Predicting fluctuations in annual risk of Lyme disease would be useful in focusing public health efforts. However, several competing hypotheses have been proposed that point to weather variables, acorn production, or mouse abundance as important predictors of Lyme disease risk. We compared the ability of acorn production, mouse density, and four relevant weather variables to predict annual Lyme disease incidence (detrended) between 1992 and 2002 for Dutchess County, New York, and seven states in the northeastern United States. Acorn production and mouse abundance measured in Dutchess County were the strongest predictors (r ≥ 0.78) of Dutchess County Lyme disease incidence, but the increase in mouse abundance from 1991 to 1992 was contrary to a decrease in Lyme disease in the following years. The Palmer Hydrologic Drought Index (PHDI) was a significant positive predictor of Lyme disease incidence two years later for three states (0.58 ≤ r ≤ 0.88), but summer precipitation was generally negatively correlated with Lyme disease incidence the next year (-0.79 ≤ r ≤ 0.02). Mean temperatures for the prior winter or summer showed weak or inconsistent correlations with Lyme disease incidence. In four states, no variable was a statistically significant predictor of Lyme disease incidence. Synchrony in Lyme disease incidence between pairs of states was not significantly concordant with synchrony in any weather variable that we examined (0.02 ≤ r ≤ 0.21). We found that acorns and mice were strong predictors of Dutchess County Lyme disease incidence, but their predictive power appeared to be weaker spatially. Moreover, evidence was weak for causal relationships between Lyme disease incidence and the weather variables that we tested. Reliable prediction of Lyme disease incidence may require the identification of new predictors or combinations of biotic and abiotic predictors and may be limited to local scales.