Reading test data from the National Assessment of Educational Progress (NAEP) were scaled using the unidimensional item response theory model. Data were collected for students aged 9, 13, and 17. To determine whether the responses to the reading items were consistent with unidimensionality, four different methods were applied: (1) principal component analysis of phi and tetrachoric correlation matrices; (2) principal component analysis of the image correlation matrix, a method based on the work of Guttman; (3) R. D. Bock's full-information factor analysis; and (4) P. R. Rosenbaum's test of unidimensionality, monotonicity, and conditional independence. Balanced incomplete block (BIB) spiralling was used with this year's NAEP to assign test items to booklets. This permitted the estimation of inter-item correlations, but resulted in an unusual pattern of missing data. Results from the analyses conducted for each age group were different from the analysis of the 25 items administered in all three samples. It was concluded that it was not unreasonable to regard the reading items as measures of a single dimension of reading proficiency. (Author/GDC)