Unpacking the Cognitive Map: The Parallel Map Theory of Hippocampal Function

Abstract

In the parallel map theory, the hippocampus encodes space with 2 mapping systems. The bearing map is constructed primarily in the dentate gyrus from directional cues such as stimulus gradients. The sketch map is constructed within the hippocampus proper from positional cues. The integrated map emerges when data from the bearing and sketch maps are combined. Because the component maps work in parallel, the impairment of one can reveal residual learning by the other. Such parallel function may explain paradoxes of spatial learning, such as learning after partial hippocampal lesions, taxonomic and sex differences in spatial learning, and the function of hippocampal neurogenesis. By integrating evidence from physiology to phylogeny, the parallel map theory offers a unified explanation for hippocampal function.

Full text

Unpacking the Cognitive Map: The Parallel Map Theory
of Hippocampal Function
Lucia F. Jacobs
University of California, Berkeley
Franc¸oise Schenk
University of Lausanne
In the parallel map theory, the hippocampus encodes space with 2 mapping systems. The bearing map
is constructed primarily in the dentate gyrus from directional cues such as stimulus gradients. The sketch
map is constructed within the hippocampus proper from positional cues. The integrated map emerges
when data from the bearing and sketch maps are combined. Because the component maps work in
parallel, the impairment of one can reveal residual learning by the other. Such parallel function may
explain paradoxes of spatial learning, such as learning after partial hippocampal lesions, taxonomic and
sex differences in spatial learning, and the function of hippocampal neurogenesis. By integrating
evidence from physiology to phylogeny, the parallel map theory offers a unified explanation for
hippocampal function.
The cognitive map theory articulated by John O’Keefe and Lynn
Nadel in 1978 not only was the first unified theory of hippocampal
function but also has been the most influential (Best & White,
1999). This theory postulated that the hippocampus creates a
mental representation of allocentric space. This representation, the
cognitive map, is more flexible than other mental representations
of space and allows the navigator to create novel routes between
familiar sites.
O’Keefe and Nadel’s (1978) cognitive map theory was sup-
ported by a diversity of empirical results, such as the impairment
of spatial navigation by hippocampal lesions (Jarrard, 1983; Mor-
ris, Hagan, & Rawlins, 1986). It was also supported by two
remarkable discoveries. The first was the elucidation of the hip-
pocampal place unit by O’Keefe (e.g., O’Keefe & Dostrovsky,
1971). O’Keefe found that activity of these hippocampal pyrami-
dal cells was localized to specific locations in a test environment
and that they retained their specificity even in the absence of visual
input (O’Keefe & Conway, 1980). This provided concrete evi-
dence for the role of the hippocampus in coding locations in space.
A second discovery was the demonstration of a hippocampal
mechanism for rapid, long-lasting, synapse-specific associative
learning. This is the process of long-term potentiation (LTP; Bliss
& Lomo, 1973). LTP mediation by the N-methyl-
D-aspartate
(NMDA) receptor provided a physiological theory of the Hebbian
synapse (Bliss & Collingridge, 1993). Evidence that spatial learn-
ing and hippocampal LTP are impaired by NMDA-receptor antag-
onists provided new support for the cognitive map theory (Morris
et al., 1986).
Twenty years later, however, the cognitive map theory remains
controversial. Although it laid the foundation for current theories
of how the hippocampus encodes space (McNaughton et al., 1996;
O’Keefe & Burgess, 1996; Redish, 1999), the present theories
differ from one another, and no one model of spatial encoding has
received universal acceptance. There are also critics of the cogni-
tive map theory itself (Eichenbaum, Dudchenko, Wood, Shapiro,
& Tanila, 1999). The disagreement about hippocampal function is
fueled in part by different approaches to the question (e.g., non-
human vs. human studies) and in part by paradoxical results.
Results from the rodent literature include the recovery of place
learning after hippocampal lesions (Whishaw, Cassel, & Jarrard,
1995), the tuning of hippocampal cells to nonspatial information
(Wood, Dudchenko, & Eichenbaum, 1997), the ameliorative ef-
fects of maze pretraining on place learning in the presence of
blocked synaptic plasticity (Bannerman, Good, Butcher, Ramsay,
& Morris, 1995; Saucier & Cain, 1995), and other dissociations of
synaptic plasticity and spatial learning (Abeliovich et al., 1993;
Huang et al., 1995; Nosten-Bertrand et al., 1996; Richter-Levin,
Thomas, Hunt, & Bliss, 1998).
Research guided by questions of hippocampal function in hu-
mans has focused on the hippocampus’s role in universal cognitive
processes, such as intermediate-term memory (Rawlins, 1985),
declarative memory (Squire, 1992), contextual processing (Gluck
& Myers, 2001), episodic memory (Burgess, Maguire, & O’Keefe,
2002; O’Keefe & Nadel, 1978; Squire & Zola-Morgan, 1991;
Vargha-Khadem et al., 1997), or relational processing (Cohen &
Eichenbaum, 1993; Eichenbaum et al., 1999). For example,
Eichenbaum et al. proposed that the hippocampus’s role in spatial
Lucia F. Jacobs, Department of Psychology, University of California,
Berkeley; Franc¸oise Schenk, Institute of Physiology, University of Lau-
sanne, Lausanne, Switzerland.
This research was supported by a Prytanean Faculty Award and by
grants from the University of California, from the J.D. French Alzheimer’s
Foundation to Lucia F. Jacobs, and from the Swiss National Science
Foundation to Franc¸oise Schenk. For comments, suggestions, encourage-
ment, and argument, we thank Kim Beeman, Andrea Chiba, Jack Cowan,
Brian Derrick, Michael Dickinson, Jack Gallant, Michael Bastiani, Leslie
Kay, Bruce Miller, Richard Morris, Lynn Nadel, Jacques Paillard, Jon
Seger, Matthew Shapiro, Miriam Smolover, David Steinsaltz, Georg
Striedter, and Matthew Wilson.
Correspondence concerning this article should be addressed to Lucia F.
Jacobs, Department of Psychology, University of California, Tolman Hall,
Berkeley, California 94720-1650. E-mail: squirrel@socrates.berkeley.edu
Psychological Review Copyright 2003 by the American Psychological Association, Inc.
2003, Vol. 110, No. 2, 285–315 0033-295X/03/$12.00 DOI: 10.1037/0033-295X.110.2.285
285

learning is an application case of its function for a more general
learning process, the learning of relations. Here, spatial represen-
tations do not constitute a map of space but instead contribute to
the general principle of “linking events within episodes” (Eichen-
baum et al., 1999, p. 216). As a consequence, this “memory space”
(Eichenbaum et al., 1999, p. 218) codes spatial and nonspatial
relations among events, processing spatial relations for navigation
and serial relations for solving more abstract, nonspatial stimulus
relations, such as those found in transitive inference (Dusek &
Eichenbaum, 1997).
Spatial models of the hippocampus, in contrast, often take a
bottom-up, computational approach to the question of how the
rodent hippocampus maps space. These models have included new
generation models from O’Keefe and Burgess (1996), the multiple
chart theory of McNaughton et al. (1996; Samsonovich & Mc-
Naughton, 1997), and the multiframe theory of Redish and
Touretsky (1997). For example, the Redish and Touretsky model
on spatial navigation combines four different navigation systems
(taxon, praxic, locale, and route) and five spatial representations
(local view, head direction, path integrator, place code, and goal
memory). The anatomical realization of the model is provided by
the connections among the several brain structures assumed to
mediate these representations (Redish, 1999). We cannot review
all of the models in detail, and the interested reader is referred to
the above citations for more information.
At present, no one theory, whether spatial or nonspatial, quali-
tative or quantitative, modeling human or nonhuman behavior, has
gained universal acceptance. Instead, the different approaches ap-
pear to be developing in parallel (Best & White, 1999; Cohen &
Eichenbaum, 1991; McNaughton, 1996; Nadel, O’Keefe, Shapiro,
McNaughton, & Disterhoft, 1998). It is clear that the rodent
hippocampus plays some critical role in spatial orientation and that
this role is not fully understood. If neurobiology cannot be under-
stood except in light of behavior (Shepherd, 1994) and, as
Dobzhansky (1951) argued, biology cannot be understood except
in light of evolution, then to understand the hippocampus, we must
understand both its evolution and its role in spatial navigation. We
begin with a reexamination of this behavior.
The Nature of Spatial Navigation
We start by defining spatial navigation and discussing what is
commonly accepted as spatial information. We do this because
there is much confusion over terms and it is important for us to
specify exactly what we mean by a word such as landmark.We
then outline a new theory of navigation based on the idea that the
cognitive map is composed of parallel component maps. We then
address the origins of the dual nature of navigation in vertebrates
and the development of a more powerful representation of space,
the cognitive map, in birds and mammals. Because we propose that
different classes of maps are constructed by the mammalian hip-
pocampus and that these maps rely on different cues, we are also
required to introduce new terminology to define experimental
conditions precisely in terms of such cues.
Navigation can be defined as purposive movement through
absolute space. For absolute space, we ascribe to O’Keefe and
Nadel’s (1978) original definition:
Absolute space embodies the notion of a framework or container
within which material objects can be located but which is conceived
as existing independently of particular objects or objects in general.
Objects are located relative to the places of the framework and only
indirectly, via this framework, to other objects. Movement of a body
(including the observer) changes its position within the framework but
does not alter the framework or the relationship of other objects to the
framework. In contrast, relative space designates a set of relationships
amongst objects or sensory inputs which in themselves are inherently
nonspatial. Objects are located relative to other objects and relative
space does not exist independent of the existence of objects. (p. 7)
Further, O’Keefe and Nadel distinguished between the mental
representations of psychological space and its complement, phys-
ical space, defined as “any space attributed to the external world
independent of the existence of minds” (pp. 6–7).
Thus, there is physical space, and there are mental maps of that
space. A map can be defined in diverse ways; we ascribe to
Neisser’s (1976) broad definition: “not pictures in the head, but
plans for obtaining information from potential environments” (p.
131). Hence, the defining characteristic of the class of mental
representations known as a map is that it provides a navigator,
moving in space, with an expectation of the sensory input given a
certain movement in a certain direction at a certain speed or around
a certain obstacle. With a cognitive map (i.e., a mental map of
absolute space; O’Keefe & Nadel, 1978, p. 2), the navigator can
place itself within the framework and navigate directly among
objects.
This ability is also known as place learning. This differs from
orientation in relative space, such as egocentric or orientation to a
simple cue. In an experimental setting, an animal is said to dem-
onstrate place learning when it automatically creates a new route to
the goal, and this new route is constructed from new egocentric
relations to objects. Thus, a rat that can spontaneously chart a
direct course in the water maze, regardless of its release position in
the pool, is said to demonstrate place learning (Morris, 1984).
Sources of Spatial Information
As an animal explores, it uses internal and external cues to relate
its current position to its start point in the environment. Internal
cues such as self-generated movement cues inform the navigator
how far and in which direction it has moved from a given position.
External cues such as landmarks can be used in two different ways,
both for direction and for position (Collett, Cartwright, & Smith,
1986; Leonard & McNaughton, 1990), as illustrated in Figure 1c
and 1e.
Self-Generated Movement Cues
Locomotion generates a dynamic sensory flow in diverse mo-
dalities (proprioceptive, tactile, auditory, olfactory, and visual).
The navigator integrates some or all of this information to update
the current position relative to the start point. Path integration is
the outcome of the process that regularly updates a directional
vector. The vector is generated by the animal’s movement during
an exploratory bout and is based on this dynamic sensory flow and
the efferent copy of the intended action. The path integration
vector encodes the distance and direction from the start point of
exploration, where the vector is apparently reset. Thus, path inte-
gration allows the navigator to beeline to its most recent start
position at any time. It appears to be an ancient component of
286
JACOBS AND SCHENK

spatial orientation, as it is found throughout invertebrate and
vertebrate taxa (Leonard & McNaughton, 1990; Maaswinkel &
Whishaw, 1999; Wehner, Michel, & Antonsen, 1996).
We point out one caveat: Although we can define path integra-
tion, it is much less clear when and how it is combined with other
spatial information. We suggest there is an important distinction
between pure path integration and the reliance of the hippocampus
on path integration. Perhaps path integration itself (a single, one-
dimensional gradient produced from vestibular and external sen-
sory feedback) is simply a vector that is exported to the hippocam-
pus, which then assigns meaning to this vector. In this case, path
integration would not be a property of the hippocampus but a
process whose output is used by the hippocampus in constructing
one-dimensional (1-D) and two-dimensional (2-D) maps. The vec-
tor obtained from path integration could be a primitive working
memory representation, one that is reset at the start of every
exploratory bout. It might then acquire more dimensions when it is
associated with external points, such as an identifiable start posi-
tion. This association of the working memory vector with external
landmarks would lead to a richer representation of space, one that
Figure 1. Determining direction and position with spatial cues. a: Gradient—A field of graded intensity. b:
Compass mark—A distal landmark that can provide only directional information; it is too distant to provide
positional information. c: Array geometry—Direction deduced from the polarization of an array of positional
landmarks. d: Two-dimensional (2-D) gradient map—A 2-D map constructed from directional cues. e: 2-D
topographic map—A 2-D map constructed from positional cues.
287
PARALLEL MAP THEORY

cannot be computed without the path integration process. The
properties of such a representation would exceed those provided
by pure path integration.
Directional Cues: Gradients and Directional Landmarks
In general, directional cues polarize the navigator’s environment
rather than identifying a specific position in space (see Figure 1a–
1c). An example of a visual directional cue is a compass mark,
such as a mountain range: It is a landmark that is too distant to
provide accurate positional information but can nonetheless pro-
vide an accurate direction (Leonard & McNaughton, 1990). Gra-
dients of distributed cues (e.g., odor, sound, polarized light, mag-
netic fields) emanating from a source are also directional cues. A
directional cue may be static or dynamic. When the navigator is
not moving, a directional cue provides static directional informa-
tion when its size and aspect do not change. It is also static when
the cue is too distant for the size or aspect to change with the
navigator’s movement.
Directional cues can be provided by the geometrical shape of the
contiguous visual space or panorama (e.g., experimental room,
cage, or forest clearing) or by the arrangement of salient objects
(see Benhamou & Poucet, 1998). Asymmetrically shaped spaces
provide directional information. In contrast, symmetrical or regu-
lar shapes (e.g., square, rectangle, equilateral triangle) can be
spatially ambiguous (Benhamou & Poucet, 1998; Cheng, 1986).
These details of the cue environment have important implications
for the encoding of space in our theory of parallel maps.
Positional Landmarks
In contrast to directional landmarks, a positional landmark is a
local object that can be used to deduce position from the relative
distances and positions of objects within an array (see Figure 1e).
Here, movement of the navigator causes the cue’s appearance to
change quickly, allowing the navigator to deduce the distance
between landmarks (or between a landmark and the navigator). In
contrast, a directional cue generally does not change with small
movements of the navigator. For example, a pigeon walking on the
ground might perceive a tree at a 100-m distance as a directional
landmark but treat it as a positional landmark if the pigeon is flying
100 m above the same tree.
Positional landmarks in an array can be processed separately, as
unique objects. In an array composed of multiple positional cues,
however, each cue has a spatial relation to at least one other cue.
This relationship forms the basis of relational coding. When dif-
ferent objects form a symmetrical geometrical figure, the figure is
identifiable even if the identity of each component object is not
learned (Benhamou & Poucet, 1998). This creates ambiguity
among symmetrical positions in a configuration, even when each
corner is uniquely identified by local cues (Cheng, 1986; Cheng &
Gallistel, 1984).
In contrast, rodents pay attention and learn the identity of a
single directional cue if it is part of a symmetrically shaped array
and if an additional directional cue is placed on the periphery of
the arena (for studies of rats, see Poucet, 1989; for studies of
hamsters, see Poucet, Chapuis, Durup, & Thinus-Blanc, 1986).
Thus, it appears that objects can either provide directional infor-
mation as a part of a geometrical figure or be recognized as unique
objects, but not both. Hence, the information that is extracted from
a landmark depends critically on the observer and the context and
implies that direction and position are processed individually,
according to context. This interpretation has important conse-
quences for the analysis of place units, as we discuss later.
Evolution of Spatial Orientation Mechanisms
“Just as the human body represents a whole museum of organs,
each with a long evolutionary history, so we should expect to find
that the mind is organized in a similar way. It can no more be a
product without history than is the body in which it exists” (C.
Jung, 1964, p. 67). We begin with a discussion of the evolution of
spatial navigation.
The environment in which the hippocampus evolved was not
that experienced by terrestrial mammals. The vertebrate forebrain
evolved underwater, with limited capacity for visual or auditory
processing (Northcutt, 1996). The sensory modalities that were
well developed depended instead on distributed fields and gradi-
ents of stimuli, such as chemical, light, or magnetic gradients
(Northcutt, 1996). Such gradients are used for orientation by all
mobile animals, even unicellular species (Dusenbery, 1992;
Scho¨ne, 1984). In a sensory world composed of such gradients,
location is defined not by discrete landmarks but by changes in the
amplitude or sign of a sensory input (i.e., the input becomes larger,
brighter, louder, etc; see Figure 1a–1b). Of these, characteristics of
the earth’s geomagnetic field (polarity, inclination, intensity) are
perhaps most universally exploited, and orientation to magnetic
fields is found in insects (Collett & Baron, 1994; Frier, Edwards,
Smith, Neale, & Collett, 1996), fish (Walker et al., 1997), amphib-
ians (Phillips & Borland, 1994), reptiles (Light, Salmon, & Loh-
mann, 1993; Lohmann & Johnsen, 2000; Lohmann & Lohmann,
1996a), birds (Wiltschko & Wiltschko, 1996), and rodents (Kimchi
& Terkel, 2001; Olcese, 1990). Magnetic fields have also been
shown to influence spatial learning ability in rodents (Kavaliers,
Eckel, & Ossenkopp, 1993; Kavaliers et al., 1996; Levine & Bluni,
1994).
Using Gradients
Orienting to simple gradient maps, such as to a magnetic field,
has both advantages and disadvantages. The animal must move up
and down the gradient to construct its crude representation of
space with repeated sampling and by knowing its rate of move-
ment. The animal must precisely calibrate changes in the single
perceptual dimension (i.e., the polarity of sensory input, whether
increasing or decreasing) to its own rate of movement (Scho¨ne,
1984). Once an animal can do this, it can predict the sensory input
that it will experience at a future location. Thus, as it travels up or
down this gradient, it creates a 1-D map, as the term was used by
Neisser (1976; see Figure 1a). With careful and discrete sampling
of the gradient (e.g., sniffs) and precise knowledge of its own
movements, an animal can extrapolate a vector through unexplored
territory and continue to navigate accurately on the basis of this
simple 1-D map. An animal can also use this gradient map to
calculate distance from its knowledge of time spent traveling.
The abstract nature of the gradient map, which allows animals to
navigate accurately through unknown terrain, also has its disad-
vantages. If the gradient is unevenly distributed or the animal loses
288
JACOBS AND SCHENK

track of its rate of sampling or its rate of movement, the map is no
longer reliable, and there is little opportunity for self-correction.
The value and reliability of a simple 1-D map can be enhanced,
however, when it is combined with other 1-D maps, the intersec-
tion of which yields a Cartesian coordinate (i.e., 2-D) map (see
Figure 1d). Once this 2-D map has been created, it can be used to
predict the length and orientation of a vector through untraveled
space.
Evidence for Gradient Maps
The hypothesis that vertebrates create 2-D maps from distrib-
uted stimuli has generated much discussion but few rigorous tests
(Wallraff, 1996). There is indirect support for this hypothesis,
however, from studies on orientation to plumes of olfactory cues
by homing pigeons (Papi, 1992) and magnetic field orientation in
sea turtles (Lohmann & Lohmann, 1996b).
Green sea turtles (Chelonia mydas) migrate as adults to their
place of hatching, Ascension Island in the southern Atlantic
Ocean, a site thousands of kilometers from the location to which
they initially dispersed. They navigate through the southern At-
lantic Ocean using traditional paths, even on their first return to the
island after many years (Luschi, Hays, Del Seppia, Marsh, & Papi,
1998). Therefore, on the basis of little experience and in the
absence of obvious landmarks, the turtles solve an oceanic water
maze, returning to a tiny island in the middle of the southern
Atlantic. A population genetic analysis of Ascension Island turtles
has shown that their genotype is unique to this island. Because no
turtle of a different genotype has ever been found on the island,
this suggests that the population has been genetically isolated,
perhaps by the unique navigational algorithm they use to return to
the island (Lohmann, Hester, & Lohmann, 1999). Such genetic
programming of long-distance orientation has also been demon-
strated in migrating birds (Helbig, Berthold, & Wiltschko, 1989;
Wiltschko & Wiltschko, 1996).
Lohmann et al. (1999) and others (e.g., Akesson, 1996) have
suggested that the mystery of this precise orientation lies in the
turtles’ ability to decode the geomagnetic map in this locale. The
angle of inclination of the earth’s magnetic field and the geomag-
netic field strength are close to orthogonal in the south Atlantic.
Ascension Island’s location can therefore be specified with some
precision by the intersection of these magnetic gradients (Loh-
mann et al., 1999). In the laboratory, loggerhead sea turtles can
deduce direction both from magnetic inclination and from field
strength (Lohmann & Lohmann, 1996a). Hence, it is feasible that
wild turtles can read location from the bicoordinate grid formed by
this intersection (Lohmann et al., 1999). As in other cases of
extraordinary migration, such abilities have probably evolved
slowly, as animals adapt their movements to slowly shifting pat-
terns of resource distribution (Alerstam, 1990).
Despite the widespread use of orientation to distributed stimuli
in animals (Dusenbery, 1992), this class of stimuli has been absent
from previous models of spatial navigation in mammals and from
models of hippocampal function. To anticipate our later argument,
we note that the hippocampus’s ability to create vectors from
distributed stimuli is at the heart of its ability to encode the
cognitive map. Our proposition that the cognitive map is based on
such directional, 1-D maps forms the basis of our theory. Only by
encoding gradients can the hippocampus create the mental repre-
sentation of a novel short cut between locales that have not been
previously connected in the animal’s spatial experience. We begin
by discussing hippocampal function among vertebrates.
The Evolution of the Hippocampus
The hippocampal formation (HPF) is the mammalian homo-
logue of the medial pallium, one of three regions (dorsal, medial,
and lateral) of the vertebrate telencephalon or pallium (Northcutt,
1995). The medial pallium is found in all jawed vertebrates and,
hence, is a remarkably conserved structure (Bruce & Neary, 1995).
The homology of structures derived from the medial pallium in
birds, mammals, and reptiles has been established by converging
lines of evidence, including patterns of embryology, connectivity,
histochemical boundaries, and homeotic gene expression (Fernan-
dez, Pieau, Reperant, Boncinelli, & Wassef, 1998; Medina &
Reiner, 2000). Although a medial pallium homologue can also be
defined in fish and amphibians (Butler & Hodos, 1996), there are
no studies of its functional or behavioral significance in these
groups. Thus, at present, we can discuss medial pallium function
only in those taxa for which studies of spatial learning exist:
reptiles, birds, and mammals. Our goal here is to present a brief
synthesis of the vertebrate literature, tempering our conclusions
with the knowledge that much research remains to be done.
Cognitive traits leave few fossils. The accepted method of
elucidating the function of an ancestor is to compare function and
structure across extant taxa. If contemporary groups share a com-
mon trait, one can parsimoniously conclude that the similarity
arises from common descent. Hence, the shared trait may be
ancestral and may have been present in the common ancestor
(Harvey & Pagel, 1991). The common ancestor of birds and
mammals, for example, existed over 200 million years ago (Ro-
mer, 1977). It is possible that the similarities seen in extant species
are thus homologies of structure and function.
Despite other differences in telencephalon structure, in both
birds and mammals the relative size of the hippocampus is pre-
dicted by the spatial behavior of the species under natural condi-
tions. For example, in both songbirds and rodents, the fitness of
individuals of certain classes (e.g., females vs. males, scatter
hoarding species vs. larder hoarding species) may depend more
heavily on spatial memory or spatial exploration. These individu-
als have larger hippocampi, relative to the remaining telencepha-
lon, than do individuals not subject to these selection pressures
(Jacobs, Gaulin, Sherry, & Hoffman, 1990; Jacobs & Spencer,
1994; Sherry, Forbes, Khurgel, & Ivy, 1993; Sherry, Jacobs, &
Gaulin, 1992; Sherry, Vaccarino, Buckenham, & Herz, 1989).
Lesions of the hippocampus in birds and mammals also produce
similar deficits in locating a place in an array of distal cues
(Bingman, 1990; Bingman, Bagnoli, Ioale´, & Casini, 1989; Morris
et al., 1986; Sherry & Vaccarino, 1989; Strasser & Bingman,
1997). Thus, the function and physiology of the medial pallium
homologue appear to be similar in birds and mammals.
At first glance, the function of the medial pallium homologue,
the medial cortex, in reptiles appears to parallel that seen in birds
and mammals. The relative volume of the medial cortex, for
example, is larger in a lizard species that forages actively for prey
compared with a species that waits for prey to arrive (Day, Crews,
& Wilczynski, 1999). As we discuss below, a more recent behav-
289
PARALLEL MAP THEORY

ioral study by Day, Crews, and Wilczynski (2001) paints a slightly
different picture of the role of the medial cortex.
Homology of Structure
The evidence for homologies at the level of hippocampal sub-
fields—that is, the dentate gyrus (DG) or the hippocampus proper
(HP)—is somewhat speculative at this point. Comparative studies
of medial cortex structure in reptiles suggest that the small-celled
area of the ventral medial cortex is homologous to the DG, the
ventral-most region of the mammalian hippocampus (Hoogland,
Martinez-Garcia, Geneser, & Vermeulen-Vanderzee, 1998). Like-
wise, the reptilian dorsomedial cortex may be homologous to the
entorhinal cortex (EC) and the subiculum (Hoogland &
Vermeulen-VanderZee, 1990; Martinez-Garcia & Olucha, 1990).
Subfield homologies are less clear in birds. Szekely (1999) has
concluded, on the basis of a comparison of intra-and extrahip-
pocampal projections between birds and mammals, that the HP
homologue in birds is ventral to the DG homologue. Because this
dorsal–ventral orientation is the opposite of that seen in mammals
and reptiles, we suspect the final definition of subfield homologies
in vertebrates awaits further research, preferably informed by
patterns of embryology (Striedter, 1997) and gene expression
(Fernandez et al., 1998; Medina & Reiner, 2000).
Because the phylogenetic history of birds, reptiles, and mam-
mals separated hundreds of million of years ago, one could spec-
ulate that the spatial function of the medial pallium predated the
separation of these lineages and, hence, can be back dated at least
this far. This does not mean that all hippocampal function is
homologous in birds and mammals. Because cognitive mapping
ability requires association structures (e.g., mammalian neocortex)
and these areas are not homologous in birds and mammals, the
ability to construct a cognitive map must be the result of conver-
gent evolution in these groups. What may be homologous among
birds, mammals, and reptiles is, instead, the role of the medial
pallium in allocentric orientation to distributed stimuli. The simi-
larities among these taxa thus would have arisen from this ances-
tral trait, even if the expression of this trait has diverged widely
among vertebrates with the further evolution and specialization of
the forebrain. It is possible that all the similarities are the result of
convergent evolution, although this is not the most parsimonious
explanation in light of the evidence for homology of structure. This
question can only be resolved with further comparative studies of
medial pallium structure and function.
Homology of Function: Spatial Learning in Reptiles
Despite patterns of medial pallium allometry (i.e., size relative
to telencephalon) that are similar to those seen in birds and
mammals, new studies by Day et al. (1999) suggest that reptiles
may not use visual cues for spatial orientation in the same way as
do birds and mammals. Reptiles, of course, represent a diverse
group that includes the order of turtles and the suborders of snakes
and lizards (Romer, 1977). There have been few studies of spatial
orientation in reptiles. Three studies published recently represent
almost the entire body of work on the use of cues during spatial
navigation in reptiles (Day et al., 2001; Holtzman, Harris, Aran-
guren, & Bostock, 1999; Lopez et al., 2000). It is unfortunate that
these new studies were each conducted by a different laboratory,
using a different task and studying a different reptile group (snake,
turtle, and lizard). On the other hand, the studies do have an
important feature in common, as each task measured reference
memory for a single location.
In the first study, by Holtzman et al. (1999), corn snakes (Elaphe
guttata guttata) searched for an escape hole among a ring of
nonescape holes in a reptilian version of the Barnes (1979) maze,
a hippocampal task in rodents. The circular arena was surrounded
by a high wall, and the only landmark was a single, conspicuous
cue card; the pathway of the snake from a central release point was
used as evidence that it had learned the correct location for escape.
The snakes did orient more quickly to the escape hole over
repeated trials. Although this was interpreted as true place learning
(Holtzman, 1998; Holtzman et al., 1999), in the published search
paths, all initial orientations of the snakes were first directed
toward the single cue card. Thus, the snakes’ performance could
also be ascribed to an egocentric encoding of a cue location rather
than to true allocentric place learning.
Day et al.’s (2001) study of two species of lacertid lizard
(Acanthodactylus boskianus, Acanthodactylus scutellatus) also
failed to find evidence for orientation to a landmark array. In this
study, the goal was a single heated rock in a ring of seven unheated
rocks. Although lizards decreased their latency to the goal, their
movements were not affected by any landmark rotation or re-
moval, as assayed by probe tests with eight cold rocks. Medial
cortex lesions did not alter the lizards’ nonresponse to landmarks,
as might be expected. Thus, lizards did not orient in the same way
as birds and rodents given similar experimental conditions (Su-
zuki, Augerinos, & Black, 1980; Vander Wall, 1982).
In contrast to the snake and lizard studies, allocentric place
learning was recently demonstrated in turtles (Pseudemys scripta).
In this study, the goal was a food bait in one arm of a water-filled
plus maze (Lopez et al., 2000). Experimental groups learned to
associate either a cue or a place with the bait. Probe trials were
conducted in which part or all of the maze was curtained, blocking
distal cues. When distal landmarks (e.g., lab furniture, distinctively
colored room walls) were blocked by a curtain, only turtles in the
place group showed a decline in performance.
The negative findings from the lizard study contrast sharply
with the positive findings from the turtle study, which the authors
rightly described as the first demonstration of unequivocal place
learning in reptiles. Rather than concluding that this ability is
special to turtles, however, we suggest an alternative explanation
that also explains the results from the other groups. We propose
that in all studies, reptiles oriented primarily to directional, not
positional, cues, whether visual or nonvisual. In the snake study,
the evidence is their initial orientation to the single cue card. In the
lizard study, the authors concluded that the lizards were learning to
orient to a goal but that the experiment did not control the cues the
lizards were using (Day et al., 2001). We suggest that the lizards
could have been orienting to directional cues, such as magnetic
fields or gradients of auditory or light cues, as these types of
stimuli are used by reptiles.
Results from the turtles orienting on a plus maze are also
consistent with this hypothesis. There are two ways to solve a plus
maze: choose the correct arm, as defined by an array of positional
landmarks (see Figure 1e), or choose the correct direction, as
defined by directional landmarks (see Figure 1a–1c). Thus, accu-
rate performance on a plus maze could be the result of orientation
290
JACOBS AND SCHENK

to a directional cue or to an array of positional landmarks. In this
study, the walls of the experimental room differed in color, and
these distinctively colored walls could have functioned as coarse
directional landmarks. Turtles could thus have oriented using this
directional information and accurately chosen the correct arm, with
no knowledge of place (a correct choice was scored as entry into
the arm). As we discuss later, this interpretation may also explain
the residual spatial learning seen after genetic lesions of the
hippocampus in mice when orienting in a water-filled plus maze
(Silva, Paylor, Wehner, & Tonegawa, 1992).
Role of the Septum
In all jawed vertebrates, the medial pallium receives significant
input from the septum (Butler & Hodos, 1996; Swanson & Risold,
2000). Because of extensive homologies between the septum in
mammals and in reptiles (Font, Lanuza, Martinez-Marcos,
Hoogland, & Martinez-Garcia, 1998), theories of the behavioral
function of the septohippocampal connections in mammals may be
relevant to spatial navigation in reptiles. In particular, Numan’s
(2000) theory of septal function dissociates the contributions of the
septohippocampal system from that of the corticohippocampal
system in spatial navigation. He concluded from physiological and
behavioral evidence that “it is possible that the hippocampus and
its connections with surrounding cortical areas encode the relations
between external stimuli, and that the hippocampus and its rela-
tions with the septum encode and maintain the self-motion cues”
(Numan, 2000, p. 316) We suggest that the septohippocampal
system, which is highly developed in reptiles (Hoogland &
Vermeulen-Vanderzee, 1990), plays a critical role in their spatial
navigation. Like a mammal with lesions in the corticohippocampal
system, intact reptiles may rely heavily on egocentric orientation to
cues, orienting primarily to directional cues such as gradients and
visual beacons but not encoding the relationship among external
stimuli.
In conclusion, we suggest that studies of spatial orientation in
reptiles are consistent with the hypothesis that reptiles orient to
directional cues but not to landmark arrays. Striedter (1997) has
shown that the medial cortex in adult reptiles is structurally similar
to the hippocampus of mammals or birds at an earlier embryolog-
ical stage. This is interesting in light of the striking similarity in
behavior between juvenile rodents and adult reptiles. Both show a
dependence on directional cues (Schenk, Grobety, Lavenex, &
Lipp, 1995), and this may reflect a rough equivalence of hip-
pocampal development. More research is needed to test this hy-
pothesis of the role of the medial cortex in directional cue-based
orientation in reptiles. If confirmed, studies of medial cortex
function would support our hypothesis for the ancestral gradient-
encoding function of the medial pallium.
The Parallel Map Theory
The term cognitive map was first introduced by E. C. Tolman
(1948) to describe what he saw as mental representations in the
rodent, and it is a term he coined for his bold reply to the tenets of
strict behaviorism. The cognitive map is currently understood as
the mental representation that conveys the ability to compute
shortcuts through untraveled terrain (Gallistel, 1990; O’Keefe &
Nadel, 1978). To date, cognitive maps have been found only in
groups that have evolved significant associational structures (i.e.,
birds and mammals) but not in reptiles (as reviewed above) or
insects (Dyer, 1994). Within mammals, hippocampal organization
increases in complexity in groups with greater development of the
neocortex (Schwerdtfeger, 1984; West, 1990; West & Schwerdt-
feger, 1985).
The concept of the cognitive map is usually assumed to be a
unitary mental representation. We propose instead that the cogni-
tive map is constructed from two parallel maps that, when inte-
grated, allow the navigator to calculate cognitive map shortcuts.
These parallel maps differ in how they represent space, what cues
are used to represent space, and what hippocampal structures are
involved in creating the representation (see Table 1). We call this
new formulation of the cognitive map the parallel map theory
(PMT) of spatial navigation.
The Maps
The first parallel map is the bearing map. It is constructed from
the integration of self-movement cues and directional cues. These
cues are used to form a mental representation of a 1-D vector. The
bearing map is created from the intersection of vectors, which
forms a 2-D coordinate system (see Figure 2a). Because the
bearing map can be based entirely on gradients, the navigator can
use a simple movement algorithm to accurately navigate over long
Table 1
The Parallel Components of the Integrated Map
Parallel component
Hippocampal structure
(must be intact)
Environmental stimuli
(must be present)
Bearing map Subcortical channel Directional cues
Dentate gyrus Compass mark, or
CA3 Gradient of distributed cues, or
Medial septum Asymmetric room or arena, or
Fimbria fornix Polarized landmark array
Sketch map Cortical channel Positional cues
CA1 Array of perceptually unique local landmarks
NMDA receptor in CA1
Integrated map Both channels intact Both directional and positional cues available
Subiculum
Note. CA ⫽ subfields of the hippocampus proper; NMDA ⫽ N-methyl-D-aspartate.
291
PARALLEL MAP THEORY

Figure 2. A schematic of the parallel map theory. a: Bearing map—This map is constructed from directional
cues such as compass marks and gradients. Here, the bearing map is formed from the transection of two gradient
maps: a chemical gradient based on odorant concentration and a visual gradient based on a distant compass mark.
Arrows indicate the axis of the gradient map. The solid triangle represents a visual beacon, the smaller shaded
cloud in the lower left represents a source of odorant, and the larger shaded cloud represents allocentric space
that is mapped by the bearing map. b: Sketch map—This topographic map is constructed from the relative
position of fixed positional landmarks. Solid and patterned shapes represent unique positional landmarks. c:
Integrated map—This map is constructed from the integrated bearing and sketch map. By linking all sketch maps
onto the single bearing map, the rat can compute novel routes among them. The solid triangle represents a
directional landmark such as a visual beacon, the smaller shaded cloud in the lower left represents a directional
cue such as an odor source, solid and patterned shapes represent positional landmarks, and the larger shaded
cloud represents the boundaries of the integrated map.
292
JACOBS AND SCHENK

distances. The navigator must only calibrate self-motion cues with
changes in the intensity of a distributed cue to extrapolate its future
position in the coordinate system. Thus, the bearing map allows
the navigator to maintain an accurate representation of its position,
even in unknown territory. Starting with a primitive gradient
algorithm, the bearing map creates a coarse-grained mental repre-
sentation of space that nonetheless provides a powerful tool for
spatial navigation, particularly long-distance navigation.
The second parallel map is the sketch map. It is constructed from
an arrangement of positional cues. These unique local landmarks
are encoded relative to each other as a topographic map (see
Figure 2b). The sketch map differs in fundamental ways from the
bearing map. First, all landmarks in the map must be individually
learned, and thus, there can no extrapolation or generalization
across novel terrain, as in the bearing map. The relations (distance,
direction) among landmarks must also be learned. The sketch map
is thus a fine-grained mental representation that is best suited for
local navigation. The positional codes within sketch maps are
allocentric, as each cue refers to another component of the sketch
map.
PMT can be described at several levels of analysis. What we
have just described is the first level, the conceptual account of two
maps. In this account, one map is based on directional cues, some
of which may be extrapolated, and the other is based on a set of
memorized positional cues. The concept of the bearing map has no
real predecessors in theoretical models of spatial navigation in
mammals, including previous formulations of the cognitive map.
In contrast, the second, topographic sketch map has much in
common with previous cognitive map theories, as we describe
below. PMT thus describes two new concepts, the bearing map and
the idea that the cognitive or integrated map is composed of
parallel maps (see Figure 2c). We use the term integrated map
instead of cognitive map because the original term was never
precisely defined by Tolman in 1948, which has led to different
usages in different disciplines (see review by Bennett, 1996).
Anatomy of the Parallel Maps
A second level of analysis is the physical scaffold underlying
the three mental representations (bearing map, sketch map, and
integrated map). Despite recent progress (Amaral & Witter, 1995;
Deadwyler & Hampson, 1999; Witter, Wouterlood, Naber, & van
Haeften, 2000), the full complexity of hippocampal anatomy has
yet to be mapped onto a complete theory of its function. What
Swanson, Ko¨hler, and Bjo¨rklund said in 1987 remains true:
It appears safe to say that nowhere is the gap between structure and
function greater than in this region [hippocampus]. The major reason
for this is that while the hippocampal formation contains the simplest
cortical fields from an anatomical point of view, it receives, processes,
and transmits the most complex array of information of any cortical
region. (p. 126)
Keeping this in mind, we present a working model of the
hippocampal structures underlying the encoding and use of the
parallel map system (see Figure 3). We propose that the maps are
mediated by different neural structures: The bearing map is medi-
ated by subcortical hippocampal channels projecting to the DG and
the CA3 subfield of the HP, the sketch map is mediated by the
CA1 subfield and its cortical connections, and the integrated map
is mediated by the synchronization of activity between these two
channels. The anatomical basis of the parallel maps is shown
schematically in Figure 3.
The assignment of map to structure is an important feature of
PMT, as is its emphasis on separate functions of the DG and the
HP, which create parallel and independent maps. Although the
hippocampal subfield functions have received some attention, hav-
ing been modeled (e.g., Deadwyler & Hampson, 1999; Granger,
Wiebe, Taketani, & Lynch, 1996) and empirically disassociated
(Gilbert, Kesner, & Lee, 2001; McNaughton, Barnes, Meltzer, &
Sutherland, 1989; Mizumori, McNaughton, Barnes, & Fox, 1989),
these studies have proposed complementary functions rather than
parallel ones. We are taking this line of reasoning in a new
direction, first by proposing that DG and CA1 mediate parallel
maps, and second by adding the concept of the gradient map. PMT,
based on parallel maps and representations of gradients, thus leads
to a set of unique predictions (see Tables 2 and 3).
In summary, the integrated map relies on the presence of two
parallel map processes. It emerges from the simultaneous activa-
tion of the parallel bearing and sketch maps and, hence, from the
synchronization of physiological activity between the DG and the
HP. When both maps are active and accessible, the navigator can
create the integrated map by double labeling positional cues within
and between sketch maps, as outlined in Table 4. The emergent
properties of the integrated map supply an ability that neither map
alone can accomplish: navigation among familiar locations that
involves movement across unfamiliar terrain. With this, the navi-
gator can extrapolate its movement beyond the knowledge of
memorized landmarks. The strength of this system lies in its
redundancy, a design feature found in spatial navigation through-
out the animal kingdom (Keeton, 1974). Because the maps work in
parallel, if one is impaired then the navigator may continue to
orient accurately, using the residual ability afforded by the remain-
ing map. It cannot, of course, use the integrated map, but it may
retain a considerable ability to navigate, as is apparent from the
literature on partial lesions of the hippocampus. Because the inte-
grated map is an emergent property of the coactivation of the
bearing and sketch maps, all three maps can be isolated with a
variety of techniques (e.g., cue manipulation, lesion, unit activity),
allowing each element of the theory to be tested (see Tables 2
and 3).
The final level of analysis for PMT lies beyond experimental
tests of its predictions, however, and that challenge is to map
concept to structure with computational models. PMT narrows the
search for such a model by mapping specific functions to different
subfields, and now feasible and realistic computational models of
the parallel maps need to be developed from knowledge of the
anatomy and topography of the HPF (Amaral & Witter, 1995;
Witter et al., 2000). We are currently working with collaborators to
develop a model of a hippocampal parallel mapping process.
We now discuss each map in turn, describing its characteristics
and its physiological properties and reviewing the evidence for the
effect of its experimental dissociation on spatial navigation in the
laboratory rodent.
Predictions From a Parallel Map System
Our theory proposes that spatial abilities depend not only on the
integrity of two independent brain mapping systems but also on the
293
PARALLEL MAP THEORY

type of information that is available in a given environment. This
leads to several corollaries. If a class of orienting cues is missing,
then the map that requires that class will be impaired. Alterna-
tively, if one mapping system is more developed than the other,
then the navigator’s orientation will be biased toward using that
system. If the anatomical structures underlying one map are le-
sioned or otherwise impaired, the corresponding map will be
impaired. Impairment of one map should reveal residual spatial
learning by the other map. Finally, cognitive mapping abilities
(e.g., one-trial spatial learning in working memory tasks, new
detour strategies in reference memory tasks) should be impaired
following damage to any of the structural channels underlying
these mapping systems. Hence, the resulting behavioral phenotype
is the consequence of two factors: data coming in from the envi-
ronment, and the capacity of the hippocampus to process different
types of data (e.g., directional vs. positional cues).
The corollaries of PMT lead to testable predictions, according to
the certain variables (see Tables 2 and 3). These are (a) cues
available in the experimental room or task environment, defined
below; (b) the state of development of different hippocampal
structures; (c) the state of impairment, due to age, lesion, or
pharmacological treatment; and (d) potential for synaptic plasticity
(e.g., NMDA receptor availability).
Task Environment
We define task environment as the environmental stimuli that
are available to the navigator at a given time to solve a particular
task (see Table 1). This concept is critical in PMT because the
bearing and the sketch maps use different types of cues. Thus,
maps (and, hence, behavioral phenotypes) are controlled and dis-
sociable by task environment. For example, a curtain surrounding
an arena or maze eliminates visual directional landmarks and
should selectively impair the bearing map. If positional landmarks
are removed or rearranged, however, the sketch map is lost,
altered, or impaired, and now the rat appears amnesiac for position.
If one structural component of each map is impaired, the integrated
Figure 3. Major intrinsic connections of the hippocampal formation, adapted from Amaral and Witter (1995).
The direction of connections leads to the stepwise assembly of the integrated map. Fiber tracts are shown in
italics. A unidirectional connection is indicated with a single-headed arrow, and a reciprocal connection is
indicated with a double-headed arrow. Circular fields represent the hippocampal structures. Rectangular fields
represent extrahippocampal structures. Dark shading indicates a structure involved in the bearing map, no
shading represents structures involved in the sketch map, and light shading represents a structure involved in the
integrated map. MS ⫽ medial septum; LS ⫽ lateral septum; DG ⫽ dentate gyrus; EC ⫽ entorhinal cortex; CA1
and CA3 ⫽ fields of the hippocampus proper; PaS ⫽ parasubiculum; PrS ⫽ presubiculum; SUB ⫽ subiculum.
From “Hippocampal Formation,” by D. G. Amaral and M. P. Witter, in The Rat Nervous System (p. 449), 1995,
San Diego, CA: Academic Press. Copyright 1995 by Academic Press. Adapted with permission.
294
JACOBS AND SCHENK

map becomes impaired, which eliminates cognitive mapping abil-
ity (see Tables 2 and 3).
Map Phenotype
By knowing the task environment (i.e., the input to the mapping
system) and the locus of impairment (i.e., the residual mapping
system), we can predict the behavioral phenotype of the navigator,
or map phenotype. We define this as an animal’s spatial strategy in
a given experimental condition. A particular strategy may be intact
or absent; if it is the remaining intact strategy, it will underlie any
residual spatial learning observed (see Tables 2 and 3).
As discussed earlier, if directional cues are removed, even in an
intact animal, the bearing map is impaired. The map phenotype for
this navigator is then a residual sketch map (rS; see Table 2). As
illustrated with the example of the Morris water maze task in
Figure 4, the rS rodent should concentrate its search in the correct
quadrant of the water maze but should not be able to organize
straight trajectories toward the platform from any starting position.
In contrast, removing positional landmarks impairs the sketch map
and hence should result in the expression of the residual bearing
map phenotype (rB; see Table 3). The rB rodent should search for
the platform by moving along parallel transects but should express
a reduced accuracy in locating the platform location (see Figure 4).
Sex Differences in Spatial Strategy: A Natural
Dissociation of Bearing and Sketch Maps
Females and males of several polygamous species, such as
humans (Kimura, 1999; Moffat, Hampson, & Hatzipantelis, 1998;
Sandstrom, Kaufman, & Huettel, 1998; Sherry & Hampson, 1997),
laboratory rats (Williams, Barnett, & Meck, 1990), and kangaroo
rats (Langley, 1994), differ in spatial navigation strategy. They
appear to use different sets of visual cues to orient. In laboratory
rats, male performance on the radial arm maze is severely dis-
rupted if the maze is curtained, even if positional landmarks are
still visible. Female performance is only slightly impaired under
this condition. In contrast, if the positional landmarks are removed
or randomized, female performance is severely impaired, whereas
this manipulation has less effect on male performance (Williams et
al., 1990; Williams & Meck, 1991, 1993). As Williams and Meck
(1993) have discussed, sex differences in spatial behavior are small
in comparison with other sexually differentiated behaviors, such as
reproductive behavior or learned bird song. The sex differences
may also disappear if the task is well learned, and they are difficult
to detect if the task is too simple (Williams & Meck, 1991).
Experimental evidence for the sexual differentiation of the hip-
pocampus and spatial strategy (Isgor & Sengelaub, 1998) contin-
ues to accumulate, however, lending strong support to the hypoth-
esis that sex differences in spatial behavior are the result of
hormonal mechanisms similar to those seen in behavior classically
defined as reproductive.
Recouched in terms of PMT, females and males rely more
heavily on different mapping systems, with males extracting their
primary orientation information from the bearing map and females
extracting this information from the sketch maps. In a typical
laboratory test room, the bearing map is constructed from room
geometry, distal cues, and stimulus gradients. The sketch map is
based on unique positional landmarks. Therefore, even with the
Table 2
Predictions of the Parallel Map Model When the Bearing Map Is Impaired (rS Phenotype)
Cause of reduced representation
Experimental conditions during water maze
testing Behavioral phenotype
Impaired DG, CA3, medial septum,
or fimbria fornix
Open room: Local and directional cues present. Unable to organize search from new release point. Swims
in a disorganized pattern until it enters the correct
quadrant, then searches locally, producing daisy loop
pattern.
Open room: Platform is visible or cued. Rapid learning of direct path to platform; no impairment.
Curtained maze: Local cues are available
within curtain; no directional cues.
Intermediate performance, which is impaired if local cues
are rearranged. With additional training, performance
again reaches an intermediate level.
Note. rS ⫽ residual sketch; DG ⫽ dentate gyrus; CA ⫽ subfields of the hippocampus proper.
Table 3
Predictions of the Parallel Map Model When the Sketch Map Is Impaired (rB Phenotype)
Cause of reduced representation
Experimental conditions during water
maze testing Behavioral phenotype
Impaired CA1 or blockade of NMDA receptors Open room: Local and directional cues
available.
Swimming is organized into long transects
across pool, with little or no local searching.
Latency decreases slowly with extended
(more than 2 days) training. Latency
increases if directional cues are manipulated.
Open room: Platform is visible or cued. Rapid learning of direct path to platform.
Curtained maze: Nonvisual directional
cues available.
Intermediate performance, which is impaired if
directional cues are manipulated.
Note. rB ⫽ residual bearing; CA ⫽ subfields of the hippocampus proper; NMDA ⫽ N-methyl-D-aspartate.
295
PARALLEL MAP THEORY

same task environment, females and males may differ in how they
map the space: Females learn and remember the relations among
unique positional landmarks, whereas males perceive positional
landmarks as points in the shape of an array. Perhaps for this
reason, reducing the ambient light after acquisition produces an
immediate and severe impairment in female but not male orienta-
tion on the radial arm maze (Williams & Meck, 1993). Thus, even
in the same environment, there are two ways to encode the space,
with females and males showing a natural dissociation of these two
strategies and of the parallel mapping systems.
This sex difference has certain implications for studies of spatial
navigation in polygamous rats and mice. If a study uses all male
subjects, because males rely more heavily on distal cues and room
geometry, this could lead to a bias in defining spatial navigation in
that paradigm as highly dependent on such directional cues. This
might be particularly evident in studies in which cues are in
conflict. In such studies, a male’s use of cue hierarchy should show
greater reliance on directional cues. For example, a recent study of
cue use in male rats found a significant preference for directional
over positional landmarks in orientation. The rats were also more
likely to extract directional than positional information from the
configuration of positional landmarks. They oriented first to the
geometry of the landmarks and only then to the identity of the
positional landmarks (Benhamou & Poucet, 1998). In this case,
positional objects appear to serve as directional landmarks. In
another study using all males, rats used the geometrical relation-
ship of proximal landmarks to deduce direction (Greene & Cook,
1997). We do not question the validity of these results but only
point out that it is possible that the same study repeated with
females might find a different set of rules, such as orientation
primarily to landmark identity and only secondarily to array
geometry.
Dissociating Parallel Maps
We now turn the remaining discussion to experimental evidence
in support of PMT. We first discuss the functional and structural
characteristics of each map, its development and plasticity. We
then discuss how task environments and specific lesions of hip-
pocampal structures should result in predictable map phenotypes.
Finally, we will discuss patterns of unit activity in the hippocam-
pus and their relation to PMT.
The Bearing Map
Oliver Wendell Holmes (1889) once wrote, “I find the great
thing in this world is not so much where we stand, as in what
direction we are moving” (p. 127). It is no less true for spatial than
for conceptual navigation. The bearing map supplies the direction
and the global location. It is the locus of primary mapping, and the
anatomical channel underlying it is the permanent scaffold of this
primary map. Because it is the scaffold, we postulate that it
necessarily increases in size and complexity as an animal matures,
explores new territory, and adds knowledge about new gradients
and directions in its environment.
The bearing map is constructed from simple 1-D gradients.
Structures in the bearing map channel create a 2-D coordinate
system from this 1-D input (see Figure 2a and Table 4). These
external gradient maps are calibrated to internal gradient maps
supplied by path integration. The 2-D coordinate system of the
bearing map becomes the scaffold not only for all directional
information but also for encoding the relative position of sketch
maps. Once a sketch map has been defined in relation to the
bearing map, the distance and direction of vectors that connect
disparate sketch maps can be computed. This is the integrated map
Table 4
Steps in the Construction of the Integrated Map
Step and locus Process Representation
1. MS Self-movement cues calibrated with theta rhythm as pacemaker.
Derives rate of movement and location relative to prior point.
1-D algorithm, stored in working memory.
2. DG 1-D map calculated from medial septum input and sensory input
from entorhinal cortex.
1-D map, stored in working memory.
Autoassociative network within DG recalls 2-D bearing map and
adds new 1-D map, creating a newly expanded 2-D bearing
map through pattern completion. Transmits updated bearing
map to CA3.
2-D bearing map, stored in reference memory.
3. CA3 Localizes current position on the 2-D map from DG by matching
pattern with current EC input. Transmits position to
subcortical structures through bilateral projection to lateral
septal nuclei and to cortical structures through Schaffer
collaterals to CA1.
Local aspect of the bearing map, stored in working memory.
4. CA1 Creates a minimap of the local panorama, computing within-
sketch vectors on the basis of head direction input.
Sketch map, stored in working memory, and integrated map,
also stored in working memory.
Receives bearing map position from CA3; localizes current
sketch map on bearing map. Transmits this part of the newly
computed integrated map to subiculum.
Integrated map, an emergent map from the integration of
bearing and sketch maps, stored in working memory.
5. Subiculum Updates reference memory of integrated map with new
information from CA1. Computes final position of this
fragment on the integrated map. Transmits integrated map to
associative cortices.
Current integrated map, stored in reference memory.
Note. This is a proposed account of how the parallel map theory could be assembled by components of the hippocampal formation. The schema must
be considered speculative, as we cannot yet even incorporate the CA2 field. MS ⫽ medial septum; 1-D ⫽ one-dimensional; DG ⫽ dentate gyrus; CA ⫽
subfields of the hippocampus proper; 2-D ⫽ two-dimensional; EC ⫽ entorhinal cortex.
296
JACOBS AND SCHENK

representation, and, using this map, the navigator can set out on
novel routes among known locales (see Figure 2c).
Structural Components of the Bearing Map
Navigators create gradient maps when they can calibrate their
movements relative to regular changes in stimulus intensity. This
requires three components: an input from internal movement (e.g.,
vestibular input), a sensory input, and a neural structure that can
create 1-D maps and 2-D maps from such gradient information.
We propose that these roles are mediated in stepwise fashion by
the septal nuclei, DG and CA3. Rather like an assembly line, each
component adds a new piece to the bearing map (see Figure 3 and
Table 4). The medial septum (MS) coordinates the input from
internal movement by supplying the pacemaker needed to assess
self-movement along a gradient. The pacemaker of this process
may be the hippocampal theta rhythm, a rhythmical, slow activity
pattern (Hasselmo, 2000). Theta occurs during active locomotion,
and its frequency and amplitude are related to the many parameters
associated with movement, such as speed (O’Keefe, 1993), antic-
ipation of movement (Morris & Hagen, 1983), and type of body
movements (Whishaw & Vanderwolf, 1973). Theta ceases when
the animal is immobile or engaged in other behaviors that do not
involve movement in a trajectory (Vanderwolf, 1969). Given its
close relationship to movement, it is not surprising that the MS and
its modulation of theta are important in spatial navigation (Whi-
shaw, 2000).
Given the pacemaker’s input, the next step is the integration of
external sensory input in the DG. The DG is the locus of a
convergence of subcortical and cortical (i.e., EC) inputs. It also has
a complex internal structure, in which projections from the poly-
morphic layer synapse within the molecular layer, which contains
projections back to the polymorphic layer (Amaral & Witter,
1995). This complex internal structure could underlie the function
of the DG to create the 2-D bearing map. After the sensory input
from EC is integrated with self-movement by calibration to the
theta rhythm to create the 1-D map, this autoassociative structure
in the DG completes the pattern of partially intersecting 1-D maps,
creating a 2-D representation. This is then used to create or update
the bearing map. This new information is then projected out of the
DG to the CA3 through the mossy fiber projection.
Figure 4. Predictions of the parallel map theory for the Morris water maze. The four patterns of spatial
performance result from the presence or absence of the parallel maps. Residual learning, resulting from the loss
of a single map (bearing or sketch), allows the rat to solve the maze, using transects when the bearing map is
intact and local loops when the sketch map is intact. The residual sketch map phenotype is expressed in two
stages: initial impairment of orientation and recovery to local loops with additional training. With both maps
impaired, the rat shows a permanent loss of spatial orientation, represented by global loops; performance
improves only with the development of nonspatial algorithms. With both maps intact, the rat encodes an
integrated map, which allows it to choose a direct path to the platform, regardless of start point. The solid circle
represents a hidden platform; the thick line represents a predicted swim trajectory.
297
PARALLEL MAP THEORY

CA3 has long been a focus of theoretical models because of its
unusual autoassociative architecture (Kali & Dayan, 2000; Marr,
1971; McNaughton & Morris, 1987; Rolls, 1996). We speculate
that one function of this architecture could be to convert gradient
map information to data that can be incorporated into a topo-
graphic map (see Figure 1d–1e). This process may be necessary to
integrate the bearing and sketch maps. Because CA3 is the conduit
between DG and CA1, this could be a role it plays. CA3 would
thus translate the current location on the bearing map into a code
that could be read by CA1. For example, the navigator’s location
could be recoded as a new object in a topographic map (i.e., on the
current sketch map). This allows CA1 to calculate the new topol-
ogy of the relationship among the navigator and the surrounding
objects.
To create an integrated map, however, CA1 also needs direc-
tional information from the bearing map, and, hence, CA3 must
transmit the location on the bearing map as well. We suggest that
CA3 could do this by segregating its output to CA1 into gradient
and topographic information projections. The area in CA3 that is
the most likely locus for the projection of gradient map data to
CA1 is the locus of heaviest projection from DG to CA (i.e., the
proximal third of CA3; also the more septal area). This proximal
third has also been called the projection zone (Ishizuka, Weber, &
Amaral, 1990) because the axons project directly to distal CA1
with few signs of synaptic contacts enroute. We speculate that the
locus for the projection of topographic data to CA1 could take
place in the middle and distal areas of CA3 (midseptotemporal),
which has been called the association zone on the basis of the
density of within-field projections and its input from the EC
(Ishizuka et al., 1990).
Finally, because all CA3 cells project both to CA1 and to the
lateral septum (Swanson, Sawchenko, & Cowan, 1980), projec-
tions from proximal CA3 and middle–distal CA3 might work in
parallel, allowing the updated position to be supplied as an afferent
copy to the bearing map channel (i.e., septal nuclei) and the sketch
map channel (CA1). In such a scenario, the CA3 would project
both location (for further encoding) and expected direction (for
current action). If so, then the lateral septum should show place-
specific activity involved in executing the next movement on the
planned trajectory. There is some evidence for this. The lateral
septum is considered to be the medial territory of the striatum
(Swanson & Risold, 2000). Characteristics of a structure involved
in place learning include LTP-like responses with fornix stimula-
tion in the mouse (Garcia, Vouimba, & Jaffard, 1993) and place
responses that have been recorded in freely moving subjects, both
in rats (Bezzi, Leutgeb, Treves, & Mizumori, 2000) and primates
(Ono & Nishijo, 1999). Because the subiculum also projects to the
lateral septum (LS; Amaral & Witter, 1995), we speculate that
movement may be driven by instructions either from the bearing
map (through CA3) or from the integrated map (through the
subiculum). Hence, the parallel map structure could well be mir-
rored in parallel output commands to the motor system, a hypoth-
esis that deserves more discussion than is possible here.
In summary, we propose that the bearing map is created in the
DG and projected through the mossy fiber projection to CA3,
which then transmits current location to CA1, in terms of both the
gradient and the topographic location, and to the LS, which me-
diates movement along the trajectories set up in the bearing map.
Development of the Bearing Map
As Tinbergen (1963) argued, to understand a behavior, one must
take into account not only its physiological mechanism but also its
evolutionary history, its current ecological significance, and its
development. Having discussed the history and structure of the
bearing map, we now address its development and plasticity, on
the basis of studies of the laboratory rat.
Structure and activity. The hippocampus is a remarkably late-
maturing brain structure, with final maturation occurring at
about 21 days after birth in the rat (Bayer, 1982). In some sense,
HPF development does not stop but simply proceeds through
different phases, in which structure, physiology, and behavior
continue to change.
Not only do hippocampal subfields develop late relative to other
telencephalon structures, but they also develop in a mosaic pattern.
The adult cytoarchitectonic pattern in DG develops by postnatal
(PN) Day 20, though mossy fibers continue to develop in adult-
hood (Amaral & Dent, 1981) and neurogenesis continues in the
granule layer throughout life in rodents and primates (Bayer,
Yackel, & Puri, 1982; Gould, Beylin, Tanapat, Reeves, & Shors,
1999). CA1 develops later, and significant dendritic arborization in
the stratum lacunosum occurs in the 2nd PN month (Pokorny &
Yamamoto, 1981). In female rats, CA1 dendritic arborization
fluctuates naturally with levels of estrogen, as during estrus
(Gould, Wooley, Frankfurt, & McEwen, 1990). Because of the
close connections between HPF and the neocortex, such plasticity
may result in as yet undetected structural changes that may con-
tinue throughout life (Alvarado & Bachevalier, 2000). We discuss
the role of neurogenesis in a later section.
Metabolic activity also develops in a mosaic pattern. Once
again, DG precedes CA1, with a marked increase in DG activity
occurring around weaning and CA1 increasing a month later
(Glick, Weaver, & Meibach, 1980; Meibach, Ross, Cox, & Glick,
1980). DG activity is correlated with gamma-aminobutryic acid
(GABA)-ergic inhibition, which does not mature until the end of
the 1st PN month in the rat (Swann, Brady, & Martin, 1989),
although activity similar to LTP is found after 2 PN weeks (Bat-
tistin & Cherubini, 1994).
Behavior. Like the hippocampal subfields, spatial learning in
rats develops in different stages during the 4th PN week, even in
simple tasks such as spontaneous alternation (Blozovski & Hess,
1989; Douglas, Peterson, & Douglas, 1973; Egger, 1973; Waters,
Klintsova, & Foster, 1997). By the time rats are weaned and have
begun exploring their environment (PN Day 20–26), they are able
to orient to a single goal, whether it is a water maze platform or an
escape hole in an open field (Chevalley & Schenk, 1987; Rudy,
Stadler-Morris, & Albert, 1987). This change in navigational strat-
egy suggests that adult spatial behavior appears at the end of the
postweaning period, around 24 days PN (Rudy et al., 1987).
However, an earlier appearance of a spatial bias toward the
training position has been observed after particular types of train-
ing. When training occurs in one day with long intertrial intervals
and the pups’ internal temperature is carefully controlled, a sig-
nificant bias toward the training quadrant appears as early as PN
Day 19 (Brown & Whishaw, 2000). When training takes place in
a small pool with highly salient visual cues in the immediate
vicinity of the pool, a significant bias is observed at PN Day 20–22
(Carman & Mactutus, 2001).
298
JACOBS AND SCHENK

Given juvenile rats’ poor long-term memory and poor ability to
thermoregulate, these special procedures may be critical in allow-
ing the expression of a significant spatial bias toward the training
position. For example, a panorama around the water maze that is
particularly rich in visual cues (Carman & Mactutus, 2001) or a
highly salient cue such as an illuminated hanging cup (Rudy et al.,
1987) placed a short distance from the hidden platform may result
in the expression of different spatial strategies and lead to a
significant spatial bias. In fact, analysis of the approach trajectories
of immature rats when a cue is hanging in the next quadrant
reveals a progressive transformation. The juveniles start by detour-
ing under the cue but end by making straight approaches from the
start point to the platform (Chevalley & Schenk, 1987) . Moreover,
juvenile rats aged 26–28 days trained on a cued-place task express
a significant degree of overshadowing that is not evident in young
adults (Schenk & Brandner, 1995), which suggests that they may
be developing particular spatial strategies at this age.
Hence, the absolute timing of cue and place learning appears to
be highly sensitive to experimental conditions. We suggest that
this is because different task environments are recruiting different
parallel maps. Further, the maps are developing at different rates,
in tandem with underlying hippocampal structures. Hence, the
bearing map should appear before the sketch map in development,
and, thus, we predict that juveniles, such as the PN Day 19
juveniles in the Brown and Whishaw (2000) study, encode the
place differently than do adults, relying more heavily on the
bearing map until much later in development.
The Ontogeny of the Parallel Mapping System
To test this hypothesis, we need to know precisely how the
hierarchy of cue use develops in the rat. There is clearly a transi-
tion from simply approaching a cue associated with the goal to a
guidance system in which a distal landmark (or landmarks) can be
used to determine the correct vector to the goal, whether it appears
at PN Day 19 (Brown & Whishaw, 2000) or later (Rudy et al.,
1987). If juvenile rats rely primarily on the bearing map, they
should orient well to distributed cues, such as distant visual cues or
orthogonal gradients of olfactory cues, but poorly to arrays of
positional landmarks.
There is evidence that such dissociation can occur. Juvenile rats
can learn the spatial position of an escape hole (homing table;
Schenk, 1989) in controlled cue conditions in a warm, dry, open
arena in which there is no risk of hypothermia. Using this task,
Schenk found that there was a marked interaction between age and
use of cues by type. During training, immature (PN Day 26–28)
and adult rats were given either discrete visual or distributed
olfactory information. Olfactory information was provided in a
radial pattern of scented strips, none of which was close to the
goal. The visual cues were an asymmetric configuration of three
discrete lights, as in a previous study (Rossier, Grobety, & Schenk,
2000). Only the adults could discriminate places purely on the
basis of such visual cues, despite normal visual acuity in the
immature rats (Salazar, Rossier, & Schenk, 2000). The immature
rats performed well only when odor cues were present (Schenk,
Jacobs, Rossier, & Kiraly, 2001).
We suggest that the immature rats relied on their relatively
mature bearing map and hence were more attentive to salient
distributed cues, such as odor gradients emanating from the
scented strips. As the sketch map matured, the rats began to rely
preferentially on the visual cues. This suggests that the sketch map
is slowly calibrated by the bearing map during the 2nd PN month
and may not become autonomous until the 3rd month, coinciding
with hippocampal maturation. The integrated map should then
develop after natal dispersal and mature further during adolescence
(Schenk et al., 2001). This hypothesis is confirmed by the obser-
vation that PN Day 48 rats can rely on the visual cues to solve the
task only if they have experienced the conjunction of the olfactory
and visual cues, as though the presence of both is necessary for
such a dual coding phase (Schenk et al., 2001).
Hippocampal Sex Differences
The sex differences in navigation strategy described earlier in
the laboratory rat develop during the period of DG maturation.
These differences between females and males are not fixed genet-
ically but rather appear as a consequence of exposure to gonadal
hormones during the PN period (Williams et al., 1990). Thus, the
differential reliance by females and males on the bearing map
appears to be the result of differential exposure to hormones; the
differences are eliminated or reversed with a perinatal treatment of
estrogen or its metabolite, testosterone (Isgor & Sengelaub, 1998;
Roof & Havens, 1992; Williams et al., 1990).
This sex difference should be reflected in the structure of male
and female HPF; we predict greater development of bearing map
components in males. Sexual dimorphism in the entire hippocam-
pus (HPF volume relative to telencephalon volume) has been
found both in wild rodents (Jacobs, 1995, 1996) and in wild birds
(Reboreda, Clayton, & Kacelnik, 1996; Sherry et al., 1993). In
both groups, the sex that relies more heavily on spatial exploration
during the breeding season has a relatively larger hippocampus.
Because both laboratory rats (Rattus norvegicus) and mice (Mus
musculus) are derived from species in which males roam more
widely than females, we predict that the hippocampus should also
be sexually differentiated in these species. This is confirmed by
several measures: Male mice have more DG granule cells (Wimer
& Wimer, 1985); male rats have a thicker DG granule layer (Roof
& Havens, 1990), which is associated with a greater cell density
(Loy, 1986), and more dendritic branching points in the DG
(Juraska, 1984; Juraska, Fitch, Henderson, & Rivers, 1985). Male
rats also have more mossy fibers projecting from DG to CA3 and
more synapses in this projection than do females (Madeira &
Paula-Barbosa, 1993; Madeira, Sousa, & Paula-Barbosa, 1991).
Thus, in all sexual dimorphisms that have been described for DG
or CA3 in laboratory rats and mice, there is greater structural
development in males than in females. Because this correlates with
the male rat’s reliance on directional cues, this supports the hy-
pothesis that the bearing map is mediated by DG and CA3.
Hippocampal Neurogenesis
A unique property of the bearing map is its role as the perma-
nent frame or scaffold for the integrated map. Further, we propose
that the bearing map is both consolidated and stored in the DG. As
the map itself becomes larger and more complex with new spatial
exploration, however, it needs a mechanism to add more informa-
tion. In addition, if the bearing map is a permanent reference
system, then it should have a high threshold for synaptic change,
299
PARALLEL MAP THEORY

in contrast to the ephemeral sketch map, which should rapidly
acquire new data, overwriting the old. For the hippocampus, unlike
the visual system, in which the rules of the visual environment
must be organized early in development (Shatz, 1992), the spatial
environment that is experienced by a navigator changes throughout
life. Incorporating new spatial data in the bearing map thus may
require different levels of neural plasticity—not only synaptic
plasticity but the addition of new structural elements to increase
storage and computational capacity.
One mechanism that serves this end is adult neurogenesis, which
both increases storage and acts as a primitive form of memory
consolidation. Adult neurogenesis is common in all vertebrates but
mammals (Perez-Canellas, Font, & Garcia-Verdugo, 1997). It is
found throughout the brain in fish (Birse, Leonard, & Coggeshall,
1980), amphibians (Polenov & Chetverukhin, 1993), and reptiles
(Garcia-Verdugo, Llahi, Ferrer, & Lopez-Garcia, 1989; Lopez-
Garcia, Martinez-Guijarro, Berbel, & Garcia-Verdugo, 1988; Por-
tole´s, Dome´nech, Martin Pe´rez, & Garcia Verdugo, 1988). In
mammals, adult neurogenesis occurs primarily in the olfactory
bulb and the granule layer of the DG (Bayer, 1985). Why neuro-
genesis is more locally restricted in mammals than other verte-
brates is an interesting question in itself (Perez-Canellas et al.,
1997). Another question is why the local restriction involves the
hippocampus. One explanation may be the importance of signifi-
cant structural plasticity for encoding and consolidation of the
bearing map.
The recruitment of new granule cells to the DG occurs through-
out life, resulting in an ever greater number of granule cells, which
appear in organized growth rings (Bayer, 1982; Kaplan & Hill,
1985). This orderly process could be the result of new exploration,
requiring new structure in the form of such growth rings. If so,
then the rate of growth should be proportionate to the amount of
new exploration that requires an updating of the bearing map.
Indeed, recent studies have found a relationship between new
exploration and the rate of neurogenesis and neuron recruitment.
Mice that engaged in hippocampal-dependent tasks such as water
maze navigation or trace conditioning showed a significant in-
crease in the recruitment and survival of new neurons in the DG,
relative to mice performing versions of these tasks that did not
require the hippocampus (Gould et al., 1999). Mice that ran on
exercise wheels also showed increased recruitment of new neurons
to the DG (van Praag, Kempermann, & Gage, 1999). Wheel-
running experience not only predicted increased neurogenesis but
was also correlated with improved performance in the water maze
and enhanced LTP in the DG. In contrast, LTP in CA1 was
unaffected by the wheel-running experience (van Praag, Christie,
Sejnowski, & Gage, 1999).
We suggest that these patterns of neurogenesis could be inter-
preted as the selective activation of the bearing map and, hence, of
plasticity in the DG during this activity. Wheel-running behavior
produces movement of apparent linear progress. It occurs under
the unusual circumstance, in the evolutionary history of the mouse,
of linear movement in the absence of concurrent change in the
appearance of positional landmarks. The bearing map channel,
constructed from the calibration of self-movement data to sensory
input, should be highly stimulated by wheel running. Because all
positional landmarks remain fixed in their aspect during wheel
running, however, the navigator must assume that such landmarks
are directional, not positional. As the mouse continues to move
forward, it should therefore continue to interpret the sensory input
as a constantly expanding bearing map. These conditions should
increase the activation, expansion, and neural plasticity of the
bearing map channel. Therefore, wheel-running behavior should
result in an increased rate of neurogenesis in the DG, enhanced
navigation ability using the bearing map, and enhanced synaptic
plasticity in the DG—predictions consistent with the published
results of van Praag, Christie, et al. (1999; van Praag, Kemper-
mann, & Gage, 1999). These hypotheses need be tested further
with explicit manipulation of the sensory environment during
running.
Testing the Theory
To test PMT, we need to isolate the parallel maps. This can be
done in two ways: restrict the task environment to one class of cues
(directional or positional) and observe the change in behavior of
the animal or patterns of neural activity, or impair the channel for
one map but not the other. Despite tens of thousands of publica-
tions on the physiology and function of the hippocampus (see
reviews in O’Keefe & Nadel, 1978; Redish, 1999), few studies
have provided these types of tests in enough detail to test PMT.
Either the lesions are not specific to subfield, or the task environ-
ment contains both types of cues, or the assay of spatial learning
is not precise enough to detect the presence and nature of residual
spatial learning. Therefore, although the studies we discuss may
appear to be a highly selected subset of the literature, in fact there
are very few studies that meet our criteria.
Once the maps have been isolated by task environment or
through the use of specific lesions or receptor blockades, residual
learning must be quantified. Figure 4 illustrates the predictions of
PMT for one task, the Morris water maze. The four possible
experimental conditions (rB, rS, both intact, both impaired) result
in different search algorithms. The analysis of these algorithms
requires new image processing techniques to analyze the shape of
swim paths, as standard measures (latency to platform, length of
search path) do not reveal map phenotype. A recipe, therefore, for
testing the predictions of PMT is to apply the methods proposed in
Tables 2 and 3 to dissociate the maps and then measure the
outcome of such manipulations in the water maze, as illustrated in
Figure 4.
Experimental Impairment of the Bearing Map
The impairment of the bearing map should produce an rS map
phenotype. These subjects should be highly dependent on unique
positional landmarks. We discuss the evidence for impairment for
each component of the bearing map, following the assembly line
order outlined in Table 4.
Medial Septum
On the basis of his studies of working memory, Numan (2000)
recently postulated a dichotomy between the hippocampus–
cortical connections that encode “external stimuli” while the sep-
tohippocampal system “encode[s] and maintain[s] the self-motion
cues“ (p. 316). This interpretation is consistent with our dichotomy
of the bearing and sketch maps; it suggests that septal lesions
should impair a sense of direction and lead to an rS phenotype.
300
JACOBS AND SCHENK

This is confirmed by lesion studies in which MS-lesioned rats
have a specific deficit in sense of direction (Kelsey & Landry,
1988). In a study by Brandner and Schenk (1998), MS-lesioned
rats were highly dependent on the use of a beacon to orient,
presumably because they could no longer maintain an internal
sense of direction. Finally, temporary inactivation of the MS with
lidocaine produces a specific loss of spatial specificity to CA3, not
CA1 (Mizumori et al., 1989). This is a precise characterization of
the rS phenotype, in which the loss of encoding is specific to the
bearing map channel and the sketch map channel is spared.
Fimbria Fornix
The fimbria fornix (FF) projection carries the major subcortical
input to the HPF and hippocampal output to various areas, includ-
ing the projection from MS to DG (Amaral & Witter, 1995). A
high proportion of fibers afferent to the CA1 travel through the
cingular bundle and thus remain unaffected by FF transection
(Amaral & Witter, 1995). Therefore, FF transection should selec-
tively impair the bearing map.
This prediction has been borne out in a series of experiments by
Whishaw and colleagues (1995). Here, the FF transection appeared
to eliminate the rat’s directional orientation, but without impairing
its recognition of the goal location. Whishaw described this as the
dissociation of a sense of direction and of place. We interpret this
as an rS phenotype, in which the rat’s orientation is based on the
residual ability to encode unique positional landmarks.
Whishaw and Jarrard’s (1995) interpretation of these results as
a dissociation between knowing the place and being able to orga-
nize a trajectory to the place is entirely consistent with PMT. The
bearing map system provides a link for efficient approach trajec-
tories, which also allows immediate reversal learning. This capac-
ity is irrevocably lost following the FF transection (Whishaw &
Tomie, 1997).
Rats with FF transection also exhibit a change in response to
cues. Place units in such rats appeared no longer to distinguish
directional landmarks but relied heavily on positional landmarks in
a radial maze. The spatial correlates of hippocampal units were
severely disrupted by rotation of the maze and maintained some
significant firing in a particular arm, most likely on the basis of
olfactory cues (Miller & Best, 1980; Shapiro et al., 1989). This,
too, we interpret as an rS phenotype, characterized by overreliance
on unique positional (i.e., local, not distal) cues.
Dentate Gyrus
Several early studies suggested that the role of the DG in spatial
navigation could be dissociated from that of the HP. For example,
the colchicine-induced lesion of the granule layer in rats was found
to impair spatial learning, as expected, but it did not affect the
spatial selectivity of pyramidal cells (McNaughton et al., 1989).
The authors interpreted this as evidence that spatial information
can be conveyed to the pyramidal cells in the absence of granule
cells, a controversial conclusion at that time but a result that is
consistent with PMT. Another intriguing result from this study is
a trend toward a decrease in directional bias in lesioned animals, a
result that is expected in an rS phenotype.
Gilbert et al. (2001) recently effected the double dissociation of
DG and CA1 function. With selective DG lesions, rats could not
orient accurately to a cued location in an open arena when a
second, decoy cue was placed a close distance from the original
target. The impairment disappeared at greater separations between
target and decoy. Moreover, this impairment was not seen in
CA1-lesioned rats (Gilbert et al., 2001).
One interpretation of this impairment is that the DG is needed to
disambiguate panoramic views associated with certain positional
landmarks. If the DG lesion produces an rS map phenotype, then
rats are limited to the use of sketch maps. Hence, they are forced
to encode the location of objects using local cues only. In an open
arena in which local cues are situated at some distance from the
goal, these cues provide only a coarse map for mapping the test
objects. Such a coarse map should suffice when the objects are far
apart, as they can be associated with unique local landmarks. But
without the bearing map, objects that are close together cannot be
disambiguated by their position on vectors to distal objects. There-
fore, the rS phenotype rats should be impaired at small but not
large separation of objects.
CA3
To date, there have been no clear behavioral dissociations of
CA3 function. In an early study by Jarrard, there was no effect of
CA3 lesion on a place-learning task (Jarrard, 1983). As in most
studies, however, this lack of effect could have been masked by
residual spatial learning (i.e., rS phenotype).
As described earlier, one of the most intriguing findings is that
CA3 place units lose spatial specificity with MS inactivation,
whereas CA1 units are unimpaired (Mizumori et al., 1989). This
finding has been used as evidence for understanding the hippocam-
pus as a series of independent channels (Amaral & Witter, 1995).
The Sketch Maps
Under natural conditions, the physical space in which mammals
navigate is not a carpentered test room but a complex world, in
which new panoramic views are learned piecemeal, in linked
fragments (Bovet, 1992; Huxter, Thorpe, Martin, & Harley, 2001;
Jacobs & Shiflett, 1999; Schenk, Grobety, & Gafner, 1997). These
fragments are obtained in discrete exploratory episodes (Eilam &
Golani, 1988). Thus, encoding of space must link exploratory
bouts within the same region of space. We propose that the sketch
map evolved to code such a series of topographic maps, whereas
the spatial relation among sketch maps was obtained from the
directional bearing map (e.g., information combining direction of
movement relative to the start of the journey and orientation
relative to a given compass mark). Like a theatrical stage, if the
bearing map is the bare stage, which serves as the framework for
a changing display, the sketch maps are the rapidly changing sets,
each defining a new location. With new exploration, a new version
of the sketch map is rapidly formed and is indexed in similar
coordinates (as the previous sketch map) to the bearing map.
Hence, the sketch map is not a cognitive map, nor is it a single,
permanent topographic map. It is, in fact, a sketch of a new locale.
It can be thought of as a minimap of a subspace rather than as the
map of all space. It encodes the chains or chunks of disconnected
local views from a particular point in space and is based on the
locations of unique positional landmarks. As the navigator moves
through the environment and comes to new vantage points with
301
PARALLEL MAP THEORY

new panoramic views, a new sketch map is formed. Thus, under
natural conditions, as a mammal explores new terrain, it creates
multiple, disconnected sketch maps. Some maps may be linked to
each other by common positional landmarks. Yet even with links
through common landmarks, this representation alone (i.e., multi-
ple sketch maps) cannot pass the test of the cognitive map: con-
structing a novel route between previously disconnected sketch
maps.
As a caveat, we note that the spatial experience of the laboratory
rat may be qualitatively different from that of the wild rodent. A
lab rat exploring a single open arena possesses at most two or three
sketch maps (i.e., at least two panoramic views of the task envi-
ronment, differing by 180° in direction). In contrast, a wild rodent
may encode tens or hundreds of sketch maps over a period of
weeks or months. In fact, wild gray squirrels show an increase in
brain and hippocampal volume during the fall harvest of nuts
(Lavenex, Steele, & Jacobs, 2000), a strategy that may require the
spatial encoding of many thousands of spatial locations (Jacobs &
Liman, 1991). The frame of reference used by wild squirrels to
orient to a goal in their home territory, however, does appear to be
similar to that used by the laboratory rat in a test room: an
extramaze, allothetic frame of reference (Jacobs & Shiflett, 1999).
Once the sketch map is encoded, the next step is to encode the
orientation of the sketch maps within the bearing map and to each
other, creating the integrated map (see Table 4). Whether or not a
particular sketch map is integrated, however, it continues to remain
active and independent until it is no longer accurate. If there is a
change in the location or identity of landmarks in the topographic
array, then the sketch map is overwritten by a new sketch map. If
there are no changes in the array, then the map is maintained in
memory and could serve some function in long-term storage.
Under natural conditions, a rodent navigator should have a single,
permanent but expanding bearing map, whereas the coding of
positional landmarks and sketch maps may be ephemeral. We
suspect that most sketch maps have a short life in the HPF. They
may be stored elsewhere as spatial objects, possibly chunked to
one another.
Although sketch maps are independent and dissociable from
each other and from the bearing map, it is possible that they can be
linked to each other by associations between common positional
landmarks during an exploration. In this way, chains of sketch
maps may be formed, creating a route. In our view, creating a route
is radically different from the simple route learning due to the
systematic repetition of the same trajectory. Normal rodents may
spontaneously generate and learn efficient routes, whereas subjects
with selective hippocampal lesions have to be specially trained to
follow and learn these routes, as in the experiments cited earlier by
Whishaw et al. (1995) in the discussion of the FF.
How Spatial Is the Sketch Map?
The function of a sketch map is to represent the identity and
location of a group of landmarks, the positions of which are coded
relative to each other. This group is defined by the animal’s
exploratory strategy. It is also determined by the proximity among
salient features. It even can be defined as a spatial object or
configuration. Our conception of the sketch map differs from a
single visual local view, however, because it integrates different
viewpoints and is based on multiple sensory modalities. Poucet
and Benhamou (1997) have proposed a mechanism for processing
the fragments of a map. The resulting supramodal panoramic local
view can thus be activated by priming from any one of these
sensory modalities. This is similar to our concept of the sketch
map.
Because the aspect and appearance of a positional landmark
change with the position and movement of the observer, the
processing of a sketch map must integrate the ongoing body
position of the observer. Thus, information about head direction is
critical to the formation of a sketch map. Path integration, how-
ever, although it is a main source of direction for the bearing map,
is not necessary. As a consequence, sketch maps are spatial ob-
jects, because the relational coding within a sketch is purely
allocentric. The observer’s head orientation only allows it to code
the relations among the sketch map components and is not stored
as such. In this way, within-sketch map relations are spatial (allo-
centric), whereas between-sketches map relations are secondary
either to their link with a common bearing map or to a simple
sketch-to-sketch relation based on locomotion.
Thus, in spite of their spatial construction, sketch maps are
disconnected from absolute locations in space, unless they are
linked to all other sketch maps through the bearing map. In the
absence of the latter, sketch maps can also be linked to one another
in specific routes, particularly if subjects are repeatedly trained to
follow systematic approach routes. This is why we suggest that the
special training procedure used by Whishaw et al. (1995) can lead
to apparently normal place memory. This also emphasizes the role
played by the bearing map during exploration in ordering sketch
maps within the spatial integrated map.
Because a sketch map is based on positional landmarks and
because positional landmarks have both spatial and nonspatial
properties, a sketch map may encode both spatial and nonspatial
properties of an array. These properties of the sketch map—
encoding nonspatial data in unique spatial arrays and linking
sketch maps in a unique temporal sequence—set the stage for
encoding of unique temporal events or episodes.
Overall, our concept of a sketch map has much in common with
other models of hippocampal function. It may function as a tem-
porary memory store, as postulated by Rawlins (1985); it may
create maps through the “automatic encoding of attended events,”
as proposed by Morris and Frey (1997, p. 1489), and sketch maps
may be linked to each other by the fragment fitting process
outlined by Worden (1992); they also have some qualities of the
scenes proposed by Gaffan (1991). Moreover, sketch maps must
code the nonspatial properties of the objects or positional land-
marks from which they are composed, as in Eichenbaum et al.’s
(1999) hypothesis of memory space.
Finally, we predict that a given sketch will be recognized as
such in spite of changes in size, although this might induce a
generalized activity response (see Poucet & Benhamou, 1997). On
the other hand, changes in the configuration of the components
should trigger reexploration of the modified component (novelty
or change reaction).
Structure and Plasticity of the Sketch Map
The structure that mediates the sketch map, CA1, is the hip-
pocampal component that has been studied in the most detail in
electrophysiological recording. Thus, it is no surprise that our
302
JACOBS AND SCHENK

description is similar to other models in which CA1 encodes a
topographic allocentric representation. Our model differs in that a
CA1-mediated representation is not necessarily the cognitive map;
it may be a temporary sketch map that is later incorporated into the
integrated map.
Because the sketch maps are temporary representations, they
should be served by a more rapid form of plasticity than that in the
bearing map. For example, rapid updating of the sketch map
should occur when an animal attends to a change in positional
landmarks. This can occur when a new area is explored or when
positional landmarks have been shifted or replaced in a familiar
area.
Another implication of the sketch map is for the role of the
hippocampus in reducing interference (Shapiro & Olton, 1994).
The precise remapping of sketch maps to unique positional land-
marks should be correlated with a reduction in interference. We
suggest that this is because it is difficult to differentiate (and,
hence, not overwrite) sketch maps without knowing their relative
position on the bearing map.
There is currently a general agreement that the major form of
synaptic plasticity for CA1 function is NMDA-mediated LTP
(Morris, 1990). This form of plasticity is appropriate for the rapid
encoding and reconfiguration required for constructing and updat-
ing a sketch map. In fact, a rat’s spatial exploration of new terrain
has been shown to reverse NMDA-mediated LTP in the rat (Xu,
Anwyl, & Rowan, 1998). CA1 is also sensitive to changes in
positional landmarks. With c-fos response as an assay, CA1 was
selectively activated after a rat viewed familiar objects in a novel
scene. After it viewed the novel scenes, c-fos response increased in
postrhinal areas and CA1 but decreased in DG and the subiculum
(Wan, Aggleton, & Brown, 1999). This differential activation of
DG and CA1 supplies additional evidence for the independent but
complementary nature of these two structures, which is appropriate
for the major components of the parallel maps.
There are predictable sex differences in CA1 plasticity. Males
appear to rely preferentially on the bearing map, whereas females
rely preferentially on the sketch map, and females should be more
sensitive to changes in the topology of positional cues. Females do
express an immediate reaction to the change of positional land-
marks in spite of the fact that the directional information contri-
bution to the bearing map has not changed (Williams & Meck,
1993). Hence, we expect that plasticity in the sketch map might be
a hallmark of the female hippocampus. Circulating levels of es-
trogen have a significant effect on CA1 structure and function in
adult female rats. Naturally occurring levels of estrogen cause
significant dendritic branching in the estrus female (Gould et al.,
1990). This is accompanied by increased levels of LTP in the
presence of natural peaks in estrogen concentration (Warren, Hum-
phreys, Juraska, & Greenough, 1995).
Experimental Impairment of the Sketch Map
Many studies have examined the role of CA1 function and the
NMDA receptor in spatial learning. Few studies have attempted to
isolate the specific function of CA1, however. PMT predicts that
CA1 lesion or the blockade of NMDA-mediated LTP should not
abolish spatial learning but instead should reveal the residual
learning capabilities of the bearing map (i.e., an rB phenotype).
CA1
The behavioral dissociation of DG function by Gilbert et al.
(2001) was one half of a double dissociation between DG and
CA1. The task that created an impairment in CA1-lesioned rats
was quite different: It was the temporal ordering of visits to arms
of the radial arm maze, a task that Chiba, Kesner, and Reynolds
(1994) had previously shown to be impaired in hippocampal-
lesioned rats. Yet even though CA1-lesioned and hippocampal-
lesioned rats were impaired on this task, the DG-lesioned rats were
not (Gilbert et al., 2001). This result is consistent with an rB
phenotype—a selective loss of the ability to link different pan-
oramic views or sketch maps. The loss of temporal order is also
predicted from the lesion of the sketch map channel, because one
operation possible with sketch maps is to link and sequentially
activate them.
The Role of the NMDA Receptor
The dissociation of NMDA-mediated LTP and spatial learning
has posed a serious challenge to the theory of the hippocampus as
a cognitive map (Bannerman et al., 1995; Saucier & Cain, 1995).
As we interpret these results, because this receptor is found in high
concentration in CA1, the loss or blockade of these receptors
should produce the rB phenotype. In contrast, techniques that are
selective for the receptors of transmitters found in greater abun-
dance in the DG should produce the rS phenotype. Such paradox-
ical dissociations of LTP and spatial learning thus may be the
outcome of residual learning by the unaffected channel.
If the NMDA blockade selectively impairs CA1, it should
produce the rB phenotype (see Table 2). Therefore, rats experi-
encing the NMDA blockade should be able to navigate using
directional cues. The water maze is a task that requires both
bearing and sketch maps and particularly requires a high-
resolution sketch map for high spatial accuracy, as accurate ap-
proach trajectories must be computed to hit the platform or to
correct misses. Therefore, NMDA blockade and, hence, impair-
ment of the sketch map should produce a greater deficit in the
water maze than in the radial arm maze, which relies more heavily
on directional cues. If the rat knows how to solve the water maze
using its bearing map, an accurate performance later under NMDA
blockade should be possible because of the residual bearing map.
We suggest that this accounts for the effect of pretraining on
NMDA blockade (Bannerman et al., 1995; Saucier & Cain, 1995).
Our interpretation is as follows: Intact rats with prior training had
had experience solving the task using the full complement of
spatial representation (bearing, sketch, and integrated maps). In
addition, the rats were male and, hence, were likely to have relied
on the bearing map. When the NMDA-blockade selectively im-
paired the sketch map, these pretrained rats were able to recall their
strategy of using a bearing map. This change in swim behavior
(illustrated in Figure 4) should be detectable with quantitative
analyses of swim path shape but also can be seen in the raw data
published by Bannerman et al. (1995). Finally, although the rats
showed the rB phenotype, this representation had only a low
resolution and was inadequate for unimpaired performance in the
water maze, and, hence, the rats’ performance rose only to an
intermediate level.
The pretraining effect was abolished, however, when the pre-
training occurred in a curtained maze. Here, NMDA blockade
303
PARALLEL MAP THEORY

during the second maze abolished accurate orientation (Banner-
man et al., 1995). Our interpretation is that the curtained maze
eliminated directional cues during pretraining and, hence, pre-
vented the rats from learning to solve the water maze using the
bearing map. Therefore, these rats were prevented from learning
during NMDA blockade.
Our prediction that the NMDA blockade after pretraining should
produce a lesser impairment in the radial arm maze is supported by
two studies (Caramanos & Shapiro, 1994; Shapiro & O’Connor,
1992). We would interpret these results as the rat’s successful
reliance on a residual bearing map to solve the highly directional
radial arm maze.
Genetic engineering techniques can also be used to selectively
lesion neurotransmitter receptors in hippocampal subfields. Re-
sults from two such studies are consistent with PMT predictions of
residual learning and plasticity in the remaining channel. Knockout
of CA1-localized NMDA-mediated LTP impaired the spatial se-
lectivity of place cells (McHugh, Blum, Tsien, Tonegawa, &
Wilson, 1996). This could be interpreted as need for the NMDA-
mediated tuning in the high-resolution sketch map. Even though
these mice should have shown the rB phenotype, the place spec-
ificity of the sketch map would be impaired.
Residual learning was also observed when the knockout of a
neuronal glycoprotein eliminated LTP in the DG but did not
eliminate CA1 LTP (Nosten-Bertrand et al., 1996). We suggest
that in this study, the rS phenotype produced by the knockout
procedure allowed the animals to recover to normal performance
levels. In contrast, in the study by McHugh et al. (1996), the rB
phenotype would not allow recovery of place specificity in dorsal
CA1 pyramidal cells.
Harking back to spatial learning by turtles, we note that mice
with calcium-calmodulin-kinase II (CCKII) mutations showed dif-
ficulty orienting in an open water maze but could nonetheless
solve a water-filled plus maze (Silva et al., 1992). In our view, the
mice had an rB phenotype and could not locate a place, only a
direction (i.e., our interpretation of the performance by turtles in a
water-filled plus maze). We suggest that orientation by CCKII
mice and turtles was mediated by their homologous brain struc-
tures: the small-celled area of the medial cortex in the turtles, and
the DG in the mice. It would be interesting to test this proposition
by combining selective lesions with precise behavioral assays of
the bearing map in mice and turtles. These results point, however,
to the great potential of these genetic techniques (e.g., a double
dissociation of receptor types and, thus, potentially of the parallel
maps).
The Integrated Map
The integrated map emerges from the coding of the sketch maps
in relation to the bearing map (see Table 4). Once this occurs,
novel routes and shortcuts can be calculated from the resultant
integrated map. The navigator can also chunk elemental sketch
maps to form larger sketch maps. Thus, the integrated map pro-
vides two important cognitive capacities. It permits the navigator
to induce the length and direction of the paths needed to travel
from one sketch map to another, on the basis of their relative
location on the bearing map. It emerges from the interaction of the
two maps and allows long-term storage of chunks of sketch maps
(or, potentially, episodes) for later retrieval from long-term mem-
ory. By this means, it is possible to derive the direction of move-
ment that is needed to move from one sketch map to another, even
if this movement is discontiguous. In fact, it is the function of the
integrated map to compute this vector. It allows the navigator to
compute the shortest route between two sketch maps that have no
positional landmarks in common.
Coactivation
The fact that discrete sketch maps (or chains of sketch maps)
have been stored with an index of their situation on a bearing map
allows them to be positioned relative to each other in a single
integrated map. This spatial index for the sketch maps is due to the
coactivation of the two maps. This facilitates the direct access (or
recall) during later movements in that locale. And because sketch
maps are assigned to unique locations on the single bearing map,
it is possible to distinguish them unambiguously from each other.
Structural Components of the Integrated Map
Forming an integrated map requires two components: a repre-
sentation of the bearings with the environment, and a representa-
tion of the relationships among discrete objects. The integrated
map also requires a reference memory representation of the current
or most recent integrated map of the environment. To have full
integrated map capacity, all components must therefore be func-
tional: the components of the bearing map (MS, DG, CA3), the
sketch map (CA1), and the subiculum, which mediates the refer-
ence memory of the integrated map. (There are no doubt upstream
structures projecting to these components, particularly subcortical
areas, that are important for map integration.) The two component
maps (bearing and sketch) must also have intact afferent projects,
such as the FF projection from MS to DG, the mossy fiber
projection from DG to CA3, and the Schaffer collaterals from CA3
to CA1 (see Figure 3).
Map Assembly
The neural architecture of the HPF is uniquely characterized by
one-way connections from CA3 to CA1 to the subiculum (Amaral
& Witter, 1995). We interpret this architecture as the assembly line
necessary to construct the integrated map (see Table 4). We
propose that 1-D maps that are initially encoded in DG are then
combined there to form the 2-D bearing map. This information is
then projected to CA3. CA3 projects the location as a position on
the bearing map (and as a discrete object) to CA1. CA1 integrates
the location on the bearing map with the current sketch map and
thereby calibrates the sketch map with the bearing map.
Once the integrated map has been formed, CA1 extracts new
routes from the integrated map. CA1 is a complex and topograph-
ically organized structure, with direct projections from sensory
areas such as the EC, olfactory bulb, and neocortical areas as well
as thalamic nuclei (Amaral and Witter, 1995). This complex ar-
chitecture may reflect CA1⬘s role as the constructor of sketch maps
and the locus of the integration of these maps with the bearing
map. In this case, CA1 topology reflects two functions: its role as
the encoding of new sketch maps as well as the locus of the current
status of the integrated map. From the integrated map (probably
dorsal CA1), the role of this area is to extract the temporal order
304
JACOBS AND SCHENK

of positional landmarks on novel routes, calculated from the inte-
grated map.
Consolidation of the Integrated Map
Both bearing and sketch maps must be stored in long-term
memory to be retrieved later, albeit in different locations. We
suggest that the storage of the bearing map is within the DG and,
hence, that the DG provides the index to all allocentric encoding
(see Teyler & DiScenna, 1986, for an earlier proposal of indexing
by the hippocampus). In contrast, the sketch maps are most likely
stored in the temporoparietal neocortex. The question of whether
the integrated map is stored as such is less clear. This depends on
the definition of what is a single sketch map and whether there is
overlap among sketch maps encoded for a given environment. If
sketch maps are coded when directional information is available,
the relation among sketch components may be associated with at
least one major gradient and with the vector from path integration.
Such dual coding of successive sketch maps would become more
frequent as the environment is intensively patrolled, leading to the
long-term storage of linked sketch maps. This may be why over-
trained memory for familiar places can be retrieved even after
hippocampal damage (Teng & Squire, 1999). Although we do not
suggest that a global integrated map is stored in long-term mem-
ory, we predict that consolidation may stabilize the relationships
among sketch maps if they have been linked by locomotion.
However, preoperative training in the water maze, although it
facilitates acquisition, does not seem to facilitate task reacquisition
and has little effect on probe trials (Morris, Schenk, Tweedie, &
Jarrard, 1990). This suggests that very little of the integrated map
has been transferred out of the hippocampus during consolidation.
Reference Memory Representation of the Integrated Map
After the integrated map is created by the hippocampus, it is still
necessary to have a reference memory representation to make the
map available for use and updating. Unlike DG and HP, the
subiculum has reciprocal connections with hippocampal, thalamic,
and cortical areas. We suggest, therefore, that it might store the
reference memory representation of the integrated map that has
been constructed by the hippocampus. If so, lesions of the subic-
ulum would differ in subtle ways from hippocampal lesions: They
would impair reference memory for the integrated map but not
working memory. Place units in the subiculum should also differ
from those in the hippocampus, as we discuss later.
Experimental Impairment of the Integrated Map
Damage to one channel should simultaneously impair the rele-
vant map and the integrated map.
Subiculum
A selective lesion of the subiculum should impair the reference
memory of the relationship among sketch maps (i.e., the integrated
map). Hence, result of this should be an rB phenotype. The animal
should show partial recovery, because it could learn to navigate
with the remaining bearing map. Such qualitative differences be-
tween subiculum-and hippocampus-lesioned groups have been ob-
served. Rats with subicular lesions (rB) appeared to swim around
the entire pool. This swim pattern was quite different from the
progressively adjusted loops commonly seen in rats with an rS
phenotype (e.g., rats with septal lesions). Moreover, the rats with
a subiculum lesion seemed to have little memory between sessions
(i.e., impaired reference memory), although they were capable of
expressing memory for a new platform position after four trials in
a match-to-place test (i.e., intact working memory; Morris et al.,
1990).
On the basis of the foregoing discussion, a pure subicular lesion
(rB phenotype) should impair learning of the new location but
leave residual capacities based on the bearing map intact. This
should result in a residual ability to learn new positions based on
the bearing map, in which new locations can be encoded. Consis-
tent with this interpretation, subiculum-lesioned rats showed im-
provement in platform location within trial blocks (Morris et al.,
1990). This suggests that they could find the platform by swim-
ming transects, orienting with the bearing map (see Figure 4).
Hippocampus Lesions Sparing the Subiculum
The challenge to any spatial theory of hippocampal function is
to explain residual spatial learning after a complete hippocampal
lesion, defined as the absence of cell bodies and/or all neural tissue
of the DG and HP. Yet learning by hippocampal-lesioned rats can
occur under certain conditions, and, with additional training, a
hippocampal-lesioned rat can show significant recovery. The ex-
planation for this may lie in the nature of the training. Recovery
does not occur quickly but only with special training, such as
alternating trials with a visible and an invisible platform at the
same position (Whishaw et al., 1995; Whishaw & Gorny, 1999;
Whishaw & Jarrard, 1995). The logistical difficulty with this
lesion is that hippocampal-lesioned rats cannot generate controlled
trajectories (i.e., trajectories associated with predictable changes in
sensory input, based on the 2-D bearing map). They therefore
cannot form sketch maps of positional cues, which are scarce in a
swimming pool when no salient cue is provided in the vicinity of
the pool. The solution of attracting the rat to the visible platform
to facilitate learning of these cues is an elegant one. Training the
rats in this way allows them to link egocentric views that could
then become an effective substitute for true allocentric place
learning.
For example, under such prolonged training, a secondary system
may be slowly brought online to compensate for the loss of
hippocampal function. The parietal cortex appears to act as a
spatial map, associating objects and places (Burgess, Jeffrey, &
O’Keefe, 1999; Long & Kesner, 1998). Therefore, hippocampal-
lesioned rats could, with special training, slowly develop a parietal
map to navigate to the platform. The subiculum could also play a
role in this type of learning, as it could encode a new reference
map for the environment, constructed from this limited parietal
input. Neither of these strategies produce intact learning ability,
but, with enough training, there should be some recovery of
function. If either of these components (parietal, subiculum) were
missing, however, there would be no recovery.
This prediction is confirmed in a study of the locus of ibotenic
lesions on spatial learning: Rats whose lesion included both the
hippocampus and some part of the subiculum showed no recovery
(Morris et al., 1990). This would be expected if the rat had lost not
only its sense of direction (lesioned bearing map channel) and its
305
PARALLEL MAP THEORY

ability to calculate a new position (lesioned sketch map channel)
but also the ability to retain any new learned spatial relations
(lesioned subiculum).
Lesion placement. In addition to complete hippocampal le-
sions, removing tissue from either dorsal or ventral hippocampal
regions produces different patterns of impairment. Dorsal lesions
produce a greater impairment than do ventral lesions (E. Moser,
Moser, & Andersen, 1993; M.-B. Moser & Moser, 1998; M.-B.
Moser, Moser, Forrest, Andersen, & Morris, 1995). In addition, the
degree of impairment is proportionate to the amount of tissue lost;
small dorsal lesions impaired the probe test performance, whereas
large (more than 20% of the hippocampus) lesions impaired la-
tency to the platform (E. Moser et al., 1993). We suggest that the
small dorsal lesions selectively impaired the CA1 and the sketch
map and, hence, the ability to encode the local position of the
platform. In other words, the dorsal lesion produced an rB pheno-
type, and the ventral lesion produced an rS phenotype. The dorsal
lesion (rB phenotype) allows escape latencies to remain low or
intermediate, as in the case of the pretraining effect (Bannerman et
al., 1995). Ventral lesions (rS) should have less effect, because of
the greater use of the sketch map in the water maze. With larger
lesions, however, both maps are lost, and escape latencies increase.
This prediction could be tested by correlating the loss of tissue for
each subfield (DG, CA3, or CA1) in a dorsal or ventral lesion of
different magnitudes with quantitative analyses of swim paths
produced by the rat.
Hippocampal residual learning and path integration. In our
view, path integration is not a hippocampus-dependent represen-
tation but instead is a source of data for the bearing map, providing
it with a default direction value. PMT therefore predicts that rats
with hippocampal lesions are not impaired in path integration (a
1-D representation) but instead are impaired in integrating path
integration into a spatial representation (a 2-D representation).
Thus, the key question is whether lesioned rats are impaired in path
integration or in orienting with a vector derived from path inte-
gration. This prediction is supported by evidence that hippocampal
rats are only impaired in tasks requiring a spatial representation but
not in pure path integration studies (Alyan & McNaughton, 1999).
In another study, hippocampal rats could only home through path
integration using combined self-movement, surface, and visual
cues but were unable to do so with only self-movement cues
(Maaswinkel, Jarrard, & Whishaw, 1999). Such learning, depend-
ing on the training, could be residual learning by the subiculum,
even in the absence of CA1. Hippocampal lesions also abolish the
rB phenotype, which is indispensable for orienting in a dark region
of an open field with no olfactory cues (Schenk et al., 1995).
Entorhinal Cortex
The EC occupies a special role in hippocampal function. It
projects to every subfield and has reciprocal connections with CA1
(Witter et al., 2000). It is thus an active partner for both sketch and
bearing maps, and lesions of the EC alone should reveal significant
residual capacities, similar to lesions of the DG and HP. Indeed,
like rats with hippocampal lesions (Morris et al., 1990), EC-
lesioned rats show stereotyped swim loops that become progres-
sively more likely to cross the platform location (Schenk & Morris,
1985). The actual expressed phenotype critically depends on the
task and the tissue remaining in the subiculum, if in fact there is a
residual capacity by the subiculum and parietal to develop sketch-
like maps. In fact, EC-lesioned rats do show similar residual
capacities to hippocampal-lesioned rats, in contrast to rats with a
combined EC and subiculum lesion, which show no recovery
(Schenk & Morris, 1985). This also raises the question of the
significance of the direct EC–CA1 connection (Witter, Griffioen,
Jorritsma-Byham, & Krijnen, 1988). This connection may play an
important role for sketch map reorganization, as is required in the
acquisition of learning a new place in a familiar environment.
Hippocampal Unit Activity
Neural activity in specific HPF areas, as measured by unit or
field recordings (and functional imaging, once finer spatial reso-
lution has been achieved), should be activated differentially de-
pending on the map that is being constructed or accessed for a
particular task.
Predictions
The parallel maps may be active individually at any time,
though, under most conditions, they should be active simulta-
neously (see Table 4). For example, new exploration along gradi-
ents should selectively activate the bearing map channel. In con-
trast, exploration into arrays of new positional landmarks or the
introduction of a new positional landmark should selectively acti-
vate the sketch map channel. Rearrangement of these landmarks or
other mismatch detection (O’Keefe, 1976) should also activate the
sketch map channel units. Activation of one component may then
require an update of the integrated map, which should activate the
whole assembly line (see Table 4).
An obvious test, then, is to examine the activity of place cells in
different HPF subfields in reaction to specific task environment or
cue manipulation conditions. However, spatial selectivity should
be found in sites that encode place, which could be found in a 2-D
representation in the bearing map based on directional cues, or a
2-D representation in the sketch map constructed from positional
landmarks, or in the integrated map. Spatial selectivity in itself
may therefore not distinguish easily between mapping systems.
The critical test is instead the response of units to changes in task
environment. For example, altering directional cues should acti-
vate units in the bearing map channel, and manipulating positional
landmarks should primarily activate the sketch map channel, but
having to navigate between two discontiguous sketch maps should
activate the integrated map.
There are significant obstacles to testing these predictions with
the current unit activity literature. First, as we have just discussed,
it is not yet clear precisely how subfields differentiate by spatial
specificity. Second, the location of the unit must be precisely
defined. In addition to differences among subfields, there is also a
highly organized topography of projections among subfields. Out-
put from CA3, for example, differentiates into three zones, each of
which remains segregated in CA1 (Ishizuka et al., 1990). A similar
topography is seen in perforant path projections (Witter et al.,
2000) and in the CA1 projection to the subiculum (Amaral &
Witter, 1995). Such regular topography is likely to be involved in
functional differentiation of independent channels within the hip-
pocampus (Amaral & Witter, 1995). We suspect that this topog-
raphy is related to the segregation of bearing, sketch, and inte-
306
JACOBS AND SCHENK

grated maps in the hippocampus. Hence, unit activity may very
well differ among such segregated channels.
There is also evidence that hippocampal regions vary in unit
activity. For example, place fields in ventral HPF are sparser and
less finely tuned than are those recorded in dorsal HPF (M. W.
Jung, Wiener, & McNaughton, 1994), corresponding to the differ-
entiation of dorsal–ventral function discussed earlier. A new focus
of physiological studies has been the recording of hippocampal
ensembles, with the goal of discerning topographical patterns of
activity. This has met with mixed results, perhaps because of
methodological differences among studies (Hampson, Simeral, &
Deadwyler, 1999; Redish et al., 2001). At this point, however,
there is no clear consensus on the relationship between unit loca-
tion and the pattern of activity.
Task environment is also critical in the evaluation of unit
activity in relation to PMT. Single-track environments (e.g., plus
maze, radial arm maze), favored in some studies (McNaughton et
al., 1996; Mizumori, Ragozzino, Cooper, & Leutgeb, 1999;
O’Keefe & Speakman, 1987; Redish et al., 2001), provide a test
environment that is clearly organized into directional vectors. Rats
in such an environment, particularly male rats, might naturally rely
more heavily on the bearing map. These rats might therefore
recruit a different pattern of cell ensembles than might rats navi-
gating in less directional environments, such as the round cylin-
drical arena with a single cue card, the task developed by Muller,
Kubie, and Ranck (1987; Ranck, 1973). The card in this task
environment provides a simple directional landmark and perhaps
positional landmarks from the edges of the card. Because other
directional cues are generally missing in this environment, how-
ever, unit recordings might biased toward activity in the sketch
map in the cylindrical, cue-card arena and the bearing map channel
in the radial arm maze.
Given these caveats, some unit activity studies may shed some
light on the mechanism underlying the parallel mapping systems.
We address two categories of unit activity, place units and head
direction units.
Head Direction Units
As we discussed earlier, orienting in a sketch map requires a
different type of directional information than does orienting in the
bearing map. Mapping positional landmarks requires three types of
information. Visual input on the size, shape, and aspect of the
object contributes to its identification. Vestibular information me-
diates the integration of the changes in aspect with changes in
egocentric movements. Finally, knowledge of the direction in
which the observer is looking (i.e., head direction) facilitates the
computation of relative position. Processing the relation between
different positional landmarks thus involves motion parallax and
the information coded by the head direction units. In comparison,
path integration is related to whole body translation, and, thus,
rotation should be more critical for gradient discrimination. The
neural basis of head direction units in structures that project to the
hippocampus is becoming well elucidated (Taube & Muller, 1998;
Taube, Muller, & Ranck, 1990). We have a different interpretation
of the function of these units. We propose that the head direction
system, involving the anterior dorsal thalamus, postsubiculum, and
CA1, furnishes directional data not to the bearing map but to the
sketch map. In fact, head direction and place information converge
in CA1 (Leutgeb, Ragozzino, & Mizumori, 2000), but because
head direction units are related to major directional cues, the
bearing map may play a critical role in updating the head direction
units. This may be one way the efficacy of the bearing map affects
sketch map processing.
Place Units
We summarize relevant findings from the literature by their
relationship to different maps.
Evidence for the bearing map. We have already discussed the
dissociation of CA3 and CA1 unit activity with MS inactivation,
for which MS inactivation resulted in a loss of spatial selectivity in
CA3 but not in CA1 (Mizumori et al., 1989). This is concordant
with the critical role of the MS input for the bearing map and hence
for CA3 spatial coding but not for the sketch map, which retains its
spatial specificity.
A second observation that is consistent with activity in the
bearing map is that place cells appear preconfigured: A rat intro-
duced into a novel arena immediately shows large (i.e., low reso-
lution) place fields at the beginning of exploration (Wilson &
McNaughton, 1993). These coarse fields could be evidence for the
existing bearing map. Carrying the subject among laboratory
rooms would not disrupt such a map, although the map might be
disrupted by vestibular disorientation. This would also explain
why place cells update their location using directional information
(O’Keefe & Burgess, 1996).
Evidence for the sketch map. The sketch map allows the
navigator to refine the coarse resolution of the bearing map. If CA1
is the primary area encoding new sketch maps, then it should
rapidly encode and tune positional landmarks to create the working
memory copy of the integrated map. Several studies have demon-
strated the role of the NMDA receptor, which is found in high
density in CA1, in fine tuning place fields (Kentros et al., 1998).
The knockout of the gene for a key subunit of the NMDA receptor
in CA1 affects the spatial selectivity of CA1 place cells; mice
lacking this receptor showed reduced LTP in the CA1 and had
coarser resolution of place fields (McHugh et al., 1996).
Evidence for the integrated map. A noted characteristic of
place unit studies is the diversity of place cell spatial specificity.
For example, a study of CA1 and CA3 place units categorized
three types according to their response to landmarks: those re-
sponding to distal or local landmarks, and those responding to both
types (Gothard, Skaggs, Moore, & McNaughton, 1996). We spec-
ulate that these three types correspond to units engaged in three
maps: bearing map units encoding distal (i.e., directional) land-
marks, sketch map units encoding local (i.e., positional) land-
marks, and integrated map units encoding both. If so, then there
should be some segregation of function by subfield. CA3 neurons
should specify place in relation to the directional cues, such as the
geometry of the array, whereas CA1 units could be involved in
integrated map or sketch map representations. Therefore, CA1
units could respond to both classes of cues, though we speculate
that the area most distal to CA3, which receives input from the
CA3 projection zone, should be more responsive to directional
cues than should units in the midseptotemporal regions of CA1,
which receive inputs from the CA3 associational zone.
Another finding, related to cue control, is the differential control
of place units by the landmarks according to their position in the
307
PARALLEL MAP THEORY

environment. In these studies, landmarks at the edge of an arena
exert greater control over unit activity compared with the same
landmark placed in the center of the arena (Cressant, Muller, &
Poucet, 1997, 1999; Hetherington & Shapiro, 1997). We suggest
that the central placement of landmarks, in the absence of direc-
tional cues to inform the bearing map, leads objects in the center
of the field to be processed as a sketch map. In contrast, the same
objects in the periphery provide both directional and positional
information, allowing the rat to calculate an integrated map. Fur-
thermore, sketch maps must be encoded in unique locations on the
bearing map. With centrally placed objects and, hence, no direc-
tional reference, there is a loss of control of unit response by cues.
Here, rats cannot associate the objects (i.e., the sketch map relating
these objects to one another) with a stable bearing map, and sketch
maps are not consolidated, but, instead, different sketch maps
continue to be created. This leads to the apparent low resistance to
rotation of cues.
Alternatively, rats may link their sketch map to a distant object
that provides some directional information. In this case, place
fields would be disturbed if the configuration was rotated, because
the rat expects intra-arena landmarks to maintain a certain relation
to the extramaze world. Evidence for these alternatives may be
provided by experiments using the round cylinder paradigm by
Knierim, Kudrimoti, and McNaughton (1995, 1998). Because of
the nature of the task, rats cannot link the inside of the cylinder to
the outside through self-movement and path integration. They also
cannot see distal cues. With this absence of directional cues, it is
difficult to encode the current location (and current sketch map) in
relation to the bearing map. Head direction may also be poorly
coded when directional cues are absent. Under such conditions,
sketch maps may not be coded accurately, and, consequently, units
might be less accurately controlled by cues.
Another intriguing case is offered by place cell studies using
two identical environments in which unit activity is measured as
rats explore first one, then a second environment (Barnes, Suster,
Shen, & McNaughton, 1997; Skaggs & McNaughton, 1998;
Tanila, 1999; Wilson & McNaughton, 1993). Depending on the
experimental conditions, different responses are obtained from
ensemble recordings of pyramidal cells. For example, old rats are
less likely to retrieve the same map for a given environment than
are young rats, which is interpreted as an age-related decline in
retrieving information from an intact spatial framework (Barnes et
al., 1997). Perhaps this framework could be conceptualized as the
bearing map. If so, then this is an age-related impairment in the
binding of new sketch maps to the consolidated bearing map.
Because of the sketch map’s reliance on NMDA-mediated LTP,
this leads to the prediction that the sketch and integrated maps
should be more sensitive to the processes of normal aging than is
the bearing map.
A second interesting comparison is among adult rats experienc-
ing different experimental layouts, as in studies by Skaggs and
McNaughton (1998) and Tanila (1999). In the former study, rats
moved from one identical environment (A) along a connecting
corridor to a second environment (B). Here, CA1 place cells
showed similar patterns of activity in A and B. In the latter study,
the rat moved through a door directly into the second identical
arena. Here, CA3 place cells completely remapped in B, but, when
the rat returned to A, the pattern of activity retained the distinction
between the two environments. These two experiments differ in
many ways. For example, in the latter study, the rats only saw one
environment at a time, and this may have aided them in maintain-
ing independent representations (Tanila, 1999). Yet it is intriguing
that the recordings in the study by Skaggs and McNaughton (1998)
were located in CA1, but the recording locations in the study by
Tanila (1999) were located mostly in CA3, with some units in
CA1. As Tanila pointed out, “the only ensemble . . . that showed
preserved location of firing with respect to common visual cues
was the CA1 ensemble” (p. 244). Thus, in both studies, CA1
neurons appeared agnostic to absolute location but instead re-
sponded to the immediate visual panorama, a behavior consistent
with sketch map function. Lacking information on the relative
position of each map in absolute space, CA1 could not distinguish
between them. We interpret the responses of the CA3 units, in
maintaining their absolute location, as consistent with CA3 access
to data from the bearing map. These data would allow the CA3
ensembles to differentiate the two visually identical environments
by their unique location on the bearing map. One could test this
interpretation by recording CA1 and CA3 ensembles while sys-
tematically changing cues available in the task environment. Re-
moving directional cues (both internal and external) should de-
crease the ability of CA3 units to encode a location and to
differentiate the two environments.
Finally, we have suggested that the role played by the subiculum
is to hold a reference or efferent copy of the integrated map during
the period of memory consolidation by temporal and parietal
cortices. In this sense, place units in the subiculum should be
similar to those in CA1. There are several intriguing experiments
on cue control of subiculum units that bear on this question. The
surprising finding has been that subiculum units seem to encode a
universal map that is oblivious to changes in geometry of the arena
or visual pattern (Sharp, 1999a). This, of course, differs dramati-
cally from units in HP, which are sensitive to both parameters
(Sharp, 1997). Another difference is the response to barriers within
the arena: Hippocampal place cells can encompass such barriers,
but subiculum place fields are inhibited by such boundaries (Sharp,
1999b). A possible interpretation of these results, cast in the
framework of PMT, is that this universal map or reference memory
of the integrated map encodes a location in absolute space. In
contrast to the ephemeral, working memory sketch maps of CA1,
the subiculum reference map should be resilient to changes in local
cues. The response of the subiculum to barriers may derive from its
reliance on the continuity of sight lines, to calculate trajectories
based on the bearing map. This is highly speculative, of course,
and here, just as in the hippocampus literature, a very different
picture could emerge with comparisons of units within different
projection zones of the subiculum. The question of place specific-
ity in EC is equally important, and we suggest that, like the other
major output structure to the HPF, some units in EC should reflect
the characteristics of the reference memory copy and, hence, may
be impervious to local changes in arena geometry, as observed
(Quirk, Muller, Kubie, & Ranck, 1992).
Summary and Conclusions
The parallel map theory postulates that the two major compo-
nents of the HPF, the DG and the HP, play fundamentally different
roles in spatial navigation. This duality reflects the inherently dual
nature of spatial navigation, the need for both map and compass,
308
JACOBS AND SCHENK

for both directional and positional landmarks. These two classes of
landmarks are selectively attended to and processed to construct
parallel spatial representations, the bearing map of DG and CA3,
the sketch map of CA1. These maps are both independent and
complementary, each adding their unique contribution in an
assembly-line fashion to create a flexible and powerful represen-
tation, the cognitive or integrated map.
Testing the Theory
The value of a theory lies in its ability to predict unique and
quantifiable results. Because each parallel map is mediated by
different hippocampal channels and relies differentially on mech-
anisms of synaptic plasticity, the maps should be dissociable by
standard methods. Damaging one component of either channel
impairs both the integrated map and the map mediated by that
channel; we offer recipes for testing the theory in Tables 2 and 3.
The definitive test of the theory is a double dissociation of bearing
map and sketch map channels, combined with a precise measure of
map phenotype for each experimental group.
We think ours is a plausible and testable theory. Given the
current state of knowledge of the structure and function of many
components of the HPF, however, we fully expect to revise our
predictions as future research reduces uncertainty about hippocam-
pal connections, activity, and function. At present, the complexity
of HPF structure far outstrips our ability to understand the logic of
its design. Even with the current evidence for topographic projec-
tions (Amaral & Witter, 1995; Scharfman, Witter, & Schwarcz,
2000; Witter et al., 2000), it has been difficult, perhaps impossible,
to reconcile the complex neuroanatomy of the HPF with feasible
models of its function. For example, there are currently no theo-
retical models or experimental studies of CA2, another subfield in
HP, and therefore we have not yet included this subfield in PMT.
Given this current situation, we openly acknowledge that this
hypothesis of map assembly may be correct in outline but must be
an oversimplification of a much more complex system and, hence,
often incorrect in the details.
Given the number of methods by which the predictions can be
tested, we have a significant chance of detecting errors in the
theory, either in detail or in the whole. To date, the empirical
literature is consistent with PMT, and when subfield functions
have been dissociated, the results have confirmed our predictions.
In many cases (e.g., the dissociation of LTP and learning, associ-
ation of wheel running with neurogenesis, sex differences in spa-
tial strategy), PMT offers a single mechanism for a diversity of
unexplained results.
Conclusion
“Science is built up with facts, as a house is with stones. But a
collection of facts is no more a science than a heap of stones is a
house” (Poincare´, 1913, p. 18). The unsolved question of hip-
pocampal function remains at the forefront of cognitive neuro-
science. With thousands of published studies on hippocampal
function and physiology, this literature represents a prodigious
heap of stones. Many structures have been built from these stones
to account for the hippocampus’s role in spatial navigation and
human memory. We have described here another such structure,
the parallel map theory. We believe that, in comparison with
previous theories, our structure rests on a deeper foundation, the
evolutionary history of spatial navigation and the hippocampal
formation in vertebrates. By constructing a model from the sim-
plest units of navigation, orientation to 1-D maps from distributed
stimuli, we have progressed inexorably to the conclusion that the
hippocampus must be encoding and integrating parallel mental
representations of the external environment. If our predictions
are confirmed, PMT will have profound implications for princi-
ples of spatial navigation and the function of the mammalian
hippocampus.
References
Abeliovich, A., Paylor, R., Chen, C., Kim, J. J., Wehner, J. M., &
Tonegawa, S. (1993). PKC-gamma mutant mice exhibit mild deficits in
spatial and contextual learning. Cell, 75, 1263–1271.
Akesson, S. (1996). Geomagnetic map used for long-distance navigation?
Trends in Ecology & Evolution, 11, 398–400.
Alerstam, T. (1990). Bird migration. Cambridge, England: Cambridge
University Press.
Alvarado, M. C., & Bachevalier, J. (2000). Revisiting the maturation of
medial temporal lobe memory functions in primates. Learning & Mem-
ory, 7, 244–256.
Alyan, S., & McNaughton, B. L. (1999). Hippocampectomized rats are
capable of homing by path integration. Behavioral Neuroscience, 113,
19–31.
Amaral, D. G., & Dent, J. A. (1981). Development of the mossy fibers of
the dentate gyrus: I. A light and electron microscopic study of the mossy
fibers and their expansions. Journal of Comparative Neurology, 195,
51–86.
Amaral, D. G., & Witter, M. P. (1995). Hippocampal formation. In G.
Paxinos (Ed.), The rat nervous system (2nd ed., pp. 443–493). San
Diego, CA: Academic Press.
Bannerman, D. M., Good, M. A., Butcher, S. P., Ramsay, M., & Morris,
R. G. M. (1995, November 9). Distinct components of spatial learning
revealed by prior training and NMDA receptor blockade. Nature, 378,
182–186.
Barnes, C. A. (1979). Memory deficits associated with senescence: A
neurophysiological and behavioral study in the rat. Journal of Compar-
ative and Physiological Psychology, 93, 74–104.
Barnes, C. A., Suster, M. S., Shen, J., & McNaughton, B. L. (1997, July
17). Multistability of cognitive maps in the hippocampus of old rats.
Nature, 388, 272–275.
Battistin, T., & Cherubini, E. (1994). Developmental shift from long-term
depression to long-term potentiation at the mossy fibre synapses in the
rat hippocampus. European Journal of Neuroscience, 6, 1750–1755.
Bayer, S. A. (1982). Changes in the total number of dentate granule cells
in juvenile and adult rats: A correlated volumetric and 3H-thymidine
autoradiographic study. Experimental Brain Research, 46, 315–323.
Bayer, S. A. (1985). Neuron production in the hippocampus and olfactory
bulb of the adult rat brain: Addition or replacement? In F. Nottebohm
(Ed.), Annals of the New York Academy of Sciences: Vol. 457. Hope for
a new neurology (pp. 163–172). New York: New York Academy of
Sciences.
Bayer, S. A., Yackel, J. W., & Puri, P. S. (1982, May 21). Neurons in the
rat dentate gyrus granular layer substantially increase during juvenile
and adult life. Science, 216, 890–892.
Benhamou, S., & Poucet, B. (1998). Landmark use by navigating rats
(Rattus norvegicus): Contrasting geometric and featural information.
Journal of Comparative Psychology, 112, 317–322.
Bennett, A. T. D. (1996). Do animals have cognitive maps? Journal of
Experimental Biology, 199, 219–224.
Best, P. J., & White, A. M. (1999). Placing hippocampal single-unit studies
in a historical context. Hippocampus, 9, 346–351.
309
PARALLEL MAP THEORY

Bezzi, M., Leutgeb, S., Treves, A., & Mizumori, S. J. (2000). Information
analysis of location-selective cells in hippocampus and lateral septum
[Abstract]. Society for Neuroscience Abstracts, 26, 11.
Bingman, V. P. (1990). Spatial navigation in birds. In R. P. Kesner & D. S.
Olton (Eds.), Neurobiology of comparative cognition (1st ed., pp. 423–
447). Hillsdale, NJ: Erlbaum.
Bingman, V., Bagnoli, P., Ioale´, P., & Casini, G. (1989). Behavioral and
anatomical studies of the avian hippocampus. In V. Chan-Palay & C.
Kohler (Eds.), The hippocampus: New vistas. Neurology and neurobi-
ology (Vol. 52, pp. 379–394). New York: Liss.
Birse, S. C., Leonard, R. B., & Coggeshall, R. E. (1980). Neuronal increase
in various areas of the nervous system of the guppy, Lebistes. Journal of
Comparative Neurology, 194, 291–301.
Bliss, T. V. P., & Collingridge, G. L. (1993, January 7). A synaptic model
of memory: Long-term potentiation in the hippocampus. Nature, 361,
31–39.
Bliss, T. V. P., & Lomo, T. (1973). Long-lasting potentiation of synaptic
transmission in the dentate area of the anaesthetized rabbit following
stimulation of the perforant path. Journal of Physiology, 232, 331–356.
Blozovski, D., & Hess, C. (1989). Hippocampal nicotinic cholinergic
mechanisms mediate spontaneous alternation and fear during ontogen-
esis but not later in the rat. Behavioural Brain Research, 35, 209–220.
Bovet, J. (1992). Mammals. In F. Papi (Ed.), Animal homing (pp. 321–
361). London: Chapman & Hall.
Brandner, C., & Schenk, F. (1998). Septal lesions impair the acquisition of
a cued place navigation task: Attentional or memory deficit? Neurobi-
ology of Learning and Memory, 69, 106–125.
Brown, R. W., & Whishaw, I. Q. (2000). Similarities in the development
of place and cue navigation by rats in a swimming pool. Developmental
Psychobiology, 37, 238–245.
Bruce, L. L., & Neary, T. J. (1995). The limbic system of tetrapods: A
comparative analysis of cortical and amygdalar populations. Brain,
Behavior and Evolution, 46, 224–234.
Burgess, N., Jeffery, K. J., & O’Keefe, J. (1999). Integrating hippocampal
and parietal functions: A spatial point of view. In N. Burgess, K. J.
Jeffery, & J. O’Keefe (Eds.), The hippocampal and parietal foundations
of spatial cognition (pp. 3–29). Oxford, England: Oxford University
Press.
Burgess, N., Maguire, E. A., & O’Keefe, J. (2002). The human hippocam-
pus and spatial and episodic memory, Neuron, 35, 625–641.
Butler, A. B., & Hodos, W. (1996). Comparative vertebrate anatomy:
Evolution and adaptation. New York: Wiley-Liss.
Caramanos, Z., & Shapiro, M. L. (1994). Spatial memory and N-methyl-
D-aspartate receptor antagonists APV and MK-801: Memory impair-
ments depend on familiarity with the environment, drug dose, and
training duration. Behavioral Neuroscience, 108, 30–43.
Carman, H. M., & Mactutus, C. F. (2001). Ontogeny of spatial navigation
in rats: A role for response requirements? Behavioral Neuroscience, 115,
870–879.
Cheng, K. (1986). A purely geometric module in the rat’s spatial repre-
sentation. Cognition, 23, 149–178.
Cheng, K., & Gallistel, C. R. (1984). Testing the geometric power of an
animal’s spatial representation. In H. L. Roitblat, T. G. Bever, & H. S.
Terrace (Eds.), Animal cognition (pp. 409–423). Hillsdale, NJ: Erlbaum.
Chevalley, A.-F., & Schenk, F. (1987). Immature processes of spatial
learning in hooded rats [Abstract]. Society for Neurosciences Ab-
stracts, 17, 5.
Chiba, A. A., Kesner, R. P., & Reynolds, A. M. (1994). Memory for spatial
location as a function of temporal lag in rats: Role of hippocampus and
medial prefrontal cortex. Behavioral and Neural Biology, 61, 123–131.
Cohen, N. J., & Eichenbaum, H. (1991). The theory that wouldn’t die: A
critical look at the spatial mapping theory of hippocampal function.
Hippocampus, 1, 265–268.
Cohen, N. J., & Eichenbaum, H. (1993). Memory, amnesia, and the
hippocampal system. Cambridge, MA: MIT Press.
Collett, T. S., & Baron, J. (1994, March 10). Biological compasses and the
coordinate frame of landmark memories in honeybees. Nature, 368,
137–140.
Collett, T. S., Cartwright, B. A., & Smith, B. A. (1986). Landmark learning
and visuo-spatial memories in gerbils. Journal of Comparative Physiol-
ogy: A, Sensory, Neural, and Behavioral Physiology, 158, 835–851.
Cressant, A., Muller, R. U., & Poucet, B. (1997). Failure of centrally
placed objects to control the firing fields of hippocampal place cells.
Journal of Neuroscience, 17, 2531–2542.
Cressant, A., Muller, R. U., & Poucet, B. (1999). Further study of the
control of place cell firing by intra-apparatus objects. Hippocampus, 9,
423–431.
Day, L. B., Crews, D., & Wilczynski, W. (1999). Spatial and reversal
learning in congeneric lizards with different foraging strategies. Animal
Behaviour, 57, 393–407.
Day, L. B., Crews, D., & Wilczynski, W. (2001). Effects of medial and
dorsal cortex lesions on spatial memory in lizards. Behavioural Brain
Research, 118, 27–42.
Deadwyler, S. A., & Hampson, R. E. (1999). Anatomic model of hip-
pocampal encoding of spatial information. Hippocampus, 9, 397–412.
Dobzhansky, T. (1951). Genetics and the origin of species. (3rd ed.). New
York: Columbia University Press.
Douglas, R. J., Peterson, J. J., & Douglas, D. P. (1973). The ontogeny of
a hippocampus-dependent response in two rodent species. Behavioral
Biology, 8, 27–37.
Dusek, J. A., & Eichenbaum, H. (1997). The hippocampus and memory for
orderly stimulus relations. Proceedings of the National Academy of
Sciences, USA, 94, 7109–7114.
Dusenbery, D. B. (1992). Sensory ecology. New York: Freeman.
Dyer, F. C. (1994). Spatial cognition and navigation in insects. In L. A.
Real (Ed.), Behavioral mechanisms in evolutionary ecology (pp. 66–98).
Chicago: University of Chicago Press.
Egger, G. J. (1973). Novelty induced changes in spontaneous alternation by
infant and adult rats. Developmental Psychobiology, 6, 431–435.
Eichenbaum, H., Dudchenko, P., Wood, E., Shapiro, M., & Tanila, H.
(1999). The hippocampus, memory, and place cells: Is it spatial memory
or a memory space? Neuron, 23, 209–226.
Eilam, D., & Golani, I. (1988). The ontogeny of exploratory behavior in the
house rat (Rattus rattus): The mobility gradient. Developmental Psycho-
biology, 21, 679–710.
Fernandez, A. S., Pieau, C., Reperant, J., Boncinelli, E., & Wassef, M.
(1998). Expression of the Emx-1 and Dlx-1 homeobox genes define
three molecularly distinct domains in the telencephalon of mouse, chick,
turtle and frog embryos: Implications for the evolution of telencephalic
subdivisions in amniotes. Development, 125, 2099–2111.
Font, C., Lanuza, E., Martinez-Marcos, A., Hoogland, P. V., & Martinez-
Garcia, F. (1998). Septal complex of the telencephalon of lizards: III.
Efferent connections and general discussion. Journal of Comparative
Neurology, 401, 525–548.
Frier, H. J., Edwards, E., Smith, C., Neale, S., & Collett, T. S. (1996).
Magnetic compass cues and visual pattern learning in honeybees. Jour-
nal of Experimental Biology, 199, 1353–1361.
Gaffan, D. (1991). Spatial organization of episodic memory. Hippocam-
pus, 1, 262–264.
Gallistel, C. R. (1990). The organization of learning. (1st ed.). Cambridge,
MA: MIT Press.
Garcia, R., Vouimba, R. M., & Jaffard, R. (1993). Spatial discrimination
learning induces LTP-like changes in the lateral septum of mice. Neu-
roReport, 5, 329–332.
Garcia-Verdugo, J. M., Llahi, S., Ferrer, I., & Lopez-Garcia, C. (1989).
Postnatal neurogenesis in the olfactory bulbs of a lizard. A tritiated
thymidine autoradiographic study. Neuroscience Letters, 98, 247–252.
310
JACOBS AND SCHENK

Gilbert, P., Kesner, R., & Lee, I. (2001). Dissociating hippocampal sub-
regions: A double dissociation between dentate gyrus and CA1. Hip-
pocampus, 11, 626–636.
Glick, S. D., Weaver, L. M., & Meibach, R. C. (1980, March 7). Lateral-
ization of reward in rats: Differences in reinforcing thresholds. Science,
207, 1093–1095.
Gluck, M. A., & Myers, C. E. (2001). Gateway to memory: An introduction
to neural network modeling of the hippocampus and learning. Cam-
bridge, MA: MIT Press.
Gothard, K., Skaggs, W. E., Moore, K. M., & McNaughton, B. L. (1996).
Binding of hippocampal CA1 neural activity to multiple reference
frames in a landmark-based navigation task. Journal of Neuro-
science, 16, 823–835.
Gould, E., Beylin, A., Tanapat, P., Reeves, A., & Shors, T. J. (1999).
Learning enhances adult neurogenesis in the hippocampal formation.
Nature Neuroscience, 2, 260–265.
Gould, E., Wooley, C. S., Frankfurt, M., & McEwen, B. S. (1990). Gonadal
steroids regulate dendritic spine density in hippocampal pyramidal cells
in adulthood. Journal of Neuroscience, 10, 1286–1291.
Granger, R., Wiebe, S. P., Taketani, M., & Lynch, G. (1996). Distinct
memory circuits composing the hippocampal region. Hippocampus, 6,
567–578.
Greene, C. H., & Cook, R. G. (1997). Landmark geometry and identity
controls spatial navigation in rats. Animal Learning & Behavior, 25,
312–323.
Hampson, R. E., Simeral, J. D., & Deadwyler, S. A. (1999, December 9).
Distribution of spatial and nonspatial information in dorsal hippocam-
pus. Nature, 402, 610–614.
Harvey, P. H., & Pagel, M. D. (1991). The comparative method in evolu-
tionary biology. Oxford, England: Oxford University Press.
Hasselmo, M. E. (2000). Septal modulation of theta dynamics: What is the
function of the theta rhythm? In R. Numan (Ed.), The behavioral
neuroscience of the septal region (pp. 92–114). New York: Springer.
Helbig, A. J., Berthold, P., & Wiltschko, W. (1989). Migratory orientation
of blackcaps (Sylvia atricapilla): Population-specific shifts of direction
during the autumn. Ethology, 82, 307–315.
Hetherington, P. A., & Shapiro, M. L. (1997). Hippocampal place fields are
altered by the removal of single visual cues in a distance-dependent
manner. Behavioral Neuroscience, 111, 20–34.
Holmes, O. W. (1889). The autocrat of the breakfast-table. Boston: Hough-
ton Mifflin. (Original work published 1885)
Holtzman, D. A. (1998). From slither to hither: Orientation and spatial
learning in snakes. Integrative Biology, 1, 81–89.
Holtzman, D. A., Harris, T. W., Aranguren, G., & Bostock, E. (1999).
Spatial learning of an escape task by young corn snakes, Elaphe guttata
guttata. Animal Behaviour, 57, 51–60.
Hoogland, P. V., Martinez-Garcia, F., Geneser, F. A., & Vermeulen-
Vanderzee, E. (1998). Convergence of thalamic and cholinergic projec-
tions in the “dentate area” of lizards. Brain, Behavior and Evolution, 51,
113–122.
Hoogland, P. V., & Vermeulen-Vanderzee, E. (1990). Intrinsic and extrin-
sic connections of the cerebral cortex of lizards. In W. K. Schwerdtfeger
& W. J. A. J. Smeets (Eds.), The forebrain of reptiles (pp. 20–29). Basel,
Switzerland: Karger.
Huang, Y.-Y., Kandel, E. R., Varshavsky, L., Brandon, E. P., Qi, M.,
Idzerda, R. L., et al. (1995). A genetic test of the effects of mutations in
PKA on mossy fiber LTP and its relation to spatial and contextual
learning. Cell, 83, 1211–1222.
Huxter, J. R., Thorpe, C. M., Martin, G. M., & Harley, C. W. (2001).
Spatial problem solving and hippocampal place cell firing in rats:
Control by an internal sense of direction carried across environments.
Behavioural Brain Research, 123, 37–48.
Isgor, C., & Sengelaub, D. R. (1998). Prenatal gonadal steroids affect adult
spatial behavior, CA1 and CA3 pyramidal cell morphology in rats.
Hormones and Behavior, 34, 183–198.
Ishizuka, N., Weber, J., & Amaral, D. G. (1990). Organization of intra-
hippocampal projections originating from CA3 pyramidal cells in the rat.
Journal of Comparative Neurology, 295, 580–623.
Jacobs, L. F. (1995). The ecology of spatial cognition: Adaptive patterns of
hippocampal size and space use in wild rodents. In E. Alleva, A. Fasolo,
H.-P. Lipp, & L. Nadel (Eds.), Studies of the brain in naturalistic
settings (pp. 301–322). Dordrecht, the Netherlands: Kluwer Academic.
Jacobs, L. F. (1996). Sexual selection and the brain. Trends in Ecology and
Evolution, 11, 82–86.
Jacobs, L. F., Gaulin, S. J. C., Sherry, D. F., & Hoffman, G. E. (1990).
Evolution of spatial cognition: Sex-specific patterns of spatial behavior
predict hippocampal size. Proceedings of the National Academy of
Sciences, USA, 87, 6349–6352.
Jacobs, L. F., & Liman, E. R. (1991). Grey squirrels remember the
locations of buried nuts. Animal Behaviour, 41, 103–110.
Jacobs, L. F., & Shiflett, M. W. (1999). Spatial orientation on a vertical
maze in free-ranging fox squirrels (Sciurus niger). Journal of Compar-
ative Psychology, 113, 116–127.
Jacobs, L. F., & Spencer, W. D. (1994). Natural space-use patterns and
hippocampal size in kangaroo rats. Brain, Behavior and Evolution, 44,
125–132.
Jarrard, L. E. (1983). Selective hippocampal lesions and behavior: Effects
of kainic acid lesions on performance of place and cue tasks. Behavioral
Neuroscience, 97, 873–889.
Jung, C. (1964). Approaching the unconscious. In C. G. Jung, M.-L. von
Franz, & J. Freeman (Eds.), Man and his symbols (pp. 18–103). Garden
City, NY: Doubleday.
Jung, M. W., Wiener, S. I., & McNaughton, B. L. (1994). Comparison of
spatial firing characteristics of units in dorsal and ventral hippocampus
of the rat. Journal of Neuroscience, 14, 7347–7356.
Juraska, J. M. (1984). Sex differences in developmental plasticity in the
visual cortex and hippocampal dentate gyrus. In G. J. de Vries (Ed.),
Progress in brain research (Vol. 61, pp. 205–214). Amsterdam:
Elsevier.
Juraska, J. M., Fitch, J. M., Henderson, C., & Rivers, N. (1985). Sex
difference in the dendritic branching of dentate granule cells following
differential experience. Brain Research, 333, 73–80.
Kali, S., & Dayan, P. (2000). The involvement of recurrent connections in
area CA3 in establishing the properties of place fields: A model. Journal
of Neuroscience, 20, 7463–7477.
Kaplan, H., & Hill, K. (1985). Hunting ability and reproductive success
among male Ache foragers. Current Anthropology, 26, 223–246.
Kavaliers, M., Eckel, L. A., & Ossenkopp, K. P. (1993). Brief exposure
to 60 Hz magnetic fields improves sexually dimorphic spatial learning
performance in the meadow vole, Microtus pennsylvanicus. Journal of
Comparative Physiology: A, Sensory, Neural, and Behavioral Physiol-
ogy, 173, 241–248.
Kavaliers, M., Ossenkopp, K. P., Prato, F. S., Innes, D. G. L., Galea,
L. A. M., Kinsella, D. M., & Perrot-Sinal, T. S. (1996). Spatial learning
in deer mice: Sex differences and the effects of endogenous opioids
and 60 Hz magnetic fields. Journal of Comparative Physiology: A,
Sensory, Neural, and Behavioral Physiology, 179, 715–724.
Keeton, W. T. (1974). The mystery of pigeon homing. Scientific American,
231(6), 96–107.
Kelsey, J. E., & Landry, B. A. (1988). Medial septal lesions disrupt spatial
mapping ability in rats. Behavioral Neuroscience, 102, 289–293.
Kentros, C., Hargreaves, E. L., Hawkins, R. D., Kandel, E. R., Shapiro,
M. L., & Muller, R. V. (1998, June 26). Abolition of long-term stability
of new hippocampal place cell maps by NMDA receptor blockade.
Science, 280, 2121–2126.
Kimchi, T., & Terkel, J. (2001). Magnetic compass orientation in the blind
311
PARALLEL MAP THEORY

mole rat Spalax ehrenbergi. Journal of Experimental Biology, 204,
751–758.
Kimura, D. (1999). Sex and cognition. Cambridge, MA: MIT Press.
Knierim, J. J., Kudrimoti, H. S., & McNaughton, B. L. (1995). Place cells,
head direction cells, and the learning of landmark stability. Journal of
Neuroscience, 15, 1648–1659.
Knierim, J. J., Kudrimoti, H. S., & McNaughton, B. L. (1998). Interactions
between idiothetic cues and external landmarks in the control of place
cells and head direction cells. Journal of Neurophysiology, 80, 425–446.
Langley, C. M. (1994). Spatial memory in the desert kangaroo rat (Di-
podomys deserti). Journal of Comparative Psychology, 108, 3–14.
Lavenex, P., Steele, M. A., & Jacobs, L. F. (2000). The seasonal pattern of
cell proliferation and neuron number in the dentate gyrus of wild adult
eastern grey squirrels. European Journal of Neuroscience, 12, 643–648.
Leonard, B., & McNaughton, B. L. (1990). Spatial representation in the rat:
Conceptual, behavioral, and neurophysiological perspectives. In R. P.
Kesner & D. S. Olton (Eds.), Neurobiology of comparative cognition
(pp. 363–422). Hillsdale, NJ: Erlbaum.
Leutgeb, S., Ragozzino, K. E., & Mizumori, S. J. Y. (2000). Convergence
of head direction and place information in the CA1 region of hippocam-
pus. Neuroscience, 100, 11–19.
Levine, R. L., & Bluni, T. D. (1994). Magnetic field effects on spatial
discrimination learning in mice. Physiology & Behavior, 55, 465–467.
Light, P., Salmon, M., & Lohmann, K. J. (1993). Geomagnetic orientation
of loggerhead sea turtles: Evidence for an inclination compass. Journal
of Experimental Biology, 182, 1–9.
Lohmann, K. J., Hester, J. T., & Lohmann, C. M. F. (1999). Long-distance
navigation in sea turtles. Ethology Ecology & Evolution, 11, 1–23.
Lohmann, K. J., & Johnsen, S. (2000). The neurobiology of magnetore-
ception in vertebrate animals. Trends in Neurosciences, 23, 153–159.
Lohmann, K. J., & Lohmann, C. M. F. (1996a, March 7). Detection of
magnetic field intensity by sea turtles. Nature, 380, 59–61.
Lohmann, K. J., & Lohmann, C. M. F. (1996b). Orientation and open-sea
navigation in sea turtles. Journal of Experimental Biology, 199, 73–81.
Long, J. M., & Kesner, R. P. (1998). Effects of hippocampal and parietal
cortex lesions on memory for egocentric distance and spatial location
information in rats. Behavioral Neuroscience, 112, 480–495.
Lopez, J. C., Rodriguez, F., Gomez, Y., Vargas, J. P., Broglio, C., & Salas,
C. (2000). Place and cue learning in turtles. Animal Learning & Behav-
ior, 28, 360–372.
Lopez-Garcia, C., Martinez-Guijarro, F. J., Berbel, P., & Garcia-Verdugo,
J. M. (1988). Long-spined polymorphic neurons of the medial cortex of
lizards: A Golgi, Timm, and electron-microscopic study. Journal of
Comparative Neurology, 272, 409–423.
Loy, R. (1986). Sexual dimorphism in the septohippocampal system. In
R. L. Isaacson & K. H. Pribram (Eds.), The hippocampus (Vol. 3, pp.
301–315). New York: Plenum.
Luschi, P., Hays, G. C., Del Seppia, C., Marsh, R., & Papi, F. (1998). The
navigational feats of green sea turtles migrating from Ascension Island
investigated by satellite telemetry. Proceedings of the Royal Society of
London, Series B, 265, 2279–2284.
Maaswinkel, H., Jarrard, L. E., & Whishaw, I. Q. (1999). Hippocampec-
tomized rats are impaired in homing by path integration. Hippocam-
pus, 9, 553–561.
Maaswinkel, H., & Whishaw, I. Q. (1999). Homing with locale, taxon, and
dead reckoning strategies by foraging rats: Sensory hierarchy in spatial
navigation. Behavioural Brain Research, 99, 143–152.
Madeira, M. D., & Paula-Barbosa, M. M. (1993). Reorganization of mossy
fiber synapses in male and female hypothyroid rats: A stereological
study. Journal of Comparative Neurology, 337, 334–352.
Madeira, M. D., Sousa, N., & Paula-Barbosa, M. M. (1991). Sexual
dimorphism in the mossy fiber synapses of the rat hippocampus. Exper-
imental Brain Research, 87, 537–545.
Marr, D. (1971). Simple memory: A theory for archicortex. Philosophical
Transactions of the Royal Society of London, Series B, 262, 24–81.
Martinez-Garcia, F., & Olucha, F. E. (1990). Afferent projections to the
Timm-positive cortical areas of the telencephalon of lizards. In W. K.
Schwerdtfeger & W. J. A. J. Smeets (Eds.), The forebrain of reptiles (pp.
30–40). Basel, Switzerland: Karger.
McHugh, T. J., Blum, K. I., Tsien, J. Z., Tonegawa, S., & Wilson, M. A.
(1996). Impaired hippocampal representation of space in CA1-specific
NMDAR1 knockout mice. Cell, 87, 1339–1349.
McNaughton, B. (1996, May 30). Cognitive cartography. Nature, 381,
368–369.
McNaughton, B. L., Barnes, C. A., Gerrard, J. L., Gothard, K., Jung,
M. W., Knierim, J. J., et al. (1996). Deciphering the hippocampal
polyglot: The hippocampus as a path integration system. Journal of
Experimental Biology, 199, 173–185.
McNaughton, B. L., Barnes, C. A., Meltzer, J., & Sutherland, R. J. (1989).
Hippocampal granule cells are necessary for normal spatial learning but
not for spatially-selective pyramidal cell discharge. Experimental Brain
Research, 76, 485–496.
McNaughton, B. L., & Morris, R. G. M. (1987). Hippocampal synaptic
enhancement and information storage within a distributed memory sys-
tem. Trends in Neuroscience, 10, 408–415.
Medina, L., & Reiner, A. (2000). Do birds possess homologues of mam-
malian primary visual, somatosensory and motor cortices? Trends in
Neurosciences, 23, 1–12.
Meibach, R. C., Ross, D. A., Cox, R. D., & Glick, S. D. (1980). The
ontogeny of hippocampal energy metabolism. Brain Research, 204,
431–435.
Miller, V. M., & Best, P. J. (1980). Spatial correlates of hippocampal unit
activity are altered by lesions of the fornix and entorhinal cortex. Brain
Research, 194, 311–323.
Mizumori, S. J. Y., McNaughton, B. L., Barnes, C. A., & Fox, K. B.
(1989). Preserved spatial coding in hippocampal CA1 pyramidal cells
during reversible suppression of CA3c output: Evidence for pattern
completion in hippocampus. Journal of Neuroscience, 9, 3915–3928.
Mizumori, S. J. Y., Ragozzino, K. E., Cooper, B. G., & Leutgeb, S. (1999).
Hippocampal representational organization and spatial context. Hip-
pocampus, 9, 444–451.
Moffat, S. D., Hampson, E., & Hatzipantelis, M. (1998). Navigation in a
“virtual” maze: Sex differences and correlation with psychometric mea-
sures of spatial ability in humans. Evolution and Human Behavior, 19,
73–87.
Morris, R. G. M. (1984). Development of a water-maze procedure for
studying spatial learning in the rat. Journal of Neuroscience Methods,
11, 47–60.
Morris, R. G. M. (1990). Synaptic plasticity, neural architecture, and forms
of memory. In E. James, L. McGaugh, E. Norman, M. Weinberger, J. L.
McGaugh, N. M. Weinberger, & G. Lynch (Eds.), Brain organization
and memory: Cells, systems, and circuits (pp. 52–77): New York:
Oxford University Press.
Morris, R. G. M., & Frey, U. (1997). Hippocampal synaptic plasticity: Role
in spatial learning or the automatic recording of attended experience?
Philosophical Transactions of the Royal Society of London, Series B,
352, 1489–1503.
Morris, R. G. M., & Hagen, J. J. (1983). Hippocampal electrical activity
and ballistic movement. In W. Seifert (Ed.), Neurobiology of the hip-
pocampus (pp. 321–331). New York: Academic Press.
Morris, R. G., Hagan, J. J., & Rawlins, J. N. (1986). Allocentric spatial
learning by hippocampectomised rats: A further test of the “spatial
mapping” and “working memory” theories of hippocampal function.
Quarterly Journal of Experimental Psychology: Comparative and Phys-
iological Psychology, 38(B), 365–395.
Morris, R. G. M., Schenk, F., Tweedie, F., & Jarrard, L. (1990). Ibotenic
lesions of the hippocampus and/or subiculum: Dissociating components
312
JACOBS AND SCHENK

of allocentric spatial learning. European Journal of Neuroscience, 2,
1016–1029.
Moser, E., Moser, M. B., & Andersen, P. (1993). Spatial learning impair-
ment parallels the magnitude of dorsal hippocampal lesions but is hardly
present following ventral lesions. Journal of Neuroscience, 13, 3916–
3925.
Moser, M.-B., & Moser, E. I. (1998). Functional differentiation in the
hippocampus. Hippocampus, 8, 608–619.
Moser, M.-B., Moser, E. I., Forrest, E., Andersen, P., & Morris, R. G. M.
(1995). Spatial learning with a minislab in the dorsal hippocampus.
Proceedings of the National Academy of Sciences, USA, 92, 9697–9701.
Muller, R. U., Kubie, J. L., & Ranck, J. B. (1987). Spatial firing patterns
of hippocampal complex-spike cells in a fixed environment. Journal of
Neuroscience, 7, 1935–1950.
Nadel, L., O’Keefe, J., Shapiro, M., McNaughton, B. L., & Disterhoft, J.
(1998). Symposium: Hippocampal neurons: Place cells or more [Ab-
stract]? Society for Neuroscience Abstracts, 24, 503.
Neisser, U. (1976). Cognition and reality: Principles and implications of
cognitive psychology. New York: Freeman.
Northcutt, R. G. (1995). The forebrain of gnathostomes: In search of a
morphotype. Brain, Behavior and Evolution, 46, 275–318.
Northcutt, R. G. (1996). The agnathan ark: The origin of craniate brains.
Brain, Behavior and Evolution, 48, 237–247.
Nosten-Bertrand, N., Errington, M. L., Murphy, K. P. S. J., Tokugawa, Y.,
Barboni, E., Koslova, E., et al. (1996, February 29). Normal spatial
learning despite regional inhibition of LTP in mice lacking Thy-1.
Nature, 379, 826–829.
Numan, R. (2000). Septal modulation of the working memory for voluntary
behavior. In R. Numan (Ed.), The behavioral neuroscience of the septal
region (pp. 298–326). New York: Springer.
O’Keefe, J. (1976). Place units in the hippocampus of the freely moving
rat. Experimental Neurology, 51, 78–109.
O’Keefe, J. (1993). Hippocampus, theta and spatial memory. Current
Opinion in Neurobiology, 3, 917–924.
O’Keefe, J., & Burgess, N. (1996, May 30). Geometric determinants of the
place fields of hippocampal neurons. Nature, 381, 425–428.
O’Keefe, J., & Conway, D. H. (1980). On the trail of the hippocampal
engram. Physiological Psychology, 8, 229–238.
O’Keefe, J., & Dostrovsky, J. (1971). The hippocampus as a spatial map:
Preliminary evidence from unit activity in the freely-moving rat. Brain
Research, 34, 171–175.
O’Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map.
Oxford, England: Oxford University Press.
O’Keefe, J., & Speakman, A. (1987). Single unit activity in the rat
hippocampus during a spatial memory task. Experimental Brain Re-
search, 68, 1–27.
Olcese, J. M. (1990). The neurobiology of magnetic field detection in
rodents. Progress in Neurobiology, 35, 325–330.
Ono, T., & Nishijo, H. (1999). Active spatial information processing in the
septo-hippocampal system. Hippocampus, 9, 458–466.
Papi, F. (Ed.). (1992). Animal homing. London: Chapman & Hall.
Perez-Canellas, M. M., Font, E., & Garcia-Verdugo, J. M. (1997). Post-
natal neurogenesis in the telencephalon of turtles: Evidence for nonradial
migration of new neurons from distant proliferative ventricular zones to
the olfactory bulbs. Developmental Brain Research, 101, 125–137.
Phillips, J. B., & Borland, C. (1994). Use of a specialized magnetorecep-
tion system for homing by the eastern red-spotted newt Notophthalmus
viridescens. Journal of Experimental Biology, 188, 275–291.
Poincare´, H. (1913). Science and method (F. Maitland, Trans.). London:
Thomas Nelson and Sons.
Pokorny, J., & Yamamoto, T. (1981). Postnatal ontogenesis of hippocam-
pal CA1 area in rats. I. Development of dendritic arborisation in pyra-
midal neurons. Brain Research Bulletin, 7, 113–120.
Polenov, A. L., & Chetverukhin, V. K. (1993). Ultrastructural radioauto-
graphic analysis of neurogenesis in the hypothalamus of the adult frog,
Rana temporaria, with special reference to physiological regeneration of
the preoptic nucleus: II. Types of neuronal cells produced. Cell & Tissue
Research, 271, 351–362.
Portole´s, M., Dome´nech, J. M., Martin Pe´rez, V., & Garcia Verdugo, J. M.
(1988). A Golgi and electron microscope study of the ventral septum in
the reptile, Podarcis hispa´nica. Anatomischer Anzeiger, 165, 311–326.
Poucet, B. (1989). Object exploration, habituation, and response to a spatial
change in rats following septal or medial frontal cortical damage. Be-
havioral Neuroscience, 103, 1009–1016.
Poucet, B., & Benhamou, S. (1997). The neuropsychology of spatial
cognition in the rat. Critical Reviews in Neurobiology, 11, 101–120.
Poucet, B., Chapuis, N., Durup, M., & Thinus-Blanc, C. (1986). A study of
exploratory behavior as an index of spatial knowledge in hamsters.
Animal Learning & Behavior, 14, 93–100.
Quirk, G. J., Muller, R. U., Kubie, J. L., & Ranck, J. B., Jr. (1992). The
positional firing properties of medial entorhinal neurons: Description
and comparison with hippocampal place cells. Journal of Neuroscience,
12, 1945–1963.
Ranck, J. B. (1973). Studies on single neurons in dorsal hippocampal
formation and septum in unrestrained rats: I. Behavioral correlates and
firing repertoires. Experimental Neurology, 41, 462–531.
Rawlins, J. N. (1985). Associations across time: The hippocampus as a
temporary memory store. Behavioral and Brain Sciences, 8, 479–528.
Reboreda, J. C., Clayton, N. S., & Kacelnik, A. (1996). Species and sex
differences in hippocampus size in parasitic and non-parasitic cowbirds.
NeuroReport, 7, 505–508.
Redish, A. D. (1999). Beyond the cognitive map: From place cells to
episodic memory. Cambridge, MA: MIT Press.
Redish, A. D., Battaglia, F. P., Chawla, M. K., Ekstrom, A. D., Gerrard,
J. L., Lipa, P., et al. (2001). Independence of firing correlates of
anatomically proximate hippocampal pyramidal cells. Journal of Neu-
roscience, 21, 1–6.
Redish, A. D., & Touretzky, D. S. (1997). Cognitive maps beyond the
hippocampus. Hippocampus, 7, 15–35.
Richter-Levin, G., Thomas, K. L., Hunt, S. P., & Bliss, T. V. P. (1998).
Dissociation between genes activated in long-term potentiation and in
spatial learning in the rat. Neuroscience Letters, 251, 41–44.
Rolls, E. T. (1996). A theory of hippocampal function in memory. Hip-
pocampus, 6, 601–620.
Romer, A. S. (1977). The vertebrate body (5th ed.). Philadelphia: Saunders.
Roof, R. L., & Havens, M. D. (1990). A testosterone related sexual
dimorphism in the dentate gyrus of the rat [Abstract]. Society for
Neuroscience Abstracts, 20, 328.
Roof, R. L., & Havens, M. D. (1992). Testosterone improves maze per-
formance and induces development of a male hippocampus in females.
Brain Research, 572, 310–313.
Rossier, J., Grobety, M.-C., & Schenk, F. (2000). Spatial learning by rats
across visually disconnected environments. Animal Learning & Behav-
ior, 28, 16–27.
Rossier, J., Haeberli, C., & Schenk, F. (2000). Auditory cues support place
navigation in rats when associated with a visual cue. Behavioural Brain
Research, 117, 209–214.
Rudy, J. W., Stadler-Morris, S., & Albert, P. (1987). Ontogeny of spatial
navigation behaviors in the rat: Dissociation of “proximal”-and “distal”-
cue-based behaviors. Behavioral Neuroscience, 101, 62–73.
Salazar, R., Rossier, J., & Schenk, F. (2000). Delayed maturation of spatial
mapping: Eight-week old rats are unable to integrate discrete controlled
visual cues in a configuration. European Journal of Neuroscience,
12(Suppl. 11), 86d.
Samsonovich, A., & McNaughton, B. L. (1997). Path integration and
cognitive mapping in a continuous attractor neural network model.
Journal of Neuroscience, 17, 5900–5920.
Sandstrom, N. J., Kaufman, J., & Huettel, S. A. (1998). Males and females
313
PARALLEL MAP THEORY

use different distal cues in a virtual environment navigation task. Cog-
nitive Brain Research, 6, 351–360.
Saucier, D., & Cain, D. P. (1995, November 9). Spatial learning without
NMDA receptor-dependent long-term potentiation. Nature, 378, 186–
189.
Scharfman, H. E., Witter, M. P., & Schwarcz, R. (2000). The parahip-
pocampal region: Implications for neurological and psychiatric dis-
eases. New York: New York Academy of Sciences.
Schenk, F. (1989). A homing procedure for studying spatial memory in
immature and adult rodents. Journal of Neuroscience Methods, 26,
249–258.
Schenk, F., & Brandner, C. (1995). Enhanced visuspatial memory follow-
ing pre and postnatal choline treatment. Psychobiology, 23, 302–312.
Schenk, F., Grobety, M. C., Lavenex, P., & Lipp, H.-P. (1995). Dissocia-
tion between basic components of spatial memory in rats. In E. Alleva,
A. Fasolo, H.-P. Lipp, & L. Nadel (Eds.), Behavioural brain research in
naturalistic and semi-naturalistic settings (Vol. 82, pp. 207–223). Dor-
drecht, the Netherlands: Kluwer Academic.
Schenk, F., Jacobs, L. F., Rossier, J., & Kiraly, M. (2001). Steps in the
development of spatial cognition in rats: Evidence for a dual mapping
system [Abstract]. Society for Neuroscience Abstracts, 27, 1436.
Schenk, F., & Morris, R. G. M. (1985). Dissociation between components
of spatial memory in rats after recovery from the effects of retrohip-
pocampal lesions. Experimental Brain Research, 58, 11–28.
Scho¨ne, H. (1984). Spatial orientation: The spatial control of behavior in
animals and man. Princeton, NJ: Princeton University Press.
Schwerdtfeger, W. K. (1984). Structure and fiber connections of the
hippocampus. Berlin: Springer-Verlag.
Shapiro, M. L., & O’Connor, C. (1992). N-methyl-D-aspartate receptor
antagonist MK-801 and spatial memory representation: Working mem-
ory is impaired in an unfamiliar environment but not in a familiar
environment. Behavioral Neuroscience, 106, 604–612.
Shapiro, M. L., & Olton, D. S. (1994). Hippocampal function and inter-
ference. In D. L. Schacter & E. Tulving (Eds.), Memory systems 1994
(pp. 87–117). Cambridge, MA: MIT Press.
Shapiro, M. L., Simon, D. K., Olton, D. S., Gage, F. H. D., Nilsson, O., &
Bjo¨rklund, A. (1989). Intrahippocampal grafts of fetal basal forebrain
tissue alter place fields in the hippocampus of rats with fimbria-fornix
lesions. Neuroscience, 32, 1–18.
Sharp, P. E. (1997). Subicular cells generate similar spatial firing patterns
in two geometrically and visually distinctive environments: Comparison
with hippocampal place cells. Behavioural Brain Research, 85, 71–92.
Sharp, P. E. (1999a). Complimentary roles for hippocampal versus subic-
ular/entorhinal place cells in coding place, context, and events. Hip-
pocampus, 9, 432–443.
Sharp, P. E. (1999b). Subicular place cells expand or contract their spatial
firing pattern to fit the size of the environment in an open field but not
in the presence of barriers: Comparison with hippocampal place cells.
Behavioral Neuroscience, 113, 643–662.
Shatz, C. J. (1992). The developing brain. Scientific American, 267, 61–67.
Shepherd, G. M. (1994). Neurobiology (3rd ed.). New York: Oxford
University Press.
Sherry, D. F., Forbes, M. R. L., Khurgel, M., & Ivy, G. O. (1993). Greater
hippocampal size in females of the brood parasitic brown-headed cow-
bird. Proceedings of the National Academy of Sciences, USA, 90, 7839–
7843.
Sherry, D. F., & Hampson, E. (1997). Evolution and the hormonal control
of sexually-dimorphic spatial abilities in humans. Trends in Cognitive
Science, 1, 50–56.
Sherry, D. F., Jacobs, L. F., & Gaulin, S. J. C. (1992). Adaptive special-
ization of the hippocampus. Trends in Neurosciences, 15, 298–303.
Sherry, D. F., & Vaccarino, A. L. (1989). Hippocampus and memory for
food caches in black-capped chickadees. Behavioral Neuroscience, 103,
308–318.
Sherry, D. F., Vaccarino, A. L., Buckenham, K., & Herz, R. S. (1989). The
hippocampal complex of food-storing birds. Brain, Behavior and Evo-
lution, 34, 308–317.
Silva, A. J., Paylor, R., Wehner, J. M., & Tonegawa, S. (1992, July 10).
Impaired spatial learning in alpha-calcium-calmodulin kinase II mutant
mice. Science, 257, 206–211.
Skaggs, W. E., & McNaughton, B. L. (1998). Spatial firing properties of
hippocampal CA1 populations in an environment containing two visu-
ally identical regions. Journal of Neuroscience, 18, 8455–8466.
Squire, L. R. (1992). Memory and the hippocampus: A synthesis from
findings with rats, monkeys, and humans. Psychological Review, 99,
195–231.
Squire, L. R., & Zola-Morgan, S. (1991, September 20). The medial
temporal lobe memory system. Science, 253, 1380–1386.
Strasser, R., & Bingman, V. P. (1997). Goal recognition and hippocampal
formation in the homing pigeon (Columba livia). Behavioral Neuro-
science, 111, 1245–1256.
Striedter, G. F. (1997). The telencephalon of tetrapods in evolution. Brain,
Behavior and Evolution, 49, 179–213.
Suzuki, S., Augerinos, G., & Black, A. H. (1980). Stimulus control of
spatial behavior on the eight-arm maze in rats. Learning and Motivation,
11, 1–18.
Swann, J. W., Brady, R. J., & Martin, D. L. (1989). Postnatal development
of GABA-mediated synaptic inhibition in rat hippocampus. Neuro-
science, 28, 551–561.
Swanson, L. W., Ko¨hler, C., & Bjo¨rklund, A. (1987). The limbic region. I:
The septohippocampal system. In A. Bjo¨rklund, T. Ho¨kfelt, & L. W.
Swanson (Eds.), Handbook of chemical neuroanatomy, integrated sys-
tems of the CNS, Part I (Vol. 5, pp. 125–136). Amsterdam: Elsevier.
Swanson, L. W., & Risold, P.-Y. (2000). On the basic architecture of the
septal region. In R. Numan (Ed.), The behavioral neuroscience of the
septal region (pp. 1–14). New York: Springer.
Swanson, L. W., Sawchenko, P. E., & Cowan, W. M. (1980). Evidence that
the commissural, associational and septal projections of the region
inferior of the hippocampus arise from the same neurons. Brain Re-
search, 197, 207–212.
Szekely, A. D. (1999). The avian hippocampal formation: Subdivisions and
connectivity. Behavioural Brain Research, 98, 219–225.
Tanila, H. (1999). Hippocampal place cells can develop distinct represen-
tations of two visually identical environments. Hippocampus, 9, 235–
246.
Taube, J. S., & Muller, R. U. (1998). Comparisons of head direction cell
activity in the postsubiculum and anterior thalamus of freely moving
rats. Hippocampus, 8, 87–108.
Taube, J. S., Muller, R. U., & Ranck, J. B. (1990). Head-direction cells
recorded from the postsubiculum in freely-moving rats. Journal of
Neuroscience, 10, 420–435.
Teng, E., & Squire, L. R. (1999, August 12). Memory for places learned
long ago is intact after hippocampal damage. Nature (London), 400,
675–677.
Teyler, T. J., & DiScenna, P. (1986). The hippocampal memory indexing
theory. Behavioral Neuroscience, 100, 147–154.
Tinbergen, N. (1963). On the aims and methods of ethology. Zeitschrift fur
Tierpsychologie, 20, 410–433.
Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological
Review, 55, 189–208.
Vander Wall, S. B. (1982). An experimental analysis of cache recovery in
Clark’s nutcracker. Animal Behaviour, 30, 84–94.
Vanderwolf, C. H. (1969). Hippocampal electrical activity and voluntary
movement in the rat. Electroencephalography and Clinical Neurophys-
iology, 26, 407–418.
van Praag, H., Christie, B. R., Sejnowski, T. J., & Gage, F. H. (1999).
Running enhances neurogenesis, learning, and long-term potentiation in
314
JACOBS AND SCHENK

mice. Proceedings of the National Academy of Sciences, USA, 96,
13427–13431.
van Praag, H., Kempermann, G., & Gage, F. H. (1999). Running increases
cell proliferation and neurogenesis in the adult mouse dentate gyrus.
Nature Neuroscience, 2, 266–270.
Vargha-Khadem, F., Gadian, D. G., Watkins, K. E., Connelly, A., Van
Paesschen, W., & Mishkin, M. (1997, July 18). Differential effects of
early hippocampal pathology on episodic and semantic memory. Sci-
ence, 277, 376–380.
Walker, M. M., Diebel, C. E., Haugh, C. V., Pankhurst, P. M., Montgom-
ery, J. C., & Green, C. R. (1997, November 27). Structure and function
of the vertebrate magnetic sense. Nature, 390, 371–376.
Wallraff, H. G. (1996). Seven theses on pigeon homing deduced from
empirical findings. Journal of Experimental Biology, 199, 105–111.
Wan, H., Aggleton, J. P., & Brown, M. W. (1999). Different contributions
of the hippocampus and perirhinal cortex to recognition memory. Jour-
nal of Neuroscience, 19, 1142–1148.
Warren, S. G., Humphreys, A. G., Juraska, J. M., & Greenough, W. T.
(1995). LTP varies across the estrous cycle: Enhanced synaptic plasticity
in proestrus rats. Brain Research, 703, 26–30.
Waters, N. S., Klintsova, A. Y., & Foster, T. C. (1997). Insensitivity of the
hippocampus to environmental stimulation during postnatal develop-
ment. Journal of Neuroscience, 17, 7967–7973.
Wehner, R., Michel, B., & Antonsen, P. (1996). Visual navigation in
insects: coupling of egocentric and geocentric information. Journal of
Experimental Biology, 199, 129–140.
West, M. J. (1990). Stereological studies of the hippocampus: A compar-
ison of the hippocampal subdivisions of diverse species including hedge-
hogs, laboratory rodents, wild mice and men. In J. Storm-Mathisen, J.
Zimmer, & O. P. Ottersen (Eds.), Progress in brain research (Vol. 83,
pp. 13–36). Amsterdam: Elsevier.
West, M. J., & Schwerdtfeger, W. K. (1985). An allometric study of
hippocampal components: A comparative study of the brains of the
European hedgehog, the tree shrew, and the marmoset monkey. Brain
Behavior and Evolution, 27, 93–105.
Whishaw, I. Q. (2000). The septohippocampal system and path integration.
In R. Nunan (Ed.), The behavioral neuroscience of the septal region (pp.
270–297). New York: Springer.
Whishaw, I. Q., Cassel, J.-C., & Jarrard, L. E. (1995). Rats with fimbria-
fornix lesions display a place response in a swimming pool: A dissoci-
ation between getting there and knowing where. Journal of Neuro-
science, 15, 5779–5788.
Whishaw, I. Q., & Gorny, B. (1999). Path integration absent in scent-
tracking fimbria-fornix rats: Evidence for hippocampal involvement in
“sense of direction” and “sense of distance” using self-movement cues.
Journal of Neuroscience, 19, 4662–4673.
Whishaw, I. Q., & Jarrard, L. E. (1995). Similarities vs. differences in place
learning and circadian activity in rats after fimbria-fornix section or
ibotenate removal of hippocampal cells. Hippocampus, 5, 595–604.
Whishaw, I. Q., & Tomie, J. A. (1997). Perseveration on place reversals in
spatial swimming pool tasks: Further evidence for place learning in
hippocampal rats. Hippocampus, 7, 361–370.
Whishaw, I. Q., & Vanderwolf, C. H. (1973). Hippocampal EEG and
behavior: Changes in amplitude and frequency of RSA (theta rhythm)
associated with spontaneous and learned movement patterns in rats and
cats. Behavioral Biology, 8, 461–484.
Williams, C. L., Barnett, A. M., & Meck, W. H. (1990). Organizational
effects of early gonadal secretions on sexual differentiation in spatial
memory. Behavioral Neuroscience, 104, 84–97.
Williams, C. L., & Meck, W. H. (1991). The organizational effects of
gonadal steroids on sexually dimorphic spatial ability. Psychoneuroen-
docrinology, 16, 155–176.
Williams, C. L., & Meck, W. H. (1993). Organizational effects of gonadal
hormones induce qualitative differences in visuospatial navigation. In
M. Haug, R. E. Whalen, C. Aron, & K. L. Olsen (Eds.), The development
of sex differences and similarities in behavior (pp. 179–185). Dordrecht,
the Netherlands: Kluwer Academic.
Wilson, M. A., & McNaughton, B. L. (1993, August 20). Dynamics of the
hippocampal ensemble code for space. Science, 261, 1055–1058.
Wiltschko, W., & Wiltschko, R. (1996). Magnetic orientation in birds.
Journal of Experimental Biology, 199, 29–38.
Wimer, R. E., & Wimer, C. (1985). Three sex dimorphisms in the granule
cell layer of the hippocampus in house mice. Brain Research, 328,
105–109.
Witter, M. P., Griffioen, A. W., Jorritsma-Byham, B., & Krijnen, J. L. M.
(1988). Entorhinal projections to the hippocampal CA1 region in the rat:
An underestimated pathway. Neuroscience Letters, 85, 193–198.
Witter, M. P., Wouterlood, F. G., Naber, P. A., & van Haeften, T. (2000).
Anatomical organization of the parahippocampal-hippocampal network.
In H. E. Scharfman, M. Witter, & R. Schwarcz (Eds.), The parahip-
pocampal region: Implications for neurological and psychiatric diseases
(Vol. 911, pp. 1–24). New York: New York Academy of Sciences.
Wood, E. R., Dudchenko, P. A., & Eichenbaum, H. (1997). Nonspatial
correlates of hippocampal cellular activity in rats performing an olfac-
tory continuous nonmatching to sample task on an open field [Abstract].
Society for Neuroscience Abstracts, 23, 502.
Worden, R. (1992). Navigation by fragment fitting: A theory of hippocam-
pal function. Hippocampus, 2, 165–188.
Xu, L., Anwyl, R., & Rowan, M. J. (1998, August 27). Spatial exploration
induces a persistent reversal of long-term potentiation in rat hippocam-
pus. Nature, 394, 891–894.
Received February 20, 2001
Revision received March 14, 2002
Accepted April 7, 2002 䡲
315
PARALLEL MAP THEORY