Autism has been linked to various speech and language deficits, including difficulties in immediate sentence repetition. In this study, we adopted the sentence repetition paradigm to investigate the effects of mode of presentation (speech versus song), semantic content (news-like, story-like, and non-semantic), and listening background (quiet versus noisy), on immediate sentence repetition in a group of 30 Mandarin-speaking autistic children and 29 non-autistic children. Results showed that repetition accuracy was poorer in the autistic group, although the group difference may be partially the product of variation in verbal ability and age. Both groups performed worse for sung as opposed to spoken sentences, and for news-like and non-semantic relative to story-like sentences. We discuss how factoring in multiple variables, including individual differences, can further our understanding of sentence repetition ability in autism.