In the field of cognitive psychology, multiple studies have shown that humans seem able to recall a large amount of information with great detail. Thus, scientists are particularly interested in understanding the exact mechanisms by which memories are encoded in the human brain. Current models of semantic knowledge origin can be classified into two families: semantic network models and embodied cognition models. My experiment provides a way to distinguish between these two models by examining which type of interferences (semantic vs. sensorimoter) has a larger effect on semantic memory retrieval. Thirty participants were asked to perform a computer task, during which they chose the picture of an object that matched the target word from an array of objects that containd different distractors. There are three conditions, corresponding to the different types of distractor objects: unrelated distractors, sensorimotor distractors, and semantically associated distractors. Greater interference from distractors reflects greater similarity in concept representation, and thus results in slower response times. Our results show that the mean reaction time of semantically associated distractor condition is the largest. The mean reaction time of unrelated distractor (control) condition is the second largest and that of sensorimotor distractor condition is the smallest. The results support the semantic network models and indicate the facilitation of sensorimotor related objects. The results from these comparisons can not only provide us with a general idea of how the embodied cognition model and semantic network models differ from each other, but also provide us with potential applications to everyday life. First, for example, knowing the patterns from which people build concepts in mind inspires people to create memorization strategies, leading to the improvement of memorization. Second, as a growing area in cognitive psychology, embodied cognition is related to not only memory, but also emotion, perception, and language, etc.