Robinson's (2001) Cognition Hypothesis claims that more complex tasks promote interaction and language development. This study examined the effect of task complexity in the learning of request‐making expressions. Task complexity was operationalized as [+/− reasoning] following Robinson's framework. The study employed a pretest–posttest research design and was conducted over 6 weeks. Korean junior high school students from 3 classes (N = 73) were assigned to one of the following groups: simple, complex, or control. Both task groups performed a pretest, 2 collaborative tasks, and 2 posttests, whereas the control group performed the pre‐ and posttests only. Learners’ oral interaction during tasks was audiorecorded and analyzed by the number of pragmatic‐related episodes (PREs). Learners’ knowledge of request expressions was measured by a discourse completion test (DCT). The results indicated that task complexity levels influenced the occurrence of PREs, but no difference was found in the quality of task outcome between the simple and complex groups. In terms of learning outcomes, both task groups outperformed the control group, but no difference was found on the immediate posttest. However, the complex group maintained its gain on the delayed posttest.