The findings indicate that the popular assumption of a positive, unchanging link between city size and crime rates may be misleading. Controlling for extraneous variables and allowing for a squared relationship between crime and city size suggests either no significant association or a changing association between city size and crime rates. As city population size increases over time, population size is related to crime rates, but only up to a point; then as cities reach sizes beyond that point, crime rates become lower. For three crimes--homicide, burglary, and robbery--any changes in population size, whether up or down, is likely to result in increases in crime rates. The study concludes that observed differences between cross-sectional and longitudinal patterns of city size and crime rate association do not constitute an "anomaly." Rather, it is an inevitable consequence of comparing static and dynamic features of the same phenomenon, much like comparing a snapshot of a fixed moment with a motion picture of changing observations over time. The cities included in this analysis were part of a larger set of 584 used by Miethe et al. (1991) in their study of the effect on crime rates of changes over two decades in variables representing routine activities and social disorganization theories. Although not a random sample, the cities all had populations over 25,000 at each of the 4 data points, and among them the independent and dependent variables all had substantial variation. Dependent variables were homicide, burglary, assault, robbery, rape, and auto theft known to police for 1960-1970, 1070-1980, and 1980-1990. Independent variables were the population of the cities and population changes. 7 tables and 72 references