Background: Sentence repetition is gaining increasing attention as a source of information about children's sentence-level abilities in clinical assessment, and as a clinical marker of specific language impairment. However, it is widely debated what the task is testing and therefore how informative it is. Aims: (1) To evaluate the effects of different types of long-term linguistic knowledge on immediate recall, (2) to assess age sensitivity of repetition tasks designed to evaluate these effects, and (3) to establish if the effects are similar across typologically different languages. The study also considers the implications of the findings for the use of sentence repetition as a research and clinical assessment tool. Methods & Procedures: Participants were 50 English-speaking and 50 Czech-speaking typically developing 4-5-year-olds. Children's ability to recall sequences of items was compared in seven linguistic conditions ranging from fully well-formed sentences to sequences of non-words. In each condition, children repeated blocks of successively longer stimuli to establish their span. Outcomes & Results: Results showed significant but differential effects of all linguistic factors in both languages. While syntactic violations and presence of non-words dramatically reduced children's span, semantic implausibility and the removal of sentence prosody played a significant but much smaller role. Familiarity of function words was more important than familiarity of content words. The effects of different linguistic factors on spans were the same for both languages and did not change between 4 and 5 years, although average spans increased over this age range. Conclusions & Implications: Children's ability to repeat sentences is more dependent on their familiarity with morphosyntax and lexical phonology than semantics or prosody, with function words of particular importance. Findings have implications for the use of recall in clinical assessment and as a research tool.