Representing Objects and Features in Long-Term Memory: A Case for Direct Feature-Feature Binding

Objects are typically remembered with their accompanying features, suggesting that objects and features are bound together in memory. Yet, there is an ongoing debate about the configuration of such bindings, particularly whether multiple features of an object are directly linked or are only indirectly linked via the object. Here we set out to test the binding configuration of object-feature representations in long-term memory in three experiments, bridging research from the domains of object and source memory. In two experiments, participants were presented with object icons with a specific color and orientation and later performed a recognition test in which feature combinations were presented with old or new objects. Participants had to indicate both whether the object shown was old or new and which out of four feature combinations they had previously seen. In a third experiment, participants were presented with only single new objects at test and had to indicate whether the feature combination shown was old or new. This procedure allowed us to control for the presence of object information at test, which constituted a major confound in previous research. Using multinomial processing tree modeling to disentangle cognitive processes, we found evidence for a stochastic dependence of the retrieval of object features in the absence of object information. This suggests that different features of an object can be directly linked in long-term memory, challenging theoretical accounts positing that features are only indirectly linked via common links to the object and instead providing proof of concept for direct feature-feature binding.