How Eye Read: A Social Network Approach

The aim of the current paper is to offer a unique perspective on eye movement analysis in reading research by applying techniques from social network analysis to examine integration processes between sentences during reading. In a first step, we explored how network measures relate to the often-used duration measures in reading research in order to examine whether there is an additional value in using network measures. In a second step, we further explored how differences in network measures are related to text (i.e., topic structure) and reader characteristics (i.e., WMC). Thirty-one participants read three short expository texts. Four network measures at the sentence level were calculated for the three texts: strength, betweenness centrality, harmonic centrality, and local clustering coefficient. Correlations were computed between first-pass reading time and second-pass reading time and the network measures. Network measures were analyzed with (generalized) linear mixed-effects models. The results show that strength is strongly correlated to second-pass reading time. Betweenness, harmonic centrality, and the local clustering coefficient are not related to these often-used duration measures and thus capture aspects of integration processes that cannot be captured with duration measures. The results demonstrated that strength and betweenness centrality are related to reader’s WMC. It was also shown that strength, harmonic centrality, and local clustering coefficient were related to the topic structure of the text. This study demonstrates that a social network approach offers a novel perspective on moment-to-moment integration processes during reading.