Temporally delayed linear modelling (TDLM) measures replay in both animals and humans

Abstract

There are rich structures in off-task neural activity which are hypothesised to reflect fundamental computations across a broad spectrum of cognitive functions. Here, we develop an analysis toolkit – Temporal Delayed Linear Modelling (TDLM) for analysing such activity. TDLM is a domain-general method for finding neural sequences that respect a pre-specified transition graph. It combines nonlinear classification and linear temporal modelling to test for statistical regularities in sequences of task-related reactivations. TDLM is developed on the non-invasive neuroimaging data and is designed to take care of confounds and maximize sequence detection ability. Notably, as a linear framework, TDLM can be easily extended, without loss of generality, to capture rodent replay in electrophysiology, including in continuous spaces, as well as addressing second-order inference questions, e.g., its temporal and spatial varying pattern. We hope TDLM will advance a deeper understanding of neural computation and promote a richer convergence between animal and human neuroscience.

Model-based aversive learning in humans is supported by preferential task state reactivation

“…Applying this classifier to task data provides an index of reactivation likelihood at each time point in every trial. Using a form of temporally delayed linear modeling ( 13 , 14 , 28 ) of lagged cross-correlations between state reactivation vectors for different states, we then determine whether reactivations occur in a sequence consistent with task structure, where positive values indicate forward replay and negative values indicate reverse replay.…”

“…While prior studies identified evidence for sequential replay averaging across the entire trial ( 15 , 18 , 28 ), here, we adopted a related approach used in our prior work ( 13 , 29 ). This approach measures sequenceness within sliding windows that quantify fluctuations in evidence for replay within a trial enabling it to uncover time-varying patterns of replay, rather than assume a constant level of replay across the entire trial.…”

“…To detect sequential reactivation of state representations in these data, we used a cross correlation method ( 13 , 14 , 29 ). While this method does not allow us to investigate forward and reverse replay independently because of the high autocorrelation present in the signal (instead, we use the difference between these metrics to obviate this autocorrelation), the cross-correlation method is more amenable to windowed analyses than general linear modeling approaches ( 28 ). Here, we took the product of the state × time reactivation matrix and the task transition matrix, where the time course of each state’s reactivation probabilities was correlated with the time courses for all other states, lagged by up to 20 time points (representing 200 ms) in steps of one.…”

Impaired neural replay of inferred relationships in schizophrenia

“…Recent advances now enable detection of spontaneous neural replay in humans noninvasively, using magnetoencephalography (MEG) ( Kurth-Nelson et al., 2016 ; Liu et al., 2021a , 2019 ; Wimmer et al., 2020 ). This allows researchers to pose new questions relating to abstract and non-spatial forms of cognition in humans, which are difficult to address in rodents, in addition to pursuing pressing questions related to neural processes underlying neuropsychiatric conditions.…”

“…We test a hypothesis of abnormal replay and cognitive map construction in schizophrenia, leveraging methodological advances that enable both detection of fast spontaneous neural replay ( Kurth-Nelson et al., 2016 ; Liu et al., 2021a , 2019 ) and the representational content of neural responses ( Barron et al., 2020 ; Deuker et al., 2016 ; Diedrichsen and Kriegeskorte, 2017 ; Luyckx et al., 2019 ) in humans. Using a non-spatial structure-learning task ( Liu et al., 2019 ) and neural data derived from MEG, we measured spontaneous replay of inferred task structure during a post-learning awake rest period.…”