Significance: Accurately determining the percentage of the U.S. population that meets the physical activity (PA) recommendation has been critical for public health research and policy decision-making. In the past, the decisions were made mainly based on self-reported questionnaire data. Very recently, the National Health and Nutrition Examination Survey (NHANES) started to use Actigraph, a type of accelerometer, to objectively measure PA. There is inconsistency, however, in setting cut-off scores when determining moderate- and vigorous-intensity PA (MVPA) due to different Actigraph thresholds being employed in practice (Ham et al, 2007). Thus, there is an urgent need to investigate the classification agreement when different methods are used in determining people's status on meeting PA recommendation or not. The purpose of this study was to compare agreement among five different methods (three Actigraph-counts-based methods and two survey-based cutoffs) in classifying whether U.S. adults meet or do not meet the PA recommendation. Design: Adults' demographic, PA monitor and questionnaire information from the NHANES 2003-2004 data sets were used, which included 5,041 participants (48% were males and 52% females). Three Actigraph cutoff points of 1,952, 190.7 and 574 counts min-1 (Freedson et al., 1998; Hendelman et al., 2000; Swartz et al., 2000) and two survey-based methods of 150 minutes MVPA per week and 3 kcal/kg/day (Corbin et al., 1994) were used to classify participants into 2 categories: Meet or do not meet the PA recommendation. Agreement coefficients and Cronbach's alpha among the 5 methods were computed. SAS 9.1 survey procedures were employed for the data analyses. Results: The proportion of American adults meeting the PA recommendation varied considerably according to different classification methods, ranging from 0.88% to 64%! The lowest proportion was from the “1,952” Actigraph cutoff and the highest from the “150 minutes MVPA per week” survey-based method. The agreement between different classification methods ranged from .29 to .90, among which, methods of using cutoffs of 150 minutes MVPA per week and activity counts of1952 min-1 reached the lowest agreement (.29), the two survey-based methods produced relatively high agreement (.77) and the agreement of classifications between methods of Freedson et al. (1998) and Swartz et al. (2000) was the highest (.90). Cronbach's alpha was.52 for the five classification methods, and was .40 for Actigraph thresholds and .79 for survey-based methods, respectively. Conclusions: Method choice decides what proportion of the US population meets the PA recommendation. Validation study to select the appropriate method is needed.Keyword(s): adult physical activity/fitness, assessment, disease prevention/wellness