This study had two main objectives. One was to quantify persistence of 700 mb heights from one month to the next using various antecedent period lengths. The second was to determine predictive skill for calendar month 700 mb heights using antecedent and simulated prognostic height information. The data consisted of standardized 700 mb heights at 126 grid points over the Northern Hemisphere in the period 1948–80. Persistence was quantified by computing pattern correlations using mean heights over various period 1engths (ranging from 1 to 28 days) ending on a forecast day (assumed to be the 28th of the month) on the one hand and the following calendar month mean heights on the other. These calculations showed that optimum period length and its corresponding average pattern correlation vary widely over the course of the year. The correlations are largest in summer and winter (maximum of 0.24), while the optimum period length ranges from 1 day (March–April) to 28 days (May–June and July–August). Regression analysis was used to determine predictive skill of calendar month 700 mb heights on a grid point by grid point basis, stratified by calendar month, using antecedent heights (optimized period length), simulated prog heights, and a combination of both as predictors. Simulated prog heights were produced by using an exponential damping function on observed heights in the first ten days of the prog period. In all, seven different regression trials were performed. The skill of models using simulated prog heights was much superior to those using antecedent heights, though it is possible that in low latitudes in the warm months antecedent heights add important information. These trials suggest further that local simulated prog heights are most important. As usual, much spatial as well as temporal variation of skill is evident so the forecaster should consult the maps of skill for the particular time of year.