If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Cattle are gregarious animals able to form social relationships. Dominance is one of the most widely studied social behaviors of dairy cattle, especially cows confined indoors. However, much of the past dairy cattle research has used an unstandardized approach, differing in definitions and conceptual understanding of dominance, as well as their methods of data collection and dominance calculation. The first of the 3 aims of this review is to evaluate how dominance relates to the social behavior of housed dairy cows. Cows engage in agonistic interactions to establish and reinforce dominance relationships. An individual's dominance is influenced by intrinsic characteristics, such as personality, and extrinsic factors, including group composition. When competing for resources, agonistic interactions can also be influenced by individual motivational differences, such as hunger, which may diminish the role of dominance in regulating competition. Our second aim is to critically review methods used to assess dominance in cows. This includes discussions on the effect of time and location of data collection on measured values as well as the viability and limitations of some dominance calculation methods. We propose that different methodologies lend themselves to different types of research questions. For example, the use of data stream-based methods that consider the sequence of interactions are useful for estimating how dominance fluctuates with changing conditions and can be used in a dynamically changing group. In contrast, matrix-based methods that aggregate social interactions may be best for identifying the social position of individuals and understanding how social characteristics influence the attributes of a stable hierarchy. Our third aim is to discuss the future of dominance research. We use a flowchart to illustrate guidelines for a more standardized approach to measuring dominance in cattle. We also identify areas in need of further conceptual clarification, suggest practical applications of dominance when managing dairy cattle, and discuss some limitations of dominance research.
A dominance relationship can be described as a “pattern of repeated, agonistic interactions between two individuals, characterized by a consistent outcome in favor of the same dyad member and a default yielding response of its opponent rather than escalation” (
). For instance, non-nepotistic (e.g., ruminant) and kinship-based (e.g., many Old World nonhuman primate species) societies involve different means of attaining social position (e.g.,
who introduced the terms hook order and bunt order to describe the social hierarchy arising from agonistic interactions of horned and dehorned cattle, respectively.
noted that when the horns were removed, the qualities of the boss cow shifted from horn-facilitated characteristics to ones of strength and tenacity. The first systematic exploration of dominance in cattle was undertaken by
, who compiled the outcomes of agonistic interactions on pasture into a winner-loser matrix, noting that the resulting distribution reflected a pattern more likely to represent dominance than chance alone. Dominance of confined cattle was first investigated by
; involving lactating cows); both groups of researchers developed an ethogram for capturing agonistic behaviors of confined cattle. Much of the research on dominance in dairy cattle, in line with typical housing practices, has since been conducted on confined groups composed only of females.
Differences in how scholars conceptualize and operationalize dominance has led to spirited debate. For instance,
critiques of the mathematical approaches and discussions regarding the relationship between aggression and dominance in previous research (including work by
, who noted that social stability and agonistic relationships are influenced by factors including social mixing, stocking density, and housing design. Approximately 2 decades later,
) summarized many of the conceptual conundrums present in past research, and showed how different methodologies for calculating dominance can lead to different outcomes. More recently,
) revisited the discussion of dominance in cattle, calling for practical solutions to overcome issues concerning data collection and calculation methods, but many questions remain unanswered (
This review has 3 aims: first, to investigate contemporary understanding of the concept of dominance in dairy cattle, focusing primarily on cattle kept in confined housing systems; second, to critically review methods commonly used to assess dominance in cattle; and third, to discuss future directions for dominance research, including ways of measuring dominance, the need for further conceptual clarification, and ways of applying the measures and concepts to improve cattle management and welfare.
Because no human or animal subjects were used, this analysis did not require approval by an Institutional Animal Care and Use Committee or Institutional Review Board.
THE CONCEPT OF DOMINANCE
Dominance is generally accepted to be multidimensional and can be considered on 3 levels (
): (1) the asymmetric agonistic relationship between a pair of animals (dyad); (2) the overall network of dyadic-dominance relationships between individuals within a group (dominance hierarchy;
Dyadic relationships can be classified as one-way (one animal always dominates the other) or 2-way (both animals dominate the other on at least one occasion, with varying degrees of asymmetry;
). Affiliative behaviors may also play a role in reinforcing dominance relationships by improving familiarization between individuals, creating positive social bonds that encourage resource sharing (
). Dominance establishment and reinforcement within a hierarchy are influenced by self-organizing social dynamics such as bystander effects (i.e., third-party observation of interactions between groupmates) that can cause the relationships that form within a group to differ from those same relationships when the dyad is isolated (
). However, compared with groups on pasture, when cattle are confined indoors, hierarchies appear to be less linear (i.e., lower proportion of transitive of relationships;
). For instance, cattle are suggested to more frequently engage in patterns of close-competition (i.e., smaller rank differences between individuals in agonistic interactions;
). Socio-positive bonds such as friendship may encourage resource sharing in which case dyadic agonistic interactions may generally be mutual and less aggressive, leading to overall small differences in dominance. We encourage further research to investigate both affiliative and agonistic bonds in groups of cattle with varying levels of familiarity to disentangle these effects.
Individual Level
An individual's dominance is a representation of their experience within the hierarchy and is generally described in 1 of 2 ways: an individually determined cardinal dominance score, or an ordinal rank that denotes the animal's position within the hierarchy relative to others (
). In the absence of knowledge of a group's social dynamics, or dominance scores, ordinal ranking risks sacrificing information about factors, such as the hierarchy distribution (
). Dominance scores offer particular benefits as they can be used to determine ranks, whereas ranks cannot be used to determine scores; thus, ranks and scores provide different and complimentary information. When used together, ranks and scores likely offer a more complete interpretation of social dynamics and individual experience than either method alone. Individual-level dominance is often expected to be temporally stable and consistent across varying circumstances (
Although dominance can only be expressed in a social environment, it includes individual characteristics independent of group dynamics and competitive context (
suggested that dominance relationships and hierarchies form on the basis of pre-existing attributes that determine an individual's dominance ability (i.e., resource holding potential;
); an animal that succeeds/fails in obtaining a reward (e.g., resource access) following engagement in agonistic behavior is more/less likely to engage in those behaviors in the future (
), and experiences may cause animals to become more or less dominant based on the social strategies of others within the environment (e.g., game theory;
). Similarly, increased group size can cause individual recognition of other group members to become more difficult and reduce the frequency of interactions between individuals (
Engagement in agonistic interactions appear to follow the predictions of state-dependent cost-benefit analysis that consider potential costs to the animal of engaging in the interaction (e.g., energetic expense, potential injury) as well as the potential benefits (e.g., resource access;
). Variation in motivation (or other state-dependent factors) among individuals in the group can alter the value of resources. Individuals with increased motivation for a resource may become more competitively successful, potentially resulting in agonistic interactions contrary to established dominance relationships (
). Changes in motivation can occur on different temporal scales and to different magnitudes. These changes are likely most common in agonistic interactions for access to resources that are limited or change in quality with time, such as fresh feed (
). Similarly, as competition for available resources increases (due to limited resource quantity or quality, increased stocking density, or increased need) there is an increase in the number of agonistic interactions between cows at the resource (feed bunk,
). Increased competition may reflect changes in motivation that potentially disrupt social stability and reduce the role of the dominance hierarchy at regulating agonistic interactions (
The most common form of dominance assessment relies only on agonistic behaviors that begin with a physical altercation and end with a clear winner or loser (e.g.,
Presence of cues from stressed conspecifics increases reactivity to aversive events in cattle: Evidence for the existence of alarm substances in urine.
) and may also be unrelated to dominance (e.g., indiscriminate social competition). The varied dominance-value of different types of interactions, and the factors that may influence these differences, has not been systematically considered in cattle dominance research.
In intensive dairy cattle systems, most agonistic interactions occur at the feeding area (i.e., feed bunk;
), possibly overshadowing any differences in the relationships associated with other resources. Differences in motivation may result in differing relationships between individuals by location (
Resource-dependent motivational fluctuations may also be temporally disparate and, therefore, when data are collected may affect which relationships are observed and to what degree confounding factors contribute. Agonistic interactions for dominance calculations have been primarily measured in 1 of 2 ways: (1) single or repeat measures of the same time periods (e.g., the hours after fresh-feed delivery;
). Measurements during these time periods may emphasize behavioral differences that reflect changes in motivation (e.g., satiation), limiting inferences from these studies. Conducting observations for dominance calculation throughout the day at multiple resources compared with a single resource (e.g., feed bunk) may reduce some potential biases, but further research is required to substantiate this claim. How time in relation to resource availability (e.g., changes in competition) affects dominance has not been experimentally explored in cattle.
CALCULATION METHODS
Multiple methods have been used for calculating dominance in dairy cattle. Most methods rely on winner-loser matrices (or actor-reactor matrices) of agonistic interactions (e.g.,
). We limit our discussion to methods that provide individuals with a cardinal score rather than a rank. In the section that follows, we discuss the strengths and limitations of 4 of the most commonly used methods for quantifying dominance in cattle, as well as 2 methods newer to cattle research.
Commonly Used Methods
One widely used method that has been proposed as a dominance index (e.g.,
) calculates an individual's score as the proportion of successes in all agonistic interactions (for equation, see Table 1). However, dominance is a relationship between 2 individuals and the identity of the participants in each interaction matters (
For each method, we outline equations and advantages and disadvantages when used to measure dominance. We include some examples from the cattle literature where the methods were used. This table was adapted from Hubbard et al. (2021a).
Calculation method
Advantages and disadvantages in dominance calculation
Advantages: can be used to reflect agonistic success when it is not possible to distinguish between individuals. Can be used continuously in groups of varying size and composition. Disadvantage: does not consider the identities of both individuals or dyadic relationships and, therefore, is not a true measure of dominance. More likely to be influenced by aggressiveness than other methods.
)
(“Dominated” indicates that the focal cow won more than they lost against an individual.)
Advantages: becomes increasingly robust to chance with more information. Determines dominance relationship by proportional comparison. Disadvantages: risks overestimating the importance of single interactions or the asymmetricity of dyadic relationships. Cannot calculate dominance when group composition changes.
)
(i.e., summed total for all interaction partners j for focal cow i)
Advantage: becomes increasingly robust to chance with more information. Disadvantages: risks overestimating the importance of single interactions or the asymmetry of dyadic relationships. Not calculated as a proportion or normalized to group size. Cannot calculate dominance in a group of changing composition.
Advantage: determines dominance relationship by proportional comparison. Disadvantages: risks overestimating the importance of single interactions or the asymmetricity of dyadic relationships. Cannot calculate dominance in a group of changing composition. Less robust to chance with more information.
)
where N = group size; DS = (w + w2) − (l + l2); w = the sum of all Dij for cow i; l = the sum of all Dji values for cow j; w2 = w weighted by the w of interaction partners; l2is l weighted by the l of interaction partners.
Dji = 1 − Dij; Pij = the proportion of interactions won by cow i against j; nij = the number of interactions between i and j.
Advantages: contains a weighting factor to limit the influence of bullying on dominance measures. Becomes increasingly robust to chance with more information. Disadvantages: cannot calculate dominance in a group of changing composition.
) WinnerRatingnew = WinnerRatingold + (1 − p) × k, LoserRatingnew = LoserRatingold − (1 − p) × k, where p is the expectation that the winner would replace the loser (see
Advantages: contains a weighting factor to limit the influence of bullying on dominance measures. Temporal bias for observing how dominance changes with time. Alterable interaction constant can be used to change the value of different types of interactions at the researcher's discretion. Can be used continuously in groups of varying size and composition. Disadvantages: does not determine dominance using the outcomes of dyadic relationships, so the results may be skewed if an actor repeatedly defeats the same reactor in multiple interactions.
) Same formula as above Elo-rating, but with order of interactions randomized instead of in temporal sequence.
Advantages: contains a weighting factor to limit the influence of bullying on dominance measures. Increasingly robust to chance with more information. Alterable interaction constant can be used to change the value of different types of interactions at the researcher's discretion. Can be used continuously in groups of varying size and composition. Disadvantages: does not determine dominance using the outcomes of dyadic relationships, so the results may be skewed if an actor repeatedly defeats the same reactor in multiple interactions.
This method has not yet been used in cows.
1 For each method, we outline equations and advantages and disadvantages when used to measure dominance. We include some examples from the cattle literature where the methods were used. This table was adapted from
Three other indices commonly used in cows all rely on actor-reactor matrices and account for identities of both individuals involved: the Lamprecht index, the Kondo–Hurnik index, and the Mendl index (for equations, see Table 1). These methods interpret dominance as a function of success in dyadic agonistic relationships. However, by retaining only binary information at the dyadic level (i.e., dominated or not), these methods neglect information about the individual's experience. Thus, these measures may overestimate the importance of single interactions or the degree of asymmetry of dyadic relationships (
) use different approaches but do overcome some of the limitations associated with the more commonly used indices described above. Both have been found to represent dominance reliably under varying circumstances (as validated using simulated data sets,
). To calculate dominance, each individual is given a weighting value based on the summed total of their agonistic success against each individual. The normalized David's score is then calculated by subtracting the unweighted and a weighted sum of an individual's proportion of losses to the sum of wins and losses from the unweighted and a weighted sum of their proportion of wins for each dyad (
). The normalized David's score corrects for the likelihood that the interaction occurred by chance (a function of the number of interactions between the 2 animals), and is normalized to the number of animals in the group (Table 1; for further description, see
). This method retains more information about the individual experience than the methods described above and can also be modified to include prior knowledge about dominance status (i.e., “informed David's score”;
). The normalized David's score becomes increasingly robust to variation as the number of interactions increase, creating a measure of dominance that is robust to temporal variation (
). The normalized David's score is recommended when seeking to calculate a stable and robust value of dominance for each individual in situations where there are few changes over time.
Elo-Rating.
The Elo-rating was initially developed as a way of rating chess players (
). This method initially assigns all individuals the same score or different scores if there is prior knowledge about dominance status (i.e., Informed Elo;
). After each agonistic interaction, this score is updated as a function of an individual's probability of winning (based on score differences between the individuals) multiplied by a constant, k (for further description, see
). Importantly, this method does not determine dominance relationships at the dyadic level and, therefore, does not align with the widely-accepted dominance definition in
. This may result in a greater influence from some factors on Elo-rating than on dyadic-level methods, such as an actor defeating the same reactor many times in a group. However, by using the temporal sequence of interactions (in contrast to matrix-based methods), Elo-rating can be used to estimate dominance in a group of changing size and composition (Table 1;
), where other methods cannot be used. The Elo-rating retains more information about the individual experience than the commonly used methods described above and is sensitive to temporal variation (
; Table 1) is calculated using the same method as the Elo-rating but with the sequence of interactions randomly ordered (i.e., not chronological). As this latter method does not have the temporal sensitivity of Elo-rating, it estimates dominance with similar robustness to variation as the normalized David's score, while still being usable for a group of changing size and composition (for examples, see
). These methods are time-consuming, and thus have resulted in measurements that are usually limited to short periods of time (a few hours), often with long intervals (days to weeks) between reassessments (e.g.,
). Future research should seek to validate how an individual's dominance changes with time. The assumption of stability may be especially problematic when animals are kept in a dynamic social environment (e.g., changing group composition), which is common practice in dairy-farming systems. Currently, little is known regarding the effect of cattle group characteristics, such as size and composition, on calculation method performance. However, larger groups may yield less steep dominance hierarchies (e.g.,
. These authors simulated data sets of agonistic interactions for individuals with assigned dominance ranks and then compared the performance (e.g., rank correlation, repeatability) of dominance calculation methods across different scenarios with varying hierarchy steepness and interaction propensities. The authors concluded that a minimum of 10 (and ideally 20) times the number of interactions as the number of animals in the group were required to reliably assess dominance when at least moderate underlying hierarchy steepness can be assumed. Unfortunately, many of the studies cited above did not collect sufficient observations to meet this minimum threshold, suggesting that their conclusions should be viewed with caution. One way to facilitate the collection of larger data sets is to record agonistic interactions automatically using data from electronic feed and water bins (
). Although, this approach also facilitates longitudinal assessments, automatic measurements of agonistic interactions at a resource can be influenced by variation in internal states (e.g., hunger), and include limitations based on assumptions (e.g., dominance being resource independent) and uncertainties during validation that should be considered and communicated transparently (e.g.,
To aid future researchers interested in studying dominance among dairy cattle, we have provided a flowchart that outlines requirements and suggestions for measuring cattle dominance (Figure 1). This 5-step approach begins by considering the research question. Second, we outline that data collection for dominance requires that both individuals engaging in an agonistic interaction (actor and reactor) must be identified by the researcher (
). If these data are not available (and more data collection is not possible), we suggest communicating the individual's score as their index value (e.g., cow A had a normalized David's score of 20), and acknowledging that the amounts of data were insufficient to determine dominance. Fourth, consider confounding factors (e.g., motivation to access a resource) that may influence measures of dominance. In this regard, further empirical research is required to determine how and whether dominance is manifested differently at different resources, levels of competition, and times of day. Finally, when calculating dominance, we encourage researchers to use methods that have been assessed for reliability and are appropriate for the intended research question. Even though the Lamprecht index, the Kondo–Hurnik index, and the Mendl index are viable methods for calculating dominance, the normalized David's score requires the same input information (i.e., an interaction matrix) and accounts for more sources of variability, likely making it the more reliable method for determining dominance in a consistent group. If the group varies in size or composition, dominance can be estimated using the temporally robust randomized Elo-rating, or temporally sensitive Elo-rating.
Figure 1A flowchart of guidelines for measuring dominance in cattle. This is intended as a living document that should evolve with findings from future research.
All dominance relationships are governed by the same fundamental principles as animals weigh costs and benefits of social interactions with reference to internal states (e.g., hunger) and external stimuli (e.g., groupmates), although costs and benefits vary by species and context (
), studying captive cattle offers some advantages compared with free-roaming and wild populations for investigating the concept of dominance. For instance, conducting social experiments is more accessible with captive animals, especially those in production systems (such as dairy cows) that regularly experience social and environmental changes during standard on-farm management practices (
). This may be useful to develop or test ethological theories, such as those relating to transitive inference (an individual inferring their likelihood of success in an unknown interaction based on observed interactions between groupmates;
). However, the effect of common dairy cattle management practices, such as sexual segregation, on social behavior is not clearly understood and should be considered when using dairy cattle as a model for social behavior research.
Some Practical Applications of Dominance
Measuring dominance may be valuable for understanding and mitigating some welfare concerns for group-housed cattle; however, methods of determining dominance may differ as to when they are best applied. For instance, temporally sensitive methods (e.g., Elo-rating) are likely ideal for observing how hierarchies develop after mixing or environmental change. Conversely, temporally robust methods (e.g., normalized David's score) may be best for determining how individual characteristics influence the attributes of a stable hierarchy. Further, recent work has highlighted the importance of the social environment and individual experience for welfare in animals (e.g.,
); therefore, longitudinal and temporally sensitive methods for monitoring changes in dominance may be useful for the early identification of compromised welfare. Moreover, subordinate cows may be particularly vulnerable (
). Temporally robust methods of calculating dominance are likely best for identifying cows that are consistently subordinate, enabling better monitoring and more timely intervention in welfare concerns.
Some Limitations of Dominance
Dominance does not encapsulate an individual's entire social experience and is limited when determined using only agonistic interactions. For example, dominance relationships may vary in their association with affiliative behaviors (e.g.,
), which may limit the reliability of measurements for dominance calculations. Moreover, estimating dominance often includes the categorization of continuous information. These categorizations occur when collecting data (e.g., definition of agonistic interaction), calculating scores, ranking, and assigning rank or dyad-associated titles (e.g., dominant or subordinate). Each categorization increases overall measurement uncertainty (
), increasing the risk of generating unreliable dominance estimates that poorly represent the social dynamics.
CONCLUSIONS
Dominance is an aspect of the social experience of cattle, the effect of which is likely to vary depending upon social and environmental context. The methods used to determine dominance have implications for how dominance is represented. Following a standardized approach for estimating dominance in cattle may be beneficial for future research.
ACKNOWLEDGMENTS
This research has been funded, in part, by a Natural Science and Engineering Research Council (NSERC; Ottawa, ON, Canada) Discovery Grant RGPIN-2021-02848 awarded to MvK. JK was funded by a NSERC postgraduate masters scholarship. General funding for the UBC Animal Welfare program comes from the NSERC Industrial Research Chair awarded to MvK and DMW with industry contributions from Dairy Farmers of Canada (Ottawa, ON, Canada), Alberta Milk (Edmonton, AB, Canada), Saputo (Montreal, QC, Canada), British Columbia Dairy Association (Burnaby, BC, Canada), Merck (Kirkland, QC, Canada), British Columbia Cattle Industry Development Fund (Kamloops, BC, Canada), Boehringer Ingelheim (Burlington, ON, Canada), Semex Alliance (Guelph, ON, Canada), Lactanet (Sainte-Anne-de-Bellevue, QC, Canada), Dairy Farmers of Manitoba (Winnipeg, MB, Canada), and SaskMilk (Regina, SK, Canada). We thank Sky Sheng (University of British Columbia, Vancouver, Canada) and Claire Lloyd (Vancouver, Canada) for helpful comments on an earlier draft of this paper. The authors have not stated any conflicts of interest.
REFERENCES
Albers P.C.H.
De Vries H.
Elo-rating as a tool in the sequential estimation of dominance strengths.
Presence of cues from stressed conspecifics increases reactivity to aversive events in cattle: Evidence for the existence of alarm substances in urine.
00760-3&id=gr1.jpg){kind=link}