119 community clusters assembled from contiguous census tracts are investigated in an ecologic study, using as the dependent variable age-adjusted lung cancer in both sexes, 1970 (CA); independent variables are: population density (PD), % white (Anglo) (WH), % Spanish surname (SP), % black (BL), % Oriental (OR), mean income per capita (IN), % of dwelling units with 1+ persons/room (CR), % annoyed by air pollution 1956 (AP), and male/female population ages 55-64 (M/F). Except for CA and AP all data are from 1970 census, compiled by John Chapman and Anne Coulson. Variables given in percent except for AP are arcsine transformed in order to more closely approximate Gaussian normality. AP is approximately normal already. The association of CA with AP is negative and small, and inclusion of other variables and structures never leads to an interesting or significant association. Positive associations with CA are with PD (r = 0.300) and WH (0.228). M/F has the highest negative association with CA (-0.285), and is positively associated with WH (0.273) and IN (0.388). M/F has a strong negative association with PD (-0.549). CR is negatively associated with CA (-0.173). In testing for the independent contribution of PD, we use the strategy of testing whether divergent methods give convergent results. Path analysis is used to reflect a number of structural assumptions among independent variables, and the association of CA and PD remains consistent and positive. After recoding the variables PD, WH, SP, BL, OR, IN, and CR into five classes of 20 and one of 19, a multiple classification analysis (MCA) is performed, using MF as a continuous co-variable. Although the unadjusted PD has a slightly greater eta than the other variables, WH and IN have a larger beta after adjustment, and the multiple R is similar to that found by path analysis, reflecting explained variance of about 0.5. The deviations of PD, CR, and BL are not monotonic but those of WH and IN are. Analysis is repeated for the 20 communities with highest IN and for the 20 with lowest IN, and the correlations (r) of CA with PD are 0.561 and 0.376 respectively. For the 20 communities with highest proportion SP, the association is a non-significant 0.108, for the 20 with highest BL, it is 0.196 and for the the 40 highest WH, the CA association with PD is 0.772. Data for the 40 most white communities are presented in a path analysis.(ABSTRACT TRUNCATED AT 400 WORDS)