We report a study where we investigated the effect of low or high phonically-decodable texts on young children learning to read. Two parallel series of 12 instructional reading books were used with 36 children in three schools. These books were purposely created so that each parallel book, in sequence, introduced the same number of new words. Children were randomly assigned to a condition in each classroom using a split-cluster design. Prior to reading the books, children played associated games to introduce the new vocabulary. Children were assessed at pre and post-intervention using standardised measures of word reading and comprehension. Our results demonstrate a statistically significant difference and large effect size for reading comprehension in favour of the low phonically-decodable texts. The findings challenge the assumption that children find highly decodable text easier to read, and may have implications for reading policies and classroom practice.