Abstract： When the associations among objects in the scene tend to remain unchanged as the time progressed, the repeated associations would guide attention to the target’s location more efficiently compared with the new context that changed across blocks, which is known as the contextual cuing effect. Though the process of the spatial layout has been widely interpreted, some studies that has investigated the role of object’s identities in contextual cuing effect were limited to the Euclidean property. The topological property, one of the most important objects’ identities referring to visual perception, was largely neglected. In this study, we aimed to manipulate configurations with topological property, Euclidean property, combined property as well as random configurations to test whether the predictability of the target associated with the topological property has the superiority relative to the Euclidean property. In Experiment one, a classic contextual cuing task was performed. Four types of configurations were randomly presented in the experiment. Experiment two was divided into two sessions, the studying session and the testing session. In the studying session, 24 configurations were repeated throughout the entire session, which was used to develop the learning effect. In the testing session, the previous 24 configurations were transformed into three groups, the topological repeated configurations, the Euclidean repeated configurations and the combined configurations. Meanwhile, eight random configurations were introduced as the baseline to measure the contextual cuing effect. In Experiment three, after the regularities of contexts had been learned, the topological properties of the target (experiment 3a) or distractors (experiment 3b) had been changed respectively. We explore whether topological changed configurations could capture attention by generating “new object” or lift the bound between topological regularities of the context and corresponding spatial layouts. In Experiment one, the main effects and the interaction between configuration and epoch were significant, indicating that all the three different repeated configurations obtained a remarkable contextual cuing effect. Further analysis showed that the reaction time in topological repeated configuration was faster than that in the random configuration in the 1st epoch, while the Euclidean repeated configuration had faster RTs than the random configuration from the 2nd epoch. In Experiment two, only the main effect of epoch was significant for the studying session, revealing an obvious learning effect. After configurations transformed, compared to the matched configurations in the learning session, RTs in both the topological repeated configuration and the Euclidean repeated configuration were significantly increased. Furthermore, the RTs of the topological repeated configuration were faster than the random configuration, while the RTs of the Euclidean repeated configuration were slower than the random configuration. The results demonstrated that the object’s property played an important role in contextual cuing effect, and the stability of topological-target associations made a greater contribution than Euclidean-target associations did in maintaining the contextual cuing effect. In Experiment three, both sub-experiments showed a significant learning effect in the studying session. For the testing session of Experiment 3a, the reaction time was not affected when the topological property of the target has changed. However, the accuracy of the topological changed configuration was significantly decreased than the topological repeated configuration of the Experiment 3b. Thus, Experiment three clarified the increased reaction time in the Euclidean repeated configuration, suggesting that contextual regularities of topological properties were bound to corresponding spatial layout. When topological regularities distorted, the "contextual confusion" came forth and made participants unable to utilize the context to guide attention to the target location effectively. For the first time, we have verified that the associations between objects’ topological property and the target could produce behavioral benefit than the Euclidean associations do. The association could probably be regarded as an informative cue to guide attention to the target location more efficiently. Nevertheless, the predictability of topological configuration takes priorities over Euclidean configuration during the learning course, and the association between objects’ topological property and the target has a more important significance in maintaining the contextual cuing effect.