The Dynamic Nature of Cognition During Wayfinding: Hugo J. Spiers, Eleanor A. Maguire

ARTICLE IN PRESS
Journal of Environmental Psychology 28 (2008) 232–249

www.elsevier.com/locate/jep
The dynamic nature of cognition during wayfinding

Hugo J. Spiers, Eleanor A. Maguire
Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, 12 Queen Square, London WC1N 3BG, UK
Available online 21 February 2008
Abstract
Much of our day-to-day wayfinding behaviour takes place in familiar large-scale urban environments, yet there is a dearth of studies
examining how wayfinding unfolds on a second-by-second basis in this context. Here we used a retrospective verbal report protocol, eye
tracking and a highly accurate virtual reality simulation of a real city (London, UK) to examine this issue. Subjects, who were taxi
drivers, were able to produce extremely detailed accounts of what they had been thinking during wayfinding, which were validated by
independent eye-tracking data. There was a high degree of consistency in the types of thoughts across subjects, permitting classification
into a number of distinct categories. Moreover, it was possible to quantify the number of thoughts in each category, their durations and
temporal order. Detailed analysis of the verbal reports provided new insights into the processes and strategies involved, and highlighted a
greater range of thoughts than has previously been reported in studies of wayfinding. By analysing the temporal order of thoughts it was
possible to identify specific relationships between categories. Some of these relationships were predicted by current cognitive models of
wayfinding, others were novel, thus shedding new light on how navigation unfolds in a busy city.
r 2008 Elsevier Ltd. Open access under CC BY license.
Keywords: Navigation; Virtual reality; Verbal reports; Planning; Spatial memory; Routes
1. Introduction relationships between locations/landmarks) (Siegel &

White, 1975; Thorndyke & Hayes-Roth, 1982). It is
Finding one’s way in spatially extended environments is thought that survey-like knowledge, also referred to as
essential for survival and requires a wide range of cognitive cognitive maps (Tolman, 1948), endows us with the ability
abilities. Pre-eminent among these is the ability to make use to take detours and shortcuts in familiar environments.
of long-term spatial memory to guide wayfinding. An In addition to the use of static tasks, such as sketching
understanding of how we might mentally represent large- maps, examination of in situ ongoing behaviour during
scale space has been gained from a variety of tests such as wayfinding has provided additional insights into the use of
sketching maps, direction pointing and giving route spatial representations, and identified factors affecting
descriptions (Chase, 1982; Garling & Garling, 1988; performance. Such factors include subjects’ spatial proces-
Giraudo & Peruch, 1988a, 1988b; Golledge, 1999; Pailhous sing abilities, their familiarity with the environment, the
1970, 1984; Peruch, Giraudo, & Garling, 1989; Thorndyke density/salience of landmarks, and the layout of the
& Hayes-Roth, 1982). These studies suggest that multiple, environment (Foo, Warren, Duchon, & Tarr, 2005; Gillner
qualitatively different types of spatial representations can & Mallot, 1998; Golledge, 1999; Holscher, Meilinger,
support wayfinding, including route knowledge (knowing Vrachliotis, Brosamle, & Knauff, 2007; Janzen, Schade,
the directions to turn at locations/landmarks) and survey- Katz, & Hermann, 2001; Newman et al., 2006; Thorndyke
like knowledge (an integrated knowledge of the spatial & Hayes-Roth, 1982; Wiener, Schnee, & Mallot, 2004). In
particular, and of primary interest here, investigating how
Corresponding author. Current address: Department of Psychology,
routes are chosen or planned has proved useful for
Institute of Behavioural Neuroscience, University College London, 26 identifying a number of wayfinding strategies (Conroy-
Bedford Way, London WC1H 0AP, UK. Tel.: +44 20 7679 7553;
fax: +44 20 7813 1420.
Dalton, 2003; Golledge, 1995; Hochmair & Frank, 2002;
E-mail addresses: [email protected] (H.J. Spiers), Pailhous, 1970, 1984; Wiener & Mallot, 2003; Wiener et al.,
e.maguire@fil.ion.ucl.ac.uk (E.A. Maguire). 2004). For example, Pailhous (1970, 1984) found that
0272-4944 r 2008 Elsevier Ltd. Open access under CC BY license.

doi:10.1016/j.jenvp.2008.02.006
ARTICLE IN PRESS
H.J. Spiers, E.A. Maguire / Journal of Environmental Psychology 28 (2008) 232–249 233
Parisian taxi drivers tended to make use of a familiar nature, verbal report protocols can only give insight into
primary network of streets to facilitate navigation. By the processes which subjects are aware of and are able to
contrast, in another strategy known as the least-angle verbalize. Nevertheless, the reports gained can help provide
strategy, paths are chosen that minimize deviation from the evidence to distinguish between competing models seeking
angle pointing directly to the goal (Conroy-Dalton, 2003). to explain cognition during a task (see Ericsson & Simon,
Different again is the fine-to-coarse planning heuristic that 1980; Jack & Roepstorff, 2003).
argues routes are planned in fine detail in the currently Passini (1981, 1984, 1992) was the first to critically assess
occupied region, but only coarsely when planning naviga- the wayfinding process with a verbal report protocol. His
tion between regions (Wiener & Mallot, 2003). Finally, the results, in agreement with other models, provided evidence
least-decision-load strategy specifies that subjects will often for two core stages in wayfinding: formulating the route
choose the path with the least number of possible decision plan and executing the plan. Analysis of the verbal reports
points (Wiener et al., 2004). also identified a hierarchical organization in the route
To account for this range of empirical findings, a number plans, with decisions at the top relating to over-arching
of cognitive models of wayfinding have been proposed goals, and decisions at the bottom related to sub-tasks.
(Garling, Book, & Lindberg, 1984; Kuipers, 1978; Kuipers, Route planning elements always contained two parts: a
Tecuci, & Stankiewicz, 2003; Passini, 1981, 1984; Peponis, behaviour component (e.g. turn left) and a location/
Zimring, & Choi, 1990; Stern & Portugali, 1999; Timpf, landmark specifier (e.g. at the central square). Thus,
Volta, Pollock, & Egenhofer, 1992). All models distinguish Passini argued that wayfinding could be conceptualized
between the processes of: (a) planning the route and (b) as a process in which route plans are initially set up and
executing the plan, and generally contain a hierarchy of executed by actions at the appropriate place and time
staged processes that unfold sequentially and iteratively leading finally to the goal. In his model, plans generate
during wayfinding. The models differ in the features they expectancies to find particular locations/landmarks in
emphasize and situations they cover. Some deal with order to perform the actions. These actions are triggered
selecting the route to multiple destinations (Garling et al., when a match occurs between a mental image of the
1984), others with navigating in unfamiliar environments expected locations/landmark and the experience of seeing
(Garling et al., 1984; Peponis et al., 1990) or city streets it. When there is no such match a new plan must be
(Garling et al., 1984; Kuipers, 1978; Kuipers et al., 2003; formulated to solve the problem.
Timpf et al., 1992). For example, Kuipers’ models Since Passini’s (1981, 1984, 1992) seminal studies, verbal
(Kuipers, 1978; Kuipers et al., 2003) build on the findings reports been used to examine how a number of factors
of Pailhous (1969, 1984) by using a primary network of affect wayfinding cognition and strategy use. These factors
roads to facilitate navigation. Timpf et al.’s (1992) model have included the use of maps (Gerber & Kwan, 1994),
deals with the translation from route plans into the actions subjects’ navigational aptitude (Holscher et al., 2007; Kato
necessary to perform driving manoeuvres. & Takeuchi, 2003), familiarity with the environment
Whilst observing in situ wayfinding has generated (Chebat et al., 2005; Holscher et al., 2007), the layout of
empirical and theoretical advances, one significant limita- the environment (Holscher et al., 2007) and wayfinding in
tion is that mere observation can make it difficult to the context of shopping (Chebat et al., 2005; Dogu &
determine the reasons or intentions behind some of the Erkip, 2000; Titus & Everett, 1996). Findings from these
different actions performed. For example, stopping at a studies have generally agreed with Passini’s (1981, 1984)
junction might be motivated by the need to reorient, look model, and have helped to identify the use of a number of
ahead, make a decision, or any number of other reasons. In additional wayfinding strategies (Gerber & Kwan, 1994;
order to dissociate such possibilities several studies have Holscher et al., 2007; Kato & Takeuchi, 2003). For
employed verbal report protocols (Chebat, Gelinas-Che- example, Holscher et al. (2007) found some subjects
bat, & Therrien, 2005; Dogu & Erkip, 2000; Gerber & completely planned the whole route before setting off,
Kwan, 1994; Holscher et al., 2007; Kato & Takeuchi, 2003; while others would formulate only partial plans. The latter
Passini, 1981, 1984; Titus & Everett, 1996). These protocols was found to be associated with poorer performance. Kato
involve subjects either ‘thinking aloud’ during the task and Takeuchi (2003) found that some subjects were able to
(concurrent verbal reports) or describing what they flexibly use a range of information to navigate, whereas
remember thinking during a structured interview after the others employed ineffective strategies, such as using parked
experimental task (retrospective verbal reports) (Ericsson cars as landmarks.
& Simon, 1980). The contents of the verbal reports are While these studies have shed some light on the cognitive
analysed with a pre-determined classification scheme processes during wayfinding, several important issues
usually developed from pilot studies. These classification remain unexplored. The environments of the previous
schemes make use of commonalities in the statements studies that used verbal reports were either indoors
contained in the verbal reports to derive a number of (Chebat et al., 2005; Dogu & Erkip, 2000; Holscher
relevant ‘thought’ categories. Examining the categories and et al., 2007; Passini, 1981, 1984; Titus & Everett, 1996) or
their contents can then be used to understand the cognitive unfamiliar (Gerber & Kwan, 1994; Kato & Takeuchi,
task in question and potential strategies employed. By their 2003). Thus, despite the fact that much of our daily
ARTICLE IN PRESS
234 H.J. Spiers, E.A. Maguire / Journal of Environmental Psychology 28 (2008) 232–249
wayfinding occurs in familiar outdoor environments, the use of VR London made it possible to assess navigation in
cognitive processes involved in navigating them has yet to a familiar yet complex large-scale environment in a
be examined. Prior studies have generally focused on controlled way that simply would not be possible with in
decision making or strategy use, ignoring other compo- situ navigation in this city. In addition, use of VR
nents such as the visual processing of the environment or permitted recording of subjects’ performance and allowed
the entities within it. No study has examined how for collection of eye-tracking data (for more on VR versus
wayfinding unfolds on second-by-second basis in a large- real environments see Morganti, Carassa, & Geminani,
scale environment. Appreciating the temporally fine- 2007; Peruch & Gaunet, 1998; Richardson, Montello, &
grained behaviour during navigation could give important Hegarty, 1999; Stanton, Wilson, & Foreman, 2002). In the
insights into the underlying spatial representations and the virtual London game used in this experiment over 110 km
wayfinding processes that act upon it. Related to this, (70 miles) of driveable roads have been accurately
quantitative examination of the temporal relationships recreated from Ordinance Survey map data, covering
between different types of thoughts during wayfinding has approximately 50 km2 (20 square miles) of the city centre.
also been largely ignored. Finally, there has been little The game designers decided to truly recreate the city and a
attempt to use the information in verbal reports to large team of photographers walked the streets of central
critically evaluate different cognitive models of wayfinding London for two years recording details of the city.
(Garling et al., 1984; Kuipers, 1978; Kuipers et al., 2003; Buildings, shops, the one-way systems, working traffic
Passini, 1981, 1984; Peponis et al., 1990; Stern & Portugali, lights, the busy London traffic and an abundance of
1999; Timpf et al., 1992). Londoners going about their business are all included. The
Here we addressed these outstanding issues by using a ‘Free Roaming’ mode of the game was used, permitting
retrospective verbal report protocol, eye tracking and a free navigation with the normal game scenarios suspended.
highly accurate virtual reality simulation of a real city Subjects moved through the environment in a virtual
(London, UK) to explore the second-by-second thoughts London taxicab controlled using a game controller,
accompanying wayfinding. We also used the information consisting of two joysticks providing analogue control of
obtained from the verbal reports to examine temporal acceleration, braking and steering left and right. The
contingencies in the thoughts, and to evaluate cognitive virtual taxi had a higher degree of controllability than
models of wayfinding. many of the other vehicles in the game, and its low top
virtual speed could be maintained easily. Even slight
2. Materials and methods contact between a subject’s vehicle and other vehicles
could result in a crash which would prematurely terminate
2.1. Participants the experiment. Therefore a ‘cheat’ modification was
employed using Action Replay Max software (r Datel
Twenty healthy right-handed male licensed London taxi Design and Development Ltd., 2003) which prevented such
drivers participated in the experiment (mean age: 49.8 crashes. Subjects were instructed to drive ‘legally’ as they
years, SD: 8.5 years, range: 27–59 years). Taxi drivers were would in actual London. To assess the ecological validity
used in order to investigate wayfinding in subjects familiar of the virtual simulation, during the pilot testing, subjects
with the environment, and to provide a consistent level of were asked to comment on the similarity of the game to the
performance. The average time spent working as a taxi real city, and their sense of ‘presence’. All remarked that
driver was 18.3 years (SD: 10.9 years, range: 1–38 years). Central London (where the experiment was set) in
All had either lived in London their entire lives or for the particular was highly accurate and detailed with a high
vast majority of it and were naı̈ve to the stimuli used in the level of presence. Following the experiment proper subjects
experiment. All subjects gave informed written consent to were asked how similar the simulation and task was their
participation in accordance with the local research ethics everyday experience. All of the taxi drivers confirmed that
committee. the game was very reminiscent of their experience of
driving in central London.
2.2. The virtual environment
2.3. Pre-test training and familiarization
The video game ‘The Getaway’ (r Sony Computer
Entertainment Europe, 2002) run on a Sony Playstation2 Two weeks prior to testing, subjects were given over 2 h
(r Sony Computer Games Inc.) was used to present of practice with the game controls by asking them to
subjects with a ground-level interactive first person navigate to various locations in areas of environment not
perspective view of a simulation of central London, UK used in the experimental task. To avoid waiting for long
(see Fig. 1, and Spiers & Maguire, 2006a for a movie of periods at red traffic lights, subjects were familiarized with
navigation through the environment). While desk-top treating all traffic lights as green, but were otherwise
virtual reality simulations have their limitations, such as required to comply with all other road traffic regulations in
a lack of peripheral and binocular visual information, there the UK. Thirty minutes before testing subjects were again
are numerous advantages to using VR. In the current study given further practice in an area not used in experimental
ARTICLE IN PRESS
Fig. 1. Virtual London (UK). Panel (A) shows a map of the region of the simulation of London that was used in the navigation task (not all the minor
streets shown were included in the video game). Reproduced by permission of Geographers’ A-Z Map Co. Ltd. r Crown Copyright 2005. All rights
reserved. Licence number 100017302. Coloured lines indicate examples of typical routes driven by subjects to each of seven destinations during the
navigation task. Panel (B) shows example views from within the video game ‘The Getaway’ r 2002 Sony Computer Entertainment Europe. Left image
shows a view at Trafalgar Square, right image shows a view at Piccadilly Circus. These images are reproduced with the kind permission of Sony Computer
Entertainment Europe.
tasks. During this practice session subjects were trained to blocks where subjects responded to customers’ requests
respond to a set of recorded customers’ requests to take (heard via head phones) by delivering them to their
them to destinations in London. Prior to testing, subjects destinations. During each block one route was tested.
were told the locations they would be starting from in the Seven routes were included which were (in order): Kings
experimental tasks, but not the order. Cross to the Middlesex Hospital (initially to Guy’s
Hospital), Trafalgar Square to the Royal Courts of Justice
2.4. Experimental task (initially to the Old Bailey), St. Giles Circus to Peter Street
in Soho (initially to Paddington Station), St. Martin’s Place
The experimental tasks took place during functional to Leicester Square (initially to Covent Garden), Piccadilly
magnetic resonance imaging (fMRI) brain scanning. De- Circus to Leicester Square (initially to St. James’s Square),
tails of the scanning parameters and related fMRI analyses Buckingham Palace to the American Embassy (initially to
can be found in Spiers and Maguire (2006a, 2006b, 2007a, Shepard’s Market) and the American Embassy to Cavend-
2007b). In the experimental task, navigation was tested in ish Square (initially to Manchester Square). When the
ARTICLE IN PRESS
game came on the screen, subjects were given between 3 Ericsson and Simon (1980), the interviewer followed a
and 5 s to orient themselves in the environment. Following predetermined protocol during the interview. A subject’s
this they heard a customer request a destination (mean report was interrupted as little as possible, the interviewer
duration 2.0 s). For all routes, at some point during intervening only to improve the subject’s specification of
navigation the subjects heard customers request a change the onset and duration of thoughts where possible, and on
of destination (mean duration 3.0 s). For three of the routes occasion where clarification was required to later aid
an additional request to avoid a location or go via a analysis. The mean duration of the collection of the verbal
location was made by the customer (mean duration 3.7 s). reports was 108.9 min (SD: 16.9 min).
Requests for a change of destination or requests to go via
another location occurred at time points that varied 2.7. Analysis of the retrospective verbal reports
along each route and were unknown to the subject but
predetermined by the experimenters. Two subjects com- Anonymized audio information from the verbal report
pleted only four routes, in one case due to discomfort, the interviews was transcribed by a professional transcription
other due to a technical problem. Each block of navigation agency who were blind to the purpose of the experiment. By
ended when either the subject reached the destination or comparing the transcript with the time stamp from the
when a predetermined period of time elapsed. The original performance video, information about the timing of
maximum time taken to get to each of the destinations the thoughts was incorporated into the transcripts and any
was between 3 and 6 min (mean: 3.8 min, SD: 1.1 min). errors or unclear statements rectified. Each statement in the
Each block of navigation was separated by a period of rest transcript was then classified into one of a set of categories,
in which the subjects viewed a blank white screen for 60 s. and where appropriate sub-categories (see Table 1), and its
Total mean time for the experimental task was 31 min 35 s onset and duration recorded to create a segmented timeline
(SD: 4 min 9 s). of every subject’s experience of every route (see example in
Fig. 2). Unambiguous categories were predetermined by
2.5. Video recording analysis of common repeated statements in the verbal reports
of four subjects who took part in an in-depth pilot study
In order to debrief subjects and create an independent prior to the main experiment. This process was also guided
record of eye tracking, two videos were recorded during the by findings from previous studies employing verbal report
experimental task. Video output from the Playstation 2 was protocols and wayfinding tasks (Chebat et al., 2005; Dogu &
split into three ways: (1) to a projector presenting stimuli to Erkip, 2000; Gerber & Kwan, 1994; Holscher et al., 2007;
the subject on a screen in the MRI scanner (view angle of Kato & Takeuchi, 2003; Passini, 1981, 1984; Titus & Everett,
27.61), (2) to a VHS video recorder for debriefing and (3) to 1996). The pilot study also served to create a guide to aid
a video mixer to create an eye-tracking video. Video output classification and generate a protocol for the initial collection
going into the video mixer was combined with camera of the verbal reports. The independent eye-tracking video
footage of a stopwatch manually synchronized with the was used to aid the identification of onsets and durations
time stamp on the debriefing video. Gaze position cross where the subjects reported looking at a feature in the
hairs collected via an ASL504LRO infra-red eye-tracker environment, and so served as an external measure to
(Applied Science Laboratories, Bedford, MA) were over- validate the procedure. Environmental features consisted of
laid onto the video sent to the mixed video recording in 14 static structures (e.g. buildings and statues) and moving
subjects. Accurately calibrated eye-gaze tracking was objects (e.g. vehicles and pedestrians). For each report of
achieved in nine subjects. looking at a feature in the environment, the eye-tracking
video corresponding to that time point was examined to
2.6. Verbal report protocol determine when, to the nearest second, a saccade to the
feature was made and how long fixation of the feature
Immediately after the experimental task the subjects occurred in seconds. To assess the degree to which the verbal
were taken to a separate room where they were given a reports could be reliably categorized, a second experimenter
surprise debriefing with a verbal report protocol (Ericsson classified the verbal reports contained in 36 random extracts.
& Simon, 1980). In this debriefing, subjects watched the Of note, it was not just taxi drivers who were able to produce
video of their performance during the experimental task. detailed retrospective verbal reports. Several non-taxi driver
They were carefully instructed to describe what they subjects who navigated in the game through areas of the city
remembered thinking, step-by-step, during their original they were familiar with were able to produce reports as
performance. The interview proceeded at a pace deter- accurate as those of the taxi drivers.
mined by the subject, with the video being paused and
rewound by the interviewer where necessary to capture the 2.8. Analysis of the temporal precedence of
details provided by the subject. A new copy of the original thought categories
video was recorded during the interview with the voices of
the subject and interviewer collected by a microphone In order to understand the temporal relationships
overlaid. In accordance with the methods described by between categories, we examined whether the thoughts
ARTICLE IN PRESS
Table 1 Table 1 (continued)

Category definitions and example extracts from verbal reports
Extract 6
Route planning: Planning after a request from a customer to be Sub. 7 I was going to use Brunswick Square and
initial taken to a specific destination Guildford Street. Well, I wasn’t too sure whether
Extract 1 to come a bit low with it and go Montague Street,
Customer Please could you take me to the Old Bailey? Great Russell and then go up Earnshaw. That was
Sub. 1 I thought, great, turn right, straight down the a bit long winded I thought. Then I thought well
Strand, straight down Fleet Street and then left up we can get Judd Street and we can go down
to the Old Bailey Tavistock Place. There was three different options
Experimenter Did you picture the Old Bailey? came in my head
Sub. 1 No. I don’t think I did. I just know where it is
Route planning: Altering the current route plan
Extract 2 spontaneous—
Customer Please could you take me to Guys’ Hospital? re-planning
Sub. 13 As soon as he said Guy’s Hospital it’s like I’m Extract 7
sitting here in the cab, I know Guy’s Hospital is Sub. 12 And then I think hold on, there’s another way. It’s
there (subject points in the correct direction to like I said before, there always seems to be an
Guy’s Hospital). I know it’s in that direction. So I option comes in your way. Um, so I thought well,
know Guy’s is now south east. In my body it’s okay then, if I can turn right, then I’m going to go
behind me and to the east and to the south. up through the back of there and across the
Straight away I can seey it’s like I can see church
through all the buildings. It’s like if I said to you
picture from here toy the other room where we Extract 8
went, you can picture the whole surroundings, Sub. 9 Here, I’ve thought oh no, I’ve gone past
can’t you? Marlborough Gate, but I thought well, it doesn’t
Experimenter Yes mattery I’ll just do Piccadilly
Sub. 13 And the door and the hallway. That’s how it is
Action planning Planning future movements with the vehicle
with the roads
Extract 9
Extract 3 Sub. 2 I just thought, well, I’ll go through there and once
Customer Sorry, no, I need to get to Peter Street, the one off I turn right I’ll stay on the right hand-side
Wardour street
Extract 10
Sub. 3 I’m thinking where the hell is it in Wardour street
Sub. 6 I’m thinking, I must get over to the right now
(emphatic tone). So then, well I know instinctively
I’ve got to turn left cos it’s—I know where it isn’t. Extract 11
I know one little street that I’m not sure ofy it Sub. 9 I’m just thinking about getting over onto the left
must be that one, that was me instinct hand lane to turn left
Experimenter Did you picture Wardour street?
Sub. 3 I had to picture the whole length of Wardour Coasting Navigating automatically without any directed
street in my head very quickly just knocking off thoughts
the streets that I knew and then Peter Street was Extract 12
this little street. I kinda figured it might be that Sub. 1 You go into automatic pilot
one cos I knew it wasn’t this one, this one, this Experimenter Were you were in automatic pilot here?
one, or this one Sub. 1 Once I’d seen the underpass, yes. Because
Extract 4 everything is mapped out and I know which way
Customer I’d like to go via Glasshouse Street, please I’m going
Sub. 14 Well, I knew that’s just by Piccadilly Circus, Extract 13
which is over to our right. At this stage I don’t Sub. 3 I’ve kinda switched off here, just keeping the car
think I thought east or west. I knew it was like sort straight really. I use the term switched off, you
of two o’clock from here don’t actually switch off, but you switch off
Experimenter Was there any plan? thinking about your direction and maybe thinking
Sub. 14 Well, I knew I was in a one way street so I just had about what’s going on
to carry on
Extract 14
Route planning: Planning the next stage of the journey Sub. 11 It’s as if someone else is doing it for you, all you
spontaneous— are is a computer and you just program in and off
filling-in it goes
Extract 5
Sub. 3 So as I turn then I think right now—okay when I Extract 15
get to the top there’s a one-way system so yo’re Sub. 20 I’ve got a dead straight run for quite a bit, so you
forced left and right, left and right, there’s a don’t need to think much at all
multitude of ways you can go from here so—but I
Expectation Looking out for the next expected environmental
know I’ve got to go to the end of the street
feature
because fundamentally I don’t want to take a left
before that Extract 16
Sub. 1 Now I’m looking for the Euston underpass
ARTICLE IN PRESS
Table 1 (continued) Table 1 (continued)
Extract 17 I’m looking at the cabs over the other side in the
Sub. 2 I was expecting to see ity about now bus lane
Extract 32
Extract 18 Sub. 9 What I thought of here was: I’ll let that car go and
Sub. 6 Now I’m going to look out for it (Mortimer street) I was very relieved that it accelerated away from
Experimenter What was it that told you to start looking out for me
it?
Sub. 6 I’m thinking I’ve traveled a long way along here Theory of mind Thinking about other people’s thoughts/
now intentions
Extract 19 Extract 33
Sub. 21 Should be going downhill now, that’s what I was Sub. 11 I mean he [the customer] may be thinking I’m
thinking going to go down to Piccadilly Circus, back up
Shaftsbury Avenue and in through that way, I
Expectation Detecting the presence of an expected don’t know, you know, or he may have got
confirmation environmental feature
Glasshouse Street and Brewer Street mixed up
Extract 20
Extract 34
Sub. 1 And then all of a suddeny a little bit further on Sub. 12 [a car pulls out in front of him and the subject has to
you go through these lightsy and there it is. break to avoid a collision] I thought what’s he
There’s the cab shelter on the left-hand side doing?.. Why would he do that? There’s a line
Extract 21 there, he’s got to stop, you know, and thaty I’ve
Sub. 7 But then, that’s it, I knew that was Glasshouse got preference
Street because that’s the shape of it
Rule related Thinking about the one-way system and road
Extract 22 traffic rules
Sub. 9 And then when that yellow van comes up, yes,
that’s Orion House Extract 35
Sub. 3 And then I realised, no it’s one-way, I can’t go
Expectation Detecting the absence of an expected that way
violation environmental feature
Extract 36
Extract 23 Sub. 11 And then there, I thought to myself: don’t park on
Sub. 6 I thought, oh, bloody hell, they’ve blocked it off
the zigzags [road markings indicating no stopping]
Extract 24
Extract 37
Sub. 7 And then we come round here and then I got to Sub. 9 I thought no, no, no, it’s a one way street I won’t
about there and I thought oh, no, please no do that
(emphatic tone). And then it was confirmed it was
closed Emotions— Change in emotional state to being happy
happy
Sub. 14 Well, now I knew that I was coming up to
Sub. 6 I’m very happy with the way I’ve driven it now.
Tottenham Court Road, I wanted to go straight I’m chuffed to bits
on and here I saw that I couldn’t
Emotions— Change in emotional state to being angry
Visual inspection Visual inspection of an environmental feature angry
Sub. 3 As I’m going along here, I thought oh look that’s Sub. 21 I was getting annoyed, I was actually getting
Dwight [Statue of General Eisenhower] annoyed
Extract 27 Emotions— Change in emotional state to being anxious
Sub. 11 I was looking at it and I thought that’s the
anxious
American Embassy
Extract 40
Extract 28
Sub. 9 And I’m scared, like a horse, frightened by the
Sub. 17 As I turn, I’m looking at Boots [chemist shop] and railings
the other shops
Extract 29 Statements in square brackets and italics provide additional information
Sub. 7 I’m just checking it [the street] and I could see the about the environment or people referred to by the subject. Statements in
no entry arrow there parentheses and italics provide additional information about the subject’s
gestures and voice tone.
Monitoring Watching moving traffic in the environment
traffic
Extract 30
Sub. 5 Well, the bus is a nuisance. You can’t see in front
of it. It’s in the way
Extract 31
Sub. 7
ARTICLE IN PRESS
Fig. 2. The route driven by an example subject (subject 3) from Charing Cross Road to Peter Street in Soho is shown segmented into different colour-
coded thought categories (see explanatory key on the right-hand side) derived from the subject’s verbal report. Note that turning, not in the key, is shown
in black on the route. The route is indirect due to the one-way road system in Soho.
belonging to some categories preceded the thoughts Maguire, 2006a, 2006b, 2007a, 2007b). We now report new
belonging to other categories more or less often than analyses focused on the separate issue of understanding
would be expected by chance. This was done by construct- cognitive process during wayfinding, and how they inform
ing a contingency table for each subject which was a models of wayfinding.
transition matrix of all the possible pairings of categories.
This table contained the number of observed occurrences in 3.1. Behavioural performance and verbal reports
which events of each category directly preceded the events
of each other category. For each pair we also calculated the All subjects completed the task successfully with a mean
number of occurrences that would be expected by chance. of 94% (SD: 9%) of their routes being efficient. An
This was done by multiplying together the number of efficient route was one where the subject moved continually
events in each category and then dividing by the total closer to the goal destination given the constraints of
number of events across all categories. We then collapsed London’s one-way system and occasional obstructed
across subjects by summing the number of observed streets that were included in the game (see Fig. 1). Using
occurrences and summing the number of expected occur- mapping software (Map24(UK): http://www.uk.map24.
rences in each subject’s table to create one table. A w2 test com) it was possible to determine that subjects travelled a
was used to test whether the number of observed mean total distance of 16.9 virtual km (SD: 3.4) during the
occurrences provided a good fit to the number of expected experimental task.
occurrences in this table, with the degrees of free- Subjects were able to produce detailed accounts of what
dom ¼ (number of columns 1) (number of rows 1). they had been thinking during wayfinding. There was a
If a significant result was obtained, a w2 test was used for high degree of consistency in the types of thoughts across
each category to test whether events from this category the 20 subjects. Reviewing the transcriptions of all subjects,
preceded events in each of the other categories more or less and aided by a classification guide developed during the
often than was expected from chance. If a significant result piloting phase, 12,484 thoughts were classified into distinct
was obtained in any of these w2 tests, sign tests were then categories (see Table 1 for examples). Not only was the
used to probe individual category pairs. Sign tests level of detail in the retrospective verbal reports very high,
examined whether there were significantly more subjects subjects reviewing the video of their wayfinding perfor-
for whom the number of observed occurrences was greater mance were quite clear about when exactly they had
or less than the number of expected occurrences. experienced particular thoughts and in what order. This
enabled a complete specification of the wayfinding
3. Results experience in terms of the onsets, durations and temporal
order of thoughts for each subject. The precision of the
Aspects of the findings from this rich and flexible data set timings was further tested using the independent eye-
involving fMRI have been reported elsewhere (Spiers & tracking data acquired during the experimental task.
ARTICLE IN PRESS
In those subjects with accurate calibration, 94% (SD: 6%) direction to the location must be determined. This process
of reports of looking at a feature in virtual London was reported on by most subjects. A few subjects’
whilst navigating were corroborated by a saccade to its descriptions were in terms of cardinal directions (see
location at (or very near, 72 s) the time retrospectively Extract 2, Table 1), but most described it in an egocentric
reported. In addition, agreement was found between the reference frame (see Extract 4, Table 1 and Extract 7,
two experimenters for 93% of classifications arising from Appendix A). For many subjects, determining the direction
the random sample of statements that were dual-classified is the most important aspect of the process (see Extracts 6
(k ¼ 0.91). and 8, Appendix A). The next stage of route planning
involves street selection, and was more diverse across
3.2. Overview of how wayfinding unfolds in a familiar city routes and individuals. This ranged from only selecting
the next street, to completely specifying the streets to
Our analysis of the verbal reports permitted a detailed the destination. Some subjects reported thinking of the
breakdown of how wayfinding unfolds in a familiar large street names, while others did not (see Extracts 10–13,
city (see Table 1, Fig. 2). In summary, we initially plan our Appendix A). Most often the streets selected were
route to a destination; en route we might adjust our plan described starting with the nearest street and ending with
because of new opportunities or obstructed paths. Some- streets close to the destination (see Extract 1, Table 1 and
times we plan our route only to an intermediate point and Extract 5, Appendix A), but occasionally there were
once reached, we then fill in the rest of the route plan. reports of starting from the destination and working
Often within a familiar environment we are almost on backwards (see Extract 21, Appendix A). Street selection
automatic pilot, ‘coasting’ along without thinking. We also appeared to be affected by the distance or number of
set up expectations, waiting to see the next junction or a streets to the goal destination. In addition, when selecting
landmark to confirm we are on the right track, occasionally the streets subjects often accounted for the one-way road
inspecting the city around us as we travel through it. We traffic system. The time of day and day of the week were
monitor the surrounding traffic to achieve safe passage to also important considerations as they related to road
our destination, and have to be able to plan our own congestion and the obstruction of streets due to street
actions, such as staying in a traffic lane or changing lanes. markets. Strategies were varied, one of the most common
Sometimes our emotional state might change, or we pause being to initially plan a route that would end up facing in a
to consider the thoughts of the people we encounter. It is direction towards the destination, and then filling in the
notable how recognisable the above aspects are to our rest of the route (see Extracts 7 and 8, Appendix A).
everyday experience of urban wayfinding, and yet many of Another strategy was to plan a route to a particular region
them have gone largely undocumented. What one truly or access point and then plan the next part of the route
appreciates from the retrospective verbal reports is how, from there (see Extracts 23 and 24, Appendix A). When
second by second, the underlying cognition can change choosing between options, some subjects reported attempt-
(Fig. 2). We now consider each of the thought categories in ing to keep it simple rather than choosing intricate
turn, and provide qualitative characterization of their routes (see Extract 6, Table 1). Some subjects reported
salient features. planning a few different route options in case of problems
en route, and also reported re-assessing the route plan after
3.3. Route planning it had been selected (see Extracts 17, 25 and 26, Appendix
A). A few subjects described imagining driving down the
Route planning is a crucial stage in wayfinding and it route at high speed in their ‘mind’s eye’ as part of a
occurred more frequently than any other category (see checking process (see Extracts 17, 18 and 22, Appendix A).
Fig. 3). In Appendix A we provide detailed extracts from Reports of imagery were not confined to ground-level
verbal reports describing route planning. It was possible to views, but also very occasionally included aerial views
identify three sub-categories of route planning: initial (see Extract 19, Appendix A), but no imagery of a map was
route planning, filling-in and re-planning (see Table 1). In reported.
the current study, initial route planning was driven by The verbal reports revealed that route planning rarely
either the customers’ requests at the start of each route stops after the initial plan. En route subjects often filled-in
or the customers’ requests for a change in destination a plan for next stage of the journey or re-planned part of
en route. No differences were discernable in the verbal the route. These two other types of route planning were
reports of route planning following these two causes of subject to similar considerations and strategies involved in
initial planning. Across subjects, the occurrences and initial route planning, but typically fewer streets and
durations of both filling-in and re-planning were similar directions were specified. Filling-in was typically precipi-
(see Fig. 4). tated by having finally reached a street facing towards the
Examination of the verbal reports shows that initial destination, reaching the end of a previously planned route
route planning involved a number of distinct stages segment, or reaching a new region. Often the filling-in
(see Table 1 and Appendix A). In the first instance, the commenced when the subject had finished turning into the
destination’s location must be retrieved. After this, the street they planned to reach, rather than at the moment of
ARTICLE IN PRESS
Fig. 3. Frequency of thoughts and their durations across subjects for each category: (A) the mean number of occurrences for each thought category, (B)
the mean duration of each period for each thought category and (C) the mean total duration across all routes of each thought category.
spotting the street (see Extract 5, Table 1). Re-planning 3.4. Action planning, monitoring traffic and rule
often occurred because a street was obstructed (see Extract related thoughts
7, Table 1). Several subjects commented that this was quite
a common occurrence in London. Re-planning was also Once route planning is complete we need to make the
caused by subjects spontaneously realizing there was a appropriate actions to reach our goal. Many actions were
better route available, or very occasionally by accidentally made without subjects reporting any prior planning, such
passing the street they intended to use (see Extract 8, Table as keeping the vehicle in a straight line. However, other
1). There was a lack of mental imagery in descriptions of actions were reported to be preceded by a plan. These
filling-in and re-planning. action plans were distinguished from route plans by the
ARTICLE IN PRESS
Fig. 4. Route planning and emotions: sub-categories. (A) The total duration of each sub-category of route planning for each subject (rank ordered by total
number of events). (B) The total duration of each sub-category of emotion for each subject (rank ordered by total number of events).
fact that they specified a single action rather than a 3.5. Coasting
sequence of actions, and the action was to be executed at a
location currently in view. Action planning was the second Whilst route and action planning are the thought
most common category after route planning (see Fig. 3), categories that occurred most frequently, examining total
indicating that thinking about future events comprises a amount of time subjects spent in each cognitive state
considerable portion of the wayfinding experience. Many revealed a different story. For a good deal of the time spent
action plans involved thoughts about turning into a street wayfinding in this familiar city subjects reported ‘coasting’
that was currently in view and part of the route plan. along without any directed thoughts (see Fig. 3). This
Planned actions also included: changing lanes, making typically occurred on long stretches of road where subjects
U-turns and manoeuvring round moving vehicles. Action knew they did not need to make any further decisions
planning involving other traffic also involved monitoring (see Extracts 12 and 15, Table 1). Several subjects used the
the movements of these vehicles. Subjects also reported terms ‘automatic pilot’ or ‘switched off’ to describe the
monitoring the movement of vehicles on other occasions, experience.
such as when looking out for a street, to see if it was
accessible. Buses were often a concern for the subjects, 3.6. Visual inspection, expectation and confirmation or
since they provided the greatest risk of collision (see violation of expectations
Extract 30, Table 1). Like route planning, action planning
was also affected by road traffic rules, such as the one-way During any wayfinding excursion there are moments
streets and road markings (see Table 1). Often these related when we need to look around to acquire information about
to inhibiting particular actions, such as complying with ‘no the environment. Four distinct categories related to this
stopping’ signs (see Extract 36, Table 1). were identifiable in the verbal reports. Some periods of
ARTICLE IN PRESS
wayfinding were spent visually inspecting fixed features in Table 2

the environment, while on other occasions inspection Results from the temporal precedence analysis
occurred with the expectation of seeing a specific feature Category Sign test p-value
not yet in view (see Table 1 and Fig. 3). Visual inspection w2-score, p-value
could occur out of a general interest in the local
environment (e.g. Extracts 26, 27 and 28, Table 1) or in Route planning
135.4, po0.001
order to extract useful information from it, for example,
Was more often followed by:
from road signs (e.g. Extract 29, Table 1). Expectation Coasting 0.002
periods, by contrast, involved looking out for the next Rule related 0.004
turning to execute the route plan or a landmark to confirm Visual inspection 0.004
the subject was on the right path. Expectation periods often Was not often followed by:
Expectation confirmation 0.001
commenced after turning into a street, but also after certain Expectation violation o0.001
amount of time had elapsed driving down a street (see
Extract 18, Table 1). Often a period of expectation would Coasting
75.3, po0.001
be followed by confirmation that the feature sought had
been spotted. Alternatively on occasion the expectation –
was violated when, for example, the street was blocked off Was not often followed by:
or the expected feature was visually altered (e.g. scaffolding Route planning o0.001
covering up a landmark). Action planning o0.001
Rule related 0.003
3.7. Theory of mind and emotions Expectation violation 0.021
Action planning
In addition to planning, coasting and visual processing,
92.7, po0.001
subjects also reported thinking about the thoughts and Was more often followed by:
intentions of other individuals (known as ‘theory of mind’) Coasting 0.002
and also reported changes in their own emotional state. Was not often followed by:
Theory of mind thoughts related to moving agents (such as Expectation confirmation 0.003
Expectation violation o0.001
pedestrians and fellow road users), the customers, and the Emotions 0.003
experimenters. Some thoughts were concerned with actions Visual inspection 0.008
made, or the intentions of the customer (see Table 1), or
Expectation
thinking about what others might think of their own
1048.0, po0.001
behaviour (see Spiers & Maguire, 2006b for more details). Was more often followed by:
Subjects also reported changes in their own emotional state Expectation confirmation o0.001
which fell into three sub-categories: happy, anxious and Expectation violation o0.001
angry (see Table 1). Anxious and angry thoughts domi- Monitoring traffic 0.019
Was not often followed by:
nated, with some subjects identifiably more ‘emotional’ Route planning o0.001
than others (see Fig. 4). A wide variety of causes could Action planning o0.001
induce the reported transitions into emotional states, such Coasting o0.001
as a road being blocked off (angry), getting to the Rule related o0.001
destination (happy) and nearly crashing (anxious), see Visual inspection 0.039
Table 1. Visual inspection
121.6, po0.001
3.8. Temporal precedence Was more often followed by:
A temporal precedence analysis was used to test for any Emotions o0.001
temporal relationships between the thought categories. We
Expectation confirmation
found that the events in some categories preceded the events
243.0, po0.001
in other categories either more or less often than was Was more often followed by:
expected by chance (w2 ¼ 2402.9, df ¼ 90, po0.001). Sub- Action planning o0.001
sequent analyses revealed that the observed temporal Monitoring traffic 0.002
relationships were significantly different from the pattern Was not often followed by:
Expectation violation o0.001
predicted by chance for all categories except theory of mind. Expectation 0.021
These results and the results of post hoc sign tests Rule related 0.039
are reported in Table 2. To summarize, occurrences of
expectation confirmation and expectation violation were
significantly more likely to be preceded by expectation. A
significantly greater number of action planning events were
ARTICLE IN PRESS
Table 2 (continued ) thoughts. The high degree of correspondence between

verbal reports of looking at features in the environment
Category Sign test p-value
w2-score, p-value and independent eye-tracking measurements provided
support for the validity of the verbal reports. A statistical
Expectation violation analysis of temporal precedence of thought categories
251.2, po0.001 allowed us to identify their temporal relationships.
Route planning o0.001
Combining this information with a detailed qualitiative
Rule related 0.007 examination of the verbal reports, we now evaluate the
ability of extant cognitive models to capture the rich
–
wayfinding process we have revealed, as well as comparing
Monitoring traffic our findings with those from previous wayfinding studies
338.6, po0.001 that employed verbal report protocols.
Was more often followed by: Cognitive models of wayfinding share a number of core
Action planning o0.001 features. Our finding of distinct route planning and action
planning processes, and that these two categories occurred
Emotions 0.001 most frequently, lends support to the notion that these two
Visual inspection 0.006 stages form a central core of wayfinding cognition (Garling
Rule related
et al., 1984; Kuipers, 1978; Kuipers et al., 2003; Passini,
40.3, po0.001 1981, 1984; Peponis et al., 1990; Stern & Portugali, 1999;
Was more often followed by: Timpf et al., 1992). Examination of the verbal reports of
Action planning o0.001 initial planning revealed that it was composed of three
Was not often followed by: sequential stages: (1) retrieving the location of the
Coasting 0.021
destination, (2) determining the direction to the destination
Emotions and (3) retrieval/selection of the streets to form the route.
45.2, po0.001 Sequential processing of information as part of initial route
Coasting 0.021
planning is a common feature in the cognitive models.
Was not often followed by: However, some models generally ignore the first two
Visual inspection 0.021 stages, focusing instead on the selection of places and
Theory of mind
paths (e.g. Kuipers, 1978; Kuipers et al., 2003; Timpf et al.,
Not significant 1992). By contrast, the models of Garling et al. (1984) and
Passini (1992, 1984) both involve retrieval of information
Degrees of freedom for w2 tests ¼ 9. prior to planning. However, the stages in Garling et al.’s
(1984) model differ from those we found. In their model the
first stage involves the retrieval of all the relevant
preceded by monitoring traffic, rule-related thoughts and information (destination and paths), and subsequent stages
expectation confirmation than would occur by chance. In deal with the selection of places and the paths linking them.
contrast, route planning was only preceded by one thought This difference may relate to the fact that Garling et al’s
category, expectation violation. Both action planning and (1984) model was devised to deal with the problem of
route planning frequently preceded coasting periods. Mon- navigating to a sequence of multiple destinations, where
itoring traffic often preceded expectation confirmation and the order in which the destinations can be visited needs
often occurred after expectation and expectation confirma- planning. In this context it is useful to retrieve the locations
tion periods. Several categories occurred less frequently after of all destinations and possible paths first in order to decide
others than would be predicted by chance, and generally which paths are optimal. Our data suggest that when single
follow a natural logic. For example, expectation confirma- destinations are considered the location of the destination
tion thoughts were not likely to be immediately followed by is retrieved first and then the subsequent stages occur
periods of expectation violation. subsequently.
Another feature of route planning under emphasized in
4. Discussion many models is the requirement to determine the direction to
the destination. Our subjects often noted that this was a very
In this study we used a retrospective verbal report salient component of initial route planning. This makes sense
protocol, eye tracking and a highly accurate virtual reality if we consider that wayfinding will fail without it, and in
simulation of a real familiar city to explore cognition some situations knowing only the direction may be sufficient
during wayfinding on a second-by-second basis. By to reach the goal (see Spiers & Maguire, 2007b for further
classifying statements in subjects’ verbal reports into a discussion). The lack of emphasis of getting the right
number of pre-determined thought categories we were able direction in verbal reports of route planning in previous
to characterize their wayfinding experience in terms of the studies may relate to the use of indoor environments, where
diversity, frequency, duration and temporal order of the next waypoint may be visible from the outset.
ARTICLE IN PRESS
The final core stage of initial planning, retrieving/ was observed in descriptions where subjects opted for the
selecting the streets, contributed most to subjects’ route simpler route when choosing between two similar route
planning descriptions and it also features most prominently options. The fine-to-coarse heuristic (Wiener & Mallot,
in wayfinding models (Garling et al., 1984; Kuipers, 1978; 2003) was also evident in the reports of subjects who first
Kuipers et al., 2003; Passini, 1981, 1984; Peponis et al., selected the region in which the destination lay and then
1990; Stern & Portugali, 1999; Timpf et al., 1992). The planned the route to that region, and the partial planning
models of Kuipers (1978, 2003) and Timpf et al. (1992) was for routes covering large distances. Strategies we
consider route planning as the construction of a set of identified that are not obviously specified in models
instructions for movement between places along paths in a included re-checking the route once it has been selected,
cognitive map stored in memory. In agreement with such using ground-level mental imagery of moving through
models, the verbal reports often contained a specification the streets to see if it is correct, and the planning of multiple
of the paths (streets) and directions (left, right) necessary to route options in case of problems. The use of mental
reach the destination. Similar to the models, the travel imagery in navigation has been reported in some
instructions were mostly specified starting near the current studies (Gerber & Kwan, 1994; Passini, 1984), but not
location and progress, street by street, towards the goal others (e.g. Chebat et al., 2005). Gerber and Kwan’s (1994)
(Garling et al., 1984; Kuipers, 1978; Kuipers et al., 2003; observation that some subjects visualized the next five
Passini, 1981, 1984; Timpf et al., 1992). The fact that many streets ahead is consistent with the reports from some of
routes were not completely specified is also consistent our subjects. Like taxi drivers from Paris and Chicago,
with the partial planning suggested in several models London taxi drivers do not appear to use bird’s eye
(e.g. Garling et al., 1984; Passini, 1981, 1984) and similar imagery of a map for planning routes (Chase, 1982;
observations in previous studies (Holscher et al., 2007; Pailhous, 1970, 1984). Thus, if survey-like representations
Passini, 1981, 1984). Passini’s (1981, 1984) model draws the (Siegel & White, 1975; Thorndyke & Hayes-Roth, 1982;
distinction between this process of retrieving the map of the Tolman, 1948) support wayfinding on some routes in
environment to select the streets and the direct retrieval of familiar environments, they do not take the form of mental
stored instructions for very familiar routes. Our data images of maps.
generally support this distinction, in that for some routes Another strategy, deduced from the study of Parisian
no streets were specified and subjects stated they just knew taxi drivers, is the use of primary street networks to
where to go. This also agrees with previous reports of the facilitate wayfinding (Pailhous, 1970, 1984). Chase (1982)
rapid ‘automatic’ route choices made by experienced taxi found no evidence for such a strategy in studies of Chicago
drivers (Chase, 1982; Pailhous, 1970, 1984). However, our taxi drivers. Similarly in our data there is no evidence for
data also serve to highlight that even in familiar environ- an explicit use of this strategy. However, the representation
ments planning routes to some destinations can be very may be more subtle. Recently, a retired taxi driver with
detailed, with multiple factors considered and strategies damage to his hippocampus bilaterally (patient TT) was
used. tested on his navigation in the virtual simulation of
A number of strategies for aiding route planning have London used in the current study (Maguire, Nannery, &
been identified in previous studies and incorporated into Spiers, 2006). Patient TT could only navigate accurately
some cognitive models (e.g. Garling et al., 1984; Kuipers, when the destinations could be reached using predomi-
1978; Kuipers et al., 2003; Timpf et al., 1992). By strategy nantly the main arterial ‘A’ roads. Thus, it may be that
we mean the use of a particular heuristic or switching of when navigational ability is compromised, a primary
behaviour that is not a requirement for the task but coarse network of London’s major streets is retained and
improves performance or frees up cognitive resources. helps to support residual navigational ability.
Some strategies we identified could be related to those All models distinguish between the need to form an
previously described, while others were novel and not initial route plan and make further plans en route. Most
previously reported. In Passini’s (1981, 1984) model focus on re-planning caused by changes to the environ-
decisions are hierarchical, organized with over arching ment, such as a street being blocked off. Our temporal
goals at the top-level and sub-goals beneath. This was not a precedence analysis showed that indeed route planning
dominant feature of the route plans of our subjects. This events were more often preceded by expectation violation
difference may relate to the fact that Passini’s model is events than would be predicted by chance. Re-planning
based the verbal reports of navigation inside buildings could also be driven by subjects spontaneously realizing
(see also Holscher et al., 2007). However, one strategy there was a better option or by missing a turning,
reported did fit this approach: initially planning a route to additional reasons not considered in the models. A few
end up facing in the direction toward the destination before models draw the distinction between filling-in and re-
planning the full route. The layout of the environment and planning (Passini, 1981, 1984; Wiener & Mallot, 2003). Our
the route requirements have also been suggested to finding that filling-in often occurs after the subject had
influence route planning strategies (Conroy-Dalton, 2003; entered a new street rather than as soon as this street was
Holscher et al., 2007; Wiener & Mallot, 2003; Wiener et al., spotted does not feature in any model. Such a delay may
2004). The least-decision-load strategy (Wiener et al., 2004) relate to the subjects cognitive resources being taken up
ARTICLE IN PRESS
with making an appropriate action (e.g. changing lane) to identified in our study are completely absent in the models.
turn into the street. The categories emotions and theory of mind were reported
Several models highlight the importance of the link by all subjects. While these may not be necessary for
between the observation of an expected landmark/view successful wayfinding, they are not irrelevant to it. For
(expectation confirmation) and a cue to perform a instance, it is often useful to be able to predict/interpret the
particular action (action planning) (Garling et al., 1984; actions of other people, since they may know something
Kuipers, 1978; Kuipers et al., 2003; Stern and Portugali, important about the environment. For example observing
1999; Timpf et al., 1992). Our temporal precedence analysis someone reversing out of a street might lead you to wonder
provides direct support for this link. It also revealed that in why they chose to take that action—perhaps the street
the context of driving it is not just static features of the beyond is blocked. Wayfinding can be stressful, particu-
environment that can initiate this process. We found larly when it involves driving in a large city where road
monitoring traffic often preceded expectation confirma- rage is a too frequent occurrence. The observation that
tions. Examining the verbal reports suggested this was angry and anxious thoughts dominate the emotional
mostly caused by monitoring the traffic on the road ahead experience confirms that this was the case for many of
to detect the next street in the route plan and then the subjects in the current experiment.
monitoring their movements after the street had been To summarize, our findings agree with many compo-
detected to assess its accessibility. This relationship nents of current models of wayfinding, but also reveal
between monitoring traffic and navigational cues was not new features not captured by them. The sequential and
predicted by models involving vehicle driving (Kuipers, hierarchical nature of route planning outlined in models
1978; Kuipers et al., 2003; Timpf et al., 1992). was evident in our data. As were the distinctions between
The existence of periods of expectation during naviga- route planning and action planning, and between initial
tion replicates similar observations in from previous studies route planning and spontaneous route planning. We
employing verbal reports (Chebat et al., 2005; Holscher observed several previously described route planning
et al., 2007; Passini 1981, 1984). However, unlike Passini’s strategies, including the least-decision-load strategy
verbal reports, there was no indication from our subjects (Wiener et al., 2004) and the fine-to-coarse planning
that they were holding a mental image in mind during these heuristic (Wiener & Mallot, 2003). Several models pre-
periods. Our results revealed that expectation was not the dicted the sequence of thoughts: expectation, expectation
only state that subjects were occupied by during wayfind- confirmation and action planning, which were related to
ing. The most common state to be in was in fact coasting carrying out the route plan (Garling et al., 1984; Kuipers,
along, not thinking directed thoughts. Given its prevalence, 1978; Kuipers et al., 2003; Passini, 1981, 1984). Aspects of
it is surprising that this category was not identified in our data not captured in the models include certain
previous studies examining verbal reports in wayfinding components of route planning, monitoring traffic to
tasks (Chebat et al., 2005; Dogu & Erkip, 2000; Gerber & detect the next waypoint, and the under-emphasis of visual
Kwan, 1994; Holscher et al., 2007; Kato & Takeuchi, 2003; processing and other cognitive states during wayfinding.
Passini, 1981, 1984; Titus & Everett, 1996). One likely With regard to route planning, extant models fail to reflect
reason for this is that many of the previous studies used the importance of determining the direction to the
concurrent ‘think aloud’ verbal report protocols. In ‘think destination. We found no evidence for the use of a primary
aloud’ procedures, coasting periods would have been network of roads to facilitate route planning (Kuipers
observed as gaps between utterances and thus not et al., 2003; Pailhous, 1970; Pailhous 1984), but instead
necessarily classified. we found a number of other strategies including planning
Another category missing from the descriptions in the route to a location facing towards the destination
previous studies and models are thoughts relating to visual and then planning the rest of the route. Overall, our
inspection, where subjects reported looking at a particular results indicate that wayfinding involves much more than
fixed feature in the environment. After coasting, visual simply planning, and carrying out plans. It can evoke a
inspection accounted for more time during the task than range of emotions, spark interest in the surrounding
the other thought categories. Some models include the need environment and lead one to consider the thoughts of
to update information from signs or environmental fellow wayfinders.
features (Garling et al., 1984; Passini, 1981, 1984), which In conclusion, much of our day-to-day wayfinding
would fall into this category. However, this only represents behaviour takes place in familiar urban environments. By
one possible reason for visual inspection. Many of the combining interactive virtual reality and retrospective
statements involved looking at features such as statues, verbal reports, the complexity and dynamic nature of the
shops and large buildings simply out of curiosity, rather cognition behind wayfinding in a well-known busy city has
than to gain specific information. Thus, visual inspection been revealed. Our findings lend some support to extant
may relate partly to keeping oriented in the environment, models of wayfinding, but also suggest that they might
but also to generally looking at features out of interest. require revision to account for the diversity and temporal
Whilst visual processing of the environment has received order of thoughts, as well as additional wayfinding
little attention in the models, other aspects of cognition we strategies. In the future, further use and development of
ARTICLE IN PRESS
verbal reports are recommended as a potentially powerful Determining the direction

means to interpret wayfinding behaviour. Extract 6: Sub. 2
Basically, you’ve got to be like ay a weather vane.
Acknowledgements You’ve actually got to be facing in the right direction, and
that’s the art of it, really. It is an instantaneous reaction as
This work was supported by a Wellcome Trust senior to where you’re making for, really. Because you can’t sit
research fellowship in basic biomedical science to E.A.M. there wondering, getting a book out wondering whether
We are grateful to the participants for their time, patience you’ve got to turn around or do a U-turn or left. It’s got to
and good humour. We also thank the major licensed be an instant reaction
London taxi companies, publications, depots and cafes for
Extract 7: Sub. 17
facilitating subject recruitment. Thanks to P. Aston, E.
You can actually think too hard, and if you think too
Featherstone, A. McClelland, C. Freemantle, R. Davis, O.
hard—you can’t think. So what you do is you say well, the
Josephs, C. Hutton, J. Hocking, D. Tamli, G. Tan, W.
Middlesex Hospital is back that way, so you don’t panic
Penny and the FIL Imaging Support Team for advice and
the brain, you leave the brain free and then what you do is
technical assistance.
you turn it and point it in the right direction and then you
start sorting it out
Appendix A. Examples of route planning from the
verbal reports Extract 8: Sub. 18
The important thing when I start is to be heading in the
Examples of route planning from the verbal reports are right direction and then that gives you enough time to sort
given in the following table: of think as to where you go. So sometimes it’s made up as
you go along as well
Planning familiar routes
Extract 9: Sub. 1
Extract 1: Sub. 17
As I said, you tend to follow your nose and trust your
Because I’ve worked the area [Soho] so much, it’s
instinct because you know that you’re going in the general
automatic. If you get regular things it’s automatic, you
direction
don’t really have to think about it, you know it better than
your own front room. When you go in your front room Verbalizing street names
you don’t think: ‘‘I’ll go and switch the television on, where Extract 10:
is it?’’ Experimenter
And are you thinking the street names?
Extract 2: Sub. 1
Sub. 6
Sometimes when you’re out on the road, you do it [route
Yes, yes, yes. Definitely, yes
planning] instantly without even thinking
Extract 11:
Extract 3: Sub. 2
Experimenter
Well, it’s an instinct and familiarity really. You’ve done
Do you think through the names of the streets?
it so often that it doesn’t really test your brain. Obviously
Sub. 20
some do. It’s only when you get sort of difficult ones, or
Not verbally, it’s quicker than that. You haven’t got time
possibly places you’ve not been for a long while. And then
to verbalize
you’re actually thinking all the way along. I had one last
week that I’ve not been through in 20 years and the names Extract 12: Sub. 2
came back and I still found it I don’t think, oh, I’ll turn left Duncannon Street into the
Strand, left into the Aldwych, left into, er, Catherine Street.
Extract 4: Sub. 3
However, that’s how you do it at the Carriage Office
Do you want to know one of the worst things about
[examination centre for the licence] because that’s the only
driving a cab? You have a lot of time to think. Not about
way they know that you know where you are going. You
the job because that comes instinctively, to think about
don’t obviously think that way when you’re driving, you
other things. People say ‘‘oh I’m too busy at work, I
just sort of do it. It’s just automatic, really
haven’t been able to think about that’’ I have all day. I
spend more time thinking about other things, this [route Extract 13: Sub. 3
planning] clicks in very quickly I’m sure I was much more worried when I first started
and I would have probably remembered the names of all
Extract 5: Sub. 6
the streets I was going down and would be going down,
It’s virtually instanty As you change the path of that
whereas now it’s far more automatic
vehicle you’ve made your mind up, basically, for 90% of
the route. And that one’s only three streets in my book. Filling-in the next stage of the journey
Strand, Fleet Street, Ludgate Hill, I’m there Extract 14: Sub. 3
ARTICLE IN PRESS
I know I’m going in the right direction and I’ won’t place like Royal Courts of Justice, I’m thinking what
make a decision until I get closer towards the end. That’s street’s it in, that’s the first thingyThat’s how everyone
how it works sort of learns it initially, so Royal Courts of Justice is in the
Strand. That’s what I’m thinking initially and then I’m
Extract 15: Sub. 17
thinking both, what road it’s in or what it looks like. So it
What I’m actually doing is allowing myself time—this is
could be either option. The Royal Courts of Justice, I
very delicate, I never knew it was as delicate as this—but
probably thought the Strand; then I pictured it. If it doesn’t
what you’re actually doing is: instead of taking the whole
snap in to perspective straightaway you normally need a
route in one go, you’re take a safe step forward to the next
keyword, the two things go together to pick up a memory
one to reappraize where you’re going to go and then it
somewhere. When it works, instantly then you’ve got like a
gives you time to think and to relax. And then what you do
snap, if you like, flashing ziggly, zaggly line and you can fill
is you take the next stage stepping stone, but by then
in the detail as you go along. And it’s more or less right
you’ve always got it, you’ve had enough time to zoom it in
Navigating by region
Extract 16: Sub. 18 Extract 23: Sub. 1
Once you’ve got the basic plan in your head and you just There are various points in London where there are only
fill in the detail as you go along one way in and one way out and when you do the
knowledge, even though I did it 20 years ago, you learn
Mental imagery
those ways in and they stick
Extract 17: Sub. 17
And that’s all visual in my mind. When I do it here, I Extract 24: Sub. 15
zigzag visually, so I do a fast forward. The nearest you can You take a region and head for the region first and then
get to it is very, veryy less than a second, very, very fast I sort of branch out
go de-de-de-de-de that fits. So in my mind I’m going very
Considering options and re-assessing
fast visually and I’m seeing the streets, not from a plan
Extract 25: Sub. 18
bird’s eye view, but just forward fast, left, right and it’s
I pretty much know what I’m going to do and what I’m
done. It’s done as quick as that, but it’s visual. It’s
going to come across and also just bearing in mind that
checking—yes, that’s it, that’ll work
some roads are closed. I’ve got at the back of my mind
Extract 18: Sub. 6 maybe another alternative route just in case ofy the roads,
You see it likey it’s like a Schumacher run [Formula the roads are closed. I remember someone did say to me
one motor car race]. Yes. It’s just like that. And you see it in once when I first started he said just have a couple of
your mind. Jun, jun, jun, It’s done in seconds. Because of options just in case there is traffic, you can’t do a specific
the nature of the job, you’ve also got to concentrate on turn or something
controlling the vehicle and putting it in the right location to
Extract 26: Sub. 18
do that journey
Obviously once you’ve got the route, then you double
Extract 19: Sub. 17 check it several times in your head. Because sometimes, you
I’ve got an over-patched picture of Peter Street. It know, you’ve got it in place and you’re getting there and
sounds daft, but I don’t view it from ground level, it was you think, no, hang on, that’s completely wrong.
slightly up and I could see the whole area as though I was
about 50 foot up. And I saw Peter Street, I saw the market Statements in square brackets and italics provide
and I knew I had to get down to Peter Street additional information to clarify what the preceding
statements refer to. Statements in parentheses and italics
Extract 20: provide information about a subject’s behaviour.
Experimenter
Were you imagining driving all the way down to the Old
Bailey? References
Sub. 17
No, it’s too easy. It’s just the Strand, Fleet Street, do a Chase, W. G. (1982). Spatial representations of taxi drivers. In D. Rogers,
left. I didn’t have to with that one & J. A. Sloboda (Eds.), Acquisition of symbolic skills (pp. 391–405).
New York: Plenum Press.
Extract 21: Sub. 18 Chebat, J.-C., Gelinas-Chebat, C., & Therrien, K. (2005). Lost in a mall,
I’d pictured the hospital up just off Great Portland Street the effects of gender, familiarity with the shopping mall and the
then I brought it back to the Marylebone Road, then I shopping values on shoppers’ wayfinding processes. Journal of Business
Research, 58(11), 1590–1598.
turned round Conroy-Dalton, R. A. (2003). The secret is to follow your nose: Route
Extract 22: Sub. 20 path selection and angularity. Environment and Behaviour, 35(1),
107–131.
I think you do think in images. Maybe you fill in details Dogu, U., & Erkip, F. (2000). Spatial factors affecting wayfinding and
and overlay it with words and things later because, I mean, orientation: A case study in a shopping mall. Environment and
your brain works in images. What I think, when you say a Behaviour, 32(6), 731–755.
ARTICLE IN PRESS
Ericsson, K. A., & Simon, H. A. (1980). Verbal reports as data. Pailhous, J. (1984). The representation of urban space: Its development
Psychological Review, 87(3), 215–251. and its role in the organisation of journeys. In R. Farr, & S. Moscovici
Foo, P., Warren, W. H., Duchon, A., & Tarr, M. J. (2005). Do humans (Eds.), Social representation. Cambridge: Cambridge University Press.
integrate routes into a cognitive map? Map-versus landmark-based Passini, R. (1981). Wayfinding: A conceptual framework. Urban Ecology,
navigation of novel shortcuts. Journal of Experimental Psychology— 5, 17–31.
Learning, Memory and Cognition, 31(2), 195–215. Passini, R. (1984). Spatial representations, a wayfinding perspective.
Garling, T., Book, A., & Lindberg, E. (1984). Cognitive mapping of large- Journal of Environmental Psychology, 4, 153–164.
scale environments the interrelationship of action plans, acquisition, Passini, R. (1992). Wayfinding in architecture (2nd ed.). New York: Van
and orientation. Environment and Behaviour, 16(1), 3–34. Nostrand Reinhold Company.
Garling, T., & Garling, E. (1988). Distance minimization during down- Peponis, J., Zimring, C., & Choi, Y. K. (1990). Finding the building in
town pedestrian shopping. Environment and Planning A, 20(547–554). wayfinding. Environment and Behaviour, 22(5), 555–590.
Gerber, R., & Kwan, T. (1994). A phenomenographical approach to the Peruch, P., & Gaunet, E. (1998). Virtual environments as a promising tool
study of pre-adolescents’ use of maps in a wayfinding exercise in a for investigating human spatial cognition. Current Psychology of
suburban environment. Journal of Environmental Psychology, 14, Cognition, 17, 881–899.
265–280. Peruch, P., Giraudo, M.-D., & Garling, T. (1989). Distance cognition by
Gillner, S., & Mallot, H. A. (1998). Navigation and acquisition of spatial taxi drivers and the general public. Journal of Environmental
knowledge in a virtual maze. Journal of Cognitive Neuroscience, 10, Psychology, 9, 233–239.
445–463. Richardson, A. E., Montello, D. R., & Hegarty, M. (1999). Spatial
Giraudo, M.-D., & Peruch, P. (1988a). Representation de l’espace urbain knowledge acquisition from maps and from navigation in real and
et potentiel d’activite du chaffeur de taxi. Psychologie-Francaise, 33(3), virtual environments. Memory and Cognition, 27(4), 741–750.
145–150. Siegel, A. W., & White, S. H. (1975). The development of spatial
Giraudo, M.-D., & Peruch, P. (1988b). Spatio-temporal aspects of the representations of large-scale environments. In H. W. Reese (Ed.),
mental representation of urban space. Journal of Environmental Advances in child development and behavior (Vol. 10). New York:
Psychology, 8, 9–17. Academic Press.
Golledge, R. G. (1995). Path selection and route preference in human Spiers, H. J., & Maguire, E. A. (2006a). Thoughts, behaviour, and brain
navigation: A progress report. In A. Frank, & W. Kuhn (Eds.), Spatial dynamics during navigation in the real world. Neuroimage, 31(4),
information theory: A theoretical basis for GIS (COSIT’95) (Vol. 988, 1826–1840.
pp. 207–222). Berlin: Springer. Spiers, H. J., & Maguire, E. A. (2006b). Spontaneous mentalizing during
Golledge, R. G. (1999). Wayfinding behavior: Cognitive mapping and other an interactive real world task: An fMRI study. Neuropsychologia,
spatial processes. Baltimore, MD: Johns Hopkins Press. 44(10), 1674–1682.
Hochmair, H., & Frank, A. U. (2002). Influence of estimation errors on Spiers, H. J., & Maguire, E. A. (2007a). Neural substrates of driving
wayfinding decisions in unknown street networks—Analyzing the behaviour. Neuroimage, 36(1), 245–255.
least-angle strategy. Spatial Cognition and Computation, 2(4), 283–313. Spiers, H. J., & Maguire, E. A. (2007b). A navigational guidance system in
Holscher, C., Meilinger, T., Vrachliotis, G., Brosamle, M., & Knauff, M. the human brain. Hippocampus, 17(8), 618–626.
(2007). Up the down staircase: Wayfinding strategies in multi-level Stanton, D., Wilson, P., & Foreman, N. (2002). Effects of early mobility
buildings. Journal of Environmental Psychology, 26(4), 284–299. on shortcut performance in a simulated maze. Behavioural Brain
Jack, A. I., & Roepstorff, A. (2003). Trusting the subject (Vol. 1). Exeter Research, 136, 61–66.
(UK): Imprint Academic. Stern, E., & Portugali, Y. (1999). Environmental cognition and decision
Janzen, G., Schade, M., Katz, S., & Hermann, T. (2001). Strategies for making in urban navigation. In R. G. Golledge (Ed.), Wayfinding
detour finding in a virtual maze: The role of the visual perspective. behavior: Cognitive mapping and other spatial processes (pp. 99–119).
Journal of Environmental Psychology, 21, 149–163. Baltimore: Johns Hopkins University Press.
Kato, Y., & Takeuchi, Y. (2003). Individual differences in wayfinding Thorndyke, P., & Hayes-Roth, B. (1982). Differences in spatial knowledge
strategies. Journal of Environmental Psychology, 23, 171–188. acquired from maps and navigation. Cognitive Psychology, 14,
Kuipers, B. (1978). Modeling spatial knowledge. Cognitive Science, 2, 560–589.
129–153. Timpf, S., Volta, G. S., Pollock, D. W., & Egenhofer, M. J. (1992). A
Kuipers, B., Tecuci, D. G., & Stankiewicz, B. J. (2003). The skeleton in the conceptual model of wayfinding using multiple levels of abstraction. In
cognitive map: A computational and empirical exploration. Environ- A. U. Frank, I. Campari, & U. Formentini (Eds.), Theories and
ment and Behavior, 35, 81–106. methods of spatio-temporal reasoning in geographic space (Vol. 639,
Maguire, E. A., Nannery, R., & Spiers, H. J. (2006). Navigation around pp. 348–367). Berlin: Springer.
London by a taxi driver with bilateral hippocampal lesions. Brain, Titus, P. A., & Everett, P. B. (1996). Consumer wayfinding tasks,
129(Part 11), 2894–2907. strategies, and errors: An exploratory field study. Psychology &
Morganti, F., Carassa, A., & Geminani, G. (2007). Planning optimal Marketing, 13(3), 265–290.
paths: A simple assessment of survey spatial knowledge in virtual Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological
environments. Computers in Human Behaviour, 23, 1982–1996. Review, 55, 189–208.
Newman, E. L., Caplan, J. B., Kirschen, M. P., Korolev, I. O., Sekuler, Wiener, J. M., & Mallot, H. A. (2003). Fine-to-coarse route planning
R., & Kahana, M. J. (2006). Learning your way around town: How and navigation in regionalized environments. Spatial Cognition and
virtual taxicab drivers learn to use both layout and landmark Computation, 3(4), 331–358.
information. Cognition, 104(2), 231–253. Wiener, J. M., Schnee, A., & Mallot, H. A. (2004). Use and interaction of
Pailhous, J. (1970). La representation de l’espace urbain: l’example du navigational strategies in regionalized environments. Journal of
chauffeur de taxi. Paris: Presses Universitaires de France. Environmental Psychology, 24(4), 475–493.

The Dynamic Nature of Cognition During Wayfinding: Hugo J. Spiers, Eleanor A. Maguire

Uploaded by

Copyright:

Available Formats

The Dynamic Nature of Cognition During Wayfinding: Hugo J. Spiers, Eleanor A. Maguire

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Dynamic Nature of Cognition During Wayfinding: Hugo J. Spiers, Eleanor A. Maguire

Uploaded by

Copyright:

Available Formats

ARTICLE IN PRESS

Journal of Environmental Psychology 28 (2008) 232–249

The dynamic nature of cognition during wayﬁnding

1. Introduction relationships between locations/landmarks) (Siegel &

0272-4944 r 2008 Elsevier Ltd. Open access under CC BY license.

Table 1 Table 1 (continued)

Table 1 (continued) Table 1 (continued)

wayﬁnding were spent visually inspecting ﬁxed features in Table 2

Table 2 (continued ) thoughts. The high degree of correspondence between

verbal reports are recommended as a potentially powerful Determining the direction

You might also like