Bruce L. McNaughton

Why there are complementary learning systems in the hippocampus and neocortex: insights from the successes and failures of connectionist models of learning and memory.

Damage to the hippocampal system disrupts recent memory but leaves remote memory intact. The account presented here suggests that memories are first stored via synaptic changes in the hippocampal system, that these changes support reinstatement of recent memories in the neocortex, that neocortical synapses change a little on each reinstatement, and that remote memory is based on accumulated neocortical changes. Models that learn via changes to connections help explain this organization. These models discover the structure in ensembles of items if learning of each item is gradual and interleaved with learning about other items. This suggests that the neocortex learns slowly to discover the structure in ensembles of experiences. The hippocampal system permits rapid learning of new items without disrupting this structure, and reinstatement of new memories interleaves them with others to integrate them into structured neocortical memory systems.

Environment-specific expression of the immediate-early gene Arc in hippocampal neuronal ensembles

We used fluorescent in-situ hybridization and confocal microscopy to monitor the subcellular distribution of the immediate-early gene Arc. Arc RNA appeared in discrete intranuclear foci within minutes of neuronal activation and subsequently disappeared from the nucleus and accumulated in the cytoplasm by 30 minutes. The time course of nuclear versus cytoplasmic Arc RNA accumulation was distinct, and could therefore be used to infer the activity history of individual neurons at two times. Following sequential exposure of rats to two different environments or to the same environment twice, the proportion of CA1 neurons with cytoplasmic, nuclear or overlapping Arc expression profiles matched predictions derived from ensemble neurophysiological recordings of hippocampal neuronal ensembles. Arc gene induction is thus specifically linked to neural encoding processes.

Replay of neuronal firing sequences in rat hippocampus during sleep following spatial experience.

The correlated activity of rat hippocampal pyramidal cells during sleep reflects the activity of those cells during earlier spatial exploration. Now the patterns of activity during sleep have also been found to reflect the order in which the cells fired during spatial exploration. This relation was reliably stronger for sleep after the behavioral session than before it; thus, the activity during sleep reflects changes produced by experience. This memory for temporal order of neuronal firing could be produced by an interaction between the temporal integration properties of long-term potentiation and the phase shifting of spike activity with respect to the hippocampal theta rhythm.

Path Integration and Cognitive Mapping in a Continuous Attractor Neural Network Model

A minimal synaptic architecture is proposed for how the brain might perform path integration by computing the next internal representation of self-location from the current representation and from the perceived velocity of motion. In the model, a place-cell assembly called a “chart” contains a two-dimensional attractor set called an “attractor map” that can be used to represent coordinates in any arbitrary environment, once associative binding has occurred between chart locations and sensory inputs. In hippocampus, there are different spatial relations among place fields in different environments and behavioral contexts. Thus, the same units may participate in many charts, and it is shown that the number of uncorrelated charts that can be encoded in the same recurrent network is potentially quite large. According to this theory, the firing of a given place cell is primarily a cooperative effect of the activity of its neighbors on the currently active chart. Therefore, it is not particularly useful to think of place cells as encoding any particular external object or event. Because of its recurrent connections, hippocampal field CA3 is proposed as a possible location for this “multichart” architecture; however, other implementations in anatomy would not invalidate the main concepts. The model is implemented numerically both as a network of integrate-and-fire units and as a “macroscopic” (with respect to the space of states) description of the system, based on a continuous approximation defined by a system of stochastic differential equations. It provides an explanation for a number of hitherto perplexing observations on hippocampal place fields, including doubling, vanishing, reshaping in distorted environments, acquiring directionality in a two-goal shuttling task, rapid formation in a novel environment, and slow rotation after disorientation. The model makes several new predictions about the expected properties of hippocampal place cells and other cells of the proposed network.

Reactivation of hippocampal cell assemblies: effects of behavioral state, experience, and EEG dynamics.

During slow wave sleep (SWS), traces of neuronal activity patterns from preceding behavior can be observed in rat hippocampus and neocortex. The spontaneous reactivation of these patterns is manifested as the reinstatement of the distribution of pairwise firing-rate correlations within a population of simultaneously recorded neurons. The effects of behavioral state [quiet wakefulness, SWS, and rapid eye movement (REM)], interactions between two successive spatial experiences, and global modulation during 200 Hz electroencephalographic (EEG) “ripples” on pattern reinstatement were studied in CA1 pyramidal cell population recordings. Pairwise firing-rate correlations during often repeated experiences accounted for a significant proportion of the variance in these interactions in subsequent SWS or quiet wakefulness and, to a lesser degree, during SWS before the experience on a given day. The latter effect was absent for novel experiences, suggesting that a persistent memory trace develops with experience. Pattern reinstatement was strongest during sharp wave–ripple oscillations, suggesting that these events may reflect system convergence onto attractor states corresponding to previous experiences. When two different experiences occurred in succession, the statistically independent effects of both were evident in subsequent SWS. Thus, the patterns of neural activity reemerge spontaneously, and in an interleaved manner, and do not necessarily reflect persistence of an active memory (i.e., reverberation). Firing-rate correlations during REM sleep were not related to the preceding familiar experience, possibly as a consequence of trace decay during the intervening SWS. REM episodes also did not detectably influence the correlation structure in subsequent SWS, suggesting a lack of strengthening of memory traces during REM sleep, at least in the case of familiar experiences.

Interactions between location and task affect the spatial and directional firing of hippocampal neurons.

When rats forage for randomly dispersed food in a high walled cylinder the firing of their hippocampal “place” cells exhibits little dependence on the direction faced by the rat. On radial arm mazes and similar tasks, place cells are strongly directionally selective within their fields. These tasks differ in several respects, including the visual environment, configuration of the traversable space, motor behavior (e.g., linear and angular velocities), and behavioral context (e.g., presence of specific, consistent goal locations within the environment). The contributions of these factors to spatial and directional tuning of hippocampal neurons was systematically examined in rats performing several tasks in either an enriched or a sparse visual environment, and on different apparati. Place fields were more spatially and directionally selective on a radial maze than on an open, circular platform, regardless of the visual environment. On the platform, fields were more directional when the rat searched for food at fixed locations, in a stereotypic and directed manner, than when the food was scattered randomly. Thus, it seems that place fields are more directional when the animal is planning or following a route between points of special significance. This might be related to the spatial focus of the rats attention (e.g., a particular reference point). Changing the behavioral task was also accompanied by a change in firing location in about one-third of the cells. Thus, hippocampal neuronal activity appears to encode a complex interaction between locations, their significance and the behaviors the rat is called upon to execute.

Place cells, head direction cells, and the learning of landmark stability

Previous studies have shown that hippocampal place fields are controlled by the salient sensory cues in the environment, in that rotation of the cues causes an equal rotation of the place fields. We trained rats to forage for food pellets in a gray cylinder with a single salient directional cue, a white card covering 90 degrees of the cylinder wall. Half of the rats were disoriented before being placed in the cylinder, in order to disrupt their internal sense of direction. The other half were not disoriented before being placed in the cylinder; for these rats, there was presumably a consistent relationship between the cue card and their internal direction sense. We subsequently recorded hippocampal place cells and thalamic head direction cells from both groups of rats as they moved in the cylinder; between some sessions the cylinder and cue card were rotated to a new direction. All rats were disoriented before recording. Under these conditions, the cue card had much weaker control over the place fields and head direction cells in the rats that had been disoriented during training than in the rats that had not been disoriented. For the former group, the place fields often rotated relative to the cue card or completely changed their firing properties between sessions. In all recording sessions, the head direction cells and place cells were strongly coupled. It appears that the strength of cue control over place cells and head direction cells depends on the rats learned perception of the stability of the cues.

The role of medial prefrontal cortex in memory and decision making.

Some have claimed that the medial prefrontal cortex (mPFC) mediates decision making. Others suggest mPFC is selectively involved in the retrieval of remote long-term memory. Yet others suggests mPFC supports memory and consolidation on time scales ranging from seconds to days. How can all these roles be reconciled? We propose that the function of the mPFC is to learn associations between context, locations, events, and corresponding adaptive responses, particularly emotional responses. Thus, the ubiquitous involvement of mPFC in both memory and decision making may be due to the fact that almost all such tasks entail the ability to recall the best action or emotional response to specific events in a particular place and time. An interaction between multiple memory systems may explain the changing importance of mPFC to different types of memories over time. In particular, mPFC likely relies on the hippocampus to support rapid learning and memory consolidation.

“dead reckoning,” landmark learning, and the sense of direction: A neurophysiological and computational hypothesis

Behavioral and neurophysiological evidence strongly suggests that, within certain limits, rodents and humans can keep track of their directional heading relative to an inertial, and hence allocentric coordinate system. This sense of direction appears to involve the integration of angular velocity signals that arise primarily in the vestibular system. A hypothesis is proposed in which the integration process, an operation that may be difficult for neurons to implement, is replaced by a linear associative mapping, an operation that is at least theoretically easy to implement with neurons. The proposed system makes use of a set of linearly independent vectors representing the combination of the current head direction, and head angular velocity representations to recall the resulting head direction. It is then proposed that visual landmarks become incorporated into the directional system, enabling both the correction of cumulative error and, ultimately, the computation of novel, optimal trajectories between locations. According to the hypothesis, this occurs through the association of hippo-campal local-view cells (i.e., direction selective place cells) with head-direction cells located downstream in the dorsal presubiculum. The possible neurophysiological and neuroan-atomical bases for the proposed system are discussed.

Multistability of cognitive maps in the hippocampus of old rats

Hippocampal neurons provide a population code for location. In young rats, environments are reliably ‘mapped’ by groups of neurons that have firing locations (‘place fields’) that can be stable for several months. Old animals exhibit deficits in spatial memory, raising the question of whether the quality or stability of their hippocampal ‘cognitive maps’ is altered. By recording from large groups of neurons, we observed the hippocampal spatial code to be multistable. In young rats, the place field maps were reliable both within and between episodes in a familiar environment. In old rats, place field maps were accurate and stable during an episode, but frequently exhibited complete rearrangements between episodes. In a spatial memory task, both young and old rats exhibited bimodal performance, consistent with map multistability early in training. However, the performance of young rats became almost unimodal with further training, whereas that of old rats remained markedly bimodal. The multistability of the hippocampal map provides an insight into the dynamics of neural coding in high-level cortical structures and their changes during ageing, and may provide an explanation for the frequent failure of place recognition in elderly humans.