Most of the previous studies found that when learning new categories, interleaving stimuli across different categories during study enhances classification performance on novel stimuli in comparison with blocking stimuli during study. The discriminative-contrast hypothesis and the attention attenuation hypothesis interpret the interleaving effect from attention. However, the existing empirical studies cannot fully support it. Thus, the main goal of this study is to explore whether attention affects the interleaving effect. Besides, working memory (WM) influences attention. Combining the differences in individual attention under different practice schedules, it can be assumed that WM affects the interleaving effect by affecting the attention. It is the second goal that this study intends to investigate. Using the family resemblance pictures as the experimental materials, participants completed observed (Experiment 1) or feedback (Experiment 2) category learning while they performed the numerical Stroop task as a dual task. Under the observed learning condition, stimulus and the category label were presented together while under the feedback learning condition, stimulus was presented without the category label. At the same time, in order to explore whether the discriminative-contrast hypothesis and the attention attenuation hypothesis are correct, participants’ eye movement data were recorded by eye-tracking. The results reveal that: (1) when doing observed category learning (Experiment 1), compared to studying in the blocked condition, participants’ performance to the novel stimuli will be better if they study in the interleaved condition. Moreover, in the dual task condition, the interleaving effect is more obvious. At the same time, the results show that participants pay more attention to the non-diagnostic features in the blocked condition, while they pay more attention to the diagnostic features in the interleaved condition. These results are consistent with the discriminative-contrast hypothesis. In addition, the attention decreases with the increase of the position in the blocked condition while the attention does not change too much with the position in the interleaved condition. These findings provide evidence for the attention attenuation hypothesis. (2) when doing feedback category learning (Experiment 2), participants do not realize the existence of the practice schedule, thus the advantages of the practice schedules could not be exerted. Interesting, unlike in Experiment 1, results in Experiment 2 do not show difference in the attention under the two practice schedule conditions. Instead, in the case of lower WM load (single task), participants tend to classify the similarities within the same category, and they are beneficial from blocking; while the WM load is higher (dual task), participants tend to identify the difference between categories, and interleaving learning enhances their performance. The results suggest that (1) when doing observed category learning, attention influences the interleaving effect, which supports the discriminative-contrast hypothesis and the attention attenuation hypothesis. However, when doing feedback category learning, it is necessary to further explore whether attention affects the interleaving effect. (2) WM influences the interleaving effect, but combining the results of these two experiments, it at least proves that WM does not completely influence the interleaving effect by affecting attention.