Measuring behavior patterns and evaluating time-sampling methodology to characterize brush use in weaned beef cattle

With growing interest in provision of brushes for cattle and the implications of brush use for behavioral development and welfare, there is a need to validate methodology for quantifying grooming behavior. Our objectives were to characterize patterns of brush use, including bouts, diurnal activity, and individual variability over 24-h periods, and to validate time-sampling methodologies to characterize these traits, including instantaneous recording at various time intervals and continuous recording for subsets of the day. Data sets from previous experiments involving steers (experiment 1; n = 18) and heifers (experiment 2; n = 64), consisting of start and end times of brush use continuously recorded from video, were used to analyze brush use. We extrapolated data sets representative of a range of instantaneous recording intervals and compared daily brush duration and bout characteristics with corresponding values from continuous recording using linear regression. To assess validity of sampling subsets of the day, we selected 2-h time periods representative of different functional parts of the day and compared hourly brush rates with continuous data using Spearman's rank order correlation (r_s). Brush use was variable among individuals. All steers used the brush in experiment 1, but 17% (n = 11 of 64) of heifers in experiment 2 did not. Bout analysis revealed that individuals used the brush for an average of 7 to 8 brush bouts lasting 4 to 6 min, leading to an average of 24 and 36 min/d for experiments 1 and 2, respectively. Cattle used the brush mainly during daylight hours, with peaks around sunrise, sunset, and the afternoon. Instantaneous recording at intervals less than 1 to 3 min, depending on the experiment, provided good estimates of daily brush use duration (R² > 0.95 and slope and intercept not different from 1 and 0, respectively), but intervals >3 min were less reliable. For bout characteristics, the intercept of the modeled line differed from 0 for most recording intervals for both experiments, and the slope differed from 1 for recording intervals >30 s in experiment 1, suggesting that time sampling may have underestimated true values. Of the 2-h periods compared with 24 h of observation, 1800 to 2000 h was most highly correlated (r_s = 0.84) for experiment 1, and 1800 to 2000 h and 1400 to 1600 h were the most highly correlated (r_s = 0.71 and 0.74, respectively) for experiment 2 with daily values. When using time-sampling methods to characterize brush use, we suggest that the recording interval used and time of day observed should be carefully considered, as time sampling at an interval of 1 to 3 min may measure daily brush use duration, but continuous recording may be required to capture bout characteristics.