Neurophysiological Evidence for Cognitive Map Formation during Sequence Learning

Associative Learning of an Unnormalized Successor Representation

The successor representation is known to relate to temporal associations learned in the temporal context model (Gershman et al., 2012), and subsequent work suggests a wide relevance of the successor representation across spatial, visual, and abstract relational tasks. I demonstrate that the successor representation and purely associative learning have an even deeper relationship than initially indicated: Hebbian temporal associations are an unnormalized form of the successor representation, such that the two converge on an identical representation whenever all states are equally frequent and can correlate highly in practice even when the state distribution is nonuniform.

Long-Horizon Associative Learning Explains Human Sensitivity to Statistical and Network Structures in Auditory Sequences

Networks are a useful mathematical tool for capturing the complexity of the world. In a previous behavioral study, we showed that human adults were sensitive to the high-level network structure underlying auditory sequences, even when presented with incomplete information. Their performance was best explained by a mathematical model compatible with associative learning principles, based on the integration of the transition probabilities between adjacent and nonadjacent elements with a memory decay. In the present study, we explored the neural correlates of this hypothesis via magnetoencephalography (MEG). Participants (N = 23, 16 females) passively listened to sequences of tones organized in a sparse community network structure comprising two communities. An early difference (∼150 ms) was observed in the brain responses to tone transitions with similar transition probability but occurring either within or between communities. This result implies a rapid and automatic encoding of the sequence structure. Using time-resolved decoding, we estimated the duration and overlap of the representation of each tone. The decoding performance exhibited exponential decay, resulting in a significant overlap between the representations of successive tones. Based on this extended decay profile, we estimated a long-horizon associative learning novelty index for each transition and found a correlation of this measure with the MEG signal. Overall, our study sheds light on the neural mechanisms underlying human sensitivity to network structures and highlights the potential role of Hebbian-like mechanisms in supporting learning at various temporal scales.

Route sequence knowledge supports the formation of cognitive maps

In this study, we examined the extent to which knowledge about the sequence of places encountered during route learning supports the formation of a metric cognitive map. In a between subjects design, participants learned a route until they could navigate it independently without error whilst also learning information about either the identity of places along the route (Recognition Learning condition) or the sequence of places along the route (Sequence Learning condition). In a follow-up Reconstruction of Order Task, we confirmed that participants in the Sequence Learning condition had more accurate route sequence knowledge than those in the Recognition Learning condition, despite requiring the same overall number of trials to learn the route. Participants then completed a Pointing Task to assess the quality of their cognitive map of the environment. Both groups performed above chance level, showing incidental encoding of metric information, but the Sequence Learning group produced significantly lower pointing errors than the Recognition Learning group. Further, we found that route distance between pairs of places was a strong predictor of pointing error in both groups, whilst Euclidean distance between places was a significant, but weak, predictor only for the Sequence Learning condition. The results of this study demonstrate that discrete route sequence knowledge directly supports the formation of metric cognitive maps. We consider how the results are best explained by interactions between striatal route representations and hippocampal metric representations, centered around the sequence of places acting as a scaffold for the encoding of metric information.

Caught in the ACTS: Defining Abstract Cognitive Task Sequences as an Independent Process

Cognitive neuroscience currently conflates the study of serial responses (e.g., delay match to sample/nonsample, n-back) with the study of sequential operations. In this essay, our goal is to define and disentangle the latter, termed abstract cognitive task sequences (ACTS). Existing literatures address tasks requiring serial events, including procedural learning of implicit motor responses, statistical learning of predictive relationships, and judgments of attributes. These findings do not describe the behavior and underlying mechanism required to succeed at remembering to evaluate color, then shape; or to multiply, then add. A new literature is needed to characterize these sorts of second-order cognitive demands of studying a sequence of operations. Our second goal is to characterize gaps in knowledge related to ACTS that merit further investigation. In the following sections, we define more precisely what we mean by ACTS and suggest research questions' further investigation would be positioned to address.