A latent growth model was used to investigate the longer term efficacy of phonics and integrated language arts instruction as well as amount of such instruction on children's reading development, using the nationally representative Early Childhood Longitudinal Study data set (kindergarten through fifth grade). Type and amount of instruction were derived from teachers' ratings. Children's entry-level skills and ethnicity were predictors of children's reading scores at the end of kindergarten. Ethnicity and parents' education level predicted rate of growth. Type and amount of reading instruction predicted children's reading scores. However, effects for type of instruction were time-sensitive, occurring only in kindergarten and first grade. Although children benefit from instruction in decoding and comprehension skills, instruction may not be optimally tailored to children most at risk. (Contains 9 tables, 2 figures and 1 note.)