This article explores the consequences of four models of language learning and acquisition on textbook design, and the sequencing and distribution of lesson activities. The model developed by Levelt (1989) to explain oral speech production, which was later extended to second language learning by De Bot (1992), points to the importance of lexical knowledge and subconscious, automatic processing. Paul Nation (2001) cited ample evidence for the effectiveness of an equal distribution of meaningful input, language focused instruction, meaningful output and fluency practice in his four strands model. Segalovitz and Gatbonton (2005) proposed the ACCESS-Model, an acronym that stands for automatization in communicative context of essential speech segments. Merrill Swain’s model of output hypothesis, which insists that learners learn not only from input but also from output and thus pushes for learner output at an early stage, has also been the subject of much discussion. As far as I can see, none of these four models have been sufficiently reflected in terms of their consequences for textbook sequences or task and exercise design. Based on practical examples of tasks and exercises, this article will put forward a few ideas and suggestions based on these models and try to come to some general conclusions with regard to a research-based textbook design.