The objectives of this study were to evaluate heritability estimates of residual feed intake (RFI), DMI, and feeding behavior traits in growing Holstein heifers (n = 609; Initial BW = 246 ± 43 kg), and to examine corresponding genetic correlations. DMI and feeding behavior traits were measured using a GrowSafe System for 70 to 100 d (15 trials) at the Ohio Heifer Center (STgenetics research farm) while heifers were fed a corn-silage based ration. RFI was computed within trial as the difference between actual and expected DMI from linear regression of DMI on mean BW^0.75 and ADG. Six feeding behavior traits were evaluated, including frequency and duration of bunk visit (BV) and meal events, head-down duration (HDD), and the ratio of BV events per meal event. Pedigree information consisting of 10 generations of 4,513 animals from 616 founder animals were used in the genetic evaluation. Heritability estimates were obtained for each trait using multivariate mixed models which included fixed effects of trial and age, and genetic correlations estimated by pairwise bivariate linear mixed models for any 2 traits. The heritability estimates for RFI, DMI and feed:gain were 0.15, 0.26 and 0.04, respectively, which were lower than the heritability estimates observed for the 6 feeding behavior traits that ranged from 0.30 to 0.50, and for ADG (0.54). As expected, RFI was highly correlated with DMI (0.52), but not ADG (0.10), although RFI was negatively correlated with final BW (−0.33). RFI was positively correlated with HDD (0.35), BV frequency (0.72), meal frequency (0.39) and duration (0.44), and the ratio of BV events per meal event (0.31), but not with BV duration (0.01). In general, the magnitude of the genetic correlations between RFI and feeding behavior traits were higher than those between DMI and feeding behavior traits, suggesting that feeding behavior traits were more predictive of RFI than DMI. The moderate to high genetic correlations between RFI and feeding behavior traits indicate that they may be useful indicator traits for predicting feed efficiency. Accordingly, future research should further evaluate the use of feeding behavior traits in multivariate genomic prediction models to improve the accuracy of RFI predictions.