Background The international literature has presented several criteria for classifying people in terms of physical activity pattern, making it difficult to compare the results.

Objective The study aims to evaluate the reliability between two criteria used to classify physical activity pattern, in adults living in Ribeirão Preto, SP, Brazil, in 2006.

Methods Using data from a cross-sectional population-based study, the International Physical Activity Questionnaire (IPAQ) and the American College of Sports Medicine/American Heart Association (ACSM/AHA) criteria were applied. The κ statistic estimated by point and 95% CIs, stratified by gender, age, education level and income was applied to evaluate reliability between the two criteria.

Results The results showed an agreement close to the unity (κ≈1) between the two criteria, in both genders, with crude coefficients k_male=0.95 (95% CI 0.83 to 1.06) and k_female=0.93 (95% CI 0.85 to 1.01). The lowest magnitude of κ were found in the age group of 60 years and over in both genders [k_male=0.87 (95% CI 0.60 to 1.15); k_female=0.88 (95% CI 0.70 to 1.06)], and among illiterate women or with <01 year of education [k=0.72 (95% CI 0.34 to 1.09)].

Conclusions The IPAQ and ACSM/AHA criteria showed practically the same ability to classify participants in terms of physical activity levels, allowing comparison of population-based studies that have used the IPAQ instrument to obtain information about physical activity practice. The results pointed out the higher prevalence of “actives” when the second criteria was applied and “walking” was the main practice of physical activity.