Evolution of Grouping Patterns

  • Robin I. M. Dunbar
Part of the Studies in Behavioural Adaptation book series (SBA)


In Chapter 2, I argued that primate societies should be viewed as multilayered sets of coalitions based on relationships that differ in intensity, character and function. The most obvious and perhaps important respect in which these clusters of relationships differ from each other is, of course, their spatial localisation and temporal stability. Traditionally, observers have always recognised that animals of many species spend much of their time in the physical company of conspecifics rather than wandering alone. In this chapter, I shall concentrate mainly on the evolution of group-living.


Group Size Predation Risk Habitat Quality Capuchin Monkey Large Group Size 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Note that exactly the same set of curves will result if groups form for reasons of resource defence: the point is only that some factor promotes the formation of large groups because lifetime reproductive output is greater in large groups. Several candidate principles can fill this role. Thus, although the interpretation of the curves will differ, the shape will remain substantially the same. I prefer predation as the explanation because the weight of evidence favours this hypothesis. Note also that we could not use the results of this analysis to test the predation-defence hypothesis, since both the resource-defence and the predation-defence hypotheses would yield the same results.Google Scholar
  2. 2.
    For graphical convenience, I have represented the optimum group size as lying at the intersection of the relevant cost and predation graphs. In fact, with graphs of these particular shapes, the optimum will generally lie slightly to the right of their intersection, but this will not make any difference to the conclusions we draw.Google Scholar
  3. 3.
    Because of the paucity of data in the outer arms, a curvilinear regression is heavily biased by the mass of data in the mid-range and so generates a nearly linear relationship between group size and rainfall. The slope of this relationship is not, however, significant: linear regression set by least squares gives a slope of b = −0.012 (r2 = 0.039, t29 = −0.643, p > 0.30 2-tailed). But this clearly ignores a very significant negative relationship over the mid-section within the range 500–1200 mm rainfall: slope b= −0.089 (r2 = 0.435, t21= −3.793, p < 0.01 2-tailed). An N-shaped curve of this kind is often the product of an interaction between two different curves: in this case, these could be an inverted-U-shaped curve and a positively increasing exponential curve, each reflecting the influence of a different environmental variable.Google Scholar

Copyright information

© Robin Dunbar 1988

Authors and Affiliations

  • Robin I. M. Dunbar
    • 1
  1. 1.Department of ZoologyUniversity of LiverpoolUK

Personalised recommendations