sign in sign up
<<
Home , Camillo Golgi, Edvard Moser, Eleanor Maguire, Phase precession

NeuroView

The 2014 in or Medicine: A Spatial Model for Cognitive

Neil Burgess1,* 1Institute of and Institute of Neurology, University College London, 17 Queen Square, London WC1N 3AR, UK *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2014.12.009

Understanding how the cognitive functions of the arise from its basic physiological components has been an enticing final frontier in science for thousands of years. The Nobel Prize in Physiology or Medicine 2014 was awarded one half to John O’Keefe, the other half jointly to May-Britt Moser and Edvard I. Moser ‘‘for their discoveries of cells that constitute a positioning system in the brain.’’ This prize recognizes both a paradigm shift in the study of cognitive neuroscience, and some of the amazing insights that have followed from it concerning how the world is represented within the brain.

Introduction persuaded of the ‘‘neuro-ethological’’ Two of the most dramatic results of the Early ideas concerning the production of approach in which, faced by a lack of neuro-ethological approach to electro- thought were varied, unconstrained by knowledge about how the brain works, physiology are honored by the Nobel direct experimental data, e.g., Hippo- the primary aim is a broad survey of how Prize: the discoveries of place cells and crates’ theory of cognition depending on the system in question is used during of grid cells. As noted in the Nobel cita- the balance of the four humors (blood, normal behavior. This approach contrasts tion, the properties of these cells, along phlegm, and two types of bile). However, with the principles of more traditional with those of other types of spatial cell, with the advance of technology, the experimental design, with its focus on most notably head direction cells, help importance of electrical signaling in the tightly controlled testing of well-specified to define ‘‘a positioning system in the brain and nervous system slowly became hypotheses, and is well described in the brain.’’ Identification of this specific sys- accepted, from the effects of electrical book written later with , his tem is indeed a great achievement in un- stimulation on muscles noted by Luigi friend and colleague from their time at derstanding the link between mind and Galvani in the 18th century, to the cogni- McGill, The as a Cognitive brain. But perhaps even more important tive effects in the noted by Map (O’Keefe and Nadel, 1978). are the general lessons they provide for Wilder Penfield in the 1940s and 50s. O’Keefe also profited from the timely the of internal cognitive The realization of the importance of neu- advent of modern transistors, allowing constructs, and the neural mechanisms rons (the ‘‘’’) by Santiago robust differential recording of electrical of learning, representation, and . Ramon y Cajal—recipient of the Nobel activity from freely behaving animals, in Prize in 1906 with —com- which the large variations in potential Place Cells and Cognitive Maps bined with an appreciation of the role of common to nearby electrodes could be In his first experiments, published with electricity gave rise to the modern field cancelled and the remaining signals Herman Bouma in 1969, O’Keefe re- of electrophysiology as a tool to under- boosted on the animal’s head. Along corded responses in the amygdala of stand brain function. Early successes with other early pioneers such as Jim freely moving cats to ethologically rele- included an understanding of neural re- Ranck, this allowed O’Keefe to introduce vant stimuli (birdsong, mouse shapes, sponses to light in the retina, for which a new paradigm in understanding brain etc.). They found many instances of Granit, Hartline, and Wald received the function—studying neuronal activity in that appeared to be tuned Nobel Prize in 1967, leading to an under- freely behaving animals in response to to represent the presence of specific standing of the feedforward processing naturalistic stimuli or tasks. This paradigm stimuli, when presented, and some that of simple visual stimuli by neurons in the has continued to spread, slowly, over the continued to respond after the stimulus visual system of anaesthetised mammals subsequent decades, notwithstanding was withdrawn. These findings have for which David Hubel and the intrinsic validity of more traditional been echoed recently by neurons in hu- received the Nobel Prize in 1981. approaches in narrowing down potential man amygdala that respond to pictures Amidst these exciting developments in functions as each function becomes of specific animals. the 1960s, John O’Keefe was studying more completely understood. With the Both the amygdala and the neighboring his PhD at McGill University, receiving retrospect of the subsequent 40 years of hippocampus had already been impli- instruction from, among others, Donald progress, the neuro-ethological approach cated in human memory by the finding, Hebb—one of the foremost proponents of electrophysiology in freely behaving at the nearby Montreal Neurological Insti- of looking for direct analogs of learning animals appears to be a crucial ‘‘para- tute founded by Penfield, that bilateral and experience within the brain. As he digm shift’’ for cognitive neuroscience of damage or removal of these structures began his research career, O’Keefe was the kind identified by Thomas Kuhn. and surrounding cortical tissue in the

1120 Neuron 84, December 17, 2014 ª2014 Elsevier Inc. Neuron NeuroView

view of the hippocampus as part of a cir- learning and memory in humans. They cuit for emotion, as proposed by James concluded that there must be something Papez. like an innate unitary spatial framework At the invitation of Pat Wall, O’Keefe onto which experience of the world can next moved to University College London be mapped, as foreshadowed by the and into the Department of philosopher Immanuel Kant, and that its and Developmental Biology, headed by implementation in the brain should pro- J.Z. Young—another pioneer in relating vide the animal with a ‘‘’’ in behavior to neuronal activity (in this the sense argued for by Edward Tolman: case, in the squid). There he began to that is, an abstract and flexible represen- look at the hippocampus in freely moving tation of space allowing novel inferences rodents as they foraged for scattered food (e.g., directions of travel), beyond simple reward. The initial paper, written with representations of specific sensory stim- John Dostrovsky (an MSc student pass- uli, bodily responses, or associations be- ing through the lab on the way to tween them. Within this framework, the becoming a neurologist), describes 8 out place cells were held to represent the of 76 putative neurons responding to the animal’s current environmental location. location of the animal, with other neurons The initial findings aroused responding to arousal, , or move- great skepticism, both regarding the ment, or responding in inconsistent or un- basic finding and potential confounds interpretable ways (O’Keefe and Dostrov- such as uncontrolled local sensory stimuli sky, 1971). The responses of many of (e.g., odours), and the bold and highly these cells were consistent with similar abstract interpretation (cognitive maps) early findings by Jim Ranck. as opposed to simpler well-studied forms Perhaps surprisingly, given the appar- of learning such as conditioned re- ently weak initial support, O’Keefe imme- sponses. These justifiable queries were diately recognized these neurons as duly addressed, in sophisticated form. relevant to a classic question in cognitive O’Keefe first showed that the orientation —that of how you represent of the firing locations (‘‘fields’’) could be your own location relative to the envi- controlled by the constellation of distal Figure 1. Examples of the Four Main Types ronment—and christened them ‘‘place cues (O’Keefe and Conway, 1978) and of Spatial Firing Recorded from Neurons in the Hippocampal Formation of Freely cells.’’ In fact, this strong interpretation then that their firing was also robust to Moving Rats was aided by the last seven cells recorded the subsequent removal of the controlling Each example shows firing rate as a function of loca- all being place cells—indicating to cues (O’Keefe and Speakman, 1987). In tion or head direction (left; peak firing rate in Hz O’Keefe that they were the dominant the absence of controlling cues, the orien- above) alongside a plot of the rat’s path within a square box (black line) and the location of the rat response, once the electrodes had tation of the location of a place cell’s firing when an was fired (green dots;right). reached the ‘‘right’’ brain location. Subse- within the environment was unpredictable (A) Place cells, found in areas CA3 and CA1 of the quent experiments revealed that a but remained consistent with those of hippocampus proper, typically fire in a restricted portion of the environment. majority of the principal cells in regions other place cells and with the behavioral (B) Head-direction cells, found in the presubiculum CA1 and CA3 of the hippocampus that choices of the animal. In addition, and deep layers of medial , typi- are active in a given environment are O’Keefe laid out his proposal that place- cally fire for a narrow range of allocentric heading place cells, firing action potentials at a cell firing must reflect an internally gener- directions (left). (C) Grid cells, found in medial entorhinal cortex and high rate whenever the animal enters a ated spatial signal based on path integra- pre- and parasubiculum, typically fire in a regular small portion of its environment and re- tion in combination with environmental triangular array of locations. Directional grid cells maining silent elsewhere (see Figure 1A). sensory inputs (O’Keefe, 1976), which is or ‘‘conjunctive’’ cells are also found, whose grid- like spatial firing is also modulated by head direction. Inspired by the discovery of place cells, considered further in the next section. (D) Boundary cells, found in subiculum and ento- O’Keefe and Nadel wrote The Hippocam- The firing pattern of place cells is spe- rhinal cortex, typically fire at a specific distance pus as a Cognitive Map. This was a spec- cific to a given environment; place cells from an environmental boundary along a specific tacular tour de force, specifically con- may fire in one environment but not fire, allocentric direction. A623 62 cm box was used for (A) and a 1 3 1m cerning the likely functional role of the or fire with a different spatial pattern, box for (B)–(D), with a 50 cm barrier within the box hippocampus, but also providing a revo- in another (O’Keefe and Conway, 1978; for (D); successively hotter colors show quintiles lutionary manifesto for how to think about Muller and Kubie, 1987)—a phenomenon of firing rate, and unvisited locations are shown in white. Adapted with permission from Hartley brain function, from the deep analysis of known as ‘‘remapping.’’ The firing rate of et al. (2014). the function in question (representation place cells can be modulated by other fac- of space) through extensive review of tors, such as the presence or absence of treatment of epilepsy caused profound behavioral, lesion, and electrophysiolog- particular sensory stimuli (O’Keefe, 1976) (Scoville and Milner, 1957). This ical data to potential neuronal mecha- and aspects of behavior such as running work immediately revised the preceding nisms and broader implications for speed and direction, but primarily reflects

Neuron 84, December 17, 2014 ª2014 Elsevier Inc. 1121 Neuron NeuroView

the animal’s location within an environ- each firing whenever the rat faced in a given the small numbers of peaks present ment. Detailed quantification of the rela- certain direction, irrespective of its loca- within the standard-sized box used for re- tionship between place-cell firing and tion (see Figure 1B). Different cells have cordings. The final piece of the puzzle was the surrounding environment came from different preferred firing directions that supplied by Bill Skaggs, working with the New York group of Bob Muller, John are strongly controlled by distal visual Bruce McNaughton, who suggested that Kubie and Jim Ranck. Other groups cues, if present. There is a robust intrinsic they try recording in larger boxes. This followed up the nonspatial modulation organization to the population activity, simple but crucial insight produced imme- of firing, and Neal Cohen and Howard such that all preferred firing directions diate dividends. With Torkel Hafting Eichenbaum produced an account of hip- rotate together coherently with each and colleagues in their lab, they found pocampal function generalizing the flex- other. Further work, continued by Jeff that grid cells in dorsomedial EC fired ible spatial learning stressed by O’Keefe Taube, showed that these neurons with fantastically regular precision at the and Nadel to all forms of flexible relational appear to underlie our sense of direction vertices of an equilateral triangular grid ar- learning. Further technological advances and also provide the directional reference ranged across the environment, despite in isolating the electrical signals from for other spatial representations in the the highly irregular foraging movements single neurons came with the develop- hippocampal formation (Taube, 1998). of the rats (Hafting et al., 2005) (see ment of the stereotrode with Bruce Around this time, Edvard and May-Britt Figure 1C). McNaughton and Carol Barnes during Moser were beginning their PhDs in the The firing patterns of nearby grid cells their visit to the O’Keefe lab, and the lab of Per Andersen in Oslo. As well as have similar orientation and scale but tetrode with Michael Recce. The power working on the neural mechanisms vary from each other in terms of their of multielectrode recording was demon- of learning, they became interested in spatial offset. The orientation of the grid strated by Matt Wilson and Bruce the relative contributions of dorsal and pattern reflects external environmental McNaughton recording enough place ventral hippocampus to , cues, similarly to place cells, and is also cells simultaneously to be able to accu- in collaboration with Richard Morris. probably determined by head-direction rately reconstruct the location of the rat. Performing local ‘‘minislab’’ lesions at cells. The grid firing patterns recorded Evidence for a causal relationship be- different dorsoventral levels of the hippo- in dorsal locations have the smallest tween the hippocampus and the specific campus, with the anatomical precision spatial scale, with increasingly larger aspects of spatial memory that might that would become a hallmark of their grids recorded more ventrally (Hafting require a cognitive map came from so- work, they showed the increasing reliance et al., 2005). Interestingly, the grid phisticated tests of the effects of hippo- of spatial memory on dorsal rather scale appears to be quantized (Barry campal lesions. The most prominent of than ventral hippocampus (Moser et al., et al., 2007), increasing dorsoventrally in these is the Morris water maze, in which 1995). Expanding their interests to discrete jumps (Stensola et al., 2012). the animal must find a goal location (a electrophysiology, they first visited John The orientations of the grid firing patterns platform hidden under the surface of a O’Keefe’s lab and then that of Bruce of different scale, while less tightly related pool of opaque water) indicated by distal McNaughton and Carol Barnes to than those of the same scale, also tend to cues rather than any local sensory cues, become proficient in recording from the cluster around similar values (Barry et al., starting from different locations around hippocampus of freely moving rodents. 2007; Stensola et al., 2012). Unlike place the edge of the pool. Performance in The Mosers’ interest in anatomy and cells, whose firing patterns can remap this task was shown to be impaired by the dorsoventral distinction in the hippo- completely between different environ- hippocampal lesions, in a collaboration campus led them to aim their tetrodes at ments, firing patterns are basi- between O’Keefe and Richard Morris. the dorsal part of the medial entorhinal cally maintained across all environments cortex (mEC). Here, aided by the anatom- visited by the animal, with only their Grid Cells and Other Spatial Cells ical knowledge of Menno Witter, they spatial offset to the environment varying The identification of place cell firing as hoped to find the most strongly spatial (Fyhn et al., 2007). representing location within a cognitive inputs to the hippocampus, because the Grid cells, head-direction cells, and map cried out for the neural support of medial (cf. lateral) entorhinal cortex re- place cells are part of a wider spatial sys- other elements necessary for a complete ceives projections from the presubiculum, tem. Grid cells were initially described in navigational system: a representation of where head-direction cells are found, and layer II of mEC, where they are most direction and a distance metric. O’Keefe the dorsal portion of mEC projects to the densely represented, but were also found and Nadel had predicted that directional most spatial, i.e., dorsal, portion of in the deeper cell layers of mEC and in the signals were available to the hippocam- the hippocampus. There, with Marianne neighboring pre- and parasubiculum, by pus, and speculated that theta might Fyhn and colleagues in their lab in Trond- Francesca Sargolini and Charlotte Boc- signal spatial translation. The first of these heim, they found neurons with multi- cara in the Moser lab. In these other loca- to be found was the directional system, peaked spatially modulated firing (Fyhn tions, directionally modulated grid cells, with the discovery of ‘‘head-direction et al., 2004), revising earlier findings of or ‘‘conjunctive’’ cells, were also found. cells’’ by Ranck and the New York lab. simple spatially modulated responses The firing of these cells shows the These cells, initially identified in the presu- in more ventral portions of mEC. There grid-like spatial modulation characteristic biculum but later found throughout were signs that the firing peaks might be of grid cells but is also modulated by Papez’ circuit, complemented place cells, regularly arranged, but this was not clear the head direction of the animal (like

1122 Neuron 84, December 17, 2014 ª2014 Elsevier Inc. Neuron NeuroView

head-direction cells, but typically with respond to local sensory cues, the pres- et al., 2010). Thus head-direction cell broader directional tuning). ence of objects, and even the absence firing may represent a fundamental repre- To summarize, grid cells show precisely of recently present objects. sentation on which other spatial represen- regularly distributed spatial firing patterns The stage is set: the work of John tations can be built, while place-cell firing which are broadly invariant to environ- O’Keefe, Edvard and May-Britt Moser, appears not to require grid cell firing. It is mental change and are organized (across and their colleagues and collaborators possible that both environmental inputs cells) into an almost crystalline structure over the years has identified the neural (such as boundary vector cells) and inputs of quantized scales and orientations. representations comprising a spatial nav- from grid cells (which may predominantly Thus, like place cells, grid cells seem to igation system in the mammalian brain. reflect path integration, see below) pro- represent an abstract sense of location From this foundation of plentiful knowl- vide parallel routes to driving place-cell rather than specific local sensory cues. edge we can begin to ask how these firing, so that either is sufficient and both However, grid cells go far beyond place representations work together to under- can be combined when present. cells in representing an abstract spatial lie behavior, learning, and memory—an The strong internal consistency of framework, providing the most striking exciting challenge for systems neurosci- head-direction cell firing suggests that embodiment of internal cognitive struc- ence in the years to come. the dynamics of their population activity ture being applied to the external world. is best described as a line attractor. That Environmental information must play Far-Reaching Implications of the is, symmetrical recurrent connectivity a part in creating firing patterns that are Study of Hippocampal Spatial constrains population activity to coherent spatially stable in the external word, and Coding patterns representing a single head direc- the way in which this information is repre- The work of the Nobel laureates has out- tion and allows smooth transitions be- sented has begun to be understood lined the neural mechanisms supporting tween them (Zhang, 1996). The addition in recent decades. As noted above, navigation, an immense achievement in it- of asymmetric interactions dependent on place cells ‘‘remap’’ (i.e., fire in unrelated self. But the importance of their findings the animal’s angular velocity could enable fashion) in distinct environments, but can been seen even more clearly when active representation of head direction. more subtle parametric changes to an considering the profound implications Two-dimensional attractor models were environment’s shape and size can pro- they have had for our understanding then applied to place-cell firing by Bruce duce corresponding parametric changes of the neuronal mechanisms for repre- McNaughton and others, with asymmetric in place-cell firing patterns (O’Keefe and sentation, learning, and memory more connectivity between velocity-modulated Burgess, 1996), indicating the presence generally. The hippocampus and spatial place representations resulting in accu- of inputs tuned to the distance and navigation has proved an ideal model rate path integration. These models were allocentric direction of environmental system for understanding the relationship even more appropriate for grid cell firing, boundaries. The proposed input neu- between brain and behavior and the prin- whose repeating firing patterns could rons (‘‘boundary vector cells’’ or ‘‘border cipals of neural coding. reflect recurrent interactions or ‘‘Turing cells’’) were subsequently found in Implications for Intrinsically patterns’’ (reviewed by McNaughton the subiculum and entorhinal cortex, Structured Neural Representations et al., 2006). In this case, asymmetric con- reviewed in Hartley et al. (2014) (see Grid cell firing provides a spectacular nectivity between velocity modulated grid Figure 1D). example of internally generated structure, representations enables accurate path In search of the mechanism by which both individually and in the almost crystal- integration—that is, updating of an inter- place-cell firing can vary in rate, the line organization of the firing patterns of nal position estimate on the basis of self- field turned to investigate lateral EC, different grid cells. A similarly strong orga- motion cues. As with head-direction cells, led by the Mosers and also the lab of nization is seen in the relative tuning of the strong relative coherence of grid cell Jim Knierim. Significant environmental head-direction cells. This strong internal firing offers support for the symmetrical change causes place-cell remapping, structure is reminiscent of Kantian ideas interactions posited by these models. and this is accompanied by shifts in the regarding the necessity of an innate Taken together, these findings provide grid patterns in medial EC relative to the spatial structure with which to understand strong support for the general coding environment (Fyhn et al., 2007). However, the spatial organization of the world. To principal of continuous attractor dy- less significant environmental changes investigate the experience dependence namics in populations of locally tuned can produce changes in place-cell firing of the spatial firing patterns, both O’Keefe neurons. rate only (‘‘rate remapping’’), which are and Moser groups recorded from hippo- Perhaps the most intriguing aspect of not accompanied by shifts in the grid pat- campus and medial EC in preweanling the structure manifested by grid cell firing terns (Fyhn et al., 2007). With Li Lu, the rat pups as they first began to explore is its obvious potential power as a coding Mosers showed that this rate remapping outside the nest. Intriguingly, head-direc- scheme—almost as if it were designed depends on input from the lateral EC. tion firing was present at adult levels from by a mathematician or engineer. The Changes in place-cell firing rates can the first recordings, and place-cell firing grid cells are organized into modules also reflect the presence or absence of was also present, but improved with with different spatial scales (Barry et al., local sensory cues (O’Keefe, 1976), and further experience, while grid cell firing 2007; Stensola et al., 2012), so that each the Moser and Knierim labs have since was not present until after several days module provides a modulo code for loca- discovered lateral EC neurons that of experience (Wills et al., 2010; Langston tion with modulus equal to the grid scale.

Neuron 84, December 17, 2014 ª2014 Elsevier Inc. 1123 Neuron NeuroView

Together a relatively small number of parsimonious mechanism for path inte- for supporting memory and the observa- modules have a potentially exponential gration whose output could be the grid tion of dense recurrent connectivity in re- coding capacity representing a neural cell firing patterns. Such a mechanism is gion CA3 inspired David Marr (Marr, 1971) code of unprecedented power and effi- broadly supported by relationships be- to produce the canonical model of hippo- ciency (e.g., Sreenivasan and Fiete, tween running speed, grid scale and the campal memory. Recordings from the 2011). How this code is actually used frequencies of subthreshold membrane hippocampus of freely moving rats pro- in the brain remains a question of intense potential oscillations, LFP theta and burst vided a test bed, verifying many of interest and may depend critically on firing by grid cells, and by ‘‘theta cells,’’ the main predictions, including the pres- whether grid cells can provide a regular and by the elimination of grid patterns by ence of attractor dynamics and pattern grid-like representation over large-scale disruption of the theta rhythm (reviewed completion in place-cell firing. space, given that grid regularity and scale in Burgess and O’Keefe 2011), and would One intriguing aspect of the Marr model can be strongly affected by the structure complement the stability provided by is that the rapidly stored information in the of the environment (Derdikman et al., attractor dynamics. However, the inter- hippocampus is transferred to long-term 2009; Barry et al., 2007). pretation of the role of theta rhythmicity in the neocortex. Both Gyuri Buz- Implications for Oscillatory remains hotly contested, given only mixed saki and Bruce McNaughton explored the Processing and Temporal Coding support from analyses of grid scale and implications of this suggestion. With Matt Following earlier work by Case Vander- firing rhythmicity in the Moser lab, the Wilson, McNaughton searched for the wolf on the 4–10 Hz theta rhythm of the strong relationship between grid cell firing recapitulation of place-cell firing patterns local field potential (LFP), and the obser- and tonic depolarization, and the exis- from a day’s experience during subse- vation of theta-modulated neuronal firing tence of grid cell in bats in the absence quent sleep. Encouragingly, they found in the hippocampus by Steve Fox and of theta. ‘‘replay’’ of firing sequences from active Ranck, O’Keefe and Nadel assumed that The interaction between theta and exploration during subsequent slow- it reflected a key role of the hippocampus gamma oscillations opens a window into wave sleep (Wilson and McNaughton, in linking perception to action (specifically studying interregional communication 1994). Buzsaki focused on ‘‘ripples’’ as translational motion). Careful consider- and mode-switching within the hippo- the most likely means of communication ation of the timing with which place cells campal formation. Gamma is a broad from hippocampus to neocortex. These fire bursts of spikes as the animal runs higher-frequency band (25–140 Hz) asso- short bursts of high-frequency activity in through the firing field led O’Keefe and ciated with local inhibitory mechanisms the hippocampal LFP, first identified by Michael Recce to realize that the phase that is prevalent in neocortical areas but O’Keefe and Nadel (1978), co-occur with of firing relative to the LFP theta rhythm also visible within the hippocampal forma- ‘‘replay’’ events during slow-wave sleep encoded distance traveled through the tion. Following work from Gyuri Buzsaki’s and during wakefulness. Interestingly, firing field (O’Keefe and Recce, 1993), lab linking the generation of low-fre- the disruption of ripples immediately after with the place cell firing at successively quency (25–50 Hz) gamma oscillations learning disrupts subsequent spatial earlier phases of theta as the animal to the CA3 region of the hippocampus memory performance, consistent with a moves through the firing field. This finding and higher-frequency gamma oscillations role in the consolidation of hippocampal represents one of the most robust exam- to the entorhinal cortex, the Mosers information to neocortex. ples of internally generated temporal and Laura Colgin began to investigate The Hippocampus as a Cognitive Map or phase coding of a behavioral variable whether gamma might shed light on inter- by O’Keefe and Nadel (1978) also promp- in neuroscience. The consequence of regional communication with region CA1. ted a new perspective on hippocampal ‘‘phase precession’’ in individual place They found that periods of high- or low- function in human memory—specifically cells is that within each theta cycle the frequency gamma in the CA1 local field relating hippocampal amnesia (Scoville location represented by the population potential were indicative of periods of and Milner, 1957) to the idea of context- of active cells sweeps forward, potentially place cell firing being driven by EC or dependent or ‘‘episodic’’ memory intro- contributing to route planning at decision CA3, respectively, with each gamma duced by . The place cells points, as noted by David Redish. band associated with different phases of could provide a substrate for spatial O’Keefe and Recce (1993) suggested theta. Thus, different gamma frequencies context, to which O’Keefe and Nadel that phase precession results from the might allow flexible routing of information argued that temporal context would be interference of two oscillators—one re- from distinct inputs to CA1, extending the added in humans, with the role of context flected in the LFP and another with a idea of interregional communication via in learning being further developed by Na- slightly higher frequency that should in- gamma band coherence suggested by del. One specific suggestion, arising from crease with running speed. Theta phase Wolf Singer and Pascal Fries, and consis- work with O’Keefe, is that the various precession is also observed in grid cells tent with Mike Hasselmo’s suggestion spatial firing properties of neurons within (Hafting et al., 2008), whose periodic of (entorhinal-driven) versus Papez’ circuit provide the spatial struc- spatial firing is more naturally described retrieval (CA3-driven) scheduling by theta ture within which the contents of memory as an interference pattern. The idea that phase. can be imaged (Burgess et al., 2001). This speed could be coded as a frequency Implications for Memory idea received recent support in the difference, such that phase differences Scoville and Milner’s (1957) identification observation that the spatial coherence represent distance traveled, provides a of the hippocampus as a critical structure of imagery as impaired in hippocampal

1124 Neuron 84, December 17, 2014 ª2014 Elsevier Inc. Neuron NeuroView

amnesics (Hassabis et al., 2007). How- considered unreachable (e.g., work by Hafting, T., Fyhn, M., Molden, S., Moser, M.B., and ever, the role of the hippocampus in , recipient of the Nobel Moser, E.I. (2005). Nature 436, 801–806. human memory remains controversial to Prize in 1987 for work on immunology). Hafting, T., Fyhn, M., Bonnevie, T., Moser, M.B., this day. Some authors, including Nadel These experiments include the deletion and Moser, E.I. (2008). Nature 453, 1248–1252. and Morris Moscovitch, argue for a spe- of fearful via disrupted recon- Hartley, T., Lever, C., Burgess, N., and O’Keefe, J. cific role in episodic or contextual mem- solidation and the artificial creation of (2014). Philos. Trans. R. Soc. Lond. B Biol. Sci. ory and question the extent and nature memories of fear or reward. 369, 20120510. of consolidation to other areas. Others, Hassabis, D., Kumaran, D., Vann, S.D., and most notably , hold to a Conclusion Maguire, E.A. (2007). Proc. Natl. Acad. Sci. USA 104, 1726–1731. time-limited but more general role in The energy, enthusiasm, and vision of explicit or declarative memory, in line John O’Keefe, , May-Britt Langston, R.F., Ainge, J.A., Couey, J.J., Canto, with earlier definitions of amnesia. C.B., Bjerknes, T.L., Witter, M.P., Moser, E.I., and Moser, and their colleagues have left the Moser, M.B. (2010). Science 328, 1576–1580. Within the spatial domain, the discov- field of cognitive neuroscience with a eries of place cells and grid cells in ro- model system—hippocampal spatial nav- Marr, D. (1971). Philos. Trans. R. Soc. Lond. B Biol. Sci. 262, 23–81. dents have given rise to several related igation—in which much of the neuronal findings in humans. The activity and struc- code has been cracked. This tremendous McNaughton, B.L., Battaglia, F.P., Jensen, O., tural integrity of the human hippocampal Moser, E.I., and Moser, M.B. (2006). Nat. Rev. Neu- achievement provides firm foundations rosci. 7, 663–678. formation has been related to accurate for rapid progress in many directions spatial navigation (e.g., reviewed in and will be widely recognized as truly Moser, M.B., Moser, E.I., Forrest, E., Andersen, P., Burgess et al., 2002), with these experi- and Morris, R.G. (1995). Proc. Natl. Acad. Sci. USA meriting the 2014 Nobel Prize in Physi- 92, 9697–9701. ments often making use of virtual reality ology or Medicine. There may be a long adaptations of hippocampal-dependent way to go in fully understanding human Muller, R.U., and Kubie, J.L. (1987). J. Neurosci. 7, 1951–1968. tasks used with rodents, or an analog of memory and how it can fail in neurological comparative neuroethology in the study and psychiatric disorders, but the founda- O’Keefe, J. (1976). Exp. Neurol. 51, 78–109. of taxi drivers by Eleanor Maguire. Evi- tions for a bridge from molecules through O’Keefe, J., and Burgess, N. (1996). Nature 381, dence for neurons with firing resembling neurons, synapses, and networks to 425–428. place cells and grid cells has come from behavior and symptoms have been laid. O’Keefe, J., and Conway, D.H. (1978). Exp. Brain intracranial recordings in epilepsy pa- Res. 31, 573–590. tients by the labs of Mike Kahana and ACKNOWLEDGMENTS Izhak Fried. Intriguingly, the wide distribu- O’Keefe, J., and Dostrovsky, J. (1971). Brain Res. 34, 171–175. tion of locations containing grid-cell-like Many thanks to Daniel Bush, Laura Colgin, John activity suggests that they may play a Kubie, Colin Lever, and Lynn Nadel for helpful O’Keefe, J., and Nadel, L. (1978). The Hippocam- comments, and to support from the Wellcome pus as a Cognitive Map. (Oxford, UK: Oxford Uni- more general (nonspatial) role in autobio- Trust and the Medical Research Council UK. versity Press). graphical memory. As with spatial navigation, the work of O’Keefe, J., and Recce, M.L. (1993). Hippocampus REFERENCES 3, 317–330. the 2014 Nobel laureates has provided a strong foundation upon which a mecha- Barry, C., Hayman, R., Burgess, N., and Jeffery, O’Keefe, J., and Speakman, A. (1987). Exp. Brain nistic, neural-level understanding of K.J. (2007). Nat. Neurosci. 10, 682–684. Res. 68, 1–27. memory can be built. Knowledge of the Burgess, N., and O’Keefe, J. (2011). Curr. Opin. Scoville, W.B., and Milner, B. (1957). J. Neurol. relevant neural representations becomes Neurobiol. 21, 734–744. Neurosurg. Psychiatry 20, 11–21. even more powerful in combination with Burgess, N., Becker, S., King, J.A., and O’Keefe, J. Sreenivasan, S., and Fiete, I. (2011). Nat. Neurosci. an understanding of how changes in syn- (2001). Philos. Trans. R. Soc. Lond. B Biol. Sci. 14, 1330–1337. aptic connectivity between neurons can 356, 1493–1503. Stensola, H., Stensola, T., Solstad, T., Frøland, K., mediate learning via long-term potentia- Burgess, N., Maguire, E.A., and O’Keefe, J. (2002). Moser, M.B., and Moser, E.I. (2012). Nature 492, tion, a phenomenon that was also discov- Neuron 35, 625–641. 72–78. ered in the hippocampus by Tim Bliss and Derdikman, D., Whitlock, J.R., Tsao, A., Fyhn, M., Taube, J.S. (1998). Prog. Neurobiol. 55, 225–256. colleagues. Further advances, such as Hafting, T., Moser, M.-B., and Moser, E.I. (2009). the application of molecular biology and Nat. Neurosci. 12, 1325–1332. Wills, T.J., Cacucci, F., Burgess, N., and O’Keefe, J. (2010). Science 328, 1573–1576. optogenetics to manipulate hippocampal Fyhn, M., Molden, S., Witter, M.P., Moser, E.I., and representations, are beginning to allow Moser, M.B. (2004). Science 305, 1258–1264. Wilson, M.A., and McNaughton, B.L. (1994). Sci- ence 265, 676–679. investigations and manipulations of mem- Fyhn, M., Hafting, T., Treves, A., Moser, M.-B., and ory that would previously have been Moser, E.I. (2007). Nature 446, 190–194. Zhang, K. (1996). J. Neurosci. 16, 2112–2126.

Neuron 84, December 17, 2014 ª2014 Elsevier Inc. 1125