Dimensionality of Reading Skills With Elementary-School-Age Children

ABSTRACT

Confirmatory factor analyses of data from 1,501 kindergarten to 5th-grade children who completed 3 measures of decoding, 3 measures of reading comprehension, and 3 measures of listening comprehension as part of a larger study were used to identify the dimensionality of reading skills across elementary school. A 1-factor (reading) model was the best fitting model for the reading measures for kindergarten through 2nd grade, and a 2-factor (decoding, reading comprehension) model was the best fitting model for 3rd through 5th grade. Structural analyses revealed little evidence that any of the reading-comprehension measures were more or less associated with either listening comprehension or decoding than the other reading-comprehension measures, indicating that each measure assessed the same underlying construct. These results support a developmental pattern for the emergence of decoding and reading comprehension as distinctly measurable constructs, with a distinct reading comprehension construct not emerging until the 3rd grade.

Funding

This research and report was supported by grants from the Institute of Education Sciences, US Department of Education (R305F100027), and the Eunice Kennedy Schriver National Institute of Child Health and Human Development (HD052120). The views expressed herein are those of the authors and have not been reviewed or approved by the granting agency.

Notes

¹ Although it is atypical to include an orally administered measure in an analysis of reading measures, inclusion of this measure for the kindergarten sample allowed a strong test of one of our hypotheses. Tests similar in content across grades that differed in the skill necessary to complete the required test items (i.e., reading vs. listening) should behave differently in analyses designed to identify commonalities among tests. In this case, the Gates had identical content in kindergarten and first grade but differed in administration across kindergarten and the other grades in a way that changed the skill required to complete the test.

² Because the fit of the two-factor model in third to fifth grades was adequate/good on only some fit indices, we explored additional solutions to improve model fit. Results of these analyses are shown in Appendix S-C of the SOM. A two-level (classroom defined the between level) model with a single reading factor at the between level and decoding and reading comprehension factors at the within level provided good fits to the data in Grades 3, 4, and 5 across fit indices (i.e., CFIs ≥ .97, TLIs > .95, root mean square errors of approximation ≤ .06), indicating that some of the model misspecification in the one-level models was due to covariation at the classroom level. As with the one-level models, the two-level, two-factor models provided a better fit to the data than did the two-level, one-factor models for Grades 3, 4, and 5.

³ This same pattern of results was obtained using alternative approaches to testing measurement invariance (e.g., evaluating all tests in first and second grades, evaluating all tests except the Gates in kindergarten to second grades).

Additional information

Funding

This research and report was supported by grants from the Institute of Education Sciences, US Department of Education (R305F100027), and the Eunice Kennedy Schriver National Institute of Child Health and Human Development (HD052120). The views expressed herein are those of the authors and have not been reviewed or approved by the granting agency.