Abstract:Networks are a useful mathematical tool for capturing the complexity of the world. In a previous behavioral study, we showed that human adults (N=23, 16 females) 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 non-adjacent elements with a memory decay. In the present study, we explored the neural correlates of this hypothesis via magnetoencephalography (MEG). Participants 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. Significance statement We conducted a MEG study in which human adults were passively exposed to sequences of tones organized in a sparse community network structure. Despite the uniform transition probabilities between tones, participants’ brain activity exhibited sensitivity to the network structure. Notably, a consistent “deviant” response was observed at ∼150 ms when the sequence switched between communities. A long-tail exponential decay in tone representation allowed overlapping representations of successive sequence elements, facilitating long-range associative mechanisms. This binding mechanism adequately accounted for various scales of sequence learning, bridging the gap between statistical and network learning approaches.

An Oscillator Ensemble Model of Sequence Learning

Oscillatory multi-timescale mechanisms underlying multisensory sequence prediction

Coherent noise enables probabilistic sequence replay in spiking neuronal networks

LEARNING IN AN OSCILLATORY CORTICAL MODEL

Adaptive oscillators support Bayesian prediction in temporal processing

An Oscillator Neural Network Retrieving Sparsely Coded Phase Patterns

Unclocklike oscillators with frequency memory for the entrainment of biological clocks

Active oscillatory associative memory

Sequence reproduction, single trial learning, and mimicry based on a mammalian-like distributed code for time

Computation with Sequences in a Model of the Brain

Dissociated Amplitude and Phase Effects of Alpha Oscillation in a Nested Structure of Rhythm- and Sequence-Based Temporal Expectation

A Recurrent Neural Network for Rhythmic Timing

ELiSe: Efficient Learning of Sequences in Structured Recurrent Networks

Networks that learn the precise timing of event sequences

Thunderstruck: The ACDC model of flexible sequences and rhythms in recurrent neural circuits

Hebbian Imprinting and Retrieval in Oscillatory Neural Networks

Computation with Sequences of Assemblies in a Model of the Brain

Resonant synchronization and information retrieve from memorized Kuramoto network

Common Software for the ALMA project

Long-horizon associative learning explains human sensitivity to statistical and network structures in auditory sequences