Objective To examine the reliability and convergent validity of physical activity (PA) and inactivity estimates obtained with the past-week Modifiable Activity Questionnaire (PWMAQ).Design The PWMAQ, an interviewer-administered questionnaire, was administered twice, one week apart, during visits 3 and 4 of six total visits. Intra-class correlation coefficients (ICC) between administrations of the PWMAQ were used to assess the reliability of summary estimates. Spearman rank-order correlation coefficients (ρ) were used to examine the associations of PWMAQ summary estimates with temporally matched and averaged accelerometer data in all participants and then stratified by whether the data were reflective of usual PA. Setting Data were obtained from the Evaluation of Physical Activity Measures in Middle-Aged Women (PAW) study. Subjects Sixty-six women, mean age 526 (sd 54) years. Results The reliability of the PWMAQ physical inactivity estimate suggested substantial agreement over one week (ICC = 077, 95 % CI 057, 082; P < 00001). With the exception of light-intensity PA, the PWMAQ leisure PA estimate was significantly associated with averaged accelerometer data (ρ = 033-076; P < 005). For both temporally matched and averaged accelerometer data, correlation coefficients were higher between the PWMAQ estimate and moderate-walk- to vigorous-intensity PA in those who indicated that reported activity was reflective of usual PA; however, the association with moderate-lifestyle-intensity PA was higher in those reporting that data were not reflective. Conclusions The PWMAQ is a reliable and valid measure of leisure PA levels in middle-aged women and supports subsequent studies evaluating this questionnaire in other population subgroups. © 2010 The Authors.