Redefining dominance calculation: Increased competition flattens the dominance hierarchy in dairy cows.

Dominance hierarchies are known for mitigating conflicts and guiding priority of access to limited resources in gregarious animals. The dominance hierarchy of dairy cows is typically investigated using agonistic interactions, often monitored at the feed bunk right after fresh feed delivery when competition is high resulting in frequent interactions. Yet, the outcome of agonistic interactions during times of high competition may be more influenced by cows' high valuation of fresh feed than their intrinsic attributes, such that the dominance hierarchy constructed using agonistic interactions under high versus low competition times might differ. We tested how the structure of the dominance hierarchy changes in relation to different levels of competition in a dynamic group of 48 lactating dairy cows over 10 mo, with 6 cows exchanged every 16 d for a total of 159 cows. Using a validated algorithm we continuously detected the actor and reactor of replacement behaviors in 30 feed bins as cows competed for feed. We also calculated the percentage of occupied feed bins to characterize competition at the moment of each replacement. These data were combined to create hierarchies using Elo ratings, separately for 25 occupancy levels ranging from 13% to 100%. For each 1% rise in feeder occupancy, hierarchy steepness fell by 2.41 × 10-3 ± 9.71 × 10-5 (SE), and the percentage of dyads where both cows replaced each other rose by 0.13% ± 0.01%. At the highest feeder occupancy level in comparison to the lowest one, we observed 7.57% more dyads in which the dominant individual (those won more interactions at the lowest feeder occupancy) started to lose proportionally more. The magnitude of decrease in the winning rate of the dominant individual in those dyads also got amplified by 1.06 × 10-3% ± 1.37 × 10-4% (SE) for each 1% increase in feeder occupancy. These findings illustrate how inferred hierarchies vary with competition, with high competition flattening the hierarchy due to increased success of subordinate animals. We suggest that during heightened competition, increased valuation of resources can affect competitive success more than the individual's intrinsic dominance attributes. We recommend against calculating dominance hierarchies based on agonistic interactions during periods of high competition alone, and more generally urge researchers to differentiate agonistic interactions based on context when constructing dominance hierarchies.

Effects of group size on agonistic interactions in dairy cows: a descriptive study.

Group-housed cattle may engage in agonistic interactions over resources such as feed, which can negatively affect aspects of welfare. Little is known about how contextual factors such as group size influence agonistic behaviour. We explored the frequency of agonistic interactions at the feeder when cattle were housed in different-sized groups. We also explored the consistency of the directionality of agonistic interactions in dyads and of the number of agonistic interactions initiated by individuals across the group sizes. Four replicates of 50 cows each were assessed in two group-size phases. In Phase 1, cows were kept in one group of 50. In Phase 2, these same cows were divided into five groups of 10, maintaining stocking density (i.e., ratio of animals to lying stalls and feed bunk spaces). We measured agonistic replacements (i.e., interactions that result in one cow leaving the feed bin and another taking her place) at an electronic feeder using a validated algorithm. We used these data from Phase 1 to calculate individual Elo-ratings (a type of dominance score). Cows were then categorised into five dominance categories based upon these ratings. To ensure a consistent Elo-rating distribution between phases, two cows from each dominance category were randomly assigned to each small group of 10 cows. The mean ± SE number of replacements per cow was similar regardless of whether the cows were housed in groups of 50 (34.1 ± 2.4) or 10 (31.1 ± 4.5), although the groups of 10 were more variable. Further, 81.6 ± 7.7% (mean ± SD) of dyads had the same directionality across group sizes (i.e., the same individual won the majority of interactions in the dyad) and individuals were moderately consistent in the number of replacements they initiated (intraclass correlation coefficient = 0.62 ± 0.11; mean ± SD). These results indicate that the relationship between group size and agonistic behaviour is complex; we discuss these challenges and suggest new avenues for further research.

Crowd sourcing remote comparative lameness assessments for dairy cattle.

Lameness assessments are rarely conducted routinely on dairy farms and when completed typically underestimate lameness prevalence, hampering early diagnosis and treatment. A well-known feature of many perceptual tasks is that relative assessments are more accurate than absolute assessments, suggesting that creating methods that allow for the relative scoring of which cow is more lame will allow for reliable lameness assessments. Here we developed and tested a remote comparative lameness assessment method: we recruited nonexperienced crowd workers via an online platform and asked them to watch 2 videos side-by-side, each showing a cow walking, and to identify which cow was more lame and by how much (on a scale of -3 to 3). We created 11 tasks, each with 10 video pairs for comparison, and recruited 50 workers per task. All tasks were also completed by 5 experienced cattle lameness assessors. We evaluated data filtering and clustering methods based on worker responses and determined the agreement among workers, among experienced assessors, and between these groups. A moderate to high interobserver reliability was observed (intraclass correlation coefficient, ICC = 0.46 to 0.77) for crowd workers and agreement was high among the experienced assessors (ICC = 0.87). Average crowd-worker responses showed excellent agreement with the average of experienced assessor responses (ICC = 0.89 to 0.91), regardless of data processing method. To investigate if we could use fewer workers per task while still retaining high agreement with experienced assessors, we randomly subsampled 2 to 43 (1 less than the minimum number of workers retained per task after data cleaning) workers from each task. The agreement with experienced assessors increased substantially as we increased the number of workers from 2 to 10, but little increase was observed after 10 or more workers were used (ICC > 0.80). The proposed method provides a fast and cost-effective way to assess lameness in commercial herds. In addition, this method allows for large-scale data collection useful for training computer vision algorithms that could be used to automate lameness assessments on farm.