Video recordings of behavioral activities including eating, ruminating, drinking, standing, and lying were monitored to determine the minimum number of sampling days and sampling frequency required to obtain reliable estimates of these behaviors. Eight continental crossbred heifers, individually housed in a tie-stall barn with total mixed ration provided once per day, were divided by BW into two blocks and assigned to each of the two dietary treatments in a crossover design: high-forage diet (HF, forage:concentrate ratio 70:30) and low-forage diet (LF, forage:concentrate ratio 30:70). The cows were monitored continuously using a digital video recording system for 6 d in each of the two periods and a trained observer manually recorded the behavioral activities by minute. Mean time spent performing each behavior for 6 d with instantaneous samples of 1 min was compared with those obtained using fewer sampling days (1, 2, 3, 4, and 5 d) and less frequent scanning (2, 3, 4, 5, 10, 15, 30, and 60 min) using linear regression analysis, and the minimum number of sampling days or frequency was determined. Diet did not affect the accuracy and precision of predicting behavior from video recordings of the cows. When sampling days and scanning intervals were combined, the minimum recommended sampling frequency for accurately estimating a specific set of behaviors of beef heifers in tie stalls was: eating, 2 d with 4-min intervals; ruminating, 3 d with 4-min intervals; drinking, 2 d with 3-min intervals; and standing and lying, 2 d with 15-min intervals. Increasing sampling frequency beyond these minimums further enhanced the accuracy and precision of predictions. The total time of each behavioral activity was different between cows fed HF and LF diets with eating, ruminating, total chewing, and standing of heifers fed the HF diet greater (P < 0.01), but the lying (P < 0.01) and drinking time (P = 0.028) of heifers fed the LF diet greater. Meal patterns were different (P < 0.05) between the treatments except when meal size (kg) was expressed as DM or OM intake. For rumination patterns, the mean or maximum bout length (min/bout) was greater, but the maximum or minimum length of time heifers took to ruminate after eating was lower for the HF diet. From a practical standpoint, it is recommended to use 3 days of observations scanned at an interval of 4 min to reliably estimate the behavior of cows. If the focus is only lying and standing time, then 2 days with 15 min intervals can be used.