Content uploaded by William Warren
Author content
All content in this area was uploaded by William Warren on Sep 02, 2016
Content may be subject to copyright.
Active and Passive Spatial Learning in Human Navigation:
Acquisition of Graph Knowledge
Elizabeth R. Chrastil and William H. Warren
Brown University
It is known that active exploration of a new environment leads to better spatial learning than does passive
visual exposure. We ask whether specific components of active learning differentially contribute to
particular forms of spatial knowledge—the exploration-specific learning hypothesis. Previously, we
found that idiothetic information during walking is the primary active contributor to metric survey
knowledge (Chrastil & Warren, 2013). In this study, we test the contributions of 3 components to
topological graph and route knowledge: visual information, idiothetic information, and cognitive decision
making. Four groups of participants learned the locations of 8 objects in a virtual hedge maze by (a)
walking or (b) watching a video, crossed with (1) either making decisions about their path or (2) being
guided through the maze. Route and graph knowledge were assessed by walking in the maze corridors
from a starting object to the remembered location of a test object, with frequent detours. Decision making
during exploration significantly contributed to subsequent route finding in the walking condition,
whereas idiothetic information did not. Participants took novel routes and the metrically shortest routes
on the majority of both direct and barrier trials, indicating that labeled graph knowledge—not merely
route knowledge—was acquired. We conclude that, consistent with the exploration-specific learning
hypothesis, decision making is the primary component of active learning for the acquisition of topolog-
ical graph knowledge, whereas idiothetic information is the primary component for metric survey
knowledge.
Keywords: navigation, spatial cognition, proprioception, decision making
Imagine the following situation: You have been invited to a
party at a house you have never been before, so a friend offers to
drive you there. But after a nice evening, you find that your friend
has had too much to drink, and now you have to drive home. What
should you have done differently to make it easier to find the way
back home? Perhaps you should have driven the car to the party
while your friend gave you directions or found the route to the
party yourself. Or maybe you should have walked instead of
driving. It seems obvious that actively navigating in a new envi-
ronment would facilitate spatial learning. But these scenarios sug-
gest that multiple factors might contribute to active spatial learning
and factors may have different roles to play.
In the present article, we investigate the exploration-specific
learning hypothesis, which posits that specific components of
active learning during exploration differentially contribute to par-
ticular forms of spatial knowledge. Previously, we found that
idiothetic (specifically motor and proprioceptive) information
from walking was the primary contributor to metric survey knowl-
edge, whereas decision making during exploration played no role
(Chrastil & Warren, 2013). Presently, we examine their contribu-
tions to topological graph and route knowledge. Finally, a planned
companion article will report on the contribution of active atten-
tion to both types of knowledge (Chrastil & Warren, 2014a).
Components of Active Learning
First, consider possible factors that may contribute to passive
and active learning as one explores a new environment. Passive
learning is based on visual information about the layout of the environ-
ment an d t he path of self-motion that is available at the eye of
the observer, such as monocular perspective, binocular dispar-
ity, the sequence of views, the pattern of optic flow, and so on.
We define passive viewing as exposure to such visual informa-
tion alone.
Active spatial learning might be based on some combination of
one or more of the following six components (see Chrastil &
Warren, 2012): (a) efferent motor commands that determine the
path of locomotion, (b) proprioceptive information about displace-
ment with respect to the substrate (a and b together are known as
podokinetic information,Weber et al., 1998), (c) vestibular infor-
mation about head movement in an inertial frame (a– c are collec-
tively referred to as idiothetic information,Mittelstaedt & Mittel-
staedt, 2001), (d) cognitive decision making about the direction of
travel or the selected route, (e) the allocation of attention to
relevant spatial properties of the environment, and (f) mental
This article was published Online First November 24, 2014.
Elizabeth R. Chrastil and William H. Warren, Cognitive, Linguistic &
Psychological Sciences, Brown University.
This research was supported by National Science Foundation Awards
BCS-0214383 and BCS-0843940. The authors thank Henry Harrison,
Michael Fitzgerald, Joost de Nijs, Kurt Spindler, and members of the
VENLab for assistance with the research, and Mike Tarr, Rebecca Burwell,
and David Badre for their helpful comments. Portions of this article were
included as part of Elizabeth R. Chrastil’s doctoral dissertation.
Correspondence concerning this article should be addressed to Elizabeth
R. Chrastil, who is now at Boston University, Department of Psychological
& Brain Sciences and Center for Memory and Brain, 2 Cummington Mall,
Boston, MA 02215. E-mail: chrastil@bu.edu
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Experimental Psychology:
Learning, Memory, and Cognition
© 2014 American Psychological Association
2015, Vol. 41, No. 4, 1162–1178
0278-7393/15/$12.00 http://dx.doi.org/10.1037/xlm0000082
1162
manipulation of spatial information. In this study, we examine the
contributions of passive viewing, idiothetic information, and de-
cision making, with the aim of identifying their role in spatial
learning. In particular, the type of spatial knowledge that is ac-
quired may depend on which of these components are present
during active exploration.
Forms of Spatial Knowledge
Next, consider the forms of spatial knowledge that appear to
underlie human navigation. Successful navigation between known
locations might involve place recognition, reliance on beacons or
landmarks, knowledge of routes or pathways between places, and
survey knowledge of their spatial layout (Trullier, Wiener,
Berthoz, & Meyer, 1997;Siegel & White, 1975;Wiener, Buchner,
& Holscher, 2009). Beacons and landmarks are prominent features
in the environment that act as place markers (beacons) or form a
configuration that specifies a location (landmarks). Route knowl-
edge consists of a series of place–action associations, detailing a
sequence of turns at each identifiable location or decision point
(Siegel & White, 1975). Survey knowledge is configural map-like
knowledge that includes metric distances and directions between
locations in an environment. A cognitive map is often thought of
as geometrically consistent survey knowledge, such that places are
localized in a common coordinate system or “global metric em-
bedding” (Thrun, 2008).
We distinguish graph knowledge from both route knowledge,
which is somewhat weaker, and survey knowledge, which is
somewhat stronger. A topological graph structure consists of a
network of nodes linked by edges that have been traveled by a
navigator. In a “place graph” of an environment, the nodes corre-
spond to known places (including junctions), and the edges to the
known paths between them, so graph knowledge would express the
known connectivity of the environment (Byrne, 1979;Chown,
Kaplan, & Kortenkamp, 1995;Kuipers, Tecuci, & Stankiewicz,
2003;Meilinger, 2008;Trullier et al., 1997;Werner, Krieg-
Bruckner, & Herrmann, 2000). In contrast, route knowledge
merely consists of a sequence of place–action associations (Siegel
& White, 1975), and unlike a graph does not express multiple
paths intersecting at a place or multiple routes between two places.
Thus, detours present a challenge for route knowledge but can be
negotiated on the basis of graph knowledge by recombining fa-
miliar path segments into a novel route. On the other hand, survey
knowledge contains metric information about distances and direc-
tions between environmental locations, enabling novel shortcuts
(Gallistel, 1990), whereas a purely topological graph only repre-
sents connectivity. A labeled graph may incorporate local metric
information about distances between known places (edge weights)
and/or angles between known paths (node labels) in this topolog-
ical structure. Note that such local information can be very rough
and biased, and a labeled graph can be globally inconsistent,
distinguishing it from survey knowledge with a global metric
embedding.
To illustrate the distinctions between these levels of knowledge,
consider Figure 1, which shows the layout of the maze used in the
present experiment. Suppose that during the course of exploration,
a navigator had only ever traveled from the sink to the well by
going down the hallway on the right of the figure, past the clock.
If, when asked to go directly from the sink to the well, the
Figure 1. A: Outline of the maze used in the experiment. The maze included eight objects (circles, clockwise
from lower right): bookcase, well, rabbit, snowman, gear, sink, earth, clock. There were also four paintings
(rectangles) in the hallways that acted as landmarks. Participants never saw this overhead view of the maze. B:
Views from inside the maze, from the participant’s perspective. Top panel: View of one of the hallways,
including a painting. Bottom panel: View of one of the objects in the maze, the well. See the online article for
the color version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1163
ACTIVE AND PASSIVE NAVIGATION
navigator followed the same path again, then they would be
displaying route knowledge. However, if the navigator had learned
the connections between locations in the maze during exploration,
they might take a novel route to the well by traveling down the
hallway on the left, past the gear and snowman. This would imply
they had acquired graph knowledge of the maze. Finally, if the
navigator was able to make an accurate straight-line shortcut
through the hedges from the sink to the well, that would be
evidence of survey knowledge.
Survey, graph, route, and landmark knowledge form a logical
hierarchy, in which each level encompasses the levels below it.
Thus, with complete survey knowledge of the places and paths in
the environment, a navigator could also exhibit behavior consistent
with graph, route, or landmark knowledge, whereas the converse
does not necessarily hold. However, a labeled graph would not
only support efficient route-finding and novel detours, but would
also permit approximate shortcuts via integration of local infor-
mation, without the difficulties of forming a consistent global
metric embedding. Recent evidence suggests that much of human
navigation may rely on knowledge consistent with a labeled graph
(Chrastil & Warren, 2014b); presently, we focus on the basis for
learning route and graph knowledge during exploration of a new
environment.
The Exploration-Specific Learning Hypothesis
There are good reasons to expect that the components of active
learning during exploration may differentially affect the acquisi-
tion of these forms of spatial knowledge. For example, survey
knowledge depends on information about metric distances and
directions. The idiothetic systems register information related to
distances traveled and angles turned along a traversed path (Israel
& Warren, 2005), which could be used to build up such survey
knowledge (Gallistel, 1990;McNaughton, Battaglia, Jensen,
Moser, & Moser, 2006). Indeed, previous evidence supports a role
for idiothetic information in survey learning, particularly in large
or complex environments (Chance, Gaunet, Beall, & Loomis,
1998;Ruddle, Volkova, & Bülthoff, 2011a;Waller & Greenauer,
2007;Waller, Loomis, & Haun, 2004). We recently found that
idiothetic information, specifically podokinetic (motor/propriocep-
tive) information during walking, significantly contributes to sur-
vey knowledge, whereas decision making about one’s route does
not (Chrastil & Warren, 2013). The purpose of the present exper-
iment was to examine the contributions of idiothetic information
and decision making to route and graph knowledge.
Our first hypothesis is that active decision making about one’s
path of travel during exploration should significantly contribute to
route and graph knowledge, over and above passive visual expo-
sure. Given that route knowledge is believed to consist of a
sequence of turns at recognized locations, making decisions about
where and how to turn should facilitate the formation of such
place–action associations. Further, given that graph knowledge
consists of a network of paths connecting places, making decisions
about one’s path of travel should help build up such a graph.
One possible mechanism for the influence of decision making is
that it directs the navigator’s attention to the relevant aspects of the
environment, facilitating learning. Another possible mechanism is
prediction learning. On this account, a decision to turn during
exploration generates an expectation about the outcome of the
action, based on a forward model (e.g., a place–action sequence or
graph); the expected and actual outcomes are compared, and the
prediction error either reinforces or revises the model. Structures
within the medial temporal lobe have been implicated both in
spatial navigation and in prospective memory, such that memory
of previous events yields better prediction of future events (Bar,
2007;Buckner, 2010;Schacter, Addis, & Buckner, 2007), which
could contribute to graph learning. Prediction learning might also
extend to survey knowledge, whereby predictions about the dis-
tance and direction between locations are compared with the
resulting idiothetic information, reinforcing metric survey knowl-
edge. However, we found no experimental support for a contribu-
tion of decision making to survey learning (Chrastil & Warren,
2013).
Our second hypothesis is that, although idiothetic information
plays a significant role in the acquisition of survey knowledge, it
is not necessary for the acquisition of route or graph knowledge.
Given that a purely topological graph contains only connectivity
relations, without metric distances and directions, there is no
reason to expect that idiothetic information is needed to build up
graph knowledge. On the other hand, one can see how idiothetic
information might augment such route or graph knowledge. Place–
action associations could incorporate information about the angle
to turn and the distance to the next place. A labeled graph could
incorporate node labels derived from idiothetic information about
angles turned and edge weights derived from idiothetic informa-
tion about distances traveled. Passive vision might also provide
local metric information about distances and angles (Bertin, Israel,
& Lappe, 2000;Frenz, Bremmer, & Lappe, 2003), although visual
space perception is subject to significant affine distortions (Koen-
derink, van Doorn, & Lappin, 2000;Loomis, Da Silva, Fujita, &
Fukusima, 1992;Norman, Crabtree, Clayton, & Norman, 2005). A
labeled graph, even with rough local metric information, would be
sufficient for the selection of shorter routes, as well as for approx-
imate shortcuts, without a global metric embedding. We thus
investigate whether idiothetic information and decision making
contribute to route and graph learning, over and above passive
visual information.
Active and Passive Graph Learning
Previous research on active and passive spatial learning has
yielded highly inconsistent results. First, most previous research
has focused on the acquisition of survey knowledge, not route or
graph knowledge, and experiments that examined the latter have
primarily focused on the role of sensory information, not decision
making or attention. Thus, the relevant research is quite limited.
Second, the research that has investigated decision making has
tended to use a desktop virtual reality (VR) paradigm, which
removes normal idiothetic information. As reviewed in more detail
in Chrastil and Warren (2012), some experiments suggest that
making decisions contributes to graph learning, whereas some find
that idiothetic information is important for route learning, whereas
other research has yielded mixed results.
Initial efforts to distinguish active and passive graph learning in
desktop VR were inconclusive. On the one hand, Péruch, Vercher,
and Gauthier (1995; see also Tan, Gergle, Scupelli, & Pausch,
2006) reported better performance navigating between learned
places after active exploration with a joystick than after passive
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1164 CHRASTIL AND WARREN
viewing of a video. In contrast, Wilson, Foreman, Gillett, and
Stanton (1997; see also Wilson, 1999) reported no active advan-
tage for either joystick control or decision making during explo-
ration, using the same navigation task. When Wilson and Péruch
(2002) teamed up to reconcile their differences, they found some-
thing else again: Not only was there no active advantage, but
passive participants who had watched active participants explore
were more accurate than the active participants themselves. Taken
together, these results suggest that any effect of active decision
making and control on graph knowledge is small and susceptible to
minor differences in procedure.
Subsequent research suggests an active contribution to route and
graph knowledge in desktop VR. Farrell et al. (2003) reported that
participants who made decisions with a keyboard to find a se-
quence of targets in a virtual environment made fewer errors when
transferred to a real environment than control participants with no
prior exposure to the environment. In contrast, passively watching
a video of the route between targets did not yield such an improve-
ment. However, the video group only viewed the shortest route,
whereas the keyboard group explored widely for a significantly
longer time, so additional exposure to the environment might have
been a contributing factor. Recently, Voss, Gonsalves, Federmeir,
Tranel, and Cohen (2011) had participants explore a 2D grid array
of hidden objects on a computer screen either by using a mouse to
decide which objects to view or by passively watching a video.
They reported that volitional control and decision making facili-
tated later object recognition as well as recall of the spatial location
of the objects, but it is unclear whether this finding translates to a
navigational context. In sum, there is little evidence of a contri-
bution of decision making to route or graph learning. It is possible,
however, that the lack of a consistent active advantage is due to the
absence of idiothetic information in desktop VR.
Studies of route or graph learning during actual walking are
surprisingly rare. Hazen (1982) reported that children who freely
explored a real playhouse were better at reversing routes and
finding novel shortcuts than children who were led around or
carried by their parents. This result suggests that making decisions
during exploration improves children’s route and survey learning,
at least when accompanied by idiothetic information. In contrast, it
implies that idiothetic information by itself does not significantly
contribute to either.
Other research suggests that idiothetic information directly con-
tributes to route knowledge. Grant and Magee (1998) guided
participants on a prescribed route during learning, eliminating
decision making, and subsequently tested them in a real environ-
ment. Participants who walked through the real environment dur-
ing learning were faster to find locations than participants who
walked in place in a matched virtual environment, as well as those
who used a joystick in the virtual environment. Walking in place
yielded shorter paths during the test than did the joystick alone.
These results imply that idiothetic information plays a role in
graph learning, but the real-environment group also had a larger
field of view and free head movements, which might have con-
tributed to their better performance. Von Stülpnagel and Steffens
(2013) also found evidence for a motor contribution to route
learning in desktop VR; participants who controlled a mouse and
keyboard while following verbal directions for a route were better
able to retrace the route than participants who passively observed
the keyboard movements and the video.
In work closely related to the present research, Ruddle and his
colleagues investigated the contribution of idiothetic information
to route, graph, and survey knowledge, using a virtual marketplace
laid out on a grid plan and presented in a (HMD). First, in a study
of route learning, Ruddle, Volkova, Mohler, and Bülthoff (2011b)
found that participants who walked a route in a small environment
(10m ⫻14m) had fewer errors retracing and repeating the route
than participants who rotated on foot but used a joystick to trans-
late. This result suggests that idiothetic information about distance
contributes to route knowledge. Second, in a study of graph and
survey learning, Ruddle et al. (2011a) had participants learn the
locations of four target objects during free exploration, then walk
to each target and estimate the distances and directions to the
others during the test phase. Participants who walked during ex-
ploration traveled shorter routes during test than did groups who
used various combinations of walking and joystick control during
exploration—at least in a small environment (10m ⫻7m), but not
in a large environment (45m ⫻25m). In contrast, distance and
direction estimates showed the opposite effect: Idiothetic informa-
tion contributed more to survey knowledge of the large environ-
ment. Taken together, these results suggest that idiothetic infor-
mation may contribute to route and graph knowledge of small
environments but not large environments. However, this research
did not dissociate the effects of idiothetic information from those
of decision making during free exploration.
In sum, few prior experiments have tested the influence of
decision making and idiothetic information on the acquisition of
route and graph knowledge, and the existing results are mixed.
Early research was inconclusive (Péruch et al., 1995;Wilson &
Péruch, 2002;Wilson, 1999;Wilson et al., 1997): Some experi-
ments have found that decision making contributes to graph
knowledge, but they did not dissociate idiothetic information (Far-
rell et al., 2003;Hazen, 1982;von Stülpnagel & Steffens, 2013;
Tan, Gergle, Scupelli, & Pausch, 2006), and others have found that
idiothetic information contributes to route knowledge, but they did
not dissociate decision making (Grant & Magee, 1998;Ruddle et
al., 2011a,2011b). There is thus no research on the interaction of
decision making and idiothetic information in the acquisition of
route and graph knowledge.
The Present Study
Our purpose is to investigate the exploration-specific learn-
ing hypothesis by testing whether these components of active
exploration differentially contribute to particular types of spa-
tial knowledge. As described above, we previously found that
podokinetic information, but not decision making, significantly
contributes to survey knowledge (Chrastil & Warren, 2013).
The present study tests the role of idiothetic information and
decision making in the acquisition of route and graph knowl-
edge.
In this experiment, we crossed two levels of information with
two levels of decision making during exploration of a virtual
hedge maze in a between-groups design (see Table 1). In the
walking condition, participants had full access to visual and
idiothetic information, whereas in the video condition only
visual information was available (including perspective, binoc-
ular disparity, and optic flow). These conditions were crossed
with the free condition, in which participants made decisions
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1165
ACTIVE AND PASSIVE NAVIGATION
about their route during exploration, and the guided condition,
in which they were guided along matched routes. To assess
route and graph knowledge, participants were then asked to take
the shortest route between two learned locations by walking in
the maze corridors; on 40% of the trials, barriers were added to
promote novel detours. The route task might be performed by
repeating routes that had been traveled during exploration.
However, for novel routes or detours, previously traveled routes
must be integrated into a graph and their segments recombined;
taking the shortest possible path additionally requires local
metric knowledge.
On the basis of the preceding theoretical and empirical con-
siderations, we can make several predictions. First, we predict
that decision making will play a significant role in route and
graph learning, in contrast to survey learning. The decisions
made during exploration are likely to facilitate the formation of
place–action associations as well as the formation of edges and
nodes in a place graph, possibly via the allocation of attentional
resources or prediction learning. Thus, we expect that partici-
pants in the free condition will be more successful in the route
test than those in the guided condition.
Second, idiothetic information may interact with decision
making. Acquiring a topological graph or route does not require
information about metric distances and directions, so partici-
pants should show evidence of graph learning without idiothetic
information (in the free video condition). On the other hand,
such local metric information might be incorporated into a
labeled graph. In this case, participants should exhibit shorter
and more accurate paths when idiothetic information is present
during learning (in the free walking condition). Third, we
expect that participants will acquire not merely route knowl-
edge but graph knowledge, displaying evidence of novel routes
and detours during the test phase.
The results demonstrate that decision making during explo-
ration makes a significant contribution to graph knowledge, at
least in combination with idiothetic information. Participants
who made decisions during the learning phase when idiothetic
information was available (free walking condition) displayed
the greatest route and graph knowledge. In addition, partici-
pants took novel routes and detours on the majority of trials—
indicative of graph knowledge, not merely route knowledge—
and frequently chose the shortest path. The results are
consistent with the acquisition of a labeled graph of the envi-
ronment. Together with our previous work (Chrastil & Warren,
2013), the findings support the exploration-specific learning
hypothesis; whereas idiothetic information is the primary con-
tributor to survey knowledge, decision making is the primary
contributor to graph knowledge.
Method
Participants
Participants were 125 (85 female, 40 male) volunteers who were
paid for their time. Ninety-six participants (sex: 48 female, 48
male; age: M⫽23.23, SD ⫽7.64) completed the experiment.
Eighteen (12 female, 6 male) withdrew because of symptoms of
simulator sickness, 6 (4 female, 2 male) were excluded for failure
to find all of the objects during exploration, and 5 (3 female, 2
male) were excluded for experimenter error or technical problems.
Attrition for each experimental group as a result of symptoms of
simulator sickness (or failure to find all objects) was as follows:
free walk, 1 (1); guided walk, 3 (0); free video, 5 (5); guided video,
9 (0). To avoid confounding by documented sex differences (e.g.,
Moffat, Hampson, & Hatzipantelis, 1998;Waller, 2000;Wolbers
& Hegarty, 2010), the final experimental groups had equal num-
bers of men and women. All participants read and signed forms
indicating their informed consent to participate in the experiment,
in accordance with a protocol approved by the Brown University
Institutional Review Board.
Equipment
This experiment was conducted in the VENLab, a 12 m ⫻14 m
ambulatory VR facility. All participants viewed stereoscopic im-
ages presented in a Rockwell-Collins SR80A (Cedar Rapids, IA)
HMD (63° H ⫻53° V field of view, 1280 ⫻1024 pixels, complete
binocular overlap, 60 Hz frame rate). An InterSense IS900 (Bil-
lerica, MA) tracking system (60 Hz sampling rate, 1.5 mm RMS
and 0.1° RMS error) recorded head position and orientation, which
was used to update the display (50 ms latency). Participants
responded by walking to target locations and pressing a button on
a radio mouse. The virtual environment was generated and ren-
dered on a Dell XPS graphics PC (Round Rock, TX) using Vizard
software (WorldViz, Santa Barbara, CA). Naturalistic evening
sounds were presented over headphones to interfere with any
auditory location or orientation cues.
Displays
The 11 m ⫻12 m virtual maze environment (see Figure 1)
contained eight objects located in the terminal segment of branch
hallways, so they were not visible from the main corridors. They
were models of common objects, such as a sink or bookcase,
scaled to be easily visible at eye height. In addition, four land-
marks—familiar paintings by Monet, Dali, Magritte, Van Gogh—
appeared in a constant location on the walls of the maze in main
corridors to aid orientation. The ground in each corridor was a
random grayscale gravel texture with a brown earthen and green
grassy border.
Design
Four groups of participants were tested in a 2 ⫻2 (Informa-
tion ⫻Decision) design, yielding four exploration conditions (see
Table 1): (1) free walking, (2) guided walking, (3) free video, and
(4) guided video. Each group consisted of 16 randomly assigned
participants, with the restriction that the groups were evenly di-
Table 1
Experimental Design
Decision
Yes No
Information
Visual ⫹Idiothetic Free Walk Guided Walk
Visual Free Video Guided Video
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1166 CHRASTIL AND WARREN
vided between males and females, making 64 participants in the
initial experiment. On the basis of a preliminary analysis, 16
participants were added to each of the two walking groups (free
walking and guided walking), for a total of 96 participants.
Procedure
Participants were informed that they would be traveling through
hallways in a virtual hedge maze, and that the task was to find all
of the objects and learn their locations. To equate eyeheight-scaled
perspective information, participants in the video condition sat in
an adjustable chair at approximately their measured standing
height, and the virtual eye height was set to their measured
standing eye height; eye height in the walking condition corre-
sponded to the participant’s actual eye height as measured by the
head tracker.
Practice. Participants were given several minutes in a practice
maze with a different layout and different objects from the test
maze, in their assigned exploration condition. This procedure
allowed them to become familiar with VR and, in the Free Video
condition, to practice using the keyboard.
Learning phase. In the learning phase, participants were in-
formed that they had 10 min to explore the environment and learn
the locations of all the objects. They were not told how many
objects were in the maze. They were then guided to one of six start
locations, and the experimental maze appeared to start the learning
phase. Participants in the free walking condition were instructed to
explore the virtual environment by walking freely, with normal
visual and idiothetic information and decision making about the
path of exploration. In the guided walking condition, each partic-
ipant was guided by the arm along the same path as taken by a
participant in the free walking group, giving them matched visual
and idiothetic information without the decision-making compo-
nent. In the free video condition, seated participants pressed the
four arrow keys on a keyboard to turn during exploration,
analogous to desktop VR. The keys provided continuous move-
ment when pressed, with a fixed translation speed of 1.0 m/s
and rotation speed of 90°/s, similar to walking speed. Partici-
pants could press two buttons at once to enable smooth curved
movement. Thus, only visual information was available, to-
gether with decision making. In the guided video condition,
each participant viewed a video replay of exploration from a
matched participant in the free walking group, without making
decisions. Head position was recorded throughout the explora-
tion period.
Given that the experiment was designed to test the effect of
removing active components from normal walking, the paths in
both of the guided conditions were matched to paths from the free
walking group, rather than the free video group. The paths in the
two free conditions could not be matched, so any differences
between those groups might be attributable to differences between
the exploration paths not only to differences in idiothetic informa-
tion. To match the two free conditions as closely as possible, the
display speed in the free video condition was matched to the mean
free walking speed. Exploration paths were also analyzed to check
for significant differences between conditions (see the discussion
of the dependent measures).
Test phase. Route and graph knowledge were tested using a
shortest route task, in which the participant walked through the
corridors of the maze from a starting object to the remembered
location of a target object. Prerecorded instructions were presented
to the participants over the headphones and then repeated by the
experimenter. Participants then completed two practice trials with
object pairs not used during the test phase, followed by 40 test
trials. Each trial began by wheeling the participant in a wheelchair
through a virtual desert environment to the location of the
entrance of the branch hallway (where the branch hallway met
the main hallway) containing the start object, approximately 1
m from the object; the participant stood up, clicked the mouse,
whereupon the maze and start object appeared. This procedure
oriented participants within the maze while limiting spatial
learning during the test phase. The participant was instructed to
walk to the start object, at which point the target object was named
over the headphones. All objects were replaced with red blocks
after the target was named to avoid providing feedback (Figure
2A); the four landmark paintings remained visible. The participant
was then given 30 s to walk to the location of the target object and
click the mouse, taking the shortest route possible in the maze
corridors. The trial ended when the participant clicked the mouse
or the 30 s elapsed. The maze then disappeared, and the experi-
menter wheeled the participant to the starting location of the next
trial, taking a circuitous route in the desert environment to prevent
path integration between trials. Participants were not given explicit
feedback regarding whether they reached the correct target loca-
tion. Head position and orientation were recorded between the
initial and final mouse click, which served as the trial endpoint.
This task allowed us to evaluate route and graph knowledge by
comparing the paths taken during test with the routes traveled
during exploration.
On 40% of the test trials, one of the shortest paths was blocked
with a barrier and participants were instructed to take a different
route to the target object if they encountered the barrier (Figure 2B
and 2C). These barrier trials probed graph knowledge because they
led participants to quickly synthesize novel routes through the
maze. Because detour paths often required additional time and
distance, participants were given 15 additional seconds for these
trials, although they were not informed of the additional time. Note
that participants did not necessarily encounter the barriers, but they
frequently led to detours.
Participants were tested on eight pairs of start and target objects
(henceforth “targets”), with five trials for each start-target pair
(three direct and two barrier trials), for a total of 40 test trials. All
trials were presented in a random order with the exception that
targets did not repeat back-to-back. Although the experimental
groups received differing levels of information during the learning
phase, they all walked with full visual and idiothetic information in
the test phase, to avoid differences in the testing conditions that
could yield differences in performance.
Follow-up tests. Finally, participants performed several addi-
tional tests to evaluate whether the experimental groups differed in
spatial ability. First, participants drew a sketch map of the maze.
Participants were presented with a list of the names of the objects
and paintings in the virtual maze and were asked to draw a map of
the maze freehand, using a pen on a sheet of paper. The maps were
scored on a 10-point scale by a double-blind rater for general
accuracy of spatial layout (cf. Figure 1), with most importance
given to the locations of the hallways and the relative positions of
the objects and paintings. Next, they answered questions prompt-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1167
ACTIVE AND PASSIVE NAVIGATION
ing self-report of the strategies used during the both the explora-
tion and the test phases. They also completed the Santa Barbara
Sense of Direction Scale (SBSOD; Hegarty, Richardson, Montello,
Lovelace, & Subbiah, 2002) and a questionnaire including report
of current and past video game use and ratings of nausea and sense
of immersion during the experiment. We note that video game
experience has been shown to be a factor in virtual navigation
(Richardson, Powers, & Bousquet, 2011). They were also given
the Road Map Test (Money & Alexander, 1966;Zacks, Mires,
Tversky, & Hazeltine, 2000), in which participants report the
direction of each turn in a route pre-drawn on a city map, modified
to have a 20-s time limit. Finally, participants performed the
Perspective-Taking Test (PTSOT; Kozhevnikov & Hegarty,
2001), in which they view a 2D array of objects on a page and
indicate their directions from different imagined viewpoints. The
Road Map Test and Perspective-Taking Tests gauge a navigator’s
ability to process location and direction information from different
perspectives, which could be important for acquiring graph knowl-
edge during exploration or for orienting within the maze during the
test phase.
Dependent Measures and Analysis
Multiple dependent measures were taken, but a number of them
were redundant and yielded very similar results. We thus report
only the following measures: Proportion of correct trials (accu-
racy), where a trial was considered correct if the participant ended
anywhere in the branch hallway containing the target object. The
chance level was defined as the probability of randomly ending at
any one of the eight object locations, or 1/8 ⫽0.125 (participants
Figure 2. Views of the test phase. A: Red blocks replaced the objects during the test trials. B: A wall was
placed in a hallway during barrier trials. C: Overhead view of the maze with an example of a barrier trial, from
the rabbit to the earth (white rectangle is barrier). A participant taking the correct path would walk toward the
earth, encounter the wall, and then take a new path to reach the target location. See the online article for the color
version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1168 CHRASTIL AND WARREN
did occasionally end up at the starting location). Measures of
response consistency, travel time, and final distance from target
yielded similar results to proportion correct and are not reported
here. Proportion of novel routes, which is the proportion of correct
test trials on which the observed route was not previously taken
during free exploration. Specifically, the sequence of junctions
through which a participant passed on each test trial was compared
against all such sequences during their free exploration phase. If
the sequence was previously observed, in either the forward or
backward direction, the route was counted as familiar; if not, it was
counted as novel. Proportion of shortest paths is the proportion of
correct test trials in which the participant took the shortest-length
path in the maze between the start and target locations, as mea-
sured in meters (metric distance). Sketch map rating, which is the
rating, out of a maximum possible of 10, for the sketch map drawn
during the follow-up tests. Exploration behavior, including total
distance traveled, total angular rotation, and mean number of visits
per object during the 10 min exploration phase; the standard
deviation and range of the number of visits per object reflect how
evenly a participant explored the environment. Most analyses were
performed on participant means using a 2 ⫻2⫻2 (Information ⫻
Decision ⫻Sex) unequal sample size analysis of variance
(ANOVA).
1
Separate item analyses and analyses of direct and
barrier trials were also performed.
Preliminary analysis. A preliminary analysis was performed
on data from 16 participants per group. Mean proportion correct in
each group was as follows: free walking (M⫽0.607, SD ⫽0.292);
guided walking (M⫽0.459, SD ⫽0.323); free video (M⫽0.479,
SD ⫽0.311); guided video (M⫽0.384, SD ⫽0.248). A 2 ⫻2⫻
2 (Information ⫻Decision ⫻Sex) ANOVA yielded a marginal
main effect of decision making, F(1, 62) ⫽2.794, p⫽.100, p
2⫽
0.048, such that the free groups performed better than the guided
groups, but no effect of information, F(1, 62) ⫽1.960, p⫽.167,
p
2⫽0.034, and no interactions. There was also a significant effect
of sex, F(1, 62) ⫽4.949, p⫽.030, p
2⫽0.081. An analysis of
effect size revealed that the main effect of decision making stabi-
lized around a moderate value (p
2⫽0.05; n⫽12), whereas the
effect size of information remained small (stabilizing around p
2⫽
0.025). Given the large interindividual variability (standard devi-
ations on the order of 0.3), this result justified an increase in power
to further investigate the possible effect of decision making. Con-
sequently, 16 additional participants were added to both the free
and guided walking groups. Thus, the final analyses included
ANOVAs for unequal sample size on data from 32 participants in
each of the free and guided walking groups, and 16 participants in
each of the free video and guided video groups, for a total of 96
participants. We note that proceeding with a fixed sample stopping
rule after conducting preliminary analysis with a sequential stop-
ping rule can affect Type I error (Frick, 1998).
Results
Proportion Correct
We begin by analyzing the proportion of correct test trials in
each condition, which appears in Figure 3. Performance in all
groups was significantly above the chance level (0.125, indicated
by the dashed line), as determined by one-sample ttests (Free
Walking: t
31
⫽8.235, p⬍.001; Guided Walking: t
31
⫽5.008,
p⬍.001; Free Video: t
15
⫽4.550, p⬍.001; Guided Video: t
15
⫽
4.177, p⫽.001). This result indicates that visual information
alone is sufficient for significant route and graph learning, as we
previously observed for survey learning (Chrastil & Warren,
2013).
A three-way ANOVA for unequal sample size (2 Information ⫻
2 Decision ⫻2 Sex) on the proportion of correct trials revealed a
main effect of decision making, F(1, 94) ⫽4.157, p⫽.044, p
2⫽
0.045, such that participants who made decisions during explora-
tion had a higher proportion correct than those who were guided.
In contrast, there was no main effect of information, F(1, 94) ⫽
0.548, p⫽.461, p
2⫽0.006, indicating the absence of an idio-
thetic advantage, and no Information ⫻Decision interaction, F(1,
94) ⫽0.276, p⫽.601, p
2⫽0.003. There was also a main effect
of sex, F(1, 94) ⫽4.609, p⫽.035, p
2⫽0.050, such that men
performed better overall than women, but no interaction was found
between sex and information, F(1, 94) ⫽1.739, p⫽.191, p
2⫽
0.019, or between sex and decision making, F(1, 94) ⫽2.132, p⫽
.148, p
2⫽0.024, neither was there a three-way interaction, F(1,
94) ⫽0.092, p⫽.762, p
2⫽0.001.
We also computed Bayesian factors for models alternative to the
null model by taking the ratio of the likelihood of the null model
to the likelihood of the alternative model (likelihood null model/
likelihood alternative model), with smaller numbers indicating
greater support for the alternative model. Proceeding incremen-
tally, the first alternative model included only the factor of infor-
mation. This analysis yielded a Bayesian factor of 0.772, indicat-
ing that the probability of the null model given the data is
approximately 77.2% of the probability of the data given this
alternative model. In contrast, the Bayesian factor for the factor of
decision making alone was 0.070, indicating that the null model is
only 7.0% as likely as the model with just decision making added.
1
Analyses of proportion data were also performed following an arcsine
transformation, however there were no substantial differences between
those analyses and those of the untransformed data. Thus, analyses of the
untransformed data are reported here.
Figure 3. Proportion of trials ended at the correct object location for the
four experimental groups. Dashed line indicates chance level. Error bars
indicate between-subjects standard error. Significant effects are indicated
by asterisks.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1169
ACTIVE AND PASSIVE NAVIGATION
The combined information and decision-making model (two-
factor) had a Bayesian factor of 0.046. Finally, the full three-factor
model (information, decision making, and sex) yielded a Bayesian
factor of 0.0008.
A planned comparison to test the role of decision making in the
walking condition found a significant main effect of decision
making, F(1, 62) ⫽4.568, p⫽.037, p
2⫽0.071, with the free
walking group having a greater proportion of correct trials than the
guided walking group. This result indicates that decision making
together with full idiothetic information during exploration signif-
icantly improves performance. The planned comparison between
the free video and guided video groups did not find an effect of
decision making, F(1, 30) ⫽1.035, p⫽.318, p
2⫽0.036.
Although this comparison had less power, the effect size had
stabilized at a value that was half that of the walking condition.
This pattern of results supports the conclusion that decision mak-
ing significantly contributes to route and graph knowledge, at least
in combination with idiothetic information and illuminates previ-
ous equivocal results in desktop VR.
Figure 4A illustrates the individual performances in the two
walking groups (each n⫽32). Participants are plotted in rank
order by performance, from left to right. All rank-ordered mem-
bers of the free walking group have a higher proportion correct
than their peer in the guided walking group. A Mann–Whitney U
test evaluated the rankings of the proportion correct in the two
groups, and found that the rankings were significantly different
(U⫽341.5, p⫽.022). The wide performance range in both
groups highlights the individual differences commonly observed in
navigation tasks. The pattern also suggests that making decisions
during exploration did not greatly affect performance at the lowest
end of the spectrum, whereas there was a ceiling effect at the
highest end, but it increased learning 10% to 20% (by one or two
objects) in the broad midrange of spatial performance. In contrast,
Figure 4B illustrates the individual performances in the two video
groups (each n⫽16). The only difference between the free and
guided groups appears at the high end of the spectrum. A Mann–
Whitney Utest evaluated the rankings of the proportion correct in
the two groups and found that the rankings were not significantly
different (U⫽106.0, p⫽.406). These results reinforce the finding
that decision making, in combination with idiothetic information,
contributes to route and graph learning.
Direct and barrier trials. The proportion correct in each
condition is broken down by direct trials (M⫽0.45, SD ⫽0.30)
and barrier trials (M⫽0.48, SD ⫽0.31) in Figure 5. The slightly
higher accuracy on barrier trials may be due to the extra 15 s
allowed for barrier trials.
For direct trials (see Figure 5A), a three-way ANOVA yielded
a marginal effect of decision making, with the same effect size as
for all trials, F(1, 94) ⫽3.855, p⫽.053, p
2⫽0.042, but no main
effect of information, F(1, 94) ⫽0.467, p⫽.496, p
2⫽0.005, or
interactions. There was also a main effect of sex, F(1, 94) ⫽4.354,
p⫽.040, p
2⫽0.047, with men performing better than women. A
separate analysis of barrier trials (see Figure 5B) revealed a sig-
nificant main effect of decision making, F(1, 94) ⫽4.490, p⫽
.037, p
2⫽0.049, and a main effect of sex, F(1, 94) ⫽4.365, p⫽
.040, p
2⫽0.047, but no effect of information, F(1, 94) ⫽0.667,
p⫽.416, p
2⫽0.008, or interactions. These results imply that
decision making during learning improves route finding on both
direct paths and barrier trials, but it does not significantly interact
with idiothetic information.
Planned comparisons in the walking condition confirmed an
advantage of decision making, revealed especially by barrier trials.
There was a significant effect of decision making on barrier trials,
F(1, 62) ⫽6.551, p⫽.013, p
2⫽0.098, with a greater proportion
correct in the free walking than the guided walking group; this
effect was marginal on direct trials, F(1, 62) ⫽3.147, p⫽.081,
p
2⫽0.050. Thus, finding a novel detour was significantly easier
for participants who had made decisions during exploration, sup-
porting the idea that decision making contributes to graph knowl-
edge.
Target difficulty. To compare the difficulty of the eight ob-
ject pairs, we performed an item analysis. First, one-sample ttests
comparing each target to the chance level (0.125) revealed that all
targets were located significantly more often than expected by
chance (all ps⬍.001). Thus, participants acquired reliable graph
knowledge of all object locations in the maze. However, some
targets were more difficult to find than others: a repeated-measures
ANOVA on the proportion of correct trials with target as a within-
Figure 4. Individual performance. Participants in each group were ordered from least proportion correct to
greatest proportion correct. Dashed line indicates chance level, 0.125. A: Walking conditions. B: Video
conditions.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1170 CHRASTIL AND WARREN
subjects factor revealed a main effect of target, F(7, 665) ⫽8.237,
p⬍.001, p
2⫽0.086. Referring to Figure 1, post hoc tests found
that snowman-to-rabbit was the most difficult pair, with only
37.1% correct, while bookcase-to-well was the easiest, with 61.3%
correct.
To investigate factors that affected target difficulty, we com-
puted the Pearson correlation between the mean proportion correct
on each of the eight targets (direct trials only) and several mea-
sures of the relationship between start and target objects: as-the-
crow-flies distance, metric (in m) and topological distance (num-
ber of intersections or segments) of the shortest route, and the
number of turns on the shortest route (not counting turns within the
branch hallways themselves; branch hallways generally all had one
turn, thus excluding these does not substantially change the re-
sults). Of these potential factors, only the number of turns was
statistically significant overall (r
6
⫽⫺0.855, p⫽.007). Looking at
each experimental group, the number of turns was significant in the free
walking group (r
6
⫽⫺0.847, p⫽.008), the guided walking group
(r
6
⫽⫺0.869, p⫽.005), and the free video group (r
6
⫽⫺0.758,
p⫽.029), but not the guided video group (r
6
⫽⫺0.078, p⫽
.855). In contrast, in the guided video group accuracy was only
correlated with topological distance (r
6
⫽⫺0.903, p⫽.002).
Finally, in the guided walking group there was also a significant
correlation with metric distance in the maze (r
6
⫽⫺0.708, p⫽
.049). It is interesting to note that target accuracy in the guided
video group was not correlated with any other group (p⬎.50 for
all comparisons), whereas the other three groups were all intercor-
related (p⬍.01 for all comparisons). These results suggest that
completely passive learning yields somewhat different graph
knowledge than does active learning.
Learning during test. It is possible that learning continued
during the 40 test trials, which could have influenced the overall
results. A within-subject ANOVA comparing the first and last
block of 10 test trials revealed that the proportion correct signifi-
cantly increased during the test phase, F(1, 94) ⫽58.829, p⬍
.001, p
2⫽0.401. However, an analysis of just the first block of 10
trials revealed a marginal effect of decision making, F(1, 94) ⫽
3.966, p⫽.050, p
2⫽0.043, but no effect of information, F(1,
94) ⫽1.003, p⫽.319, p
2⫽0.011, and no interaction. A separate
analysis of the walking condition demonstrated that the free walk-
ing group had a greater proportion correct than the guided walking
group in the first block of 10 trials, F(1, 62) ⫽5.464, p⫽.023,
p
2⫽0.083. Overall, these results confirm a decision making
advantage even in the first 10 test trials.
Novel Routes
Route knowledge would only enable participants to use familiar
routes to the target in the test phase, whereas graph knowledge
would enable participants to recombine known path segments into
novel routes. Thus, to determine whether performance reflected
route or graph knowledge, we analyzed the proportion of correct
test trials in which participants took routes they had not traveled
during exploration. Overall, novel routes were taken on 64.5% of
correct direct trials and, even more strikingly, 88.7% of correct
barrier trials. Pure route knowledge predicts that a navigator could
never take a path that they had not experienced during exploration
because they can follow only known routes. Both of these values
differ significantly from this pure route knowledge prediction,
namely, 0 novel routes (p⬍.001 for both direct and barrier trials,
one-sample ttest against 0). This result implies that participants
did not simply learn specific routes between objects, but they
acquired a graph of connections between places by traversing
different segments at various times during exploration; they were
then able to recombine the segments to generate novel routes
during test. There were no differences between conditions in the
number of novel paths on direct trials, but for barrier trials there
was a marginal effect of information, F(1, 92) ⫽3.967, p⫽.050,
p
2⫽0.044, such that the walking groups took more novel routes
than the video groups (see Table 2). There was also a three-way
interaction, F(1, 92) ⫽5.067, p⫽.026, p
2⫽0.056.
Shortest Routes
Pure graph knowledge would support routes with the shortest
topological length (fewest number of intersections or segments),
whereas a labeled graph would enable routes with the shortest
metric length (in m). To estimate whether performance reflected
such local metric knowledge, we measured the proportion of
correct trials in which participants took the metrically shortest
Figure 5. Proportion correct of direct and barrier trials. Dashed line indicates chance level. Error bars indicate
between-subjects standard error. Significant effects are indicated by asterisks. A: Proportion correct of the direct
trials for the four experimental groups. B: Proportion correct of the barrier trials for the four experimental groups.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1171
ACTIVE AND PASSIVE NAVIGATION
available route between the start and test objects. Overall, the
percentage of metrically shortest routes was 64.6% on correct
direct trials, and 64.0% on correct barrier trials, suggesting that
metric information played a role in route selection. For direct trials
(see Table 3), a three-way ANOVA on the proportion of shortest
path among the correct trials revealed no main effects of informa-
tion, F(1, 92) ⫽0.280, p⫽.598, p
2⫽0.003, decision making,
F(1, 92) ⫽0.316, p⫽.575, p
2⫽0.004, or sex, F(1, 92) ⫽1.843,
p⫽.178, p
2⫽0.021, and no interactions. For barrier trials (see
Table 3), however, there was a main effect of decision making,
F(1, 92) ⫽6.012, p⫽.016, p
2⫽0.065, such that participants who
made decisions during exploration had a higher proportion of
shortest paths than those who were guided. There was also a
marginal effect of information, F(1, 92) ⫽3.651, p⫽.059, p
2⫽
0.041, and a main effect of sex, F(1, 92) ⫽5.071, p⫽.027, p
2⫽
0.056, such that men took more shortest paths than women. The
only interaction was a three-way interaction, F(1, 92) ⫽6.055,
p⫽.016, p
2⫽0.066. Participants who had made decisions during
exploration were able to take the shortest paths during test,
whereas participants who had been guided tended to take longer
routes, particularly on barrier trials.
However, the shortest metric and topological routes in the maze
were often the same. To dissociate them, we examined cases in
which there were alternative routes between the start and target
objects that were topologically equivalent to but metrically longer
than the shortest route. Of the eight targets, there were two that
possessed at least one such alternative route on direct trials and
three others that had alternative routes on barrier trials (see Table
4). On four of these five targets, participants took the metrically
shortest route more frequently than the topologically equivalent
alternatives combined (each p⫽.001 or better, repeated-measures
ANOVA); the fifth case was marginally significant in the other
direction. This clearly demonstrates that participants generally
selected the metrically shortest route to the target, even when there
were topologically equivalent but longer alternatives, which is
indicative of a labeled graph with local metric information. Alter-
natively, it is possible that participants used survey knowledge for
these paths; we consider this possibility in the discussion.
Map Scores
Sample sketch maps appear in Figure 6A, and the mean scores
in each condition are plotted in Figure 6B. A three-way ANOVA
found a main effect of decision making, F(1, 94) ⫽5.097, p⫽
.026, p
2⫽0.055, such that participants in the free groups drew
better maps than those in the guided groups, but there was no effect
of information, F(1, 94) ⫽2.404, p⫽.125, p
2⫽0.027, and no
interaction. There was also a significant Information ⫻Sex inter-
action, F(1, 92) ⫽4.855, p⫽.030, p
2⫽0.052, although none of
the post hoc Tukey’s tests were significant for this interaction. The
sketch maps thus confirm a decision-making advantage.
Group Differences
Due to our between-subjects design, it is possible that the effects
we observed were due to inadvertent sampling bias in the assign-
ment of participants to experimental groups rather than to the
experimental manipulations. Thus, we briefly examine other pos-
sible sources of group differences.
Spatial ability. To assess whether the spatial abilities of par-
ticipants differed between groups, we performed three-way ANOVAs (2
information ⫻2 decision ⫻2 sex) on four individual measures.
There were no statistical differences in age between the experi-
mental groups, and each group had an equal number of men and
women. Sex, rather than group, accounted for most of the signif-
icant differences in spatial ability: Men performed better than
women on the SBSOD, F(1, 94) ⫽10.792, p⫽.001, p
2⫽0.109,
the PTSOT, F(1, 94) ⫽3.974, p⫽.049, p
2⫽0.043, and the Road
Table 2
Percentage (SD) of Correct Trials in Which Novel Routes Were Taken During the Test Phase,
for Both Direct and Barrier Trials in the Four Groups
Decisions
Direct trials Barrier trials
Free Guided Free Guided
Information
Walk 62.64% (22.55) 72.84% (14.59) 90.06% (11.38) 92.49% (19.56)
Video 61.13% (29.86) 66.18% (26.15) 86.37% (17.77) 80.72% (27.50)
Table 3
Percentage of Trials (SD) in Which the Shortest Route Was Taken to the Target, for Both Direct
and Barrier Trials
Decisions
Direct trials Barrier trials
Free Guided Free Guided
Information
Walk 64.34% (27.38) 66.80% (27.78) 72.90% (19.22) 61.43% (33.05)
Video 67.13% (22.37) 58.14% (21.23) 64.87% (20.74) 49.07% (28.34)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1172 CHRASTIL AND WARREN
Map Test, F(1, 94) ⫽9.402, p⫽.003, p
2⫽0.097. The only other
observations were opposing trends for idiothetic information: On
the SBSOD, the video groups reported a marginally worse sense of
direction than did the walking groups, F(1, 94) ⫽3.9124, p⫽
.051, p
2⫽0.043, yet on the PTSOT, the video groups showed a
better perspective-taking ability than did the walking groups, F(1,
94) ⫽6.527, p⫽.012, p
2⫽0.069. Perspective-taking ability
could contribute during learning by helping to link different views
of the maze together into a larger path structure. Thus the PTSOT
difference could account for the lack of an effect of idiothetic
information, but it seems to be countervailed by the SBSOD result.
Video gaming. The three video game measures also primarily
differed by sex, including current video game use, F(1, 94) ⫽
8.394, p⫽.005, p
2⫽0.087, as well as current, F(1, 94) ⫽5.374,
p⫽.023, p
2⫽0.058, and past use, F(1, 94) ⫽6.896, p⫽.010,
p
2⫽0.073, of first-person navigational video games, with men
spending more time playing video games than women. Thus, sex
differences on spatial ability measures and video game experience
might account for the sex differences observed in our shortest
route task.
Experience in VR. It is possible that the experimental condi-
tions were associated with different personal experiences in VR.
There was only a significant effect of information condition on
nausea rating, F(1, 94) ⫽4.604, p⫽.035, p
2⫽0.050, with video
groups reporting greater nausea than the walking groups, a mar-
ginal effect of sex, F(1, 94) ⫽3.322, p⫽.072, p
2⫽0.036, and
an Information ⫻Sex interaction, F(1, 92) ⫽11.193, p⫽.001,
p
2⫽0.113. Post hoc Tukey’s tests showed that women in the
video groups gave significantly higher ratings of nausea than did
women in the walking group or men in either group (all p⬍.05).
Greater nausea ratings could imply that participants in the video
groups, especially women, were more distracted during learning.
However, there were no group differences in ratings of immersion
in the virtual environment.
Exploration behavior. It is possible that different patterns of
exploration during the 10-min learning phase might account for
Table 4
Proportion of Correct Trials in Which Participants Took the Shortest Path Compared With
Alternatives With an Equal Number of Nodes in the Graph
Trial Type (direct/barrier)
(No. of alternatives)
Proportion Taking
Shortest Path
Proportion Taking Any
Topologically Equal Alternative Np
1 (direct) (2) 0.370 0.520 68 .073
2 (barrier) (1) 0.822 0.089 45 ⬍.001
3 (barrier) (1) 0.690 0.080 44 ⬍.001
4 (direct) (1) 0.620 0.170 53 ⬍.001
7 (barrier) (1) 0.660 0.130 46 .001
Note. Some participants did not have correct trials for certain trial types and were excluded from analysis.
Figure 6. A: Examples of sketch maps drawn by participants. B: Sketch map scores. Error bars indicate
between-subjects standard error. Significant effects are indicated by asterisks.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1173
ACTIVE AND PASSIVE NAVIGATION
group differences in the test phase. The video groups (M⫽360.76
m, SD ⫽60.47) had a greater total distance traveled than did the
walking groups (M⫽321.31 m, SD ⫽56.54; F[1, 94] ⫽9.870,
p⫽.002, p
2⫽0.101) and had higher total angular rotation during
exploration (video group: M⫽46022.5°, SD ⫽8351.9; walking
group: M⫽40688.1°, SD ⫽9767.0; F[1, 94] ⫽7.600, p⫽.007,
p
2⫽0.079).
The number of visits per object during learning indicates how
thoroughly the maze was explored. There was a marginal effect of
information, F(1, 94) ⫽3.959, p⫽.050, p
2⫽0.043, such that the
video groups (M⫽3.59, SD ⫽0.65) had more visits per object
than the walking groups (M⫽3.29, SD ⫽0.70). The within-
subject standard deviation of number of visits reflects how evenly
participants explored all parts of the maze, and here there was a
main effect of decision making, F(1, 94) ⫽7.316, p⫽.008, p
2⫽
.077, and a significant Information ⫻Decision Making interaction,
F(1, 92) ⫽6.320, p⫽.014, p
2⫽0.067. Post hoc Tukey’s tests
revealed that these effects were due to participants in the free video
group, who had a higher standard deviation than the other three
groups, which had matching paths (all p⬍.05). Thus, the effects
on standard deviation of visits are actually driven by the fact that
the free walking group visited objects more evenly than the free
video group.
The main finding of the exploration analysis is that the free
walking group visited objects more evenly than the free video
group, implying that idiothetic information helps participants keep
track of the locations they have previously explored. On the other
hand, the free video group traveled farther, had greater angular
rotation, and visited each object more frequently than did the free
walking group (and hence the matched guided walking and guided
video groups), which is likely due to the ease of exploration with
the keyboard and could have undercut an idiothetic advantage in
spatial learning.
Discussion
To investigate potential components of active and passive spa-
tial learning, we tested the contributions of idiothetic information
and cognitive decision making to the acquisition of route and
graph knowledge compared with visual information alone. The
results support five conclusions about active spatial learning. First,
we found that decision making during exploration significantly
contributes to graph learning. Second, in contrast, we find no
evidence that idiothetic information alone significantly contributes
to graph learning. However, we note that the decision-making
advantage is significant in the presence of idiothetic information,
but not without it, which may help explain previous inconsistent
findings in desktop VR. Third, humans acquire not merely route
knowledge but graph knowledge of the connectivity of the envi-
ronment early in learning. Specifically, the data support a level of
spatial knowledge that incorporates local metric information, most
likely a labeled graph. This leads to our main conclusion, that the
findings support the exploration-specific learning hypothesis.
Whereas idiothetic information during exploration is the primary
contributor to survey knowledge, decision making is the primary
contributor to graph knowledge. Finally, there is a reliable sex
difference in graph learning, such that men perform the route-
finding task more successfully than do women. We discuss these
conclusions in more depth.
Contribution of Decision Making to Graph Knowledge
First, the results of this experiment demonstrate that decision
making significantly contributes to graph knowledge. Overall,
participants who made decisions during exploration (free groups)
were more accurate in the test phase, took more successful and
shorter detours, and drew better sketch maps than those who did
not (guided groups). In particular, the free walking group had a
significantly greater proportion of correct trials than the guided
walking group. This result is consistent with the hypothesis that
making decisions about one’s exploration path facilitates learning
the network of connections between places. It is possible, for
example, that making decisions directs the navigators’ attention to
aspects of the environment that are relevant to graph learning.
Decision making might aid route learning by promoting the for-
mation of place–action associations during exploration. Further-
more, it could facilitate graph learning by eliciting predictions
about the consequences of each turn, which were then consolidated
via feedback.
It is important to point out that passive visual information alone
(including optic flow, binocular disparity, and perspective) was
sufficient for some graph learning, for accuracy in the guided
video group was well above chance. Thus, although decision
making yields a significant improvement, it is not necessary for the
acquisition of some graph knowledge.
Contribution of Idiothetic Information
to Graph Knowledge
Second, we found little evidence that idiothetic information
alone significantly contributes to graph knowledge. Overall, par-
ticipants who walked during exploration did not perform better
than seated participants who viewed a video. In particular, perfor-
mance in the guided walking group was virtually identical to that
in the guided video group. The results are consistent with the
hypothesis that idiothetic information is not necessary for the
acquisition of topological graph or route knowledge. In addition,
local metric distances and angles in a weighted graph can be
estimated from visual information such as optic flow (Bertin,
Israel, & Lappe, 2000;Frenz, Bremmer, & Lappe, 2003).
This result departs from two previous experiments that exam-
ined route (but not graph) knowledge that reported an idiothetic
advantage when participants walked on a prescribed path during
learning without making decisions (Grant & Magee, 1998;Ruddle
et al., 2011b), which corresponds to the guided walking and guided
video groups in our experiment (see Figure 3). It is possible that
the performance of our video group was inflated because they had
better PTSOT scores (lower errors) than the walking group. Yet
SBSOD scores showed the opposite effect, and there were no
group differences in video game experience. The present results
thus provide no evidence that idiothetic information by itself
improves graph learning.
We note, however, that we observed a significant decision-
making effect in the walking condition when idiothetic informa-
tion was available, but not in the video condition when idiothetic
information was absent. This is best illustrated by Figure 4, which
shows a systematic decision-making advantage with idiothetic
information (left panel) but no reliable advantage without it (right
panel). The pattern suggests that graph learning may be enhanced
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1174 CHRASTIL AND WARREN
by the combination of decision making and idiothetic information,
perhaps by augmenting local metric information in a weighted
graph. However, because the interaction between decision making
and information was not statistically significant, we cannot draw a
firm conclusion. It is possible that the lack of a significant
decision-making effect in the video condition was due to insuffi-
cient statistical power, given that the video groups (n⫽16) were
smaller than the walking groups (n⫽32). However, the stable
effect size in the video condition was half that of the walking
condition. We thus believe the pattern of results is well-
represented by Figure 4, implying that any influence of decision
making in the video condition is quite weak.
This observation might help explain why we find a reliable
decision-making advantage when prior research was inconclusive
(Farrell et al., 2003;Péruch et al., 1995;Tan et al., 2006;Wilson,
1999;Wilson et al., 1997;Wilson & Péruch, 2002). These studies
were conducted in desktop VR, under conditions similar to our
video condition. A weak influence of decision making without
idiothetic information would account for these equivocal results.
Conversely, a decision-making advantage with idiothetic informa-
tion would explain Hazen’s (1982) finding that walking children
learned routes better when they freely explored than when they
were led about by their parents. It is also consistent with Ruddle et
al.’s (2011a) report that the combination of decision making and
idiothetic information contributed to graph learning, although they
did not dissociate the two. We thus suggest that previous mixed
results for decision making may be attributable to the absence of
idiothetic information in desktop VR.
Route, Graph, and Survey Knowledge
The results support a third more general conclusion, that humans
acquire not merely route knowledge but graph knowledge of the
environment early in learning. This graph also includes some local
metric information, consistent with a labeled graph. Together with
our previous findings on survey knowledge (Chrastil & Warren,
2013), this finding implies that primary spatial knowledge is best
characterized as a labeled graph.
In contrast to the route knowledge prediction, participants did
not just reproduce routes that they had traveled during exploration,
but took novel routes to the target on three-quarters of the correct
trials. This finding indicates that participants had learned path
segments during exploration and could recombine them to gener-
ate novel routes during the test phase, consistent with graph
knowledge of the network of connections between places.
Furthermore, participants took the metrically shortest route on
two-thirds of correct trials, suggesting that they had also encoded
some metric information about these path segments. Participants
took the metrically shorter route rather than a topologically equiv-
alent but longer route in four out of five cases (see Table 4),
indicating that they did not just select a path with the fewest
number of segments or intersections. This result supports the
hypothesis that the primary structure of spatial knowledge is not
purely topological but can be characterized as a labeled graph that
incorporates local metric information. Moreover, as noted above
(see Table 3), participants who made decisions during exploration
took the shortest detours more often during test (significantly), as
did those who walked during exploration (marginally). This find-
ing reinforces the conclusion that decision making facilitates graph
learning and idiothetic information contributes local metric knowl-
edge.
Using a similar experimental design, we previously tested active
contributions to survey knowledge of the same hedge-maze envi-
ronment (Chrastil & Warren, 2013). We probed survey knowledge
with a direct shortcut task, in which participants were instructed to
walk from the start object to the target object on a straight-line
path, rather than staying in the corridors of the maze. Accurate
performance required survey knowledge of the metric distances
and directions between learned locations. Although participants in
the free walking condition were above chance overall, absolute
shortcut errors averaged around 70°, and half of the participants
were near chance (90°). Although a few participants may have
learned accurate survey knowledge (12% had absolute errors
around 20°), there is little indication that this is the primary form
of spatial knowledge acquired. In contrast, in the present study
only 10% of participants in the free walking condition performed
at the chance level, whereas half were successful in route finding
on about 60% or more trials. This pattern of results implies that the
majority of participants acquired graph knowledge but not survey
knowledge, when freely exploring the same environment.
Together, these findings suggest that the primary form of human
spatial knowledge may be best described as a labeled graph that
incorporates local metric information (Chrastil & Warren, 2014b).
Such a knowledge structure would be sufficient for tasks including
route finding, taking novel detours, choosing the shortest route,
and even taking approximate shortcuts, without requiring a glob-
ally consistent metric cognitive map (Byrne, 1979;Chown et al.,
1995;Kuipers et al., 2003;Meilinger, 2008;Trullier et al., 1997;
Werner et al., 2000). Behavior that appears to implicate a global
metric map could actually be supported by a labeled graph, despite
rough and inconsistent metric information.
The pattern of errors offers some additional clues about the
information that may be encoded in a graph structure. The analysis
of target difficulty revealed that the best predictor of errors was the
number of turns in the maze between the start location and the
target. For example, referring to Figure 1, the easiest object pair
(bookcase–well) was separated by only two turns, whereas the
hardest pair (snowman–rabbit) was separated by four. Thus, dif-
ficulty in learning the graph structure is related to the number of
the turns in the trajectory, not just the number of intersections,
suggesting that turns (with angle labels) may be entered as nodes
in graph knowledge. In addition, the rabbit was located in a branch
hallway whose shape was almost identical to that of three other
objects. Indeed, on rabbit trials participants tended to go to the gear
location more frequently, for it had the same branch hallway shape
and was closer to the start location (snowman). Such errors imply
that participants may have occasionally relied on a view-based
strategy, which suggests that view information may also be incor-
porated into graph knowledge.
The Exploration-Specific Learning Hypothesis
Together with our previous results, the present findings support
the exploration-specific learning hypothesis that the components
of active learning differentially contribute to particular forms of
spatial knowledge. Specifically, decision making is the primary
component of active learning for the acquisition of topological
graph knowledge, whereas idiothetic information is the primary
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1175
ACTIVE AND PASSIVE NAVIGATION
component for metric survey knowledge. Idiothetic information
may also contribute local metric information to a labeled graph.
Chrastil and Warren’s (2013) shortcut test demonstrated that
idiothetic information (specifically podokinetic information
during walking) made a significant contribution to survey
knowledge, reflected in lower absolute errors in the direction of
shortcuts. In contrast, decision making during exploration made
no significant contribution to survey knowledge. Those findings
stand in contrast to the present results for graph knowledge, in
which decision making makes a significant contribution but
idiothetic information by itself does not. We note that the
decision-making effect was significant in combination with
idiothetic information, consistent with a contribution of local
metric information to labeled graph knowledge.
The contrasting results in these two experiments imply that
the components of active learning differentially affect particular
types of spatial knowledge. Decision making helped partici-
pants in the present experiment learn the graph structure of the
maze, presumably by facilitating acquisition of the topological
connections between locations. But it did not contribute to
survey knowledge (Chrastil & Warren, 2013), which depends
on metric information about the distances and angles between
those locations. A navigator with graph knowledge may be able
find a route through the maze to the target location without
knowing its metric location, whereas an accurate shortcut re-
quires knowing the target’s direction and distance with respect
to the starting position.
On the other hand, idiothetic information from walking spec-
ifies the metric distances and angles needed to acquire survey
knowledge. Indeed, Chrastil and Warren (2013) found that
walking during exploration led to more accurate shortcuts in the
test phase. Such metric relations may be derived from path
integration (McNaughton et al., 2006), and idiothetic informa-
tion has been shown to contribute to path integration over and
above visual information alone (Kearns, 2003;Tcheang et al.,
2011). It is possible that idiothetic information also plays a
subtle role in graph learning. We find that navigators learn not
just the connectivity of their environment, but a labeled graph
that incorporates some local metric information (Chrastil &
Warren, 2014b). This knowledge could aid route finding by
constraining the direction of initial segments that lead to a
successful path and the selection of the shortest path. If a
labeled graph is the primary form of spatial knowledge, it
would explain why we observed a decision-making advantage
in the presence of idiothetic information. Thus, graph and
survey knowledge rely on different types and amounts of in-
formation during learning and possibly different learning mech-
anisms.
If different components of active learning influence graph
and survey knowledge, it may indicate that these forms of
spatial knowledge can be learned independently. Siegel and
White (1975) proposed that landmark information is learned
first, followed by route knowledge, and finally metric survey
knowledge. However, recent evidence suggests that this order-
ing may not always hold. Ishikawa and Montello (2006) re-
ported that some people acquire survey knowledge and route
knowledge simultaneously, whereas others never acquire sur-
vey knowledge despite repeated exposure to the environment.
Our results indicate that graph knowledge is acquired early in
learning, after only 10 min of exploration, and is exhibited early
in the test phase. Neural evidence indicates that brain areas
supporting habitual route learning are active at a later stage of
learning (e.g., Hartley, Maguire, Spiers, & Burgess, 2003;
Packard & McGaugh, 1996). These findings imply that the
primary form of spatial knowledge has a labeled graph structure
and is easier to learn than route or survey knowledge, consistent
with the idea that navigators acquire different types of spatial
knowledge at different stages of learning (Tversky, 1993).
Sex Differences
We also observed a consistent sex difference in graph learning,
such that women tended to perform worse than did men, overall.
This result is particularly noticeable in the free decision-making
groups, where the difference in proportion correct was 0.205.
However, female participants reported less experience with video
games and higher ratings of nausea, especially in the video con-
ditions. Although nausea was not correlated with performance, it is
possible that women were more distracted in these conditions than
were men. Our results are consistent with previous research that
has found large sex differences in spatial navigation (Moffat,
Hampson, & Hatzipantelis, 1998;Waller, 2000;Wolbers & He-
garty, 2010), although other researchers have found few differ-
ences, particularly in route learning (Castelli, Corazzini, & Gemi-
niani, 2008;Coluccia & Louse, 2004). The present results show
that women performed lower on tests of spatial ability and graph
learning than did men, but they do not illuminate the source of this
difference. For example, spatial abilities may be susceptible to
stereotype threat (Spencer, Steele, & Quinn, 1999), where mem-
bership in a group that has stereotypically poor skills in a particular
area may adversely affect performance in that area. Women are
typically associated with lower spatial skills, and their perfor-
mance in spatial tasks has been shown to suffer depending on the
level of stereotype threat or anxiety about spatial skills (Lawton &
Kallai, 2002;Martens, Johns, Greenberg, & Schimel, 2006;Mc-
Glone & Aronson, 2006). It is possible that this was a factor in the
observed differences.
Finally, there were also large individual differences in route
finding performance. Even in the free walking group, the percent
correct ranged from nearly 0% to 100% (see Figure 4A). One
quarter of participants in the free walking group successfully
located virtually all targets, whereas 15% of participants were near
chance. This distribution of performance is similar to that reported
by Ishikawa and Montello (2006) in a more difficult survey-
learning task conducted over multiple sessions. Further analysis of
individual differences and their relationship to route and graph
performance is planned for a future article.
Conclusion
Overall, we found that decision making during exploration plays
an important role in learning the graph structure of an environ-
ment. This result is consistent with the exploration-specific learn-
ing hypothesis that making decisions helps to link paths and places
together in graph knowledge but does not contribute to survey
knowledge. It is also possible that the role of decision making is
enhanced by the presence of idiothetic information, which could
contribute local metric information to labeled graph knowledge.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1176 CHRASTIL AND WARREN
We began by posing several questions about active and passive
spatial learning in a typical scenario: Driving back from a party
after riding as a passenger on the way there. To improve your
chances of getting home, our results suggest that driving to the
party while receiving directions would not have been helpful.
Similarly, walking to the party would not have helped either. But
making decisions about the route to the party, particularly while
walking there, may be optimal. In contrast, if you want to take a
direct shortcut home as the crow flies, walking there is your best
bet.
In sum, there are systematic differences between active and
passive spatial learning. Idiothetic information and decision mak-
ing are both important components of active spatial learning, over
and above passive visual exposure. However, they make differen-
tial contributions to spatial knowledge: Decision making contrib-
utes to topological graph knowledge, whereas idiothetic informa-
tion contributes to survey knowledge and may also contribute to
local metric knowledge in a labeled graph.
References
Bar, M. (2007). The proactive brain: Using analogies and associations to
generate predictions. Trends in Cognitive Sciences, 11, 280 –289. http://
dx.doi.org/10.1016/j.tics.2007.05.005
Benhamou, S. (2010). Orientation and Navigation. In G. F. Koob, M.
LeMoal, & R. F. Thomson (Eds.), Encyclopedia of Behavioral Neuro-
science (Vol. 2, pp. 497–503). London: Academic Press http://dx.doi
.org/10.1016/B978-0-08-045396-5.00106-8
Bertin, R. J. V., Israël, I., & Lappe, M. (2000). Perception of two-
dimensional, simulated ego-motion trajectories from optic flow. Vision
Research, 40, 2951–2971. http://dx.doi.org/10.1016/S0042-6989(00)
00134-6
Buckner, R. L. (2010). The role of the hippocampus in prediction and
imagination. Annual Review of Psychology, 61, 27–, 48, C1–C8. http://
dx.doi.org/10.1146/annurev.psych.60.110707.163508
Byrne, R. W. (1979). Memory of urban geography. The Quarterly Journal
of Experimental Psychology, 31, 147–154. http://dx.doi.org/10.1080/
14640747908400714
Castelli, L., Corazzini, L. L., & Geminiani, G. C. (2008). Spatial navigation
in large-scale virtual environments: Gender differences in survey tasks.
Computers in Human Behavior, 24, 1643–1667. http://dx.doi.org/10.1016/
j.chb.2007.06.005
Chance, S. S., Gaunet, F., Beall, A. C., & Loomis, J. M. (1998). Locomo-
tion mode affects the updating of objects encountered during travel: The
contribution of vestibular and proprioceptive inputs to path integration.
Presence (Cambridge, Mass.), 7, 168 –178. http://dx.doi.org/10.1162/
105474698565659
Chown, E., Kaplan, S., & Kortenkamp, D. (1995). Prototypes, location, and
associative networks (PLAN): Towards a unified theory of cognitive
mapping. Cognitive Science, 19, 1–51. http://dx.doi.org/10.1207/
s15516709cog1901_1
Chrastil, E. R., & Warren, W. H. (2012). Active and passive contributions
to spatial learning. Psychonomic Bulletin & Review, 19, 1–23. http://dx
.doi.org/10.3758/s13423-011-0182-x
Chrastil, E. R., & Warren, W. H. (2013). Active and passive spatial
learning in human navigation: Acquisition of survey knowledge. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 39,
1520 –1537. http://dx.doi.org/10.1037/a0032382
Chrastil, E. R., & Warren, W. H. (in press). Role of attention in spatial
learning of graph and survey knowledge. Unpublished manuscript.
Chrastil, E. R., & Warren, W. H. (2014). From cognitive maps to cognitive
graphs. PLOS ONE, 9 (11), e112544.
Coluccia, E., & Louse, G. (2004). Gender differences in spatial orientation:
A review. Journal of Environmental Psychology, 24, 329 –340. http://
dx.doi.org/10.1016/j.jenvp.2004.08.006
Farrell, M. J., Arnold, P., Pettifer, S., Adams, J., Graham, T., & MacMa-
namon, M. (2003). Transfer of route learning from virtual to real
environments. Journal of Experimental Psychology: Applied, 9, 219 –
227. http://dx.doi.org/10.1037/1076-898X.9.4.219
Frenz, H., Bremmer, F., & Lappe, M. (2003). Discrimination of travel
distances from ‘situated’ optic flow. Vision Research, 43, 2173–2183.
http://dx.doi.org/10.1016/S0042-6989(03)00337-7
Frick, R. W. (1998). A better stopping rule for conventional statistical tests.
Behavior Research Methods, Instruments, & Computers, 30, 690 – 697.
http://dx.doi.org/10.3758/BF03209488
Gallistel, C. R. (1990). The organization of learning. Cambridge, MA: The
MIT Press.
Grant, S. C., & Magee, L. E. (1998). Contributions of proprioception to
navigation in virtual environments. Human Factors, 40, 489 – 497. http://
dx.doi.org/10.1518/001872098779591296
Hartley, T., Maguire, E. A., Spiers, H. J., & Burgess, N. (2003). The
well-worn route and the path less traveled: Distinct neural bases of route
following and wayfinding in humans. Neuron, 37, 877– 888. http://dx
.doi.org/10.1016/S0896-6273(03)00095-3
Hazen, N. L. (1982). Spatial exploration and spatial knowledge: Individual
and developmental differences in very young children. Child Develop-
ment, 53, 826 – 833. http://dx.doi.org/10.2307/1129399
Hegarty, M., Richardson, A. E., Montello, D. R., Lovelace, K., & Subbiah,
I. (2002). Development of a self-report measure of environmental spatial
ability. Intelligence, 30, 425– 447. http://dx.doi.org/10.1016/S0160-
2896(02)00116-2
Ishikawa, T., & Montello, D. R. (2006). Spatial knowledge acquisition
from direct experience in the environment: Individual differences in the
development of metric knowledge and the integration of separately learned
places. Cognitive Psychology, 52, 93–129. http://dx.doi.org/10.1016/j
.cogpsych.2005.08.003
Israel, I., & Warren, W. H. (2005). Vestibular, proprioceptive, and visual
influences on the perception of orientation and self-motion in humans. In
S. I. Wiener & J. S. Taube (Eds.), Head direction cells and the neural
mechanisms of spatial orientation (pp. 347–381). Cambridge, MA: MIT
Press.
Kearns, M. J. (2003). The roles of vision and body senses in a homing task:
The visual environment matters. Unpublished doctoral dissertation,
Brown University.
Koenderink, J. J., van Doorn, A. J., & Lappin, J. S. (2000). Direct
measurement of the curvature of visual space. Perception, 29, 69 –79.
http://dx.doi.org/10.1068/p2921
Kozhevnikov, M., & Hegarty, M. (2001). A dissociation between object
manipulation spatial ability and spatial orientation ability. Memory &
Cognition, 29, 745–756. http://dx.doi.org/10.3758/BF03200477
Kuipers, B., Tecuci, D. G., & Stankiewicz, B. J. (2003). The skeleton
in the cognitive map: A computational and empirical exploration.
Environment and Behavior, 35, 81–106. http://dx.doi.org/10.1177/
0013916502238866
Lawton, C. A., & Kallai, J. (2002). Gender differences in wayfinding
strategies and anxiety about wayfinding: A cross-cultural compari-
son. Sex Roles, 47(9/10): 389C1– 401http://dx.doi.org/10.1023/A:
1021668724970
Loomis, J. M., Da Silva, J. A., Fujita, N., & Fukusima, S. S. (1992). Visual
space perception and visually directed action. Journal of Experimental
Psychology: Human Perception and Performance, 18, 906 –921. http://
dx.doi.org/10.1037/0096-1523.18.4.906
Martens, A., Johns, M., Greenberg, J., & Schimel, J. (2006). Combating
stereotype threat: The effect of self-affirmation on women’s intellectual
performance. Journal of Experimental Social Psychology, 42, 236 –243.
http://dx.doi.org/10.1016/j.jesp.2005.04.010
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1177
ACTIVE AND PASSIVE NAVIGATION
McGlone, M. S., & Aronson, J. (2006). Stereotype threat, identity salience,
and spatial reasoning. Journal of Applied Developmental Psychology,
27, 486 – 493. http://dx.doi.org/10.1016/j.appdev.2006.06.003
McNaughton, B. L., Battaglia, F. P., Jensen, O., Moser, E. I., & Moser,
M.-B. (2006). Path integration and the neural basis of the ‘cognitive
map’. Nature Reviews Neuroscience, 7, 663– 678. http://dx.doi.org/
10.1038/nrn1932
Meilinger, T. (2008). The network of reference frames theory: A synthesis
of graphs and cognitive maps. In C. Freksa, N. S. Newcombe, P.
Gärdenfors, & S. Wölfl (Eds.), Spatial Cognition VI (pp. 344 –360).
Berlin: Springer. http://dx.doi.org/10.1007/978-3-540-87601-4_25
Mittelstaedt, M.-L., & Mittelstaedt, H. (2001). Idiothetic navigation in
humans: Estimation of path length. Experimental Brain Research, 139,
318 –332.
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. http://dx.doi.org/10.1016/S1090-5138(97)00104-9
Money, J., & Alexander, D. (1966). Turner’s syndrome: Further demon-
stration of the presence of specific cognitional deficiencies. Journal of
Medical Genetics, 3, 47– 48. http://dx.doi.org/10.1136/jmg.3.1.47
Norman, J. F., Crabtree, C. E., Clayton, A. M., & Norman, H. F. (2005).
The perception of distances and spatial relationships in natural outdoor
environments. Perception, 34, 1315–1324. http://dx.doi.org/10.1068/
p5304
Packard, M. G., & McGaugh, J. L. (1996). Inactivation of hippocampus or
caudate nucleus with lidocaine differentially affects expression of place
and response learning. Neurobiology of Learning and Memory, 65,
65–72. http://dx.doi.org/10.1006/nlme.1996.0007
Péruch, P., Vercher, J., & Gauthier, G. M. (1995). Acquisition of spatial knowl-
edge through visual exploration of simulated environments. Ecological Psy-
chology, 7, 1–20. http://dx.doi.org/10.1207/s15326969eco0701_1
Richardson, A. E., Powers, M. E., & Bousquet, L. G. (2011). Video game
experience predicts virtual, but not real navigation performance. Com-
puters in Human Behavior, 27, 552–560. http://dx.doi.org/10.1016/j.chb
.2010.10.003
Ruddle, R. A., Volkova, E., & Bülthoff, H. H. (2011a). Walking improves
your cognitive map in environments that are large-scale and large in
extent. ACM Transactions on Computer-Human Interaction, 18 (Art.
10), 1–20.
Ruddle, R. A., Volkova, E., Mohler, B., & Bülthoff, H. H. (2011b). The
effect of landmark and body-based sensory information on route knowl-
edge. Memory & Cognition, 39, 686 – 699. http://dx.doi.org/10.3758/
s13421-010-0054-z
Schacter, D. L., Addis, D. R., & Buckner, R. L. (2007). Remembering the
past to imagine the future: The prospective brain. Nature Reviews
Neuroscience, 8, 657– 661. http://dx.doi.org/10.1038/nrn2213
Siegel, A. W., & White, S. H. (1975). The development of spatial repre-
sentations of large-scale environments. In H. W. Reese (Ed.), Advances
in child development (Vol. 10, pp. 9 –55). New York: Academic Press.
http://dx.doi.org/10.1016/S0065-2407(08)60007-5
Spencer, S. J., Steele, C. M., & Quinn, D. M. (1999). Stereotype threat and
women’s math performance. Journal of Experimental Social Psychol-
ogy, 35, 4 –28. http://dx.doi.org/10.1006/jesp.1998.1373
Tan, D. S., Gergle, D., Scupelli, P., & Pausch, R. (2006). Physically large
displays improve performance on spatial tasks. ACM Transactions on
Computer-Human Interaction, 13, 71–99. http://dx.doi.org/10.1145/
1143518.1143521
Tcheang, L., Bülthoff, H. H., & Burgess, N. (2011). Visual influence on
path integration in darkness indicates a multimodal representation of
large-scale space. PNAS Proceedings of the National Academy of Sci-
ences of the United States of America, 108, 1152–1157. http://dx.doi
.org/10.1073/pnas.1011843108
Thrun, S. (2008). Simultaneous localization and mapping. In B. Siciliano,
O. Khatib, & F. Groen (Eds.), Robotics and cognitive approaches to
spatial mapping (Vol. 38, p. 13– 41). New York: Springer. http://dx.doi
.org/10.1007/978-3-540-75388-9_3
Trullier, O., Wiener, S. I., Berthoz, A., & Meyer, J.-A. (1997). Biologically
based artificial navigation systems: Review and prospects. Progress in
Neurobiology, 51, 483–544. http://dx.doi.org/10.1016/S0301-0082
(96)00060-3
Tversky, B. (1993). Cognitive maps, cognitive collages, and spatial mental
models. In A. U. Frank & I. Campari (Eds.), Spatial Information Theory
A Theoretical Basis for GIS (Vol. 716, pp. 14 –24). Berlin, Heidelberg:
Springer Berlin Heidelberg. http://dx.doi.org/10.1007/3-540-57207-4_2
von Stülpnagel, R., & Steffens, M. C. (2013). Active route learning in
virtual environments: Disentangling movement control from intention,
instruction specificity, and navigation control. Psychological Research,
77, 555–574. http://dx.doi.org/10.1007/s00426-012-0451-y
Voss, J. L., Gonsalves, B. D., Federmeier, K. D., Tranel, D., & Cohen, N. J.
(2011). Hippocampal brain-network coordination during volitional ex-
ploratory behavior enhances learning. Nature Neuroscience, 14, 115–
120. http://dx.doi.org/10.1038/nn.2693
Waller, D. (2000). Individual differences in spatial learning from
computer-simulated environments. Journal of Experimental Psychol-
ogy: Applied, 6, 307–321. http://dx.doi.org/10.1037/1076-898X.6.4.307
Waller, D., & Greenauer, N. (2007). The role of body-based sensory
information in the acquisition of enduring spatial representations. Psy-
chological Research, 71, 322–332. http://dx.doi.org/10.1007/s00426-
006-0087-x
Waller, D., Loomis, J. M., & Haun, D. B. (2004). Body-based senses
enhance knowledge of directions in large-scale environments. Psycho-
nomic Bulletin & Review, 11, 157–163. http://dx.doi.org/10.3758/
BF03206476
Weber, K. D., Fletcher, W. A., Gordon, C. R., Jones, G. M., & Block,
E. W. (1998). Motor learning in the “podokinetic” system and its role in
spatial orientation during locomotion. Experimental Brain Research,
120, 377–385. http://dx.doi.org/10.1007/s002210050411
Werner, S., Krieg-Bruckner, B., & Herrmann, T. (2000). Modelling navi-
gational knowledge by route graphs. In C. Freksa, C. Habel, W. Brauer,
& K. F. Wender (Eds.), Spatial Cognition (pp. 295C1–316). Berlin:
Springer-Verlag http://dx.doi.org/10.1007/3-540-45460-8_22
Wiener, J. M., Buchner, S. J., & Holscher, C. (2009). Taxonomy of
human wayfinding tasks: A knowledge-based approach. Spatial Cog-
nition and Computation, 9, 152–165. http://dx.doi.org/10.1080/
13875860902906496
Wilson, P. (1999). Active exploration of a virtual environment does not
promote orientation or memory for objects. Environment and Behavior,
31, 752–763. http://dx.doi.org/10.1177/00139169921972335
Wilson, P., Foreman, N., Gillett, R., & Stanton, D. (1997). Active versus
passive processing of spatial information in a computer-simulated envi-
ronment. Ecological Psychology, 9, 207–222. http://dx.doi.org/10.1207/
s15326969eco0903_3
Wilson, P., & Péruch, P. (2002). The influence of interactivity and atten-
tion on spatial learning in a desk-top virtual environment. Cahiers De
Psychologie Cognitive/Current Psychology of Cognition, 21, 601– 633.
Wolbers, T., & Hegarty, M. (2010). What determines our navigational
abilities? Trends in Cognitive Sciences, 14, 138 –146. http://dx.doi.org/
10.1016/j.tics.2010.01.001
Zacks, J. M., Mires, J., Tversky, B., & Hazeltine, E. (2000). Mental spatial
transformations of objects and perspective. Spatial Cognition and Com-
putation, 2, 315–332. http://dx.doi.org/10.1023/A:1015584100204
Received July 29, 2014
Revision received September 18, 2014
Accepted September 19, 2014 䡲
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1178 CHRASTIL AND WARREN
A preview of this full-text is provided by American Psychological Association.
Content available from Journal of Experimental Psychology Learning Memory and Cognition
This content is subject to copyright. Terms and conditions apply.
