The age-predicted maximal heart rate (PMHR) formula, 220--age, is frequently used for identifying exercise training intensity, as well as determining endpoints for submaximal exercise testing. This study was designed to identify variables discriminating those with actual maximal heart rates considerably above or below that predicted from the 220--age equation. Subjects included 2010 men and women ranging in age from 14 to 77 yr. Stepwise discriminant analysis was performed using maximal heart rate error groups as the dependent variable, and selected preexercise test characteristics as predictors. The HR error groups were based on the difference between the measured and PMHR as follows: below (> or = 15 beats.min-1 below PMHR), within (+/- 14 beats.min-1 of PMHR), and above (> or = 15 beats.min-1 above PMHR). A contrast of the below and above groups identified age, resting HR, body weight, and smoking status as predictors of group membership (P < 0.01) for both men and women. The overall canonical correlation was 0.282 and 0.294 for the men and women, respectively. Older age, higher resting HR, lower weight, and non-smoking were related to the above group, while the inverse was related to the below group. Standardized coefficients suggest that age and resting heart rate for the men, and age and smoking status for the women were the most potent variables for discriminating extreme deviations between measured and PMHR.(ABSTRACT TRUNCATED AT 250 WORDS)