Abstract:Relational cognition—the ability to infer relationships that generalize to novel combinations of objects—is fundamental to human and animal intelligence. Despite this importance, it remains unclear how relational cognition is implemented in the brain due in part to a lack of hypotheses and predictions at the levels of collective neural activity and behavior. Here we discovered, analyzed, and experimentally tested neural networks (NNs) that perform transitive inference (TI), a classic relational task (if A > B and B > C, then A > C). We found NNs that (i) generalized perfectly, despite lacking overt transitive structure prior to training, (ii) generalized when the task required working memory (WM), a capacity thought to be essential to inference in the brain, (iii) emergently expressed behaviors long observed in living subjects, in addition to a novel order-dependent behavior, and (iv) expressed different task solutions yielding alternative behavioral and neural predictions. Further, in a large-scale experiment, we found that human subjects performing WM-based TI showed behavior inconsistent with a class of NNs that characteristically expressed an intuitive task solution. These findings provide neural insights into a classical relational ability, with wider implications for how the brain realizes relational cognition. The ability to infer underlying relationships in the world—or "relational cognition"—is fundamental to both human and animal intelligence. Despite this importance, it remains generally unknown how relational abilities are realized in the brain. Addressing this lack of knowledge requires generating testable hypotheses regarding the activity of neurons when relationships are inferred. In this study, we discovered artificial neural networks that perform transitive inference (TI), a classic relational task that has the form "if A > B and B > C, then A > C." We found that these neural networks show behavioral patterns that match those seen in humans and animals performing TI, and, further, performed TI in different ways, thereby providing different testable hypotheses regarding neural activity. We also conducted a large-scale behavioral study in humans, which suggested systematic differences in biological accuracy across the neural networks. These findings provide new hypotheses and insights regarding how TI, and relational abilities more generally, can be realized in the brain.

Inferior parietal cortex represents relational structures for explicit transitive inference

Representation of the inferred relationships in a map-like space

Emergent neural dynamics and geometry for generalization in a transitive inference task

The Cognitive and Neural Basis of Transitive Inference

Inter-Brain Network Underlying Turn-Based Cooperation and Competition: A Hyperscanning Study Using Near-Infrared Spectroscopy

Abstract representations emerge in human hippocampal neurons during inference

An Event-related Potential Study on Neural Basis of Transitive Inference

Cognitive representation in transitive inference: a comparison of four corvid species

Learning to infer transitively: serial ordering on a mental line in premotor cortex

Relational Complexity Modulates Activity in the Prefrontal Cortex During Numerical Inductive Reasoning: an Fmri Study.

Common and distinct neural substrates of forward-chaining and backward-chaining syllogistic reasoning

When Memory Fails, Intuition Reigns

Deductive‐reasoning brain networks: A coordinate‐based meta‐analysis of the neural signatures in deductive reasoning

Neural Correlates of Causal Inferences in Discourse Understanding and Logical Problem-Solving: A Meta-Analysis Study

Differentiable cortical networks for inferences concerning people's intentions versus physical causality

Involvement of the left inferior frontal gyrus in predictive inference making.

A mathematical theory of relational generalization in transitive inference

Mapping Domain-Selective and Counterpointed Domain-General Higher Cognitive Functions in the Lateral Parietal Cortex: Evidence from fMRI Comparisons of Difficulty-Varying Semantic Versus Visuo-Spatial Tasks, and Functional Connectivity Analyses

Human Brain Activity Related to the Perception of Spatial Features of Objects

The Neural Basis for Spatial Relations

Brain Activation and Deactivation in Human Inductive Reasoning: an Fmri Study