GENERAL BACKGROUND

Aggression and violence are issues of global concern, identified by the World Health Organisation as a leading cause of death among young people and exerting considerable strain on individuals and society. Extensive research effort has been directed at understanding both the evolutionary origins of aggression and the factors affecting its development, but these two lines of research have proceeded largely independently. Developmental studies (mainly on humans) neglect how aggressive behaviour is shaped by evolutionary forces; while evolutionary studies (mainly on other animals) focus on fully developed,

adult behaviour and ignore how that behaviour develops during early life. This divergent approach

severely limits the conclusions we can draw about the root causes of aggressive behaviour. For a fuller

understanding of aggression, we urgently need an integrated approach in which developmental changes

in behaviour are seen as an outcome of evolved mechanisms organising behavioural development. My proposed project using the ALSPAC data forms part of a larger programme of research (5-year ERC Consolidator Grant application) aimed at understanding the evolution of developmental trajectories of aggression in humans and other animals.

AIMS

I plan to combine evolutionary theory with data on behavioural development to understand how natural selection has influenced developmental trajectories and how these interact with early life conditions to shape behaviour. Using the ALSPAC data, I will test some general predictions derived from a series of models of aggressive behaviour, in which individuals are uncertain of their own strength but can learn about this through their social experiences (see Fawcett & Johnstone 2010, Proc. R. Soc. B 277, 1427-1434). To refine the models and generate more specific predictions, I will incorporate data on patterns of physical development (growth in body size, strength etc.) and age-structuring of interactions in an attempt to explain age-dependent patterns of play fighting and aggressive behaviour. The underlying assumption of this work is that children learn about their physical capabilities through interactions with their peers.

HYPOTHESES

(i) In early childhood, there is little or no correlation between aggressive behaviour and proxies of physical strength.

(ii) As children mature, a positive correlation develops (at least in boys) between aggressive behaviour and proxies of physical strength: children (or at least boys) who are physically stronger than their peers tend to display higher levels of aggression.

(iii) As children mature, overall levels of aggression decline. The decline is more rapid for children (or at least boys) who are physically weaker than their peers.

(iv) Levels of aggression during adolescence are related to the timing of growth spurts: aggression increases immediately after growth spurts, particularly for those who undergo growth spurts earlier.

(v) There is a hump-shaped relationship between the age to which boys continue to playfight and their level of aggression later in life; boys who stop playfighting earlier show either a high or a low level of aggression, whereas those who continue to playfight until a later age show intermediate aggression.

EXPOSURE VARIABLE(S)

None.

OUTCOME VARIABLE(S)

Levels of aggression in interactions with peers, both in terms of play fighting and genuine fighting. Assessed through questionnaire responses about the tendency to start fights, hit/kick/bite other children, etc.

CONFOUNDING VARIABLE(S)

In the longitudinal analyses I wish to conduct, the principal confounding variables are age and collinear measurements of body size, such as height, weight and BMI. To deal with this issue I will centre the body size measurements around the mean for each age group, and compute residuals from the regression of weight on height. I will also need to control for the effects of socioeconomic status, fatness (obesity) levels and peer victimisation.

METHODOLOGY

To take account of the repeated-measures design inherent in longitudinal datasets, I will use multilevel models (Bryk & Raudenbush, 1992) with a first-order autoregressive error structure (Goldstein et al., 1994). This statistical approach retains individual-specific measures of growth and behaviour and therefore uses all of the available longitudinal information.