Abstract
A set of 6 grapheme-morpheme correspondence skills measured by the Kennedy Institute Phonics Test (KIPT) (Guthrie, 1973) was tested for goodness of fit to the Quasi-Markov Simplex model for hierarchical structures among variables (Joreskog, 1970). The hierarchical order from simple to complex was hypothesized to be: single letter production, initial letter recognition, consonant-cluster recognition, nonsense word recognition, consonant-cluster production, nonsense word production. The correlation matrices for these variables reported by Guthrie (1973) for normal and slow readers respectively, were tested for goodness of fit to the hierarchical model by means of the maximum likelihood estimation method proposed by Joreskog (1970). The Ss were 7 year old children (N = 19) and 9 year old children (N = 19), both reading at the grade 2 level. The null hypothesis of no difference between observed and estimated variance-covariance structure was not rejected for normal or slow readers using the chi-square statistic (p = .073 and p =.132, respectively) or the Mann-Whitney U test (.05 < p < .10 for normal readers, p = .10 for slow readers). The implications of a hierarchical model of reading subskills for planning reading instruction in the beginning stages were discussed.