Spatial Cognition and the Brain

NEIL BURGESS
Institute of Cognitive Neuroscience, University College London, United Kingdom

Recent advances in the understanding of spatial cognition are reviewed, focusing on memory for
locations in large-scale space and on those advances inspired by single-unit recording and lesion
studies in animals. Spatial memory appears to be supported by multiple parallel representations,
including egocentric and allocentric representations, and those updated to accommodate self-
motion. The effects of these representations can be dissociated behaviorally, developmentally, and
in terms of their neural bases. It is now becoming possible to construct a mechanistic neural-level
model of at least some aspects of spatial memory and imagery, with the hippocampus and medial
temporal lobe providing allocentric environmental representations, the parietal lobe egocentric
representations, and the retrosplenial cortex and parieto-occipital sulcus allowing both types
of representation to interact. Insights from this model include a common mechanism for the
construction of spatial scenes in the service of both imagery and episodic retrieval and a role
for the remainder of Papez’s circuit in orienting the viewpoint used. In addition, it appears that
hippocampal and striatal systems process different aspects of environmental layout (boundaries
and local landmarks, respectively) and do so using different learning rules (incidental learning and
associative reinforcement, respectively).

Key words: parietal; hippocampal; striatal; fMRI; place cells; grid cells; allocentric; egocentric;
computational modeling

Introduction research has shifted from exclusive comparisons, for

example, “Is spatial memory egocentric?” (Wang &
The explosion of interest in cognitive neuroscience Spelke 2002; Shelton & McNamara 1997), to address-
derives from the ability of recent advances in neuro- ing how the various systems combine to support behav-
science to shed new light on the mechanisms support- ior under different circumstances (Burgess 2006; Mou
ing cognition. Within this endeavor, the field of spatial et al. 2004).
cognition is particularly well placed due to the ease There has also been considerable recent progress
with which similar (spatial) experimental paradigms in identifying the specific characteristics of neural rep-
can be applied to both humans and animals. Thus resentations of spatial location, including the startling
neuroscientific findings in animals can be integrated recent discovery of an entirely new type of spatial rep-
with noninvasive (behavioral, neuropsychological, and resentation “grid cells” (Hafting et al. 2005). These
neuroimaging) findings in humans. Accordingly, re- new findings are, in turn, beginning to feedback into
cent years have seen the reasonably direct application cognitive models of spatial behavior. A crucial aid in
of insights from single-unit recording in freely moving relating neurophysiological findings to behavior is the
animals to understanding the mechanisms of human use of computational modeling, which I touch upon
cognition regarding memory for locations within large- where appropriate.
scale space. In this review I focus on the basic representations
One theme to emerge, upon which I elaborate in and mechanisms supporting spatial memory, naviga-
this article, is the growing acceptance within the field tion, and imagery. This necessarily leads to a focus on
of human spatial cognition of a long-held tenet of ani- posterior parts of the brain: parietal, retrosplenial, and
mal neuroscience (e.g., White & McDonald 2002) that medial temporal neocortex as well as the hippocampal
multiple parallel systems are at play, making use of and striatal systems. Prefrontal cortex is only briefly
a variety of reference frames. Thus the emphasis of mentioned, in the context of supplying simulated mo-
tor efference in the service of planning/imagining po-
tential movements and in strategic mediation between
Address for correspondence: Neil Burgess, Institute of Cognitive Neu- hippocampal and striatal systems in controlling be-
roscience, University College London, 17 Queen Square, London WC1N
havior. The review is structured according to differ-
3AR, U.K.
n.burgess@ucl.ac.uk ent types of representations and processes and the

Ann. N.Y. Acad. Sci. 1124: 77–97 (2008).

C 2008 New York Academy of Sciences.
doi: 10.1196/annals.1440.002 77
78 Annals of the New York Academy of Sciences

corresponding neural systems. I start at the more sen- there whose firing is modulated by the orientation of
sory end, with egocentric representations of location the monkey within the room can support translation
and their updating to accommodate self-motion, and between egocentric and allocentric representations of
move on to more abstract allocentric representations locations. These neurons potentially allow translation
of location and their updating to accommodate self- between allocentric representations of environmental
motion. These issues, their neural bases, and their re- layout in the medial temporal lobe and head-centered
lationship to memory and imagery are treated within representations required for imagining spatial scenes
the framework of a model of medial temporal–parietal in medial parietal areas, see below and (Byrne et al.
interactions in spatial cognition (Byrne et al. 2007; 2007; Burgess et al. 2001b). In addition, the anterior
Burgess et al. 2001b; Becker & Burgess 2001). The bank of the parieto-occipital sulcus, which runs be-
second part of the review concerns the relationships tween the medial parietal and retrosplenial cortices,
between hippocampal and striatal systems in spatial contains visually responsive neurons which respond to
memory and navigation: which aspects of environmen- stimuli presented at a given location irrespective of the
tal information they process; which learning rules they direction of gaze (Galletti et al. 1995).
use; and how they combine to control behavior. The Behavioral evidence for egocentric representations
discussion focuses on the application of insights from in human spatial memory includes “alignment effects”
spatial paradigms to more general issues such as the in retrieval of an array of objects that was studied from
idea of a “geometric module” in the brain and the a specific viewpoint. Thus the time taken to correctly
roles played by the hippocampal, parietal, and stri- recognize the array in photographs from other view-
atal systems in supporting declarative and procedural points around the array increases with the size of the
memory and imagery. angular difference of the test viewpoint from the en-
coding viewpoint (see, e.g., Diwadkar & McNamara
1997). Equally, if people are asked to close their eyes
Multiple Parallel Reference Frames for
and imagine being at a different location and orienta-
Location
tion and then to point to where an object in the array
Egocentric Representations would be (tasks sometimes referred to as judgments
Locations in the external world can be represented of relative direction), they are faster and more accu-
in a variety of ways. Sensory information is generally rate when the imagined viewpoint has the same direc-
acquired in the coordinates of the corresponding re- tion as the studied viewpoint (Shelton & McNamara
ceptor (e.g., retinotopic for vision and head-centered 1997). These findings are consistent with storage of a
for audition), while actions must be specified in the viewpoint-dependent representation of the array, fol-
appropriate body-centered coordinates for the cor- lowed by a cumulative process of mental movement of
responding effector. Sensorimotor integration, as in viewpoint. (Note, however, that some findings can also
reaching for a visual target, requires translation be- be interpreted as interference between the imagined
tween these various egocentric representations. Evi- perspective and the participant’s current perspective,
dence for all of these egocentric representations can be rather than the stored perspective, e.g., May 2004).
found at the level of single neurons in sensory and mo-
tor cortices, and mechanisms for translation between Self-Motion and Egocentric Spatial Updating
them are suggested by “gain field” responses of neu- An interesting puzzle implied by egocentric percep-
rons recorded in posterior parietal area 7a in primates. tual representations is that of the perceived stability
These neurons respond to visual stimuli at a specific of the external world despite the rapid and rapidly
retinotopic location, but their rate of firing is also mod- changing motion of our sensory receptors as we move.
ulated by the orientation of the monkey’s gaze relative The problem of perceptual stability is beyond the
to the head (Andersen et al. 1987), by the orientation scope of this review (see Ross et al. 2001; Bridgeman
of the head relative to the trunk, or by the orienta- et al. 1994; Melcher 2007 for more on visual stabil-
tion of the monkey within the testing room (Snyder ity). Suffice it to say, following Helmholtz (1866), that
et al. 1998), see FIGURE 1. These gain-field responses a major component of the solution appears to be the
are ideal for translating locations between the various automatic updating of sensory representations by in-
egocentric reference frames (Zipser & Andersen 1988; formation about intended movements, often referred
Pouget & Sejnowski 1997; Deneve et al. 2001). to as “motor efference copy.” In a recent parallel to
Of particular interest for the following sections, area the long history of behavioral studies of visual stability,
7a is the posterior parietal area most strongly con- electrophysiological studies in primate posterior pari-
nected with the medial temporal lobe, and the neurons etal cortex have examined the effects of eye movements
Burgess: Spatial Cognition & the Brain 79

FIGURE 1. Parietal gain fields: encoding of the retinotopic location of a stimulus modulated by the angle of gaze
relative to the head (B-C), the angle of the head relative to the trunk (D), or the angle of the trunk in the lab (not shown).
(A) Example of a neuron in primate parietal area 7a with a retinotopic receptive field for visual stimuli, showing peak
firing rate as a function of stimulus location relative to fixation at the center. (B) The angle of gaze relative to the head can
be varied, modulating the amplitude of the response to a stimulus in the same retinotopic location. (C) Arrow indicates
stimulus onset; the position of each plot and figures in brackets indicates the angle of fixation. (D) Retinotopic responses
can also be modulated by the angle of the head relative to the trunk. Some neurons show the same modulation by angle
for movement of the head or for passive rotation of the trunk (shown, cf. above and below). Others show modulation by
the angle of gaze within the room for the same angles of gaze relative to the head and of the head relative to the trunk.
Adapted from Andersen, Essick, & Siegel (1985) and Snyder, Grieve, Brotchie, & Andersen (1998).

on the responses of single neurons. Interestingly, neu- indicate spatial updating of neuronal representations
rons with retinotopic receptive fields, for example, in by motor efference copy (see e.g., Colby & Goldberg
the lateral intraparietal area, can be seen to update 1999 for a review).
their responses so as to respond to stimuli which will Of course, the other potential mechanisms for spa-
be in the receptive field after a saccadic eye movement, tial updating, aside from motor efference copy, should
even though the stimulus has actually disappeared be- not be ignored. These include the integration of
fore the completion of the movement. These results vestibular signals reflecting accelerations of the head,
80 Annals of the New York Academy of Sciences

proprioceptive information regarding actual move- memory concerns the distribution of errors in point-
ments performed, and optic flow. All of these types of ing to object locations, developed by Wang and Spelke
information can contribute to the processes of spatial (2000). In this paradigm, the participant views an ar-
updating and have been extensively studied within the ray of objects scattered throughout a room and must
more restricted context of “path integration”: updating then point to them from within a chamber in the cen-
a representation of one’s own displacement from the ter of the room (from which the objects can not longer
start of a movement trajectory by integrating the veloc- be seen). Wang and Spelke reasoned that disorien-
ities experienced along the trajectory. We initially con- tating the participant by blindfolded rotation would
sider the likely egocentric processes supporting spatial have different effects upon egocentric and allocen-
updating in parietal areas and later consider likely allo- tric representations of the object locations. Namely,
centric processes supporting spatial updating in medial updating of individual egocentric representations will
temporal areas. Note that, although spatial updating induce different amounts of error across the differ-
and path integration are regarded as egocentric pro- ent locations, while updating of an integrated allocen-
cesses in some treatments (e.g., Wang & Spelke 2002), tric representation should induce similar amounts of
categorization of these processes per se is arbitrary, error across the different locations. Consistent with
depending on whether the object or start locations their egocentric-only model (Wang & Spelke 2002),
are updated relative to the participant or whether the Wang and Spelke (2000) found that blindfolded ro-
location of the participant is updated relative to the tation increased the variance in errors across object
locations of objects, start positions, or other aspects of locations.
the environment. An interesting parallel to the effects of actual self-
Following work from Rieser (1989) and others, Si- motion on spatial representations is provided by studies
mons and Wang (Simons & Wang 1998; Wang & of imagined movement of viewpoint. In experiments
Simons 1999) introduced an elegant paradigm for on imagery, subjects study an array of objects and are
dissociating the contributions to spatial memory of subsequently blindfolded. They can then be asked to
sensory-bound representations such as “visual snap- indicate the locations of objects following either imag-
shots” from representations of locations relative to the ined rotation of the array of objects or following imag-
body which are automatically updated to accommo- ined translocation of themselves to a new viewpoint
date self-motion. In this paradigm, people see five around the array. In these situations performance is
objects on a circular table and later have to indicate superior following imagined movement of the viewer
which one has been moved. In between presentation than following an equivalent imagined movement of
and test phases, with the table out of view, it is possi- the array (e.g., Wraga et al. 2000). It is only when the
ble to have the person change his or her viewpoint, or array consists of a single location that performance for
for the table to be rotated. In four conditions, either imagined array-rotation approaches that for imagined
the person’s viewpoint or the table is rotated about movement of the viewer.
the center of the table or both are rotated, or neither. Thus, there is strong evidence that egocentric rep-
The angle of rotation is the same for the viewpoint resentations of locations are maintained in the brain
and for the table so that, if both are rotated, the ego- and that these are automatically updated by our own
centric locations of the objects relative to the viewer movements, intentions to move, or imagined move-
are unchanged. See FIGURE 2 conditions _, P, PT, and ments. However, it is also possible that, where multi-
T. The consequence of these manipulations is a 2x2 ple locations or extended layouts are concerned, it is
factorial design in which the test array is either con- more efficient to maintain a cognitive representation
sistent or inconsistent with viewpoint-dependent rep- of the world and to update our own location within
resentations of locations (consistent if both viewpoint it, rather than maintaining multiple egocentric rep-
and table are rotated together or if neither move) and resentations each of which is affected differently by
also consistent or inconsistent with representations that self-motion. In addition, having to retain information
are updated by self-motion (consistent if the viewpoint over prolonged self-motion increases the importance
alone changes or if neither move). The results indi- of knowledge of locations relative to environmental
cate a strong positive effect of consistency with rep- landmarks in avoiding the cumulative errors associ-
resentations updated by self-motion and a weaker ef- ated with egocentric updating. As we shall see in the
fect of consistency with viewpoint-dependent sensory next section, recent evidence points to the presence
representations. of these allocentric or world-centered representations
Another paradigm for investigating the presence in parallel to the egocentric representations discussed
of egocentric or allocentric representations in spatial above.
Burgess: Spatial Cognition & the Brain 81

A B +
SU condition
_ S
Card
Table + TC _
STC
EC

_ SC

1m T
_ C
VS ST
Subject +

C 1.0
Performance

0.8

0.6

0.4

0.2
_ C S SC ST STC T TC
Condition
FIGURE 2. Paradigm for dissociating spatial reference frames, based on Simons and Wang (1998).
(A) After initially seeing the array of objects and before being asked which one has moved, the position
of the person (P), the table (T), or the external cue card (C) can be rotated about the center of the
table. (B) These changes provide a fully factorial manipulation of the consistency (+ or −) of the test
array with 3 different potential types of stored representation: visual snapshots (VS), representations
spatially updated to accommodate the subject’s movement (SU), representations of location relative to the
external cue card (EC). (C) Performance benefits from consistency with any of these representations: the
more the test array is consistent, the better performance is. Adapted from Burgess, Spiers, & Paleologou
(2004).

Allocentric Representations representation of the location relative to the available

Although egocentric representations provide the ob- environmental landmarks will often be of more use
vious format for sensory perception and motor action, than egocentric representations. See Burgess (2006) for
and transformations between such egocentric repre- further discussion.
sentations suffice for short-term sensorimotor integra- The idea of parallel representations (egocentric ver-
tion, it has long been argued that memory over the sus allocentric in the above discussion) with differen-
longer term is likely better served by allocentric repre- tial utility according to the amount of self-motion be-
sentations centered on environmental landmarks (e.g., tween presentation and test was recently investigated
Milner et al. 1999; O’Keefe & Nadel 1978). Updat- by Waller and Hodgson (2006). They reinterpreted the
ing of egocentric representations of location to accom- results of Wang and Spelke (2000), described above, ar-
modate self-motion (referred to as “path integration” guing that increased variance in pointing errors after
when the location in question is the start of the path) blindfolded rotation might result from a switch from
will fall prey to cumulative error after relatively short using an accurate but transient representation to using
paths (e.g., Etienne et al. 1996). Thus, when attempting a less accurate but more integrated enduring represen-
to return to a previously visited location from a new tation. In support of their interpretation, they found
direction after more than a few minutes of self-motion, that, although variation in pointing errors increases
82 Annals of the New York Academy of Sciences

FIGURE 3. Place cells, head-direction cells, and grid cells. (A) Example of a place cell: the black line
shows the rat’s path in exploring a square enclosure, red squares show the locations at which a single
neuron fired an action potential. (B) Example of a head-direction cell, showing firing rate as a function
of head direction within the enclosure. (C) Example of a grid cell: firing occurs at an array of locations
arranged in a regular triangular grid across the environment (same color scheme as A). (D) Example of
three neighboring grid cells simultaneously recorded on the same tetrode, action potentials from the three
cells shown in red, blue, and green. Adapted from Jeffery & Burgess (2006).

after disorientation (consistent with the use of a less neurophysiological evidence for allocentric representa-
accurate representation), variation in the errors made tions in the mammalian brain and then describe some
when judging the relative direction of one object from recent experiments indicating the presence of allocen-
another actually reduced after disorientation (consis- tric representations in human spatial cognition.
tent with the use of a more integrated representation). A neural representation of the animal’s location rel-
Waller and Hodgson also found that the effect of “dis- ative to the surrounding environment can be seen in
orientation” on pointing error occurs in an all-or-none the firing of “place cells” in the hippocampus of rats
fashion after rotations of 135◦ or more, consistent with (O’Keefe 1976) and primates (Ono et al. 1991), see
a switch from one representation to the other after Muller (1996) for a review. This representation is sup-
movements of a certain magnitude. In addition, they ported by representations of orientation (Taube 1998)
found no disorientation-related increase in pointing and a grid-like representation suitable for path inte-
error variation when pointing to objects within a very gration (Hafting et al. 2005) in nearby regions, both
familiar environment, consistent with the development also environment centered. See FIGURE 3. The ori-
of more accurate enduring representations with expo- entation of the firing patterns of these cells is pri-
sure to an environment, and preferential use of them marily determined by distal visual cues, when avail-
even over short movements and timescales. Here I in- able, with place cells specifically encoding location
terpret Waller and Hodgson’s “enduring representa- relative to extended boundaries in the environment,
tion” as likely to be allocentric, see Burgess (2006) for given this orientation (O’Keefe & Burgess 1996; Hart-
further discussion. Below I briefly outline some of the ley et al. 2000). These representations appear to guide
Burgess: Spatial Cognition & the Brain 83

behavior in spatial memory paradigms in which simple ary within the testing room, are also consistent with
egocentric representations do not suffice. In these situ- allocentric representations centered on environmental
ations, behavioral responses match the firing of the cells cues. Thus, some of the effect ascribed to spatial up-
(Kubie et al. 2007; Lenck-Santini et al. 2005; O’Keefe dating may be due to the presence of allocentric rep-
& Speakman 1987). In addition, hippocampal le- resentations. To pull apart these multiple influences,
sions or inactivations impair performance in these Burgess et al. (2004) included independent manipu-
tasks (e.g., Morris et al. 1982; Packard & McGaugh lation of environmental cues: testing effects of con-
1996). sistency with viewpoint-dependent, spatially updated,
Recent investigations of the neural bases of nav- and allocentric representations within a 2 × 2 × 2 de-
igation in humans have made use of desk-top vir- sign, see FIGURE 2. In this test, people viewed an array
tual reality (VR): allowing simulation of movement of fluorescent objects with an external fluorescent cue,
through large-scale space in stationary subjects, al- in darkness, and subsequently indicated which object
beit without the vestibular and proprioceptive com- had moved. Between presentation and test, the per-
ponents. Using VR, neural responses resembling those son’s viewpoint, the array, or the external cue could be
of place cells have been found in the human brain, rotated to change the consistency of the test array with
clustered in the hippocampus (Ekstrom et al. 2003), either type of representation. In addition to replicating
while functional neuroimaging (Maguire et al. 1998a; the effects of consistency with viewpoint-dependent
Hartley et al. 2003; Iaria et al. 2003) and neuropsy- and spatially updated representations (when the cue
chological (Abrahams et al. 1997; Spiers et al. 2001a; did not move), an effect of consistency with the orien-
Spiers et al. 2001b) data confirm the involvement of tation of the external cue was also found. For example,
the human hippocampus in accurate large-scale nav- performance increased when the card and table moved
igation. In addition, Hartley and colleagues (2004) together compared to when one or other moved alone.
used VR to investigate the effect of deformation of Thus, allocentric representations of object locations
the environmental boundary on human search loca- relative to environmental cues probably exist in paral-
tions, finding results compatible with the assumption lel to egocentric representations of location relative to
that place cells guide behavior, given how place cells the subject.
respond to such manipulations (O’Keefe & Burgess The Simons and Wang–inspired paradigm of dis-
1996). sociating frames of reference by shifting the viewpoint
Direct evidence for allocentric representations in and/or the array of objects has recently been success-
human spatial cognition has come from recent fully applied to developmental psychology. Thus, rep-
paradigms designed to replicate earlier experiments resentations of locations within the testing room appear
showing evidence of egocentric representations (dis- to be present as early as 3 years and to make a greater
cussed above), but also designed to probe any allocen- contribution to behavior than egocentric snapshots at
tric representations which might exist in parallel. Thus, this age (Nardini et al. 2006). Representations of loca-
memory for locations within an array has recently been tion relative to the intrinsic frame of reference of the
found to show effects of the alignment of the testing array appear to develop between years three and six.
perspective with directions defined by aspects of the Although the relative dependence of room-related re-
external environment, as well as with those defined sponding on allocentric representations or egocentric
by the person’s initial viewpoint (Mou & McNamara spatial updating is not clear, the results demand a re-
2002; Schmidt & Lee 2006). When an array of objects think of Piagetian ideas of early egocentrism at least
contains an intrinsic axis (e.g., defined by symmetry), (Piaget & Inhelder 1956).
improved performance is found when pointing to ob-
jects from imagined viewpoints that are aligned with Temporo-Parietal Mechanisms of Spatial
this axis or aligned with environmental features such as Memory, Imagery, and Motion-Related
testing room walls (Mou & McNamara 2002; Schmidt Updating
& Lee 2006) and external landmarks (McNamara The nature of the representation of location by place
et al. 2003). cells has received much study over the several decades
The use of allocentric and egocentric represen- since their discovery. As a result, a neural-level model
tations can also be dissociated within Simons and of spatial memory has begun to emerge from these
Wang’s (1998) egocentric spatial updating paradigm findings, in combination with findings in related ar-
(see above). In the original paradigm the conditions eas and in the parietal lobe. As a starting point, we
consistent with representations updated by self-motion, briefly review some of the evidence concerning how
which all involve object locations that remain station- environmental cues determine the spatial firing fields
84 Annals of the New York Academy of Sciences

A B C

Data Model

D E
FIGURE 4. The boundary vector cell model of the sensory input to place cells. (A) Putative place cell inputs (boundary
vector cells, BVCs) are assumed to be tuned to respond to the presence of a boundary at a given distance and allocentric
direction (firing rate indicated by the bar on the left), with broader tuning at greater distances (below). (B) The firing
of a place cell in four different enclosures can be modeled by an appropriate selection of BVC inputs. (C) The model
can predict the firing of the cell in new configurations of the enclosure, including those with an internal barrier. (E) The
actual firing of the cell in these environments reasonably matches the prediction. (D) Adapted from Burgess & Hartley
(2002).

of place cells, and thus contribute to the rat’s sense of consistently when external or internal orientation is
self-location. manipulated.
The relative independence of place cell firing from In contrast to the robust effect of environmental
low-level sensory representations can be seen in the boundaries on place cell firing, discrete landmarks
independence of place cell firing from the animal’s ori- within an environment have very little effect on place
entation as it explores open environments (Muller & cell firing (Cressant et al. 1997). Equally, while remov-
Kubie 1989) and in the robustness of the response to ing individual distal cues to orientation does not have a
removal of the sensory cues controlling the orientation marked effect on place cell firing (although the overall
of the firing fields within the environment (O’Keefe & orientation of the representation may drift), removing
Conway 1978; O’Keefe & Speakman 1987; Pico et al. environmental boundaries tends to lead to destruction
1985; Fenton et al. 2000). In addition to cues to orienta- of the place cell response (Barry et al. 2006; Barry
tion (Taube 1998), place cell firing is strongly driven by et al. 2007). The BVC model has been used both to
any extended boundaries to motion within the environ- predict the pattern of place cell firing following de-
ment. O’Keefe and Burgess (1996) recorded from the formation of the environmental boundaries to make
same cells across similar (rectangular) environments an environment of different shape and size, includ-
that differed in their dimensions. They observed that ing the addition of extended walls within the maze
the location of peak firing of a given place cell typ- (Hartley et al. 2000). It has also been used to predict
ically remained in a constant position relative to the the search behavior of humans returning to a previ-
nearest walls, and in addition, several of the firing ously seen location, by assuming that they solve the
fields were stretched along the axes of the environ- task by moving to maximize the match between their
ment. They proposed that place cells received inputs current place cell representation and a stored place
that are tuned to respond to the presence of a barrier cell representation of the target location (Hartley et al.
at a given distance along a given allocentric direction, 2000; Hartley et al. 2004; O’Keefe & Burgess 1996).
with sharper tuning at shorter distances; this is the so- The BVC model, in combination with models of
called boundary vector cell (BVC) model (Barry et al. the firing properties of neurons in the parietal lobe
2006; Hartley et al. 2000), see FIGURE 4. The allocen- (Pouget & Sejnowski 1997; Salinas & Abbott 1995)
tric directions of BVC tuning are presumably defined suggest a computational model of memory and im-
relative to the head-direction cells, given that place agery for spatial scenes (Byrne et al. 2007; Burgess
cell and head-direction cells always seem to rotate et al. 2001b; Becker & Burgess 2001). In this model,
Burgess: Spatial Cognition & the Brain 85

Top Down
Bottom Up Medial Temporal

Retrospl.
PW Head BVCs
Egocentric modulated Allocentric
Frame Transformation Frame
CA3
Place
I Cells
PR
HD Object
Cells Identity

Head Rotation
Velocity
FIGURE 5. Schematic temporo-parietal model of spatial memory and imagery. Each box represents
a set of neurons in a different brain region (PW: medial parietal “window” for visual perception and
imagery; Retrospl: retroplenial cortex and parieto-occipital sulcus mediating ego-allo translation in con-
junction with area 7a; HD cells: head-direction cells; BVCs: parahippocampal boundary vector cells; PR:
perirhinal cortical encoding of visual identity). Thin solid arrows represent “bottom-up” connections from
egocentric visual perception and imagery to allocentric medial temporal memory representations; thin
dashed arrows represent “top-down” connections back again. Adapted from Byrne, Becker, & Burgess
(2007).

connections between hippocampal place cells, describe the left-hand side of an imagined view of a
parahippocampal boundary vector cells, and perirhi- famous piazza in their home town, whether imag-
nal cells encoding visual textures/features form an ining facing towards the cathedral or away from it.
associative memory. Thus, within a familiar environ- This is consistent with an intact medial temporal al-
ment, a partial cue can reactivate the hippocampal locentric representation of the whole square along
representation of occupying a single location within with damage to the parietal substrates of the egocen-
an environment, which in turn reactivates the corre- tric representation or the allo-ego translation mecha-
sponding parahippocampal and perirhinal representa- nism. In addition, experiments in which place cells are
tions of environmental boundaries and visual features recorded while visual and path-integrative information
respectively. To be able to examine the products of this are put into conflict (Gothard et al. 1996) can also be
reconstructive process in visual imagery, the allocen- simulated.
tric (North, South, East, West) parahippocampal rep- As noted earlier in the review, it is important to
resentation must be translated into an egocentric (left, bear in mind the multiple ways in which locations can
right, ahead, behind) medial parietal representation. be updated to accommodate self-motion. The model
This is assumed to occur via processing by gain-field described above uses the translation from allocentric
neurons in posterior parietal cortex and representation medial temporal representations to egocentric medial
of the intermediate stages of translation in retrosple- parietal representations to perform spatial updating of
nial cortex/parieto-occipital sulcus, making use of the the egocentric locations, and translation back to the
representation of head direction found along Papez’s medial temporal representations to make sure the al-
circuit to dereference allocentric directions into ego- locentric representation of self-location is also updated
centric ones, see FIGURE 5. appropriately. However, it is also theoretically possible
The model is able to simulate effects found in neu- to directly update the allocentric representation of self-
ropsychological and single-unit recording experiments. location given self-motion information. Again, both
For example, the effect of hemi-spatial neglect in im- processes may exist in parallel, with egocentric updat-
agery following right parietal damage can be sim- ing most useful for keeping track of small numbers of
ulated, as in the famous Milan Square experiment locations over short durations and allocentric updating
(Bisiach & Luzzatti 1978), in which patients could not most useful where one’s position must be maintained
86 Annals of the New York Academy of Sciences

relative to larger amounts of environmental informa- original references for the details of these proposals.)
tion and over longer durations. In addition, the reciprocal connections between en-
The likely importance of task demands in determin- torhinal cortex and the hippocampus might allow the
ing whether egocentric or allocentric mechanisms for place cell and grid-cell representations to combine both
spatial updating are used to control behavior is illus- motion-related inputs (to grid cells) and environmental
trated by comparison of the studies by King et al. (2002; sensory information (the BVC inputs to place cells) in
King et al. 2004) with that by Shrager et al. (2007). determining the animal’s current location (Barry et al.
In this first group of studies, participants saw objects 2007; O’Keefe & Burgess 2005; Burgess et al. 2007).
within a VR arena and were subsequently tested on A recent fMRI study in humans (Wolbers et al. 2007),
their memory for the objects’ locations from the same in which they performed a path-integration task us-
viewpoint as at presentation or from a shifted view- ing only optic flow, showed performance to correlate
point. In the studies by King et al., the new viewpoint with activation of the anterior hippocampus, possibly
was imposed abruptly and, when different numbers of consistent with a role in allocentric spatial updating.
objects were used, the trials with different list lengths
were intermingled. In these studies a developmental Parallel Hippocampo-Striatal Systems in Rats
amnesic with focal hippocampal pathology was found The place cell data, summarized briefly above, seem
to be specifically impaired from the shifted viewpoint, to indicate that the hippocampus specifically processes
interpreted as consistent with hippocampal support of the surface geometry of the rat’s environment, with
the allocentric mechanism. In the subsequent study an important role also for the head-direction system,
(Shrager et al. 2007), participants watched as the vir- governed by distal cues to orientation. Hippocampal
tual arena rotated in front of them between presenta- lesions dramatically impair performance on the classic
tion and test and performed tasks with increasing list- version of the water maze, where rats must use dis-
lengths in order (i.e., spending several trials watching tal landmark information as well as distance to the
one location rotate, then two, then three). In this study, maze boundary to locate a hidden platform (Morris
no specific effect of hippocampal damage upon perfor- et al. 1982). Interestingly, the maze walls are powerful
mance from a shifted viewpoint was found, which was cues used to locate the platform even when they are
interpreted as an absence of evidence for hippocam- transparent, illustrating the importance of continuous
pal support of allocentric updating. However, a likely boundaries for navigation (Maurer & Derivaz 2000).
alternative interpretation, to my mind, is that the later Distinct hippocampal and striatal contributions to
study was solved by egocentric mental rotation. spatial navigation can be seen in tests in the water
It may be that the grid cells recently discovered maze. Hippocampal lesions do not disrupt the ability
in medial entorhinal cortex (Hafting et al. 2005) pro- to navigate towards a location marked by a distinct
vide the neural substrate for allocentric updating of the visible landmark (or “beacon”). By contrast, striatal le-
place cell representation of one’s own position within sions impair navigation towards a location marked by
the environment, as follows. As a rat moves through a distinct visible landmark but not to an unmarked
its environment, each grid cell fires whenever the rat one defined relative to distal landmarks and bound-
enters one of several locations which fall at the vertices aries (Packard & McGaugh 1992; McDonald & White
of a regular triangular grid across the environment, 1994). When a location is defined by its distance and
see FIGURE 3. The grids of neighboring grid cells are direction from an intramaze cue (given distal orient-
simply shifted copies of each other, so that the rela- ing cues), and not by the maze boundary, hippocam-
tive positions of the firing locations of two grid cells pal damage does not impair navigation (Pearce et al.
remain constant across an environment and also re- 1998), although lesions of the anterior thalamus (with
main constant across different environments (Fyhn et presumed disruption of the head-direction system) do
al. 2007). The nearby presubiculum, which contains impair navigation (Wilton et al. 2001). Thus, the hip-
head-direction cells, projects into medial (but not lat- pocampus may define locations relative to the bound-
eral) entorhinal cortex. This projection may allow the ary, while the striatum defines locations relative to local
grid cells to perform path integration, allowing the landmarks, and the head-direction system is required
grid cell activations to be updated in correspondence to derive a heading direction from distant landmarks.
with the rat’s movement. This could occur by each cell The distinct contributions of hippocampal and stri-
passing activity on to the appropriate neighbor (Mc- atal systems to spatial cognition can also be seen in
Naughton et al. 2006; Sargolini et al. 2006; Fuhs & the plus maze. In this task, rats are trained in an ini-
Touretzky 2006) or by each cell integrating movement tial learning phase to retrieve food from the end of
information individually (Burgess et al. 2007). (See the one arm (e.g., West), starting from another arm (e.g.,
Burgess: Spatial Cognition & the Brain 87

South). This paradigm can be used to elegantly study and striatal (caudate) activation to reflect navigational
whether rats learn to navigate to the food through speed (Maguire et al. 1998a). Recent advances in imag-
learning a stereotyped response (turn left) or through ing technology, more realistic environments, and more
learning the place within the test room (presumably sophisticated analyses of behavior (Spiers & Maguire
defined by distal cues in the environment). The use 2007) have refined these interpretations further. Hart-
of a response or a place strategy can be assessed dur- ley et al. (2003) found that hippocampal activation
ing a probe trial in which rats start from a novel arm corresponded to flexible wayfinding using new paths
(e.g., North). The rat could either follow the learned through previously explored environments, while stri-
response, that is, turning left and thus searching for atal activation corresponded to following well-used
food in the East arm (response strategy), or follow a routes (explaining the correlation with speed in the
place strategy and search in the arm in the West of previous study). Wolbers and Buchel (2005) examined
the testing room. In probe trials after 8 and 16 days activation during the learning of a new environment
of training, healthy rats shifted from approaching the and found hippocampal activation to correspond to
“place” associated with food after 8 days to making increases in knowledge of the environmental layout,
the turn “response” associated with food after 16 days. while retrosplenial activation corresponded to the ab-
However, injections of lidocaine to inactivate the hip- solute level of performance.
pocampus abolished place learning, while injections Iaria et al. (2003) adapted an elegant paradigm for
into the striatum abolish response learning (Packard & identifying the use of distal cues in rat navigation for
McGaugh 1996). use with humans. In this task, subjects found objects in
We have described some of the different types of 4 arms in a virtual 8-arm maze with distal cues present
neural representation available to animals when solv- around it to provide orientation. Their memory was
ing spatial tasks. The head-direction system found then tested by asking them to revisit the same 4 arms
throughout Papez’s circuit may provide orientation, again—entering the other arms counted as an error.
while the hippocampus has been identified with In probe trials, the distal cues were removed during the
environment-centered representations of locations and test phase: an increase in the number of errors indi-
the dorsal striatum has been associated with approach cated that the subject was making use of the distal cues
responses to a single landmark. In addition, the pro- rather than, for example, remembering a sequence of
jections from the head-direction system to the nucleus turns. When the study was performed in an fMRI ex-
accumbens imply that the striatal system might also periment, Iaria et al. found that the dependence on dis-
allow navigation towards unmarked locations by using tal cues correlated with hippocampal activation, while
a visible local landmark in conjunction with external distal cue–independent responding correlated with ac-
orienting cues. In the next section we consider recent tivation of the caudate nucleus. These results are con-
evidence for a similar dissociation between right pos- sistent with hippocampal provision of an allocentric
terior hippocampal and right dorsal striatal substrates representation, requiring the distal cues, and striatal
of spatial learning in the human brain. provision of route-like egocentric responses.
It seems that the hippocampal and striatal systems
Parallel Temporo-Striatal Systems in can act cooperatively in the context of adaptation to
Humans brain damage. Voermans et al. (2004) compared the
Study of the neural bases of large-scale navigation in activation provoked by remembering routes through
humans has recently begun to take advantage of a com- houses (shown as video clips) between a group of pa-
bination of desk-top VR and functional neuroimaging. tients with Huntington’s disease and a group of healthy
Several early studies revealed activation in parietal, ret- volunteers. They found reduced caudate activation
rosplenial, and parahippocampal areas as people find corresponding to the progression of the disease (which
their way around (e.g., Aguirre & D’Esposito 1997; attacks this part of the brain) in the patient group, but
Maguire et al. 1998b; Ghaem et al. 1997), but in- also increased hippocampal activation. Thus the more
terpretation of the functions of specific subregions re- flexible hippocampal system may be able to take over
mained difficult. In further experiments, some patterns some of the function of the striatal system in remem-
began to emerge. Thus parahippocampal gyrus acti- bering routes. The extent to which the striatal system
vation may reflect sensory (Epstein & Kanwisher 1998) could accommodate for hippocampal damage is an
or mnemonic (Janzen & van Turennout 2004) process- interesting question for future research.
ing of spatial scenes and perhaps the use of peripheral As well as the above dissociation between striatal
vision in this (Levy et al. 2001), while hippocampal support of overlearned route-like responses and hip-
activation was found to reflect navigational accuracy pocampal support of more flexible navigation, the
88 Annals of the New York Academy of Sciences

FIGURE 6. Experimental paradigm for comparing the relative contributions of local landmarks and
boundaries to spatial memory. Participants play an adapted first-person perspective videogame in which
they find objects within a virtual environment comprising a circular boundary and a landmark, with distal
cues to orientation (rendered at infinity). (A) After collecting the objects in their respective locations,
each subsequent trial has a cue phase (an object is shown) and a replace phase (the subject attempts to
navigate to the object’s location and presses a button). Learning trials, but not test trials, contain feedback
from which subjects learn to improve their performance (i.e., replacement accuracy). Feedback consists
of the object appearing in the correct location and being collected by the subject. A view of the virtual
environment is shown in (B). Adapted from Doeller and Burgess (2007).

animal studies reviewed above also suggest different with the same object. Once the two cues had been
neural bases for processing locations relative to local moved relative to each other, the relative influence of
landmarks and environmental boundaries. This disso- landmark or boundary on responding was reflected
ciation has recently been examined in humans (Doeller implicitly in the distance of the response location from
et al. 2007) using a VR object-location memory task, in the locations predicted by either cue. Both cues played
which, without being distinguished by any explicit in- functionally equivalent roles in the task and were not
structions, some objects maintained a fixed relation to distinguished in the participants’ instructions, and their
the environmental boundary while others maintained relation to the distant orientation cues remained un-
a fixed relation to a single intramaze landmark. Par- changed as these were projected at infinity.
ticipants explored a VR arena bounded by a circular Participants gradually learned locations relative to
wall, containing a single landmark and surrounded by both types of cue at similar rates, with performance
distant orientation cues. Within this arena they en- increasing similarly within and across blocks (FIG. 7).
countered four objects in four different locations, see Inaccurate responses largely reflected use of the incor-
FIGURE 6. On each subsequent trial they saw a picture rect cue early in each block. Consistent with the predic-
of one of the objects (the “cue phase”) and indicated tions from animal studies, fMRI activation in the right
its location within the arena by navigating to it from a dorsal striatum during the feedback phase correlated
random start location and making a button-press re- with learning for landmark-related objects, while acti-
sponse (the “replace” phase). The object then appeared vation in the right posterior hippocampus correlated
in its correct location and was collected (the “feedback” with learning for boundary-related objects. In addi-
phase). Each set of 16 trials (four per object) composed tion, the influence of the boundary on the replacement
a block, with four blocks in the entire experiment. Crit- location correlated with right posterior hippocampal
ically, the landmark and boundary were moved relative activation, while the influence of the landmark corre-
to each other between blocks, with two objects main- lated with right dorsal striatal activation. Thus, differ-
taining their location relative to the boundary and two ential activity seen in the hippocampus and caudate
relative to the landmark. corresponded to the acquisition and expression of in-
Performance was measured in terms of the prox- formation about locations derived from environmental
imity of response location to the correct location, and landmarks or boundaries respectively.
learning during the feedback phase could be measured This analysis raises the question of what distin-
as the improvement in performance on the next trial guishes a boundary from a landmark? A simple
Burgess: Spatial Cognition & the Brain 89

A directions around the rat. Thus, for place cells, any

obstacle in the environment simply has an importance
proportional to the horizontal angle it subtends at the
rat. This potentially provides a measure of “boundary-
ness” in terms of proximity and extent. However, it is
B
not clear what aspects of environmental stimuli corre-
spond to “landmarkness” as reflected in their influence
on striatal processing.
A final consideration of this study was the nature of
the interaction between the two systems in controlling
behavior after they had been put into conflict by the
relative motion of the two cues. In this study, the levels
of activation in hippocampus and striatum could be in-
vestigated trial-by-trial as both systems contributed to
a single behavior. No evidence was found of a direct in-
teraction between activation in the two systems, other
than the variation reflected in behavior. Thus it seems
C
that both systems operate in parallel, with their activa-
tion signaling their suitability to control behavior. Such
an interpretation would be consistent with the effects of
lidocaine injections in rats (Packard & McGaugh 1996)
and extends previous studies of hippocampo-striatal in-
teraction by looking at different tasks or different stages
of the same task (Poldrack et al. 2001).

Incidental and Reinforcement Learning Rules

in Hippocampus and Striatum
So far, I have concentrated on the representations
supporting spatial cognition, but what of the rules
underlying the learning of adaptive behavior? “Re-
inforcement learning” (Sutton & Barto 1988) provides
FIGURE 7. An fMRI study of the neural bases of learning the dominant model for learning from feedback over
locations relative to a local landmark or a local boundary; multiple trials. This model, derived from the Rescorla-
the paradigm is shown in FIGURE 6. (A) After four objects Wagner law (Rescorla & Wagner 1972) and its de-
have been collected in their locations, participants com- velopment into associative learning theory (see e.g.,
pleted blocks of four trials replacing each object, with feed- Mackintosh 1975 and Dickinson 1980), associates cues
back regarding their correct location, see FIGURE 6. At the
end of each block, the landmark and boundary were moved
to a measure of predicted feedback (see also, Dayan &
relative to each other, with two objects moving with the land- Abbott 2002). This rule explains results in classical con-
mark (shown in orange), and two with the boundary (shown ditioning and also in “instrumental conditioning,” or
in green). (B) Object locations were learned within and be- learning to act, by also associating cues to appropriate
tween blocks from the feedback provided. Average replace- actions. In both cases, the learning rule for adjusting
ment error is shown separately for the landmark-related (or- the strengths of these associations depends on the dif-
ange) and boundary-related (green) objects. (C) Activation
ference between the predicted and actual feedback (the
during the feedback phase corresponded to learning (im-
provement on the next trial with that object) in the right stria-
“prediction error”). This simple and elegant model has
tum for landmark-related objects and in the right posterior provided a quantitative description of a great variety
hippocampus for the boundary-related objects. Adapted of behavioral data on learning and has received strong
from Doeller, King, & Burgess (2007). support from evidence that the prediction error sig-
nal can be seen in the striatal targets of the dopamine
potential answer comes from simulation of the fir- system (O’Doherty et al. 2004; Pessiglione et al. 2006;
ing of hippocampal place cells (Hartley et al. 2000; Schultz 2002).
Burgess & Hartley 2002). In these simulations, place However, there has been a long-standing but un-
cells effectively perform a match to the distances to proven hypothesis that spatial learning does not obey
the nearest obstacle, integrating over distances in all this type of error-correcting learning rule, but rather is
90 Annals of the New York Academy of Sciences

incidental, occurring independently of performance, (consequences of any association from the unpaired
motivation, or prediction error (Tolman 1948). This cue to the absence of reward) will be additive to the
type of learning was subsequently attributed to the blocking effect.
hippocampus by O’Keefe and Nadel (1978). However, Doeller and Burgess (2007) found that, although
despite the many studies since, aimed at proving this learning of objects to either type of cue occurred at
hypothesis, the results have been mixed, with most similar rates and with similar levels of performance,
finding results consistent with reinforcement learning. there were different blocking effects for learning to the
A major prediction of reinforcement learning con- boundary and to the landmark. When tested with the
cerns the situation where, because the association from landmark, performance was much worse for the ob-
one cue already accurately predicts feedback, there ject paired with the boundary during prelearning. By
will be no prediction error and no possibility of sub- contrast, performance when tested with the boundary
sequent learning to a second cue (learning is said to was equal for objects paired with either cue during pre-
be “blocked”). Similarly, where learning occurs to two learning. See FIGURE 8 for the results of the boundary–
cues concurrently, the learning to one will be reduced landmark blocking experiment. Consistent results were
by the extent to which the other accurately predicts found when, again in the presence of the distal orien-
feedback (learning to it is said to be “overshadowed”). tation cues, two landmarks were used as cues (each
It is possible that the confounding of hippocampal landmark blocking the other), when two opposing sec-
and nonhippocampal contributions to spatial cogni- tions of the boundary were used (neither blocked the
tion may have contributed to the previous findings of other), or when overshadowing was investigated (the
blocking and overshadowing in spatial tasks (Hamilton boundary overshadowed the landmark, but not vice
& Sutherland 1999; Chamizo et al. 2003; Pearce et al. versa). Overall, the consistent finding was that learning
2006). The above study (Doeller et al. 2007) indicates to landmarks obeyed the predictions of reinforcement
a way to dissociate the specifically hippocampal con- learning and learning to boundaries did not. Given the
tribution to spatial learning. Under this view, learning striatal and hippocampal activation corresponding to
to an environmental boundary, dependent on the hip- learning relative to landmarks and boundaries respec-
pocampus, would be incidental and would not show tively (see above), it seems that, in this task, the striatum
blocking or overshadowing, while learning to the land- supports reinforcement learning relative to landmarks,
mark, dependent on the striatum, would conform to while the hippocampus supports incidental learning to
reinforcement learning. boundaries.
Doeller and Burgess (2007) examined blocking be-
tween boundaries and landmarks within their virtual
arena. Their blocking experiment consists of three Discussion: Implications beyond
phases. In a first “prelearning” phase, participants Spatial Cognition
learn object locations while landmark and boundary
are moved relative to each other at the beginning of As noted in the Introduction, spatial cognition en-
each block: four objects maintaining a fixed location joys an advantage over some other fields of higher
relative to the landmark and four other objects main- cognition in being able to share paradigms between
taining a fixed location relative to the boundary. In human and animal research. This link allows some
a second “compound learning” phase, both the land- inferences regarding the actual neural representations
mark and the boundary remain in fixed positions, pre- and processes involved in human cognition to be drawn
dicting the position of all eight objects. During the final from invasive studies in animals. Given this advantage,
“test” phase, memory performance is tested (without are there more general implications that can be drawn
feedback) in the presence of either the landmark or the for cognition beyond the spatial domain?
boundary alone.
If a given object is paired with cue 1 during pre- Memory and Imagery: Common Processes
learning, and the subject learns to accurately replace and Neural Bases?
it on the basis of this association, then there should be Initial attempts to form a computational model of
little learning of the association to cue 2 during the memory for spatial scenes, or for the spatial context
compound learning phase. Thus performance should of an event (Burgess et al. 2001b; Becker & Burgess
be poor when tested with cue 2 alone (compared to an 2001), force a consideration of the neural mechanisms
object paired with cue 2 during prelearning). This pro- involved. I briefly review these mechanisms and then
vides a powerful test of reinforcement learning, since discuss their more general implications for memory
any effects of “learned irrelevance” or “super learning” and imagery (see also Hartley et al. 2007).
Burgess: Spatial Cognition & the Brain 91

FIGURE 8. Learning to the boundary blocks learning to the landmark, but not vice versa. Participants
learned eight object locations using the paradigm shown in FIGURE 6. During “prelearning” (eight blocks
of two trials per object on average), the landmark and boundary moved relative to each other after each
block, with four objects moving with the landmark (orange “+”; object locations are orange spots) and
four with the boundary (green circle, object locations are green spots). During ”compound learning”
(one block of six trials per object), both types of cue (landmark and boundary) remained fixed, allowing
them to become associated to the locations of objects previously paired with the other cue. Both learning
phases included feedback at the end of each trial. Test phases (four objects tested with the landmark L;
four with the boundary B) did not include feedback and showed little learning to the landmark for objects
previously associated with the boundary (Test L) but unimpeded learning to the boundary of objects
previously associated with the landmark (Test B). Adapted from Doeller and Burgess (2007).

The model by Byrne and colleagues (Byrne et al. rounding neocortical areas. Specifically, the hippocam-
2007) makes use of the idea that the pattern of ac- pus restricts retrieved subsets to be mutually consistent
tivation of place cells is constrained by the recurrent with perception from a single location. These prod-
connections within area CA3 of the hippocampus to ucts of retrieval are then capable of being put into
be consistent with the subject being at a single loca- a head-centered representation for imagery in me-
tion. Other patterns of activation, involving place cells dial parietal areas via (re)constructive mechanisms in
which normally fire in different environmental loca- retrosplenial/parieto-occipital and posterior parietal
tions, can only be transient, that is, the place cells areas, including imposition of a viewing direction onto
form a “continuous attractor” representation of lo- the allocentric medial temporal representation by the
cation (Zhang 1996; Samsonovich & McNaughton head-direction cells in Papez’s circuit. As well as pro-
1997). The activation of place cells representing a sin- viding an outline for the functional roles of the various
gle location can then reactivate the parahippocampal regions identified in episodic memory (Burgess et al.
(BVC) representation of the distances and allocen- 2001a), the model explicitly highlights the close rela-
tric directions of environmental boundaries around tionship between the mechanisms and neural bases of
that location. The retrieval of this information into memory and imagery (Becker & Burgess 2001).
visual imagery/working memory requires translation Recent work has verified the predicted link be-
into an imaginable egocentric (head-centered) repre- tween memory and imagery. Similar effects of hip-
sentation, involving retrosplenial/parieto-occipital sul- pocampal lesions (known to affect episodic memory)
cus and posterior parietal cortex, as well as the provi- have been found in spatial working memory (Hartley
sion of current heading by Papez’s circuit. et al. 2007) and novel imagery (Hassabis et al. 2007).
This model has implications for the way episodic Similar patterns of activation have also been found in
memories or “events” are retrieved, or at least how the functional neuroimaging of memory for past events
spatial context in which they occur is retrieved—a pro- and for imagined “future” events (Addis et al. 2007).
cess thought to be specifically hippocampal dependent In addition, patients with hemispatial neglect in im-
(Burgess et al. 2002; O’Keefe & Nadel 1978). Thus, agery, but not those with perceptual or motor neglect
the hippocampus provides a strong constraint on the but unimpaired imagery, have also been found to have
subsets of information retrieval from the vast amount deficits in spatial navigation in tasks resembling the
of abstract (allocentric) knowledge stored in the sur- water maze (Guariglia et al. 2005). Given the close
92 Annals of the New York Academy of Sciences

relationship between retrieval and imagery, and be- not differ between landmark and boundary learning)
tween their neural bases, it is possible that one of the stressed by the declarative and episodic theories may
hallmarks of episodic memory—that of subjective reex- be later consequences of a trial-and-error based learn-
perience (Tulving 2001)—actually in large part reflects ing rule on the one hand, and a Hebbian encoding
the success of the generation of a vivid internal image. of coincidences on the other (see also Hirsch 1974;
And such imagery is of course not necessarily restricted O’Keefe & Nadel 1978).
to memory at all, although clearly implicated in much The neuroscience literature reviewed here demon-
of the “reconstructive” process on which memory relies strates that a variety of spatial representations are avail-
(Bartlett 1932; Addis et al. 2007). able for the purposes of controlling behavior. One
Much future work is obviously required to iden- of these, supported by the hippocampus, is special-
tify the functional interactions between memory and ized for processing location relative to environmental
imagery and between medial temporal and parietal boundaries and appears to operate a distinct, inciden-
areas, but hopefully the spatial model outlined above tal, learning rule. This view of the hippocampus has
can provide some sort of initial framework. curious echoes of the idea of a “geometric module”
(Cheng 1986; Gallistel 1990): a module for processing
Learning Rules, Procedural Versus the surface geometry of the surrounding environment
Declarative Memory, and the “Geometric in an “encapsulated” way (i.e., its output concerns only
Module” this subset of the information available to the animal,
The finding of different learning rules in the process- independent from other, e.g., featural, information).
ing associated with the hippocampal and striatal sys- The geometric module was proposed on purely be-
tems (Doeller & Burgess 2007; Doeller et al. 2007) sug- havioral grounds, independently of some of the above
gests a different way of looking at the major suggested ideas of hippocampal processing already present in
divisions of long-term memory. Thus, it has been sug- O’Keefe and Nadel’s (1978) book. The main data for
gested that the acquisition of consciously retrievable the geometric module were the preferential role played
long-term knowledge (whether semantic or episodic in by environmental boundaries in re orienting a disori-
nature) depends on the hippocampus (Scoville & Mil- ented animal (or young child; Hermer & Spelke 1994)
ner 1957); this is termed “declarative” memory (Squire within a small rectangular enclosure.
& Zola-Morgan 1991), in contrast to “procedural” The generality of the idea of a geometric module
memory (e.g., habits, motor learning), important as- is restricted by its dependence on a single “reorien-
pects of which depend on the striatum (Yin & Knowl- tation” paradigm. For example, animals and young
ton 2006). Equally, it has been suggested that rapid children do appear to be able to use featural cues to
one-shot encoding of events and their contexts (Tul- reorient in slightly different reorientation experiments,
ving 1983) specifically depends on the hippocampus such as when using larger (Learmonth et al. 2002;
(Kinsbourne & Wood 1975; O’Keefe & Nadel 1978; Sovrano & Vallortigara 2006), rhombic (Hupbach &
Mishkin et al. 1997; Rugg & Yonelinas 2003; Fortin Nadel 2005), or symmetrical (Nardini et al. 2007; Mc-
et al. 2002), as opposed to slowly acquired semantic Gregor et al. 2004) enclosures, or learning over re-
knowledge acquired slowly over multiple exposures. peated trials (Cheng 1986; Vallortigara et al. 1990;
These traditional dissociations might arise from dif- Gouteux et al. 2001). They also routinely combine
ferences in the basic neurobiology of the two systems— featural cues with geometric cues to navigate when
leading each system to implement a different learn- not disoriented (e.g., Maurer & Derivaz 2000), indi-
ing rule. Thus, synaptic plasticity in the striatum may cating that the geometric module is not strictly “en-
be controlled by levels of dopamine locally released capsulated.” See Cheng and Newcombe (2005) for a
(see Redgrave & Gurney 2006) as a function of pre- recent review. The conflicting patterns of results from
diction error (Waelti et al. 2001; Montague et al. disorientation paradigms regarding both the presence
2004; O’Doherty et al. 2004). By contrast, hippocam- or otherwise of geometric modules and the applica-
pal synaptic plasticity simply reflects co-occurrence (in bility or otherwise of reinforcement learning to spatial
Doeller et al.’s study, co-ocurrence between the repre- cognition may arise in part because so many types
sentation of the object and the place cell representa- of cue interact to determine orientation, and do so
tion of location as a conjunction of bearings from the within the head-direction system (e.g., Taube 1998),
boundary). rather than the hippocampus. These include local cues,
Thus, the aspects of conscious awareness (which did distal cues, and environmental geometry, and which
not differ greatly between landmark and boundary cues actually determine orientation depends on many
learning) and rapidity of acquisition (which also did factors, including each cue’s apparent stability (Jeffery
Burgess: Spatial Cognition & the Brain 93

et al. 1997; Jeffery & O’Keefe 1999), its distal or prox- References
imal location (Cressant et al. 1997), and the subject’s
history of disorientation (Knierim et al. 1995). By con- Abrahams, S., Pickering, A., Polkey, C. E., & Morris, R. G.
trast, Doeller et al. (2007) attempted to dissociate a (1997). Spatial memory deficits in patients with unilateral
cleanly hippocampal contribution to spatial cognition, damage to the right hippocampal formation. Neuropsycholo-
which may explain their more consistent results. gia, 35(1), 11–24.
Addis, D. R., Wong, A. T., & Schacter, D. L. (2007). Re-
Conclusions membering the past and imagining the future: Common
and distinct neural substrates during event construction
and elaboration. Neuropsychologia, 45(7), 1363–1377.
The field of spatial cognition has been evolving Aguirre, G. K., & D’Esposito, M. (1997). Environmental
rapidly over the last few years, driven by convergence knowledge is subserved by separable dorsal/ventral neural
of results from the application of similar experimental areas. J. Neurosci., 17(7), 2512–2518.
paradigms in both humans and other animals. Here I Andersen, R. A., Essick, G. K., & Siegel, R. M. (1987).
have reviewed a small subset of these recent advances, Neurons of area 7 activated by both visual stimuli and
focusing on the attempt to build a mechanistic un- oculomotor behavior. Exp. Brain Res., 67(2), 316–322.
Barry, C., Hayman, R., Burgess, N., & Jeffery, K. J. (2007).
derstanding of how spatial behavior results from the Experience-dependent rescaling of entorhinal grids. Nat.
actions of networks of neurons. In addition to the tra- Neurosci., 10(6), 682–684.
ditional strengths of cognitive science, of experimental Barry, C., Lever, C., Hayman, R., Hartley, T., Burton, S.,
dissociation of processes at the behavioral level, there et al. (2006). The boundary vector cell model of place cell
have been vital contributions from other disciplines op- firing and spatial memory. Rev. Neurosci., 17, 71–97.
erating at different levels of description. These include Bartlett, F. C. (1932). Remembering, a study in experimental and
social psychology. New York: Macmillan.
the ability to link behavior to neural systems, for exam-
Becker, S., & Burgess, N. (2001). A model of spatial recall,
ple, via neuropsychology or functional neuroimaging; mental imagery and neglect. Neural Information Processing
the continued success of systems neuroscience, includ- Systems, 13, 96–102.
ing the ability to examine the actual neural represen- Bisiach, E., & Luzzatti, C. (1978). Unilateral neglect of rep-
tations at play; and the ability to quantitatively inte- resentational space. Cortex, 14, 129–133.
grate results at different levels by using computational Bridgeman, B., Van Der Heijden, A. H. C., & Velichkovsky,
modeling. B. M. (1994). A theory of visual stability across saccadic
eye movements. Behav. Brain Sci., 17, 247–292.
The picture which emerges of spatial cognition is
Burgess, N. (2006). Spatial memory: How egocentric and
one of multiple parallel systems using representations allocentric combine. Trends Cogn. Sci, 10(12), 551–557.
with different frames of reference, processing different Burgess, N., Barry, C., & O’Keefe, J. (2007). An oscillatory
aspects of the spatial scene, using different learning interference model of grid cell firing. Hippocampus, 17, 801–
rules, and supported by a corresponding set of neural 812.
systems. A current challenge for the field is to begin to Burgess, N., Becker, S., King, J. A., & O’Keefe, J. (2001).
work out how these various neural systems, including Memory for events and their spatial context: Models and
experiments. Philos. Trans. R. Soc. Lond B Biol. Sci., 356,
medial temporal, parietal, and striatal regions, com- 1493–1503.
bine to produce the rich variety of spatial behavior Burgess, N., & Hartley, T. (2002). Orientational and geomet-
underpinning our daily life. This in turn should lead ric determinants of place and head-direction. In Neural in-
to insights into the organization of other aspects of formation processing systems (Vol. 14). Cambridge, MA: MIT
cognition and into the effects or causes of damage or Press.
dysfunction observable at the various levels of neuro- Burgess, N., Maguire, E. A., & O’Keefe, J. (2002). The hu-
man hippocampus and spatial and episodic memory. Neu-
physiology, systems neuroscience, and behavior.
ron, 35(4), 625–641.
Burgess, N., Maguire, E. A., Spiers, H. J., & O’Keefe, J.
Acknowledgments
(2001). A temporoparietal and prefrontal network for re-
trieving the spatial context of lifelike events. Neuroimage, 14,
I thank C. Bird and C.F. Doeller for commenting on
439–453.
this manuscript, and the UK Medical Research Coun- Burgess, N., Spiers, H. J., & Paleologou, E. (2004). Orienta-
cil, Biotechnology and Biological Sciences Research tional manoeuvres in the dark: Dissociating allocentric and
Council and the Wayfinding project of the European egocentric influences on spatial memory. Cognition, 94(2),
Union for their support. 149–166.
Byrne, P., Becker, S., & Burgess, N. (2007). Remembering
Conflict of Interest the past and imagining the future: A neural model of
spatial memory and imagery. Psychol. Rev., 114(2), 340–
The author declares no conflicts of interest. 375.
94 Annals of the New York Academy of Sciences

Chamizo, V. D., Aznar-Casanova, J. A., & Artigas, A. A. activates the hippocampus, precuneus, and insula. Neurore-
(2003). Human overshadowing in a virtual pool: Simple port, 8(3), 739–744.
guidance is a good competitor against locale learning. Gothard, K. M., Skaggs, W. E., & McNaughton, B. L. (1996).
Learn. Motiv., 34, 262–281. Dynamics of mismatch correction in the hippocampal en-
Cheng, K. (1986). A purely geometric module in the rat’s semble code for space: Interaction between path integra-
spatial representation. Cognition, 23(2), 149–178. tion and environmental cues. J. Neurosci., 16(24), 8027–
Cheng, K., & Newcombe, N. S. (2005). Is there a geomet- 8040.
ric module for spatial orientation? Squaring theory and Gouteux, S., Thinus-Blanc, C., & Vauclair, J. (2001). Rhe-
evidence. Psychon. Bull. Rev., 12(1), 1–23. sus monkeys use geometric and nongeometric information
Colby, C. L., & Goldberg, M. E. (1999). Space and attention during a reorientation task. J. Exp. Psychol. Gen., 130(3),
in parietal cortex. Annu. Rev. Neurosci., 22, 319–349. 505–519.
Cressant, A., Muller, R. U., & Poucet, B. (1997). Failure Guariglia, C., Piccardi, L., Iaria, G., Nico, D., & Pizzamiglio,
of centrally placed objects to control the firing fields L. (2005). Representational neglect and navigation in real
of hippocampal place cells. J. Neurosci., 17(7), 2531– space. Neuropsychologia, 43(8), 1138–1143.
2542. Hafting, T., Fyhn, M., Molden, S., Moser, M. B., & Moser, E.
Dayan, P., & Abbott, L. F. (2002). Computational neuroscience. I. (2005). Microstructure of a spatial map in the entorhinal
Cambridge, MA: MIT Press. cortex. Nature, 436, 801–806.
Deneve, S., Latham, P. E., & Pouget, A. (2001). Efficient Hamilton, D. A., & Sutherland, R. J. (1999). Blocking in
computation and cue integration with noisy population human place learning: Evidence form virtual navigation.
codes. Nat. Neurosci., 4(8), 826–831. Psychobiol., 27, 453–461.
Dickinson, A. (1980). Contemporary animal learning theory. Cam- Hartley, T., Bird, C. M., Chan, D., Cipolotti, L., Husain,
bridge, UK: Cambridge University Press. M., et al. (2007). The hippocampus is required for short-
Diwadkar, V. A., & McNamara, T. P. (1997). Viewpoint term topographical memory in humans. Hippocampus, 17,
dependence in scene recognition. Psychological Science, 8(4), 34–48.
302–307. Hartley, T., Burgess, N., Lever, C., Cacucci, F., & O’Keefe,
Doeller, C. F., & Burgess, N. (2007). Distinct error-correcting J. (2000). Modeling place fields in terms of the cortical
and incidental learning of location relative to landmarks inputs to the hippocampus. Hippocampus, 10(4), 369–379.
and boundaries. Submitted for publication. Hartley, T., Maguire, E. A., Spiers, H. J., & Burgess, N.
Doeller, C. F., King, J. A., & Burgess, N. (2007). Paral- (2003). The well-worn route and the path less traveled:
lel striatal and hippocampal systems for landmarks and Distinct neural bases of route following and wayfinding in
boundaries in spatial memory. Submitted for publication. humans. Neuron, 37(5), 877–888.
Ekstrom, A. D., Kahana, M. J., Caplan, J. B., Fields, T. A., Hartley, T., Trinkler, I., & Burgess, N. (2004). Geometric
Isham, E. A., et al. (2003). Cellular networks underlying determinants of human spatial memory. Cognition, 94, 39–
human spatial navigation. Nature, 425(6954), 184–188. 75.
Epstein, R., & Kanwisher, N. (1998). A cortical represen- Hassabis, D., Kumaran, D., Vann, S. D., & Maguire, E. A.
tation of the local visual environment. Nature, 392(6676), (2007). Patients with hippocampal amnesia cannot imag-
598–601. ine new experiences. Proc. Natl. Acad. Sci. U.S.A., 104(5),
Etienne, A. S., Maurer, R., & Seguinot, V. (1996). Path 1726–1731.
integration in mammals and its interaction with visual Hermer, L., & Spelke, E. S. (1994). A geometric process for
landmarks. J. Exp. Biol., 199(Pt 1), 201–209. spatial reorientation in young children. Nature, 370(6484),
Fenton, A. A., Csizmadia, G., & Muller, R. U. (2000). Con- 57–59.
joint control of hippocampal place cell firing by two visual Hirsch, R. (1974). The hippocampus and contextual re-
stimuli. I. The effects of moving the stimuli on firing field trieval of information from memory: A theory. Behav. Biol.,
positions. J. Gen. Physiol, 116(2), 191–209. 12(4), 421–444.
Fortin, N. J., Agster, K. L., & Eichenbaum, H. B. (2002). Hupbach, A., & Nadel, L. (2005). Reorientation in a rhombic
Critical role of the hippocampus in memory for sequences environment: No evidence for an encapsulated geometric
of events. Nat. Neurosci., 5(5), 458–462. module. Cognitive Development, 20, 279–302.
Fuhs, M. C., & Touretzky, D. S. (2006). A spin glass model Iaria, G., Petrides, M., Dagher, A., Pike, B., & Bohbot, V. D.
of path integration in rat medial entorhinal cortex. J. Neu- (2003). Cognitive strategies dependent on the hippocam-
rosci., 26(16), 4266–4276. pus and caudate nucleus in human navigation: Variability
Fyhn, M., Hafting, T., Treves, A., Moser, M. B., & Moser, E. and change with practice. J. Neurosci., 23(13), 5945–5952.
I. (2007). Hippocampal remapping and grid realignment Janzen, G., & van Turennout, M. (2004). Selective neural
in entorhinal cortex. Nature, 446(7132), 190–194. representation of objects relevant for navigation. Nat. Neu-
Galletti, C., Battaglini, P. P., & Fattori, P. (1995). Eye position rosci., 7(6), 673–677.
influence on the parieto-occipital area PO (V6) of the Jeffery, K. J., Donnett, J. G., Burgess, N., & O’Keefe, J. M.
macaque monkey. Eur. J. Neurosci, 7(12), 2486–2501. (1997). Directional control of hippocampal place fields.
Gallistel, R. (1990). The organization of learning. Cambridge, Exp. Brain Res., 117(1), 131–142.
MA: MIT Press. Jeffery, K. J., & O’Keefe, J. M. (1999). Learned interaction
Ghaem, O., Mellet, E., Crivello, F., Tzourio, N., Mazoyer, B., of visual and idiothetic cues in the control of place field
et al. (1997). Mental navigation along memorized routes orientation. Exp. Brain Res., 127(2), 151–161.
Burgess: Spatial Cognition & the Brain 95

King, J. A., Burgess, N., Hartley, T., Vargha-Khadem, F., & McNaughton, B. L., Battaglia, F. P., Jensen, O., Moser, E.
O’Keefe, J. (2002). Human hippocampus and viewpoint I., & Moser, M. B. (2006). Path integration and the neural
dependence in spatial memory. Hippocampus, 12(6), 811– basis of the “cognitive map.” Nat. Rev. Neurosci., 7(8), 663–
820. 678.
King, J. A., Trinkler, I., Hartley, T., Vargha-Khadem, F., & Melcher, D. (2007). Predictive remapping of visual features
Burgess, N. (2004). The hippocampal role in spatial mem- precedes saccadic eye movements. Nat. Neurosci., 10(7),
ory and the familiarity-recollection distinction: A single 903–907.
case study. Neuropsychology, 18, 405–417. Milner, A. D., Dijkerman, H. C., & Carey, D. P. (1999). Vi-
Kinsbourne, M., & Wood, F. (1975). Short-term memory suospatial processing in a case of visual form agnosia. In
and the amnesic syndrome. In D. Deutsch & J. A. Deutsch N. Burgess, K. J. Jeffery, & J. O’Keefe (Eds.), The hippocam-
(Eds.), Short term memory (pp. 257–291). New York: Aca- pal and parietal foundations of spatial cognition (pp. 443–466).
demic. Oxford, UK: Oxford University Press.
Knierim, J. J., Kudrimoti, H. S., & McNaughton, B. L. Mishkin, M., Suzuki, W. A., Gadian, D. G., & Vargha-
(1995). Place cells, head direction cells, and the learning Khadem, F. (1997). Hierarchical organization of cognitive
of landmark stability. J. Neurosci., 15(3 Pt 1), 1648–1659. memory. Philos. Trans. R. Soc. Lond B Biol. Sci., 352(1360),
Kubie, J. L., Fenton, A., Novikov, N., Touretzky, D., & 1461–1467.
Muller, R. U. (2007). Changes in goal selection induced by Montague, P. R., Hyman, S. E., & Cohen, J. D. (2004).
cue conflicts are in register with predictions from changes Computational roles for dopamine in behavioural control.
in place cell field locations. Behav. Neurosci., 121(4), 751– Nature, 431(7010), 760–767.
763. Morris, R. G. M., Garrud, P., Rawlins, J. N., & O’Keefe, J.
Learmonth, A. E., Nadel, L., & Newcombe, N. S. (2002). (1982). Place navigation impaired in rats with hippocam-
Children’s use of landmarks: Implications for modularity pal lesions. Nature, 297(5868), 681–683.
theory. Psychol. Sci., 13(4), 337–341. Mou, W., & McNamara, T. P. (2002). Intrinsic frames of
Lenck-Santini, P. P., Rivard, B., Muller, R. U., & Poucet, B. reference in spatial memory. J. Exp. Psychol. Learn. Mem.
(2005). Study of CA1 place cell activity and exploratory Cogn., 28(1), 162–170.
behavior following spatial and nonspatial changes in the Mou, W., McNamara, T. P., Valiquette, C. M., & Rump,
environment. Hippocampus, 15(3), 356–369. B. (2004). Allocentric and egocentric updating of spatial
Levy, I., Hasson, U., Avidan, G., Hendler, T., & Malach, R. memories. J. Exp. Psychol. Learn Mem. Cogn., 30(1), 142–
(2001). Center-periphery organization of human object 157.
areas. Nat. Neurosci., 4, 533–539. Muller, R. U. (1996). A quarter of a century of place cells.
Mackintosh, N. J. (1975). A theory of attention: Variations Neuron, 17(5), 813–822.
in the associability of stimuli with reinforcement. Psychol. Muller, R. U., & Kubie, J. L. (1989). The firing of hippocam-
Rev., 82, 276–298. pal place cells predicts the future position of freely moving
Maguire, E. A., Burgess, N., Donnett, J. G., Frackowiak, R. rats. J. Neurosci., 9(12), 4101–4110.
S., Frith, C. D., et al. (1998a). Knowing where and getting Nardini, M., Atkinson, J., & Burgess, N. (2007). Chil-
there: A human navigation network. Science, 280(5365), dren reorient using the left/right sense of coloured land-
921–924. marks at 18–24 months. Cognition. Epub ahead of print.
Maguire, E. A., Burgess, N., Donnett, J. G., O’Keefe, J., & DOI:10.1016/j.cognition.2007.02.007.
Frith, C. D. (1998b). Knowing where things are: Parahip- Nardini, M., Burgess, N., Breckenridge, K., & Atkinson, J.
pocampal involvement in encoding object locations in vir- (2006). Differential developmental trajectories for egocen-
tual large-scale space. J. Cogn. Neurosci., 10(1), 61–76. tric, environmental and intrinsic frames of reference in
Maurer, R., & Derivaz, V. (2000). Rats in a transparent spatial memory. Cognition, 101(1), 153–172.
morris water maze use elemental and configural geometry O’Doherty, J., Dayan, P., Schultz, J., Deichmann, R.,
of landmarks as well as distance to the pool wall. Spatial Friston, K., et al. (2004). Dissociable roles of ventral
Cognition and Computation, 2, 135–156. and dorsal striatum in instrumental conditioning. Science,
May, M. (2004). Imaginal perspective switches in remem- 304(5669), 452–454.
bered environments: Transformation versus interference O’Keefe, J. (1976). Place units in the hippocampus of the
accounts. Cognit. Psychol., 48(2), 163–206. freely moving rat. Exp. Neurol., 51(1), 78–109.
McDonald, R. J., & White, N. M. (1994). Parallel infor- O’Keefe, J., & Burgess, N. (1996). Geometric determi-
mation processing in the water maze: Evidence for inde- nants of the place fields of hippocampal neurons. Nature,
pendent memory systems involving dorsal striatum and 381(6581), 425–428.
hippocampus. Behav. Neural Biol., 61(3), 260–270. O’Keefe, J., & Burgess, N. (2005). Dual phase and rate cod-
McGregor, A., Hayward, A. J., Pearce, J. M., & Good, M. ing in hippocampal place cells: Theoretical significance
A. (2004). Hippocampal lesions disrupt navigation based and relationship to entorhinal grid cells. Hippocampus, 15,
on the shape of the environment. Behav. Neurosci., 118(5), 853–866.
1011–1021. O’Keefe, J., & Conway, D. H. (1978). Hippocampal place
McNamara, T. P., Rump, B., & Werner, S. (2003). Ego- units in the freely moving rat: Why they fire where they
centric and geocentric frames of reference in mem- fire. Exp. Brain Res., 31(4), 573–590.
ory of large-scale space. Psychon. Bull. Rev., 10(3), 589– O’Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive
595. map. Oxford, UK: Oxford University Press.
96 Annals of the New York Academy of Sciences

O’Keefe, J., & Speakman, A. (1987). Single unit activity in Sargolini, F., Fyhn, M., Hafting, T., McNaughton, B. L.,
the rat hippocampus during a spatial memory task. Exp. Witter, M. P., et al. (2006). Conjunctive representation
Brain Res., 68(1), 1–27. of position, direction, and velocity in entorhinal cortex.
Ono, T., Nakamura, K., Fukuda, M., & Tamura, R. (1991). Science, 312(5774), 758–762.
Place recognition responses of neurons in monkey hip- Schmidt, T., & Lee, E. Y. (2006). Spatial memory organized
pocampus. Neurosci. Lett., 121(1–2), 194–198. by environmental geometry. Spatial Cognition & Computation,
Packard, M. G., & McGaugh, J. L. (1992). Double dissoci- 6(4), 347–369.
ation of fornix and caudate nucleus lesions on acquisition Schultz, W. (2002). Getting formal with dopamine and re-
of two water maze tasks: Further evidence for multiple ward. Neuron, 36(2), 241–263.
memory systems. Behav. Neurosci., 106(3), 439–446. Scoville, W. B., & Milner, B. (1957). Loss of recent mem-
Packard, M. G., & McGaugh, J. L. (1996). Inactivation of ory after bilateral hippocampal lesions. J. Neurol. Neurosurg.
hippocampus or caudate nucleus with lidocaine differen- Psychiatry, 20, 11–21.
tially affects expression of place and response learning. Shelton, A. L., & McNamara, T. P. (1997). Multiple views of
Neurobiol. Learn. Mem., 65(1), 65–72. spatial memory. Psychon. Bull. Rev., 4(1), 102–106.
Pearce, J. M., Graham, M., Good, M. A., Jones, P. M., & Shrager, Y., Bayley, P. J., Bontempi, B., Hopkins, R. O.,
McGregor, A. (2006). Potentiation, overshadowing, and & Squire, L. R. (2007). Spatial memory and the human
blocking of spatial learning based on the shape of the hippocampus. Proc. Natl. Acad. Sci. U.S.A., 104(8), 2961–
environment. J. Exp. Psychol. Anim. Behav. Process, 32(3), 201– 2966.
214. Simons, D. J., & Wang, R. F. (1998). Perceiving real-world
Pearce, J. M., Roberts, A. D., & Good, M. (1998). Hippocam- viewpoint changes. Psychological Science, 9, 315–320.
pal lesions disrupt navigation based on cognitive maps but Snyder, L. H., Grieve, K. L., Brotchie, P., & Andersen, R.
not heading vectors. Nature, 396(6706), 75–77. A. (1998). Separate body- and world-referenced represen-
Pessiglione, M., Seymour, B., Flandin, G., Dolan, R. J., & tations of visual space in parietal cortex. Nature, 394(6696),
Frith, C. D. (2006). Dopamine-dependent prediction er- 887–891.
rors underpin reward-seeking behaviour in humans. Na- Sovrano, V. A., & Vallortigara, G. (2006). Dissecting the ge-
ture, 442(7106), 1042–1045. ometric module: A sense linkage for metric and landmark
Piaget, J., & Inhelder, B. (1956). The child’s conception of space. information in animals’ spatial reorientation. Psychological
London: Routledge. Science, 17(7), 616–621.
Pico, R. M., Gerbrandt, L. K., Pondel, M., & Ivy, G. (1985). Spiers, H. J., Burgess, N., Hartley, T., Vargha-Khadem, F.,
During stepwise cue deletion, rat place behaviors correlate & O’Keefe, J. (2001a). Bilateral hippocampal pathology
with place unit responses. Brain Res., 330(2), 369–372. impairs topographical and episodic memory but not visual
Poldrack, R. A., Clark, J., Pare-Blagoev, E. J., Shohamy, D., pattern matching. Hippocampus, 11, 715–725.
Creso, M. J., et al. (2001). Interactive memory systems in Spiers, H. J., Burgess, N., Maguire, E. A., Baxendale, S.
the human brain. Nature, 414(6863), 546–550. A., Hartley, T., et al. (2001b). Unilateral temporal lobec-
Pouget, A., & Sejnowski, T. J. (1997). A new view of hemine- tomy patients show lateralised topographical and episodic
glect based on the response properties of parietal neurones. memory deficits in a virtual town. Brain, 124, 2476–2489.
Philos. Trans. R. Soc. Lond B Biol. Sci., 352(1360), 1449–1459. Spiers, H. J., & Maguire, E. A. (2007). A navigational guid-
Redgrave, P., & Gurney, K. (2006). The short-latency ance system in the human brain. Hippocampus, 17(8), 618–
dopamine signal: A role in discovering novel actions? Nat. 626.
Rev. Neurosci., 7(12), 967–975. Squire, L. R., & Zola-Morgan, S. (1991). The medial
Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlo- temporal lobe memory system. Science, 253(5026), 1380–
vian conditioning: Variations in the effectiveness of rein- 1386.
forcement and non-reinforcement. In A. H. Black & W. Sutton, R. S., & Barto, A. G. (1988). Reinforcement learning: An
F. Prokasy (Eds.), Classical conditioning II. Current research and introduction. Cambridge, MA: MIT Press.
theory (pp. 64–99). New York: Appleton-Century-Crofts. Taube, J. S. (1998). Head direction cells and the neuropsy-
Rieser, J. J. (1989). Access to knowledge of spatial structure chological basis for a sense of direction. Prog. Neurobiol., 55,
at novel points of observation. J. Exp. Psychol.-Learn. Mem. 225–256.
Cogn., 15(6), 1157–1165. Tolman, E. C. (1948). Cognitive maps in rats and men.
Ross, J., Morrone, M. C., Goldberg, M. E., & Burr, D. Psychol. Rev., 55, 189–208.
C. (2001). Changes in visual perception at the time of Tulving, E. (1983). Elements of episodic memory. Oxford, UK:
saccades. Trends Neurosci, 24(2), 113–121. Clarendon Press.
Rugg, M. D., & Yonelinas, A. P. (2003). Human recogni- Tulving, E. (2001). Episodic memory and common sense:
tion memory: A cognitive neuroscience perspective. Trends How far apart? Philos. Trans. R. Soc. Lond. B Biol. Sci.,
Cogn. Sci., 7(7), 313–319. 356(1413), 1505–1515.
Salinas, E., & Abbott, L. F. (1995). Transfer of coded infor- Vallortigara, G., Zanforlin, M., & Pasti, G. (1990). Geometric
mation from sensory to motor networks. J. Neurosci., 15(10), modules in animals’ spatial representations: A test with
6461–6474. chicks (Gallus gallus domesticus). J. Comp. Psychol., 104(3),
Samsonovich, A., & McNaughton, B. L. (1997). Path inte- 248–254.
gration and cognitive mapping in a continuous attractor Voermans, N. C., Petersson, K. M., Daudey, L., Weber, B.,
neural network model. J. Neurosci., 17(15), 5900–5920. Van Spaendonck, K. P., et al. (2004). Interaction between
Burgess: Spatial Cognition & the Brain 97

the human hippocampus and the caudate nucleus during direction” impairs performance on spatial memory tasks
route recognition. Neuron, 43(3), 427–435. in rats. Behav. Neurosci., 115(4), 861–869.
von Helmholtz, H. (1866). Handbuch der physiologischen optik. Wolbers, T., & Buchel, C. (2005). Dissociable retrosplenial
Hamburg, Germany: Leopold Voss. and hippocampal contributions to successful formation of
Waelti, P., Dickinson, A., & Schultz, W. (2001). Dopamine survey representations. J. Neurosci., 25(13), 3333–3340.
responses comply with basic assumptions of formal learn- Wolbers, T., Wiener, J. M., Mallot, H. A., & Buchel, C.
ing theory. Nature, 412(6842), 43–48. (2007). Differential recruitment of the hippocampus, me-
Waller, D., & Hodgson, E. (2006). Transient and endur- dial prefrontal cortex, and the human motion complex
ing spatial representations under disorientation and self- during path integration in humans. J. Neurosci., 27(35),
rotation. J. Exp. Psychol. Learn Mem. Cogn., 32(4), 867– 9408–9416.
882. Wraga, M., Creem, S. H., & Proffitt, D. R. (2000). Updating
Wang, R. F., & Simons, D. J. (1999). Active and passive scene displays after imagined object and viewer rotations. J. Exp.
recognition across views. Cognition, 70(2), 191–210. Psychol. Learn. Mem. Cogn., 26(1), 151–168.
Wang, R. F., & Spelke, E. (2000). Updating egocentric rep- Yin, H. H., & Knowlton, B. J. (2006). The role of the basal
resentations in human navigation. Cognition, 77, 215–250. ganglia in habit formation. Nat. Rev. Neurosci., 7(6), 464–
Wang, R. F., & Spelke, E. S. (2002). Human spatial rep- 476.
resentation: Insights from animals. Trends Cogn. Sci., 6(9), Zhang, K. (1996). Representation of spatial orientation by
376–382. the intrinsic dynamics of the head-direction cell ensemble:
White, N. M., & McDonald, R. J. (2002). Multiple parallel A theory. J. Neurosci., 16(6), 2112–2126.
memory systems in the brain of the rat. Neurobiol. Learn Zipser, D., & Andersen, R. A. (1988). A back-propagation
Mem., 77(2), 125–184. programmed network that simulates response properties
Wilton, L. A., Baird, A. L., Muir, J. L., Honey, R. C., & of a subset of posterior parietal neurons. Nature, 331(6158),
Aggleton, J. P. (2001). Loss of the thalamic nuclei for “head 679–684.