How does the brain store sequences of experience? Clues come from brain recordings of rats running along a track. The animals' memories seem to be consolidated in an unexpected way as they rest between runs.
Memories develop in several stages. After the initial encoding of new information during learning, memories are consolidated 'off-line', seemingly while not being actively thought about, through a cascade of events that is not well understood. In humans and other mammals, such an enhancement of recent memories may occur during sleep1. But on page 680 of this issue, Foster and Wilson2 show that substantial consolidation might also happen while awake during rest periods.
Insight into how sleep benefits memory consolidation has been gained by recording neural activity in the hippocampus, a brain region that is crucial for mnemonic processing3. Cells that are activated in the hippocampus during certain awake behaviours fire in the same order but faster during the subsequent slow-wave phase of sleep4,5. This reactivation of firing patterns occurs during ‘sharp waves’, excitatory waveforms that dominate hippocampal recordings throughout slow-wave sleep6. Sharp waves are accompanied by very fast oscillations (about 200 hertz) known as ripples, generated when multiple cells fire together within a narrow time window7. Co-activation of interconnected neurons during ripples may result in long-lasting modifications of the synapses in the network (that is, the communication junctions between neurons)8.
Although reactivation during sleep may provide a mechanism for consolidation of recent memories, the mystery remains as to how memories can be maintained as distinct entities for hours or days in sleep-deprived subjects, considering that the participating neurons are probably involved in myriad events before the subject is finally allowed to take a nap. One clue comes from the observation that sharp waves occur also during waking states; for example, during resting, eating, drinking and brief breaks in exploration6,9. Such ‘interleaved’ sharp waves may strengthen associations between recently activated cells only seconds after an event9.
Foster and Wilson2 provide fascinating evidence for a mechanism that could generate such associations. They studied rats running back and forth on a narrow track, and they recorded neural activity from so-called place cells10. These hippocampal cells have spatial receptive fields, so each cell responds when the animal is in a particular location. Food was placed at the ends of the track, and the animals stopped after every lap to eat. When the rat paused, sharp waves emerged in its hippocampus. During these sharp waves, the place cells from the running period were reactivated, but their order of firing was reversed with respect to their earlier order of activation on the track (Fig. 1).
But how do neurons reverse firing sequences that were just stored in forward order? This might happen in at least two ways, one depending on the rat's recent history and one reflecting its location in the environment. In the first possibility, the cells responding to place fields closest to the rest location are the first to reach the threshold for firing during the sharp wave because their synapses are still in a ‘facilitated’ state. Cells with fields that are farther away are less facilitated, so they take longer to reach the threshold. In the second option, cells fire in reverse order merely because firing probabilities of place cells increase with decreasing distance from the centre of their place fields, regardless of whether or not the rat has just passed through the fields. The latter possibility is partly ruled out because Foster and Wilson did not observe reverse replay in sharp waves recorded at the start of the session, before the rat began moving. This suggests that reverse reactivation is determined by the preceding sequence of events.
The million-dollar question, however, is what the brain gains by rewinding its neural record. At present, we do not know why sharp- wave-associated replay is forward in some circumstances (during sleep, say) and reversed in others. Foster and Wilson speculate that reverse replay has a role in reward-directed sequence learning during spatial navigation. Rewards (reinforcers) such as the food received at the ends of the track strengthen the preceding behavioural responses in a time-dependent manner such that the longer between the response and the reward, the less the behaviour is strengthened11,12. This mechanism is adaptive in evolutionary terms as it normally causes a fairly selective enhancement of those responses that generate the reward.
The authors hypothesize that the formation of associations between a reward and the representation of elements of a rat's trajectory in the immediate past is boosted during sharp-wave-associated replay by a neuromodulatory signal such as dopamine. Dopamine is a chemical released in the forebrain (in the striatum and cortex, and presumably the hippocampus) at the time of reward, especially when reward is not expected by the animal13,14,15. Because ripple trains are variable in length, the effects of the boosting signal would be most reliable if it occurred at the beginning of the sharp wave; however, an early boost could be linked to the key later elements of the preceding firing sequence only if the sequence were reactivated in reverse order, as in Foster and Wilson's study. It remains to be seen whether these speculations will stand up to experimental testing. At the moment, we do not know whether dopamine-releasing neurons fire in synchrony with hippocampal sharp waves.
If reverse replay is a mechanism for strengthening hippocampal sequence memories during goal-directed behaviour, several questions arise. For example, is the firing sequence stored as an ordered memory or as a unitary representation with a stronger representation of the later than of the earlier elements? Moreover, is reverse replay specific to sharp waves that coincide with reward? Sharp waves are observed during breaks without rewards. Do these sharp waves also exhibit reverse replay and, if so, are these associated with memory storage? Finally, can memories of events be stored without interleaved sharp waves? Whatever the answers may be, the discovery of reverse replay is bound to pave the way for more surprises.
Stickgold, R. Nature 437, 1272–1278 (2005).
Foster, D. J. & Wilson, M. A. Nature 440, 680–683 (2006).
Squire, L. R., Stark, C. E. & Clark, R. E. Annu. Rev. Neurosci. 27, 279–306 (2004).
Wilson, M. A. & McNaughton, B. L. Science 265, 676–679 (1994).
Lee, A. K. & Wilson, M. A. Neuron 36, 1183–1194 (2002).
Buzsaki, G., Leung, L. W. & Vanderwolf, C. H. Brain Res. 287, 139–171 (1983).
Klausberger, T. et al. Nature 421, 844–848 (2003).
Buzsaki, G. Neuroscience 31, 551–570 (1989).
O'Neill, J., Senior, T. & Csicsvari, J. Neuron 49, 143–155 (2006).
O'Keefe, J. & Dostrovsky, J. Brain Res. 34, 171–175 (1971).
Skinner, B. F. The Behavior of Organisms (Appleton-Century-Crofts, New York, 1938).
Hull, C. L. A Behavior System (Wiley, New York, 1952).
Berridge, K. C. & Robinson, T. E. Brain Res. Rev. 28, 309–369 (1998).
Montague, P. R., Hyman, S. E. & Cohen, J. D. Nature 431, 760–767 (2004).
Schultz, W. Annu. Rev. Psychol. 57, 87–115 (2006).
About this article
Neurobiology of Learning and Memory (2017)
Theta oscillations orchestrate medial temporal lobe and neocortex in remembering autobiographical memories
Neural circuits and temporal plasticity in hindlimb representation of rat primary somatosensory cortex: revisited by multi-electrode array on brain slices
Neuroscience Bulletin (2010)