The Development of Intraindividual Interest Profiles Across Four Domains

• In Event: Subjective Task Values and Influences on Their Development From an International Perspective

Abstract

According to the Expectancy-Value Theory of achievement choice (EVT), intra-individual belief hierarchies or comparisons are key to decision making. Person-centered quantitative analyses that examine these hierarchies across domains are rare in motivational research, but are gaining attention (Wigfield & Cambria, 2010). This analysis examines interest profiles across math, reading, music, and sports in order to address the following questions:
1) To what extent are interests more universal in younger children, and more specialized in older children (Todt & Schreiber, 1998)?
2) Group average gender differences have been found in all four of these domains (Frenzel et al., 2010). Do similar gender differences exist in interest hierarchies?
3) When each domain is examined separately, trajectories of declining interest and intrinsic motivation are often found (Fredricks & Eccles, 2002). Do changes in interest profiles over time reflect this declining interest in all domains?
Data was used from the Childhood and Beyond longitudinal study, which followed 3 cohorts of predominantly White and middle- and working-class students in Southeastern Michigan (Eccles, Wigfield, Harold & Blumenfeld, 1993). Interest was assessed with a 2-item measure in the domains of math, reading, music, and sports during 6 waves of data collection, with grades 1 through 12 (excluding 11) each represented by at least one cohort. 1134 individuals contributed data from at least one time point, and these participants were 51% female, 49% male, 32% were in the youngest cohort 356, 30% in the middle cohort, and 38% in the oldest cohort.
In order to identify common profiles of interest across grades, I-states as objects analysis (ISOA) was performed following the guidelines of Bergman, Magnusson, and El-Khouri (2003); agglomerative hierarchical clustering with Ward's method and k-means relocation was used on all 4,137 time points (“I-states”) contributed from grades 1 through 12 simultaneously. A 13 cluster solution was selected on basis of variance explained (η²= 77%), proportional reduction in error variance (PRE), and theoretical significance, and this cluster solution was adequately cross-validated with random split halves of the data (Cohen’s κ= .709, .717).
Among these 13 common profiles of interest, one profile reflects high interest in all four domains while the remaining "specialized" profiles reflect low interest in at least one subject as well as high interest in at least one subject. Older students are more likely to express these more differentiated interests, with chi square tests demonstrating that grades in 1-5 range are underrepresented and grades in 7-12 range overrepresented in the most specialized clusters. While the "High all" profile represented 41% of 1st grade students, membership in this cluster was negligible by 12th grade. However, participants did not experience declines in interest uniformly in all domains, since most 12 graders maintained high interest in at least one area. In addition, women were overrepresented in the High All, Low Math, Low Sports, and High Reading Music clusters, while men were overrepresented in the Very Low Music, High Math Sports, and High Sports clusters.

Authors

• Alanna Epstein, University of Michigan - Ann Arbor

• Jacquelynne Eccles, University of California - Irvine