Which insights can research on achievement motivation gain from network analysis? Comparing different network methods empirically.

Research on learning and achievement motivation often relies on the analysis of multivariate covariance with structural equation models. Recently, network analyses have been proposed as alternative or additional analytical methods with the potential to provide novel insights to motivation research. However, there are many different versions of network analyses, many different kinds of data and calculations that can be fed into and illustrated by a network, and it remains so far unclear which kind of network will provide which kind of insight to the psychology of motivation. In this article we, first, give an overview of different versions of network analyses and the kinds of research questions they potentially address. Second, we empirically compare four network analysis methods by analyzing one data set with these four approaches to compare which kinds of insights one method provides and the other possibly overlooks. We used self-report data of expectancy–value appraisals collected from N = 309 first-year university students. We chose six task value facets of intrinsic, attainment, and utility value (one facet each) and cost value (three facets) along with expectancy to represent aspects of achievement motivation specified in Eccles and Wigfield’s expectancy–value theory. These were analyzed as nodes in four different network analyses: (a) using zero-order correlations as edges, (b) using partial correlations as edges, (c) using intraindividual coendorsements, and (d) corejections among the facets. We demonstrate that conclusions about associations can differ between different network approaches and that the specific research question should determine which network to choose.