Comparison of convergent and independent pathways in neural networks during second-order conditioning and blocking...

This study explores the role of convergent versus independent pathways in a neural network model to simulate blocking (Blk) and second-order conditioning (SOC). Convergent connections refer to the intersection of connections from one hidden layer unit to an adjacent unit in feedforward neural networks, whereas independent pathways involve exclusive, nonconverging connections. This research compares five network architectures with varying degrees of convergent connectivity in SOC and Blk. These phenomena are relevant as they illustrate how prior reinforcement history influences learning about a second stimulus, either by enhancing or inhibiting the conditioned response to it. The findings indicate that networks with convergent connections better replicate Blk than SOC, whereas those with independent pathways more accurately model SOC than Blk. Some exceptions were observed, which may have implications for the conceptual analysis of deep learning models. Future work could incorporate theoretical insights into the underlying mechanisms of stimulus-stimulus and stimulus-response associations. Additionally, the findings from this study could inform model-driven hypotheses in neurobehavioral research. (PsycInfo Database Record (c) 2025 APA, all rights reserved).