Subjective methods to quantify temperament in beef cattle are insensitive to the number and biases of observers

Abstract Associations between excitable temperament and many economically relevant traits have been established, resulting in an increased focus on docility in breeding programs in cattle. Several methods have been proposed to quantify temperament during normal handling procedures, which appear to differ in their usefulness and accuracy of measure. The objective of this study was to determine the merit of using chute score (CS), exit score (ES), and exit velocity (EV) to evaluate temperament in a production setting, by assessing (i) the impact of experience and number of observers on inter-observer reliability, and (ii) the repeatability of these measures. Over three consecutive years, a factorial design of two measurement protocols [frequent (F), infrequent (IN)], and three recording periods, each one month apart, was used. Each year, twenty commercial Bos taurus heifers, 2-wk post weaning, were randomly assigned to each protocol. Each day of observation, heifers were moved into a squeeze chute and their heads caught. A CS from 1 (docile) to 6 (aggressive) was assigned to loosely restrained cattle by as many as six observers of varying experience. When released from the chute, EV over a distance of 2 m was obtained, and an ES of 1 (docile) to 5 (aggressive) assigned by the same observers. Heifers were evaluated repeatedly over three months, with some heifers scored on as many as nine occasions. Inter-observer reliability was calculated using the irr package in R. Kendall's coefficient of concordance (KCC) and intra-class correlation coefficients (ICC) were calculated for the average of two, three, or four experienced observers for all nine days of the study. Observers were then divided by experience [experienced (E), inexperienced (IE)] and compared to a benchmark observer using these same statistics. Repeatabilities were calculated using ASReml fitting protocol, event, and their interaction as fixed effects, body weight as a covariate, and sire and year as random effects. While reliability decreased with increasing number of observers for both measures, agreement was consistently higher when assessing ES compared to CS. Both measures had reliabilities indicative of acceptable agreement (ICC CS ≥ 0.74, ICC ES ≥ 0.90; KCC CS ≥ 0.79, KCC ES ≥ 0.93). However, inexperience reduced reliability of CS assignments, with more variation in scores than in E observers. All methods were highly repeatable. These findings demonstrate the usefulness of subjective methods for quantifying temperament on site. Their lack of susceptibility to individual bias in distinguishing behaviors makes them useful criteria for evaluating temperament in cattle.