Understanding how humans form knowledge structures from a stream of information is a question with implications for both human cognition and artificial intelligence. Structural learning is affected by the training regime, which can be either blocked into trials of similar forms or interleaved into trials of different forms. Previous studies show mixed results with some showing that blocked regimes improve different types of learning [1, 2, 3], while others show advantages of blocked regimes [4, 5, 6]. Additionally, few studies investigated the effect of regime on structural learning specifically [7,8,9]. We thus designed a study to explore which training regime best supports structural learning. We hypothesized that a blocked training regime encourages participants to rely on a working memory strategy [10], limiting them to memorizing individual configurations that may later be forgotten. In contrast, an interleaved training regime is expected to overload working memory, prompting participants to identify patterns across different configurations. This, in turn, may drive them to extract the underlying structure of the task, facilitating generalization. Our results show that participants in the blocked group performed better in training, while the interleaved group performed better in the test phase. This could be likely due to the blocked group relying more on memorization, while the interleaved group is forming a structure. To explore this further, we developed an incremental Gaussian Mixture Model (iGMM) simulating learning mechanisms in each regime [11]. The model performs better at test during the interleaved training regime, which is consistent with our behavioral data. During blocked training, the model relies more on a recency effect while during interleaved training it creates a global representation of the task structure. Our results suggest that interleaved training encourages the formation of a global representation, aligning with theories of distributed representations in hippocampal learning [8,12].