza diffused toward the east, invading through the major ports in the Western Japan, while it diffused into the urban areas, ‘moving down’ the urban hierarchy from Tokyo via Yokohama which was then the largest port in the Eastern Japan. And, two months later, the epidemic spread from the urban areas (especially Keihin region) to the Eastern Japan. Above-mentioned result differs from the common opinion, but it suggests that there is a regular spatial order in the epidemic process of Spanish influenza in Japan.