The relationships between climatic variables and the frequency of human plague cases (1960-1997) were modeled by Poisson regression for two adjoining regions in northeastern Arizona and northwestern New Mexico. Model outputs closely agreed with the numbers of cases actually observed, suggesting that temporal variations in plague risk can be estimated by monitoring key climatic variables, most notably maximum daily summer temperature values and time-lagged (1 and 2 year) amounts of late winter (February-March) precipitation. Significant effects also were observed for time-lagged (1 year) summer precipitation in the Arizona model. Increased precipitation during specific periods resulted in increased numbers of expected cases in both regions, as did the number of days above certain lower thresholds for maximum daily summer temperatures (80 degrees F in New Mexico and 85 degrees F in Arizona). The number of days above certain high-threshold temperatures exerted a strongly negative influence on the numbers of expected cases in both the Arizona and New Mexico models (95 degrees F and 90 degrees F, respectively). The climatic variables found to be important in our models are those that would be expected to influence strongly the population dynamics of the rodent hosts and flea vectors of plague.