David Foster

Introduction

David J. Foster is a neuroscientist whose recordings from the rodent hippocampus have reshaped how researchers think about memory consolidation, planning and the neural basis of spatial navigation. He holds appointments in the Department of Psychology and the Helen Wills Neuroscience Institute at the University of California, Berkeley, and retains an adjunct appointment in the Solomon H. Snyder Department of Neuroscience at the Johns Hopkins University School of Medicine. He is best known for showing, with Matthew Wilson at the Massachusetts Institute of Technology, that hippocampal place cells reactivate recent experience while an animal is awake and resting, and that much of this reactivation runs backwards in time. That observation turned replay from a curiosity of sleep into a candidate mechanism for how the brain consolidates experience and evaluates rewards.

Although Foster was trained as a physicist and as a computational neuroscientist rather than as a psychologist, his experimental work has become central to the psychology of memory and spatial cognition. It sits directly downstream of the discovery of place cells by John O'Keefe and of grid cells by May-Britt Moser and Edvard Moser, work recognised by the 2014 Nobel Prize in Physiology or Medicine. His papers on awake replay are among the most frequently cited in the modern literature on the neural basis of memory.

Early Life and Education

Foster's scientific training began in physics. He took a B.Sc. in physics at Imperial College London in 1992, then spent a year as a research consultant at BT Labs before moving deliberately toward the study of neural computation. He completed an M.Sc. in neural computation at the University of Stirling in 1994 and spent the following year as a research associate in the Department of Experimental Psychology at the University of Oxford, working with Edmund Rolls on computational models of the hippocampus.

He then moved to the University of Edinburgh, where he completed a Ph.D. in computational neuroscience in 2000. His supervision paired two rather different traditions: Richard Morris, who had developed the water-maze test that became a standard measure of spatial learning in rodents, and Peter Dayan, whose work on temporal-difference learning helped found the modern computational account of dopamine and reward. That combination set the agenda for everything that followed. During his doctoral years Foster published simulation studies of the CA3 subfield of the hippocampus modelled as an attractor network, and, with Morris and Dayan, a temporal-difference model of hippocampally dependent one-trial spatial learning. Both papers treated memory as a problem in reinforcement learning rather than as a simple storage operation, an approach that would later distinguish his experimental work.

Academic and Professional Career

After a brief postdoctoral period at the Gatsby Computational Neuroscience Unit at University College London with Peter Dayan, Foster moved to the United States. From 2000 to 2007 he was a postdoctoral associate in behavioural neurophysiology at the Massachusetts Institute of Technology with Matthew Wilson, where he learned the techniques of chronic multi-electrode recording in freely moving rats that his entire subsequent career depends on. It was there that the reverse-replay result was obtained.

In 2008 Foster joined the Johns Hopkins University School of Medicine as an assistant professor in the Solomon H. Snyder Department of Neuroscience and was promoted to associate professor in 2015. In 2017 he moved to the University of California, Berkeley, first as acting associate professor and, from 2018, as associate professor of psychology and of the Helen Wills Neuroscience Institute. At Berkeley his laboratory has continued to record from large populations of hippocampal neurons in rats performing spatial and reward-guided tasks, steadily moving from description of the replay phenomenon toward an account of what controls it and what it is for.

Major Contributions to Psychology

Awake reverse replay

Before Foster's work, replay, meaning the reactivation of sequences of hippocampal place cells during sharp-wave ripple oscillations, had been described mainly during sleep. In 2006 he and Wilson reported in Nature that replay also occurs while rats are awake and pausing, and that immediately after a spatial experience the sequences are frequently replayed in reverse order. A backward-running sequence is hard to reconcile with any simple account of replay as retrieval of the past, so Foster and Wilson proposed instead that reverse replay serves the kind of credit assignment described by reinforcement-learning models, propagating information about a reward backwards along the path that led to it. The paper has been cited well over a thousand times and is now standard reading in the literature on memory consolidation.

Theta sequences

In follow-up work with Wilson, published in Hippocampus in 2007, Foster showed that ordered sequences of place-cell firing also occur within individual cycles of the hippocampal theta oscillation while an animal is actively moving. These theta sequences compress a portion of the animal's ongoing spatial experience into a few tens of milliseconds and are closely related to the phenomenon of phase precession. The result established that compressed sequential coding in the hippocampus is not confined to quiet rest but is a continuous feature of spatial behaviour.

Forward replay and goal-directed planning

As an independent investigator, Foster and postdoctoral fellow Brad Pfeiffer reported in Nature in 2013 that before goal-directed navigation in an open arena the rat hippocampus generates brief sequences encoding spatial trajectories strongly biased to run from the animal's current location toward a known goal. Critically, these sequences predicted the animal's immediate future behaviour, even when the specific combination of start and goal locations was novel. The finding reframed replay: the same sequence-generating machinery that had been characterised in narrow linear tracks as a memory phenomenon could also support a trajectory-finding mechanism used for planning.

Reward, dopamine and the function of replay

A long-standing ambiguity in the field was whether awake replay supports learning from reward or retrieval for decision making. Foster's group addressed this directly in a 2016 Neuron study led by Ellen Ambrose, in which rats ran back and forth along a track while more than a hundred CA1 place cells were recorded and the amount of a chocolate reward at one end was increased or decreased. Forward and reverse replay behaved differently: only reverse replay increased in rate when reward increased and decreased when reward decreased, while forward replay was unchanged. The dissociation supports the view that the two directions do different jobs, with reverse replay tied to reward-based learning and forward replay possibly serving prospective evaluation of routes.

More recent work from the laboratory has begun to identify the signals that control replay. A 2025 study in eLife by Matthew Kleinman and Foster combined chemogenetic silencing of dopamine neurons in the ventral tegmental area with hippocampal recordings, and reported that dopamine signalling is required for replay to remain properly localised to the location where reward changed, particularly in novel environments.

Replay in models of disease

Foster has also used replay as a window onto psychiatric models. With collaborators including Susumu Tonegawa's group, he reported in Neuron in 2013 that a mouse model of schizophrenia shows impaired hippocampal ripple-associated replay. The result connects a systems-level memory phenomenon to a disorder in which hippocampal dysfunction is strongly implicated, and it opened a line of work on replay as a possible biomarker rather than only a mechanism.

Major Publications

Rolls, E. T., Treves, A., Foster, D., & Perez-Vicente, C. (1997). Simulation studies of the CA3 hippocampal subfield modelled as an attractor neural network. Neural Networks, 10(9), 1559–1569.
Foster, D. J., Morris, R. G. M., & Dayan, P. (2000). A model of hippocampally dependent navigation, using the temporal difference learning rule. Hippocampus, 10(1), 1–16.
Foster, D. J., & Wilson, M. A. (2006). Reverse replay of behavioural sequences in hippocampal place cells during the awake state. Nature, 440(7084), 680–683.
Foster, D. J., & Wilson, M. A. (2007). Hippocampal theta sequences. Hippocampus, 17(11), 1093–1099.
Pfeiffer, B. E., & Foster, D. J. (2013). Hippocampal place-cell sequences depict future paths to remembered goals. Nature, 497(7447), 74–79.
Suh, J., Foster, D. J., Davoudi, H., Wilson, M. A., & Tonegawa, S. (2013). Impaired hippocampal ripple-associated replay in a mouse model of schizophrenia. Neuron, 80(2), 484–493.
Ambrose, R. E., Pfeiffer, B. E., & Foster, D. J. (2016). Reverse replay of hippocampal place cells is uniquely modulated by changing reward. Neuron, 91(5), 1124–1136.
Foster, D. J. (2017). Replay comes of age. Annual Review of Neuroscience, 40, 581–602.
Kleinman, M. R., & Foster, D. J. (2025). Spatial localization of hippocampal replay requires dopamine signaling. eLife, 13, RP99678.
Mallory, C. S., Widloski, J., & Foster, D. J. (2025). The time course and organization of hippocampal replay. Science, 387(6733), 541–548.

Impact and Legacy

Foster's contribution has been to make replay measurable, dissociable and experimentally tractable. By showing that a single preparation produces both forward and reverse sequences, and that the two respond differently to reward, his laboratory provided the field with a within-animal control against the objection that replay simply increases whenever an animal is more aroused or more rewarded. His 2017 review in the Annual Review of Neuroscience, titled "Replay Comes of Age", is widely used as the field's summary statement.

The work has also attracted substantive criticism, some of it from within the field. The most persistent problem is methodological: for years there was no agreed standard for detecting sharp-wave ripples, which meant that different laboratories could report different replay statistics from broadly similar data. Foster responded by joining a multi-laboratory effort that published a consensus statement on ripple detection in Nature Communications in 2022, an implicit acknowledgement that the field's measurement practices needed to be made explicit before its claims could be compared.

A second line of criticism concerns interpretation. It remains contested whether replay is a genuine mechanism of consolidation and planning or an epiphenomenon of hippocampal network dynamics; whether reverse replay is a distinct process or an artefact of behavioural state and analysis choices; and whether the causal evidence is as strong as it looks, since the interventions used to disrupt or lengthen ripples also affect other processes. The generality of the findings is a related worry: almost all replay evidence comes from small numbers of rats and mice, usually male, performing highly simplified tasks, and direct recordings from healthy human hippocampus are not possible. Investigators must rely instead on intracranial recordings from epilepsy patients undergoing pre-surgical evaluation and on indirect imaging evidence.

Foster himself has been notably cautious on these points. Discussing the 2016 reward result, he stressed that a great deal more work was needed and that whether the mechanism applies to humans is unknown. The published peer reviews accompanying his 2025 eLife paper record reviewers' concerns about small sample sizes and variability in drug dosing, along with the authors' revisions, which illustrates how far the subfield has moved toward open methodological scrutiny rather than assertion.