Significance: Several accelerometer equations have been developed for predicting energy expenditure (EE) during ambulatory activity such as walking. In many such studies, walking movement and EE have been measured during treadmill (TM) walking rather than during overground (OG) walking. The purpose of this study was to investigate whether the relationship between speed (mph) and EE (ml/kg/min) is similar during TM and OG walking in healthy adults in order to examine the construct validity of such studies. Design: Men and women (N = 75) recruited from three sites completed three 6-minute randomly assigned slow (2.0-2.6 mph), intermediate (2.7-3.2 mph), and fast (3.3-4.0 mph) TM walking trials. Subsequently, stride rates from each TM trial were replicated using a clip-on metronome during three OG trials in an effort to reproduce TM walking speeds, following the protocol of Gordon et al. (1999). During all trials, EE was measured using indirect calorimetry and heart rate was measured every minute to determine steady state. Data from the last 2 minutes of each trial were analyzed. TM speed was determined using a tachometer and OG speed was determined from time to complete the trial and measured distance. Results: Preliminary analyses from site 1 (n = 25) indicated that mean OG walking speed was significantly (p < .05) faster than TM walking speed during slow (2.85 ± 0.29 mph vs. 2.34 ± 0.22 mph) and intermediate (3.23 ± 0.24 mph vs. 3.06 ± 0.22 mph) trials but was not significantly (p > .05) different during the fast (3.74 ± 0.35 mph vs. 3.75 ± 0.22 mph) trial. Prediction accuracy of EE from speed was significant (p < .05) from both TM walking (r = .83, SEE = 1.94 ml/kg/min) and OG walking (r = .75; SEE = 1.84 ml/kg/min). The associated regression equations indicated a similar slope (4.67 vs. 5.00) but different intercepts (– 0.31 vs. – 3.02). Interestingly, 3 METs (a commonly used EE cutoff for moderate physical activity) corresponded to walking 2.3 mph on the TM, but 2.7 mph during OG walking. In summary, this evidence implies that accelerometry equations developed from TM walking protocols may not be valid for predicting EE during OG walking.Keyword(s): exercise/fitness/physical activity, measurement/evaluation, research