Sex differences in social networks
11
friendships were calculated from reciprocated friendship and best friend nominations. Friendship
relations were identified on the basis of reciprocated nominations.
Measures and Identification of Social Networks
Mean sizes of play networks are reported for all games, team games and non-team games.
Social network membership was identified on the basis of the aggregation of the observed play
networks at each time point and included only members of the class studied. A symmetrical person-
person data matrix was derived for each time point and each cell contained the proportion of total
observations that each child in the class appeared in each target child's play network. Cells on the
diagonal were empty. Proportions, rather than frequencies, were used to take account of the slight
variation in pupil observation frequency.
The identification of network boundaries is a problem for all social network research and all
approaches require the use of a relatively arbitrary inclusion-exclusion threshold. As part of this
research and in order to identify the most applicable threshold, observers were asked to keep a field
journal on pupils’ interactions, groups, group roles and games on the playground. After the first data
collection point, these notes were used to identify the social networks that were perceived to exist
on the playground. This was then compared to the networks identified in the data firstly through the
use of a variety of thresholds for inclusion in a network and secondly through the use of the Social
Composite Map procedure from Cairns, Gariёpy, Kindermann & Leung (1989). The threshold
approach, but not the SCM approach, allows examination of emergent network structure. The
researchers’ notion of the networks corresponded closely to that resulting from the SCM procedure
and to a 25% threshold imposed on the data for inclusion in a network1. That is, children who spent
25% or more of their scans with each other were included in a network which we termed a ‘group’.
To examine the internal structure of networks and to examine how members connect together into
groups, two further thresholds were imposed on the data to provide three levels of involvement in
social networks. Children observed together for 50% or more of their combined scans were