Background: Reading comprehension is a multifactorial process, but one of its features has been relatively under-investigated--it is the strategy used when answering reading comprehension questions. In order to find the correct answer, children can either respond to questions about a text relying on their text memory or look back at the written text. This study analyses (i) which strategy, memory or look-back is more frequently adopted in primary school children according to grade level and type of text (expository vs narrative) and (ii) the preferred (most frequently used) and more efficient (higher number of correct responses) strategy for poor oral comprehenders. Poor oral comprehenders were selected for having adequate nonword decoding skills but impaired oral comprehension, compared with good oral comprehenders. Methods: The total sample comprised 1,417 primary school children. A standardised test assessing reading and comprehension of oral and written language was administered. Results: In the comparison among children attending different grades, older children showed a more frequent use of the look-back strategy for the expository text, whereas no age-related difference emerged for the narrative text. Poor oral comprehenders (n = 88) showed a different pattern, compared with matched good oral comprehenders (n = 88), for both preference and efficiency of strategies. Despite having a globally poorer performance in reading comprehension, poor oral comprehenders were more accurate than good oral comprehenders when using the look-back strategy. Conclusions: The complex pattern of results obtained reinforces the idea that intervention strategies should be personalised based on individual characteristics and specific for the type of task and text.