The label effect observed in inductive reasoning suggests that label, by providing an effective cue, such as category knowledge or feature similarity promotes the performance of inductive reasoning. Consequently, there are two kinds of hypotheses to explain the phenomenon, one is the view of category which emphasizes the category information provided by a label, and the other one is the view of similarity which focuses on the physical similarity of labels. Previous studies with unfamiliar artificial materials usually focused on the surface knowledge about category carried by a label. However in reality, a real label, say “tiger” or “lion”, usually contains deep knowledge, such as the information about body, size, behavior, living environment and others, especially in Chinese. But this kind of information carried by a label is seldom to distinguish from surface knowledge, and how the surface and deep knowledge carried by a label affect the process of inductive reasoning is still not clear. In this study, two experiments were conducted to explore the influence of knowledge on label effect during adults’ inductive reasoning. To avoid the influences of other factors, such as visual picture or verbal introduction and to investigate the label effect in a more precise and sensible way, a relationship judgment with 5 point rating was used as the inductive reasoning task. In Experiment 1, natural animal names were used as materials. To compare the influence of surface and deep knowledge carried by labels, 24 animal names were selected, which could be divided into familiar and unfamiliar groups with equal number of items, according to a prior familiarity assessment. For the familiar materials, a label of animal name would carry both surface and deep knowledge; while for the unfamiliar materials, only surface knowledge was carried by a label. In Experiment 2, artificial animal names were used as materials to further investigate the influence of surface and deep knowledge on label effect, and the labels were manipulated whether contained a natural animal character or not. In the condition with animal character, both surface and deep knowledge would be carried in the artificial label, while in the condition without animal character, only surface knowledge of category would be carried. And the label effects were manipulated by sharing common ending characters or not in both experiments. The result reveals that consistent with previous studies, label effect was observed in both experiments, and label promoted inductive reasoning. However in Experiment 1, the label effect in real natural materials decreased significantly in familiar condition, indicating deep knowledge affects the label effect during inductive reasoning. And in Experiment 2, when a label with a natural animal character, participants preferred to judge a closer relationship for the artificial label pairs, which further support the influence of deep knowledge on label effect. In both Experiments, the interaction between label effect and other factors supported the view of category, since the view of similarity could not explain the phenomenon of similar physical features in the conditions with common labels but different effect sizes of label.