When learning procedures in real life, learners generally use action atomization strategies (interleaving instructions consultation and execution) and need several repetitions to acquire the skill. However, in studies on procedural learning, delayed execution paradigms (2 separate steps consisting of instructions consultation, then execution) and/or only 1 execution of the task, are commonly used methods. The aim of the present study was to examine the effects of those two paradigms on procedural learning. The effects of the procedure repetition were also studied. Forty-two undergraduate medical students were asked to make 5 stitches, either in an atomized or a delayed manner. The hypothesis was that participants who had learned in an atomized manner would perform better than those who had learned in a delayed manner, because the former approach may have helped them reduce task-related cognitive load. Results show an effect of repetition of the procedure on participants' performance, and confirm the need to examine more than one execution of a procedure when studying procedural learning. Moreover, different outcomes in terms of suture quality and speed of instructions consultation were observed for the two learning paradigms. This implies that studying learning following a delayed execution paradigm may lead to distorted results compared to procedural learning occurring in real-life situations.