Selective breeding for over 35 generations has led to four replicate (S) lines of laboratory house mice (Mus domesticus) that run voluntarily on wheels about 170% more than four random-bred control (C) lines. We tested whether S lines have evolved higher running performance by increasing running economy (i.e., decreasing energy spent per unit of distance) as a correlated response to selection, using a recently developed method that allows for nearly continuous measurements of oxygen consumption (VO2) and running speed in freely behaving animals. We estimated slope (incremental cost of transport [COT]) and intercept for regressions of power (the dependent variable, VO2/min) on speed for 49 males and 47 females, as well as their maximum VO2 and speeds during wheel running, under conditions mimicking those that these lines face during the selection protocol. For comparison, we also measured COT and maximum aerobic capacity (VO2max) during forced exercise on a motorized treadmill. As in previous studies, the increased wheel running of S lines was mainly attributable to increased average speed, with males also showing a tendency for increased time spent running. On a whole-animal basis, combined analysis of males and females indicated that COT during voluntary wheel running was significantly lower in the S lines (one-tailed P=0.015). However, mice from S lines are significantly smaller and attain higher maximum speeds on the wheels; with either body mass or maximum speed (or both) entered as a covariate, the statistical significance of the difference in COT is lost (one-tailed P> or =0.2). Thus, both body size and behavior are key components of the reduction in COT. Several statistically significant sex differences were observed, including lower COT and higher resting metabolic rate in females. In addition, maximum voluntary running speeds were negatively correlated with COT in females but not in males. Moreover, males (but not females) from the S lines exhibited significantly higher treadmill VO2max as compared to those from C lines. The sex-specific responses to selection may in part be consequences of sex differences in body mass and running style. Our results highlight how differences in size and running speed can account for lower COT in S lines and suggest that lower COT may have coadapted in response to selection for higher running distances in these lines.