Measuring social structure: a comparison of eight dominance indices

Karen L. Bayly, Christopher S. Evans*, Alan Taylor

*Corresponding author for this work

    Research output: Contribution to journalArticlepeer-review

    57 Citations (Scopus)


    Measurement of social status is an important component of many behavioural studies. A variety of techniques have been developed and adopted, but while there have been some analyses of index properties using simulated data, the rationale for selecting a method remains poorly documented. As a first step in exploring the implications of index choice, we compared the characteristics of eight popular indices by applying each to the same data set from interactions between male fowl Gallus gallus, the system in which social hierarchies were first described. Data from eight social groups, observed over four successive breeding seasons, were analysed to determine whether different indices produced consistent dominance scores. These scores were then used in tests of the relation between social status and crowing to explore whether index choice affected the results obtained. We also examined the pattern of dominance index use over the last decade to infer whether this has likely been influenced by tradition, or by taxa of study animal. Overall agreement among methods was good when groups of birds had perfectly linear hierarchies, but results diverged when social structure was more complex, with either intransitive triads or reversals. While all regression analyses revealed a positive relationship between dominance and vocal behaviour, there were substantial differences in the amount of variance accounted for, even though the original data were identical in every case. Index selection can hence perturb estimates of the importance of dominance, relative to other factors. We also found that several methods have been adopted only by particular research teams, while the use of others has been taxonomically constrained, patterns implying that indices have not always been chosen solely upon their merits. Taken together, our results read as a cautionary tale. We suggest that selection of a dominance index requires careful consideration both of algorithm properties and of the factors affecting social status in the system of interest.

    Original languageEnglish
    Pages (from-to)1-12
    Number of pages12
    JournalBehavioural Processes
    Issue number1
    Publication statusPublished - Jul 2006


    • social status
    • methods
    • behaviour in groups


    Dive into the research topics of 'Measuring social structure: a comparison of eight dominance indices'. Together they form a unique fingerprint.

    Cite this