Você está na página 1de 12

Journal of Environmental Psychology 29 (2009) 208219

Contents lists available at ScienceDirect

Journal of Environmental Psychology


journal homepage: www.elsevier.com/locate/jep

Adaptivity of waynding strategies in a multi-building ensemble: The effects


of spatial structure, task requirements, and metric information
Christoph Holscher a, *, Simon J. Buchner a, Tobias Meilinger a, b, Gerhard Strube a
a
University of Freiburg, Centre for Cognitive Science, Friedrichstrasse 50, 79098 Freiburg, Germany
b
Max-Planck-Institute for Biological Cybernetics, Tu bingen, Germany

a r t i c l e i n f o a b s t r a c t

Article history: This study investigates the adaptivity of waynding strategies in a real-world setting of a multi-building
Available online 13 June 2008 ensemble. Familiarity with the environment, map usage and verbal vs. visual task instructions were
systematically varied. Measures included path choices, waynding performance and information usage.
Keywords: Thirty-two participants had to nd eight goals in a multi-level building ensemble consisting of two
Cognition distinctive building parts. It was tested whether the standard wall-mounted oor maps found in the
Architecture
majority of public buildings can help navigation in a complex unknown environment. Unfamiliar users
Waynding strategies
tried to make use of these plans more frequently, but were not able to compensate for spatial knowledge
Adaptivity
Representation decits compared to participants familiar with the setting. Two strategies of multi-level waynding were
compared with respect to a region-based hierarchical planning approach. Strategy selection could be
shown to be highly adaptive to spatial properties of the environment as well as characteristics of the task
instruction, i.e., spatial precision of target information. Overall, the strategy of moving horizontally into
the target building prior to vertical travel was shown to be more effective in this multi-building setting.
2008 Elsevier Ltd. All rights reserved.

1. Introduction identifying incongruent oor layouts, disorienting staircases and


lack of visual access to important level-related building features as
Finding ones way around public buildings such as airports, main causes of difculty (cf. Weisman, 1981).
hospitals, ofces, or university buildings often proves to be The current study extends this research to a setting which
a tedious and frustrating task. Waynding in a complex setting introduces an additional dimension of complexity: It comprises
with less than perfect knowledge requires decision making under both multiple levels, and multiple building parts that differ in their
uncertainty and we aim to identify the behaviors and strategies vertical and horizontal congurations. Holscher et al. (2006) have
that people employ to navigate in such environments. pointed both to the importance of familiarity with a building and
Several researchers (e.g., Garling, Lindberg, & Mantyla, 1983; the use of specic strategies to support route choice decisions in
Moeser, 1988; Montello, 1991) stressed the role of familiarity with multi-level waynding. The building setting in the current paper
a building, and illustrated how training of sequential routes or provides the opportunity to shed further light on the relationship
survey knowledge can boost orientation performance in complex between these aspects, and how the conguration of the building
buildings like hospitals. Most studies on waynding performance and task characteristics contribute to it.
and building complexity have limited themselves to investigations
in the horizontal plane of isolated oor levels. Soeda, Kushiyama, 1.1. Waynding strategies in complex buildings
and Ohno (1997) were able to show that waynding performance
in tasks involving oor level changes is largely hindered by When people have only an incomplete or uncertain represen-
disorientation during vertical travel on stairs or elevators, extend- tation of a spatial setting, they need to rely on navigation strategies
ing earlier observations (Passini, 1992). Holscher, Meilinger, to ll in the gaps. Holscher et al. (2006) identied three waynding
Vrachliotis, Brosamle, and Knauff (2006) report waynding dif- strategies that are used to support route choice decisions in three-
culties observed in a complex multi-level conference centre, dimensional multi-level buildings: First, the central point strategy of
sticking as much as possible to main corridors and main places in
the building predominantly used by rst-time visitors. Second,
* Corresponding author. Tel.: 49 761 203 4937; fax: 49 761 203 4938. the direction strategy of choosing routes that head towards the
E-mail address: hoelsch@cognition.uni-freiburg.de (C. Holscher). horizontal position of the goal as directly as possible, irrespective of

0272-4944/$ see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jenvp.2008.05.010
lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho 209

level changes (cf. least-angle strategies: Conroy Dalton, 2003; regulations specify exactly how the oor plans on display must be
Hochmair & Frank, 2002). Both proved to be inferior to a third designed. These plans are dened by law and are omnipresent. One
strategy based on forming a hierarchically organized navigation would naturally hope that these maps adequately support
plan (Hirtle & Jonides, 1985). Such a strategy seems to be based on waynding tasks in the building especially in case of an emergency.
cognitively segmenting the setting into regions which guide navi-
gation decisions (see Wiener, Schnee, & Mallot, 2004, who describe 1.4. Research questions and hypotheses
such strategies for two-dimensional outdoor settings). This strategy
reduces the complexity of planning and navigation by rst entering In the previous study (Holscher et al., 2006), region-based
the target region before starting ne-tuned search. In the Holscher navigation was clearly based on mentally segmenting the building
et al. (2006) study, this strategy of rst moving to the correct oor by oor levels (cf. Montello & Pick, 1993). But how does
in the building was characterized as a oor strategy, since in that a regionalization account of strategy choice generalize to other,
building oor levels were the predominant hierarchical aspect of more complex building settings? For the building ensemble in the
the building. present study, it is not immediately obvious how a route choice
preference would look if a hierarchy-based, region-oriented strat-
1.2. Visual vs. number-based instructions for navigation tasks egy is employed: The building ensemble consists of at least two
identiable horizontal parts which could be interpreted as regions
Besides the spatial characteristics of a building, waynding tasks by a navigator. At the same time, however, the building ensemble
can differ especially regarding the amount and type of goal directed has a multi-level structure, which may also foster a level-based
information they provide and thus inuence waynding strategy mental regionalization of the setting, just as in Holscher et al.
choices. Rooms in public buildings are generally identied by (2006). Thus, it is not clear how the participants will segment the
a room number, a propositional descriptive scheme that implies building when applying a hierarchy-based strategy. Also, it is not
information about the general region of a goal location, i.e., in a task a priori clear if individuals idiosyncratically prefer a certain strategy
like go to room no. 1215 the participant usually knows on which or if they adapt it according to external constraints such as the task,
oor he will nd the target room. The room numbering scheme in the instruction or familiarity.
this setting also provides information about the building part in The results in Holscher et al. (2006) suggest that a oor-based
which the room is located. In contrast to this, a task instruction that strategy would be (a) more successful than a direction-based
provides only directional but no propositional information, e.g., by strategy and that (b) users familiar with the building will rely on it
pointing to the goal location, may not foster a hierarchical planning to a greater extent. An alternative hypothesis, as sketched above,
process to a similar extent. In this case, people may try to minimize states the multi-building ensemble requires a highly adaptive
the angular deviation from the linear distance and follow selection of routes, reecting the possible dual hierarchy in multi-
a direction-based least-angle strategy (Conroy Dalton, 2003; building ensembles, so that a clear prevalence of the oor-based
Hochmair & Frank, 2002). Thus, the kind of information that is strategy (especially with familiar users) would not be expected.
provided in the task instruction may elicit route choice and More generally, we wish to clarify to what extent the choice of
performance differences. navigation strategies is adaptive to systematic variation of the
waynding task, both regarding information provision (instruction)
1.3. The role of maps and properties of the local setting (origindestination pairs). Gray,
Sims, Fu, and Schoelles (2006) have recently presented evidence
Thorndyke and Hayes-Roth (1982) found evidence that the that participants can be highly sensitive to small changes in the
information learned from a map is quite distinct from information required effort and in the processing costs of cognitive tasks.
learned while navigating in an environment, since the map Participants do change between competing solution strategies
provides survey knowledge rather than accurate, direct route adaptively based on these factors. While this general observation
knowledge. Garling et al. (1983) presented evidence that showing was obtained in a humancomputer interaction context it should
a oor map to participants immediately prior to testing reduced the be tested whether and if, how it translates to other real-world
effects of limited view when learning the building. Learning from domains like waynding in complex buildings. If the participants in
a map can be equally effective as having actually visited a building our study are sensitive to differences in the spatial relation of start
(Richardson, Montello, & Hegarty, 1999) or even lead to better and goal location as well as the information content provided with
performance compared to long-term users of a building (Moeser, the task instruction, they should not stick to one strategy across all
1988) at least with respect to survey knowledge. tasks, but should change strategies relative to task properties.
Yet it is not unequivocally clear that access to oor maps does Background knowledge may play a signicant role in this context.
indeed have a positive impact on real-life waynding performance. From the study by Holscher et al. (2006) one would expect
Butler, Acquino, Hissong, and Scott (1993) present evidence that participants familiar with the setting to prefer a oor-rst strategy,
you-are-here maps in a similarly complex setting had no positive unless the local properties of the building favour the opposing
effect at all, in fact, waynders attending to such wall-mounted strategy of directional navigation.
maps lost time without gaining any navigational advantage. It is also To further test the adaptive character of strategy choice, the
well-documented that using a map that is misaligned with ones present study will also systematically vary, rst, the instruction
current orientation can be detrimental (Levine, Jankovic, & Palij, between number-based presentation and visual presentation,
1982), a feature of many standard oor maps in public buildings. mimicking a situation someone is making a gesture to indicate
Independent from map alignment, information in maps has to a location. The number-based task instruction can be expected to
be transferred from survey perspective into route perspective. This foster a oor-based strategy rather than a direction-based strategy,
transfer is associated with switching costs (Lee & Tversky, 2001; because in the room number the oor information is particularly
Shelton & McNamara, 2004). Learning an environment from direct salient. Second, the availability of additional metric information
experience or an oblique perspective drawing, therefore, can lead about the building was varied by controlling access to wall-
to better performance than learning from a oor plan, or from mounted maps of the building. In general, participants familiar
a oor plan and a cross-section (Fontaine, 2001). with the building should perform better than unfamiliar ones due
We have chosen a specic type of map for our study: In all public to their enhanced knowledge about the environment. Also,
buildings in Germany as well as most of the Western world re participants using a map should perform better than those not
210 lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho

using a map. Map users receive additional information about the 2.2. Material
environment which participants not using a map dont have avail-
able. In order to build up a spatial representation of the environment The design of Freiburg Universitys main building (see Fig. 1) is
they have to rely on the knowledge they have acquired by navigating composed of two distinct building sections, built in 1911 (KG I) and
through it. However, the study by Butler et al. (1993) would suggest 1961 (KG III). This building is now used for the arts and humanities
no effect of map use. Concerning the interaction of familiarity and exclusively. The KG I section consists of a squarish main section
map use the studies of Richardson et al. (1999) and Moeser (1988) with a linear outbuilding (mezzanine, half-height level change),
prompted the hypothesis that the availability of maps will allow which at its end meets the open-space design of the KG III section.
unfamiliar users to compensate at least partially for their spatial The oors of the buildings are connected by various passages
knowledge decit compared to familiar users. Thus, unfamiliar and stairways (except 2nd oor, KG III). Due to different heights
participants in the map condition should use maps more frequently between the oors, and different numbers of oors (four at KG I
than familiar participants but with respect to their performance they compared to ve at KG III) the user has to change the oors in an
should be as successful as familiar participants. irregular and often unexpected way (see Fig. 2). The numbering
system in the building consists of four-digit room numbers. The
2. Method rst digit codes whether a room is located in the KG I or KG III part
(e.g., 1019 vs. 3047), treating the main part of KG I and its mezzanine
In the experiment, navigational performance and verbalisation oors as one. The second digit corresponds to the oor level,
of visitors with prior experience with the adjacent university counted separately for KG I and KG III, leading to a mismatch on
buildings was compared to the behavior of novices visiting for the upper levels (see Fig. 2). The wall-mounted maps in the experiment
rst time. Map usage was controlled by either allowing or denying were standard re plans that, by law, are required to be posted at
the participants access to the re plans of the building which were specied points in all public buildings. Each plan depicts the layout
located close to the various staircases in the building. of the current oor, but is limited to either KG I (for an example, see
Generally, the task consisted of nding a goal specied by a room Fig. 3) or KG III. Plans are located next to main staircases and
number. In two of the tasks, this instruction was systematically elevators on each level. Eleven out of these 20 wall-mounted re
compared to giving exact location information by showing the rooms maps were misaligned by 90 or 180 , the rest were properly
location from the outside, without mentioning the room number. aligned with a viewers facing direction.

2.1. Participants 2.3. Procedure

Participants were students from different universities and The experiment was set up as a 2  2 design with map acces-
colleges in Freiburg. One had a background in Geography, none of sibility (accessible/not accessible) and familiarity (familiar with the
them studied architecture. Sixteen women and 16 men between 18 building/not familiar) as between subjects factors. For two of the
and 34 years of age (M 23, SD 3.0) participated in the experi- tasks, a variation of instruction type (visual vs. number-based) was
ment. The 16 participants familiar with the building had visited it on introduced as a third factor.
a weekly basis for at least two semesters. The 16 unfamiliar partic- The experiment lasted about 90 min, including the instruction
ipants, by contrast, had never been to the building at all. Participants and debrieng as well as post-experimental written tests (not
were randomly assigned to the map and instruction conditions, reported here, as they are beyond the scope of this article).
yielding experimental groups of equal size. The familiarity and map Participants were instructed to think aloud and thus verbalize their
conditions were fully balanced for gender. There was a slight gender thoughts and considerations (Ericsson & Simon, 1993; Passini,
imbalance for the order of the number-based and visual task 1992); performance was video-taped. Participants practiced the
instruction, as seven women received the visual instruction (see thinking-aloud procedure on a mental arithmetic task in order to
below) on task 6 while nine women performed that task with the familiarize them with providing task-concurrent verbalization in
number-based instruction (and vice versa for the following task 7). the waynding tasks. During the waynding tasks they were

Fig. 1. Plan view of the building ensemble. The left and the central part belong to the original building from 1911, the right part was added in the 1960s.
lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho 211

Fig. 2. Conguration of oors with circulation areas. Starting points and goals of the navigation tasks are marked by numbers (example: s1 is the starting point of the rst task,
1 its goal location. As tasks are linked, 1 marks the starting point for task 2 and 2 marks its goal.). The gure also documents the mismatching oor numbering scheme across
different parts of the building.
212 lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho

Fig. 3. An example re plan, located in the KG I main section, 3rd oor.

repeatedly encouraged to continue verbalizing their thoughts while 4 to 8) were planned to equally allow for both a oor-based and
making navigational decisions.1 Half of the participants were a direction-based strategy, i.e., changing rst to the oor of the
allowed to use the oor maps of the building, whereas the other assumed goal location (oor strategy) or walking rst horizon-
half were not. Due to the size of the re plans it was not possible to tally into the direction where the goal is assumed to be located
read the maps without standing close to it, so in the no-map (direction strategy). In order to investigate the effect of number-
condition it was impossible for the participants to look at the maps based vs. visual (directional) information, the instruction was
without the experimenter taking notice. The participants main varied accordingly in two tasks: For one half of the participants it
task was to subsequently nd eight locations in the building. All was changed in task 6, for the other in task 7. This was done in
participants received the tasks in the same order, and the tasks order to balance between the directions of movement: The two
were linked such that one tasks goal location was the next tasks tasks were matched with respect to complexity of the path, level
starting point (see Fig. 2). changes and length. Instead of a number-based instruction, the
The tasks were selected in a way that their complexity increased target room was shown visually to the participant in such a way
(with respect to the number of oor changes and the requirement that the experimenter pointed to the target room, while the
to change the buildings). This was done in order to accommodate participant was standing in the door leading to the courtyard
for the participants increasing knowledge about the building while (visual instruction). In addition, the target room was marked by
they performed the tasks. In order to provide a fairly realistic a black X in the window. This type of instruction provided both
situation, the waynding tasks were initiated by verbally providing vertical and horizontal direction information non-verbally. It
the room number such as Find room number 1019.2 This type of explicitly avoided any reference to room numbering or level
instruction provides the same information as students might have information (as an instruction like it is the third window from
when they gain the room number from their timetable. Participants the bottom might have done). The order of instruction, i.e.,
were given no further explanation about the room numbering whether participants received the visual instruction on task 6 or
scheme itself. on task 7, was randomly assigned.
While tasks 13 were designed to foster only one waynding
strategy each (for details see Table 3), the remaining tasks (from
2.4. Measures

2.4.1. Performance
1
Authors like Soh and Smith-Jackson (2004) have found retrospective thinking
For each task, the shortest route as well as a list of reasonable
aloud to provide very good accounts of navigational procedures and the use of
waynding cues. For the present study, a retrospective thinking-aloud procedure route alternatives was determined beforehand. The traversed route
would not have been feasible. In contrast to Soh and Smith-Jackson (2004) par- of each participant was drawn for each task in the building plan.
ticipants perform a series of tasks in the same environment, so that in retrospective This was used to determine route distances and detour distances.
prompting participants might easily confuse their state-of-knowledge and their Stops and stop time were marked along the route at the places
planning decisions between situations of passing the same decision point in dif-
where it occurred. Navigation performance was measured by
ferent tasks. Under these circumstances, hindsight rationalizations (Ericsson & Si-
mon, 1993) might have dominated the reported planning processes. a number of variables:
Task-concurrent thinking aloud is well established for capturing active working
memory content. We are aware of the limitations of task-concurrent verbal reports  Duration: mean time to complete a task
to illicit procedural and tacit knowledge and thus base our analysis foremost on
 Stops: mean number of stops per task (other stops than those at
behavioral data, using the verbalizations as a supplementary data source.
2
The target locations were: Room 1019 (task 1), room 1192 (task 2), room 3120
maps)
(task 3), room 1486 (task 4), courtyard (task 5), room 1234 (task 6), room 3301 (task  Stop time: mean stop time per task (from stops other than those
7), room 3220 (task 8). at maps)
lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho 213

 Detour distance: the mean length of detours/task, where detour A multivariate ANOVA with factors familiarity (familiar, unfamiliar)
is dened as leaving the path and returning to it at the same and map usage4 (map used, no map used) was conducted for the
point.3 performance measures. A MANOVA was chosen, because the per-
 PAO (percentage above optimal path length): the distance (in % of formance measures are not independent from each other (median
the optimal route) that was walked additionally to the length of correlation: .55).
the optimal route Separate ANOVAs were conducted for the measures of orientation
behavior since they were considered to be independent (median
correlation: .20). An alpha level of .05 was used for all statistical tests.
2.4.2. Orientation behavior and verbalizations
The second group of measures classied the participants 3.1. Performance across tasks
orientation behavior and verbal comments. A symbol was assigned
to each category and marked at the location where the behavior Participants familiar with the building were expected to
occurred. The coding scheme was ne-tuned until all categories perform better than those who were not. Map usage was expected
could be reliably recognized with a sufcient inter-rater reliability to help unfamiliar participants more than familiar participants.
(kappa value of .70; for a detailed description of verbal data The MANOVA showed a main effect of familiarity, F(5,24) 4.84,
handling see Holscher et al., 2006) p .003, h2partial :50, but neither a main effect of map usage nor an
The following measures were collected based on video analysis interaction. Tests of the between subject effects showed that un-
combined with verbal comments of the participants: familiar participants took more time to complete the task,
F(1,28) 19.34, p < .001, h2partial :41, stopped more frequently,
 Looked at room no.: frequency of looking at room numbers in F(1,28) 14.32, p .001, h2partial :34, stopped longer, F(1,28)
order to orientate oneself 7.34, p .011, h2partial :21, walked longer detours, F(1,28) 7.37,
 Orient on outside: frequency of looking to the outside in order to p .011, h2partial :21, and had a higher PAO, F(1,28) 14.51, p .001,
orientate oneself h2partial :34 (for means and standard deviations see Table 1).
 Forgot room number: frequency of forgetting the target room As expected the familiar participants performed better than
number and asking the experimenter for it unfamiliar, presumably due to their deeper spatial knowledge
 Stops at maps: frequency of stopping at a wall-mounted map about the test environment. Map use in general did not enhance
 Time/stop at maps: average time of stopping at a wall-mounted performance. The hypothesis that unfamiliar participants are able
map to compensate for their lack of knowledge by using information
from the maps (interaction of familiarity and map use) was not
empirically supported.
2.4.3. Strategy choice Yet, not all maps were aligned with their surroundings. Could the
Strategy choice was classied based on the trajectories observed lack of performance advantages in maps be due to improper map
in the study, separately for each of the eight tasks. A coding scheme alignment, generally known to hamper performance? To answer
for mapping trajectories to strategy choice was compiled by three this, a separate analysis was conducted for tasks 3 and 7 in which
raters and revised until unanimous classication was possible. only correctly aligned maps were encountered by the participants:
A trajectory was classied based on the initial part of the journey. again neither a signicant main effect of map use, nor an interaction
The oor-rst strategy was assigned when the participant changed with familiarity was established. Thus, the lack of advantage for map
to a different oor before moving into another section of the users most likely cannot be ascribed to misaligned maps.
building. A direction strategy was assigned, if the participant did Map use did not increase performance, but how frequently and
not change the oor before either reaching the goal location or how long did participants use the maps? For those participants who
moving into the other main section of the building. If movement via had a map available a MANOVA with factor familiarity was performed
the mezzanine was ambiguous regarding these two strategies or if for the frequency of map access per task as well as the reading time of
the participant returned to the main entrance hall of the KG I maps per access and task. The overall MANOVA was signicant
section, no strategy was assigned. F(2,11) 5.54, p .022. Unfamiliar participants consulted maps more
frequently than familiar participants, F(1,12) 9.65, p .009,
3. Results
h2partial :45. The mean reading time did not differ, F(1,12) 0.08,
p .782 (for means and standard deviations see Table 1).
The waynding tasks were analyzed with respect to perfor- For familiar participants, it was not necessary to use maps very
mance measures as well as orientation behavior and corresponding often due to their elaborate knowledge about the setting. However,
verbalizations. First results are presented for the inuence of even familiar participants who visit the setting regularly performed
familiarity with the building and map usage on waynding far from perfect. Unfamiliar participants consulted the maps more
performance and orientation behavior. This is followed by results frequently than familiar, but they were not able to take advantage
about path choice strategies. of this additional information.
For the analysis of waynding performance and orientation
behavior the data of all tasks were aggregated. As in one of the two 3.2. Orientation behavior across tasks
tasks 6 and 7 the instruction was varied only the task with the
number-based instruction was included in the analysis. The effect Separate ANOVAs with factors familiarity and map use were
of instruction is considered separately and is reported later in this conducted for each category of orientation behavior.
paper.
4
Univariate tests within the MANOVA were performed as planned paired com-
parisons (LSD). For this and the following analyses, two participants (one familiar,
one unfamiliar) were moved from the map condition to the no-map condition, as
they never used a map in any of the tasks. Since they did not gain any information
3
As detour frequency correlated strongly with detour distance (r .79) only from external representations, they may be treated in the same way as participants
detour distance is reported, as it is more sensitive to the length (and thus severity) who were not allowed to use a map. Thus, the name of the factor map accessi-
of detour episodes. bility was changed to map usage.
214 lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho

Map users looked at room numbers in order to orient them- Table 2


selves less frequently than participants who never looked at a map, Means and (standard deviations) of selected performance and orientation behavior
measures in the visual and the number-based instruction (tasks 6 and 7 only).
F(1,28) 6.43, p .017, h2partial :19. The main effect of familiarity
and the interaction were not signicant. It appears that people Measures Instruction
without access to the wall-mounted maps were more prone to Visual Number-based
adopting this strategy of systematic local search. Duration [s] 235 (68) 209 (46)
Unfamiliar participants looked more frequently at the outside Stops [n] 2.97 (2.19) 2.09 (2.31)
in order to orient themselves than familiar participants, Complete stop time [s] 35 (35) 9 (11)
Orient on outside [n] 0.66 (0.90) 0
F(1,28) 5.60, p .025, h2partial :17. The main effect of map use
Looked at room number [n] 0.31 (0.82) 2.06 (2.09)
and the interaction were not signicant. Looking for landmarks Stops at maps [n] 0.14 (0.36) 1.14 (1.29)
outside the building in order to keep up ones own location in
relation to them is a common technique during navigation.
Unfamiliar participants used it more frequently presumably due to
3.4. Performance and orientation behavior with instruction
their lack of knowledge about the building.
variation
Unfamiliar participants forgot the target room number more
frequently than familiar participants, F(1,28) 8.55, p .007,
In tasks 6 and 7, half of the participants received a visual
h2partial :23. The main effect of map use and the interaction were
instruction instead of a room number. The tasks were matched (cf.
not signicant. The target room number was a four-digit number
Section 2.3) starting at the ground oor of one building. The goal
which participants were told at the beginning of the task. The task
was located on the third oor of the other building. The tasks only
to remember such a number is trivial. However, the result can be
differed in the direction of movement. Each participant received
interpreted as an indicator of increased processing load for
the visual instruction in only one of the two tasks.
unfamiliar participants.
No differences were found for the order of instructions (i.e.,
visual instruction rst vs. number-based instruction rst). There-
3.3. Gender effects fore, the data were collapsed for the further analysis. Data of the
two tasks were aggregated and a MANOVA with factors familiarity,
While gender effects on spatial performance are commonly map use and instruction type (visual, number-based) was
reported in the literature (for an overview, see Coluccia & Louse, conducted for the performance measures. As in the previous
2004), our previous studies had found hardly any gender effect for analyses there was a main effect of familiarity, F(5,52) 2.99,
indoor waynding (Holscher et al., 2006; Meilinger, Holscher, p .019, h2partial :22, but no main effect of map usage,
Buchner, & Brosamle, 2007). We conducted a MANOVA with factors F(5,52) .67, p .650. More interestingly for this analysis, there
gender, familiarity and map use for the performance measures and was a main effect of instruction type, F(5,52) 3.90, p .004,
separate ANOVAs with the same factors for the behavioral mea- h2partial :27, though no interaction of familiarity or map use with
sures. In the MANOVA neither the main effect nor any interaction of instruction type was signicant. In the visual instruction condition
gender with the other factors was signicant (all F(5,20) < 2.37, all participants stopped longer, F(1,56) 18.73, p .001, h2partial :25,
p > .08). The ANOVAs for the behavioral measures revealed no main and there was a statistical trend for them to stop more frequently,
effect of gender (all F(1,24) < 1.88, all p > .18), and no signicant F(1,56) 3.33, p .073, h2partial :06, as well as to take longer to
interactions (all F(1,24) < 3.37, all p > .08). There were no signicant complete the, F(1,56) 3.85, p .055, h2partial :06 (for means and
differences between male and female participants with respect to standard deviations refer to Table 2).
waynding performance and orientation behavior in this study. Again, separate ANOVAs with factors familiarity, map use and
For the analysis of the frequency of map use and the time per instruction type were conducted for the behavior measures. In the
map access we conducted a MANOVA with factors gender and visual instruction participants oriented more frequently on the
familiarity. There was a main effect of gender, F(2,9) 8.28, outside than those in the number-based condition, F(1,56) 10.40,
p .009, h2partial :65, though no interaction. Male participants p .002, h2partial :16 (in fact in the number-based condition no
appeared to access the maps slightly less frequently (M 0.83 participant oriented on the outside), while in the number-based
times/task; SD 0.31) than female (M 1.23; SD 0.73). Yet, this condition participants more frequently looked at room numbers for
gender effect could not be conrmed in the univariate analyses, as orientation, F(1,56) 6.50, p .014, h2partial :10. Also, participants
these tests failed to reach the signicance criterion (frequency of who had a map available consulted it more frequently in the
map access: F(1,10) 3.49, p .09, h2partial :26; time per map
access: F(1,10) 1.76, p .21). Overall there was little evidence for
a gender effect in waynding performance, orientation behavior Table 3
and map use in this experiment. Task properties (left panel) and frequencies of strategy choice (right panel).

Task properties Frequencies

Table 1 Task Instruction Floor change Building change Floor Direction Sum
Means and (standard deviations) of performance and behavior measures. required required strategy strategy
(no. of oors)
Measures Familiar Familiar Unfamiliar Unfamiliar
1 No (0) No
map no map map no map
2 Yes (1) No 25 5 30
Duration [s] 160 (20) 157 (28) 224 (50) 202 (34) 3 No (0) Yes 2 24 26
Stops [n] 1.54 (0.87) 1.83 (0.77) 3.58 (1.94) 3.29 (0.96) 4 Yes (3) Yes 18 10 28
Complete stop time [s] 10 (6) 8 (6) 22 (21) 18 (7) 5 Yes (4) Optional 20 8 28
Detour distance [m] 11 (5) 19 (7) 30 (14) 25 (18)
PAO [%] 43 (14) 52 (15) 84 (44) 98 (40) 6 X Yes (3) Yes 5 10 15
Looked at room number 1.23 (0.69) 1.82 (0.78) 1.66 (0.70) 2.56 (1.13) Room no. Yes (3) Yes 10 7 17
Orient on outside 0.13 (0.19) 0.04 (0.09) 0.27 (0.19) 0.28 (0.27) 7 X Yes (3) Yes 5 11 16
Forgot room number 0.29 (0.19) 0.43 (0.22) 0.63 (0.31) 0.69 (0.28) Room no. Yes (3) Yes 7 6 13
Stops at maps [n] 0.61 (0.29) 1.28 (0.5)
Time/stop at maps [s] 9 (3) 25 (11) 8 Yes (2) Yes 7 25 32
lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho 215

25 Table 4
Selected means (and standard deviations) of performance and behavior measures in
Strategy the analysis of the post-hoc identied strategy choices.
20 Direction
Floor Measures Strategy
Frequency

15 Floor Direction
Duration [s] 210 (66) 186 (70) *
Stops [n] 2.49 (2.30) 2.17 (2.94)
10
Complete stop time [s] 12 (20) 15 (27)
Detour distance [m] 22 (31) 13 (25) *
5 PAO 98 (109) 37 (51) *

*Indicates that p < .05.


0
X Room Number
Instruction
There was a main effect of strategy, F(5,131) 3.84, p .003,
Fig. 4. Choice of strategy (frequency) as a function of instruction. The gure shows the h2partial :13. Participants who changed the oor rst performed
data from tasks 6 and 7 combined. worse than those who changed the building rst. They took longer
to complete the task, F(1,135) 10.14, p .002, h2partial :07, they
walked longer detours, F(1,135) 9.13, p .003, h2partial :06, and
number-based condition than in the visual, F(1,24) 10.32, they had a higher PAO, F(1,135) 5.97, p .016, h2partial :04(for
p .004, h2partial :30. This illustrates that participants chose means refer to Table 4).
different external sources of information depending on the type of
information already provided with the task instruction. 4. Discussion

Major waynding performance differences were identied


3.5. Strategies
between participants who were familiar with the building and
those who were not. This nding was expected as it conrms the
The analysis of strategies in this study was twofold: On the one
nding of familiarity effects by other researchers. It demonstrates
hand, strategy choice was investigated as a function of different
how difcult indoor waynding in a realistically complex setting is
factors like task instruction and familiarity. On the other hand, the
for untrained users, the target audience of most public buildings.
success (in terms of waynding performance) of a chosen strategy
Even the familiar participants performed far from perfect in this
was tested.
setting, pointing to usability challenges in the multi-level multi-
The hypothesis stated that strategic choice would vary as
building ensemble. The observation that participants frequently
a function of task, instruction and familiarity, not by intrinsic
forgot target room numbers suggests a severe memory load
preferences of individuals. This could be clearly established (cf.
imposed by the complex building structure. This stresses a practical
Table 3).
need to better understand waynding challenges in architecture.
In the tasks where both strategies were applicable (tasks 48)
Methods for analyzing such waynding challenges have recently
only two participants chose the same strategy throughout all tasks.
been presented by Holscher et al. (2006) and Brosamle and
In general, in different tasks different strategies were applied (c2(7,
Holscher (2007).
N 221) 68.8, p < .001).5
In the following sections, the evidence for an adaptive nature of
Also, the variation of instruction type in tasks 6 and 7 led to the
waynding strategies in complex building settings is discussed
choice of different strategies (Fig. 4): If participants had to reach the
rst. We then look at further experimental results with respect to
room marked with an X (visual instruction) they more often
task instruction effects and map use and conclude with a preview
changed the building rst, whereas if they had to reach the same
to follow-up research.
room identied by its number (number-based instruction) they
more often changed the oor rst (tasks 6 and 7 added together,
4.1. Adaptive strategy choice
c2(1, N 61) 3.68, p < .028, one-tailed). Across all eight tasks,
familiar and unfamiliar participants acted very similarly: No
When people have incomplete knowledge about a path
general difference in their choice of strategy across all tasks was
network, they rely on strategies to guide their path choices. The
revealed (c2(1, N 221) 1.15, p .283). Similar to familiar and
main question of this experiment was whether participants would
unfamiliar participants, map users and participants not using a map
adapt their waynding strategies to task and environmental
also did not differ in their general strategy preference (c2(1,
characteristics. And indeed it was found that strategy choice is
N 221) 0.03, p .956), nor did women and men (c2(1,
largely inuenced by characteristics of the building, task features
N 221) 0.17, p .392). So the two main factors determining the
and the instruction rather than being an idiosyncratic preference of
choice of a strategy are task and instruction. Familiarity, gender and
the individual participant.
individual preferences play a minor role.
But which strategy was more useful? In order to answer this
4.1.1. Instruction variation
question, we performed a MANOVA on the above mentioned
With respect to the instruction variation, a striking interaction
measures with factors strategy, instruction and task. The latter two
was found between the visual vs. number-based instruction and
factors were included in order to statistically control the inuence
the preferred path choice strategy: with the visual instruction more
of task and instruction. As it was not possible to reasonably apply
people change to the other building section rst, while with the
both strategies in task 1, 2 or 3, these tasks were excluded from this
number-based instruction the oor strategy dominates. One could
analysis.
now speculate that a propositional presentation of the room
number might trigger a navigation strategy based on the oor level
5
Not all subjects could be clearly assigned to one strategy in every task. For
of the target room. But, in this setting the numbering scheme
comparing the inuence of task, both instructions in tasks 6 and 7 were added provides information on both of the hierarchy dimensions oor
together. level and building section, making this interpretation less clear-cut.
216 lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho

Fig. 5. Visual vs. number-based task instruction: The large X illustrates the position of the marked window, the circle represents the viewing location/start point of the task, dark
grey areas mark the target region likely to be inferred from the room number instruction Go to room 1234.

Looking at the information content provided with the task, there region (see also Fig. 5) is considered to be the relevant sub-goal
is not only an obvious difference in the presentation format, but for strategy selection.
also in the degree of precision our participants were able to gain If the participants knew that this target region was further away
from it. In the visual condition, it was very clear for them, that the horizontally than vertically, they rst walked horizontally changing
target location was at the other end of the courtyard, in the the building and then went for the correct oor. This was the case
opposite building part and thus horizontally much further away with the visual instruction (X in the window) in tasks 6 and 7 as well
than vertically. But if participants received the room number as in task 8 where the participants rst had to cross the whole
information, none of them knew exactly in which corridor the building KG I. If the participants otherwise had to assume that the
target room would be (especially for task 6, room 1234). From the horizontal distance was at least potentially shorter compared to
numbering scheme it was clear in which general region the room the vertical one, they rst changed the oor and only then the
must be (2nd oor of KG I), and verbal comments indicate that most building section. That was the case in tasks 4 and 5 and in the room
participants could also not eliminate the mezzanine part of the number instruction in tasks 6 and 7. In other words, if two options for
building as a potential target area (for an illustration, see Fig. 5). As a hierarchical strategy were equally available in a task, the vertical
a consequence, the rational strategy for the room-number and horizontal direction information was used to disambiguate the
instruction group was to access the potential target region as path choice.
quickly as possible, following a oor strategy. For the visual
instruction such a precaution was not called for, as the target room 4.1.3. Strategy choice and waynding performance
was pinpointed more clearly. Finally, which strategy was more successful? In the case of this
multi-level multi-building setting, the strategy of rst moving into
4.1.2. Task variation the correct part of the building before changing to the proper level
The navigation strategies followed a decisive pattern of was much more effective across a number of performance
regionalization (Wiener et al., 2004): The building can be mentally measures. On rst sight, this observation is at odds with the
divided into a dual hierarchy of building parts and oor levels. performance boost found in Holscher et al. (2006) for exerting
When only one hierarchical aspect is involved (the oor change in a oor-rst strategy in a multi-level building. But in fact, in both
task 2 or the building change in task 3), people relied very strictly studies, the effective strategy relies on hierarchical planning to
on it. Thus, they changed either the oor or the building before overcome difculties in the building. The present study extends the
searching for the rooms horizontal position. complexity of the multi-level building in Holscher et al. (2006) to
When participants had to change both buildings and oors a multi-building setting. This multi-building setting has a mis-
within a single task, the picture was more ambiguous. For an matching number of oors, so that, e.g., for the second and third
interpretation of the path choices observed here, it is important oor, changing between the buildings was not easily possible and
to consider with what level of precision the participants could the 4th oor of KG III had the same elevation as the 3rd oor of KG I.
identify the target position based on the number information Across all tasks, the direction strategy of rst changing the building
provided to them. Interestingly, familiar and unfamiliar partici- section is best suited to overcome the additional burden of the
pants overall differed rather little in their strategy choice, which present setting, namely the inconsistencies in oor numbering and
most clearly reects differences between the tasks. Familiar relative oor elevations between parts of the building.
participants had a better understanding of the overall structure One further difference between the eight tasks is the degree of
and connectivity of the building ensemble, but when participants consistency or dominance of one particular path choice: for the rst
received a room number, almost none of the participants were three tasks there is little variation, since only one type of region
sure about its precise location. After performing the rst three change is required. For tasks 47 a considerable minority (about
tasks, most participants were able to identify the general region one-third) choose the non-dominant strategy. This can be attrib-
of the building in which the target room must be located, i.e., uted to the fact that for these tasks, both strategies offer reasonably
they could identify the building section (KG I vs. KG III) as well as direct access to the target region, thus making path choices more
the oor level based on the numbering scheme. A rational ap- arbitrary. For the nal task (task 8) once again a strong dominance
proach, trying to avoid potential backtracking inside the target of one option, the direction strategy, was observed. In this task, it
region, would suggest moving to the nearest edge of the target was quite clear to the participants that their horizontal travel was
region on a straight path whenever possible (Wiener et al., 2004). longer than the vertical part. Furthermore, there was no direct
Thus, rst moving to the nearest edge of the potential target connection on the target level, making the directional strategy safer.
lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho 217

4.1.4. Strategies, familiarity and map access 4.2. Visual vs. number-based task instruction
Also in contrast to our previous ndings on strategy choice, no
general differences were found between strategy choices of par- This study has identied signicant differences between a visual
ticipants who were familiar with the building and novices, a fact and number-based task instruction both with respect to perfor-
that is likely attributable to the strong inuence of task- and in- mance measures and path choice strategies.
struction-related effects. Yet, for the one task where familiarity Despite the fact that the start and destination locations were the
differences in strategy choice were clearly identied, the familiar same in both task versions, participants who received the visual
participants denitely preferred the more successful strategy. This instruction had more problems solving the task. This is reected in
is a direct replication of the earlier nding (Holscher et al., 2006): a longer overall task duration as well as in more frequent and
experts prefer a strategy with a higher success rate. longer stops. What makes the task with a visual instruction more
As with familiarity, the availability of maps did not inuence difcult to perform correctly? Several factors are likely to play
strategy choice much. This was likely due to the fact that any given a role: Participants saw the target location from the outside of the
map only displayed a single oor within one building. The three- building and had to translate that information to their
dimensional structure of the building ensemble, the differences understanding of the internal structure of the building, i.e., they
between oors and the unusual vertical connection were not sa- need to switch from an external to an internal perspective of the
liently represented in the maps (see Section 4.3). It therefore makes building. In addition, when approaching the target location on
sense that map availability did not inform strategy choice. the inside (from the corridor) a match must be found between the
Overall, the path choice strategies observed here were highly window and its corresponding room door.6
adaptive to both the overall geometry of the building and the in- Participants clearly understood that the task variation put
formation provided with the task variants. If one takes into account different demands on their behavior and consequently used dif-
that participants for the most part knew about target regions rather ferent types of environmental information to solve the task: With
than precise room locations, it appears that they have used hierar- the visual instruction, the participants used orientation to the
chical strategies insofar as they tried to move to the proper region outside about twice as often as across all other tasks, while in the
most directly and then performed local search. For tasks which two matched number-based tasks, no orientation to the outside
supplied two equally reasonable alternative paths to the target re- was observed at all. By contrast, participants given visual in-
gion, they ne-tuned their initial path choice with a least-angle structions hardly ever referred to one of the wall-mounted maps
heuristic, not to the nal room number, but to the target region. to identify the location of their destination in the corridor system,
Thus, a nesting of direction-based and hierarchical decision ele- while participants with the number-based instruction very
ments was found in this complex building setting. This nding may heavily relied on map usage to solve their task (more than twice
help to reconcile the apparent opposition between path choice as often than on average across all eight tasks). Interestingly, no
strategies based on hierarchical representations of space (Hirtle & signicant interaction was revealed between prior familiarity
Jonides, 1985; Wiener et al., 2004) and those concentrating on with the building and the instruction variation. Consequently, the
geometric relations like least-angle (Hochmair & Frank, 2002). The different task characteristics affected familiar and unfamiliar
adaptive nature of strategy choice indicates that both types of participants in a similar fashion. Familiarity may be more limited
strategy can be present in the same person and the same setting and to everyday task demands, i.e., search a room by number, than
are triggered by specic factors of spatial relation between desti- one might have expected.
nations as well as the information detail available to the navigator. The results are also consistent with recent ndings regarding
In general, the results indicate a strong inuence of strategy the representation of complex spaces like buildings. These ndings
choice due to the spatial and information characteristics of the task indicate that such spaces are rather represented in several units
rather than a stable preference of the individual participant. This than unied in within one continuous representation (e.g.,
nding may appear to contradict studies indicating individual dif- Meilinger, Riecke, & Bulthoff, 2007; Wang & Brockmole, 2003). For
ferences in spatial ability and strategy use (e.g., Coluccia & Louse, example participants do not necessarily update the whole area
2004; Lawton, 1996). However, many results revealing ability dif- (Wang & Brockmole, 2003) and instead orient on local reference
ferences focus on object-sized spaces rather than on spaces one frames rather than on a global reference frame common for the
navigates through like in the case of the present study (cf. Hegarty whole area (Meilinger et al., 2007). Accordingly in the present study
& Waller, 2005). Also, a questionnaire constructed to capture in- participants quickly pick up relevant units of the building ensem-
dividual differences in large-scale spaces did not correlate sub- ble, i.e., oors, and adopt their strategies according to them
stantially with performance parameters in this study (Santa regarding the task requirements. Having the environment clustered
Barbara Sense of Direction Scale; Hegarty, Richardson, Montello, into relevant partitions should also allow humans to exibly adopt
Lovelace, & Subbiah, 2002). In principle, the application of a certain strategy choices according to the task affordances, as was observed.
strategy might depend on individual spatio-cognitive abilities. For Further information about the geometric ne structure within such
example, directly approaching a target on one level probably de- a proposed unit as provided by the maps did not increase perfor-
pends on the ability to encode metric relations. The FSBSOD self- mance. These units seem to be accessed more easily via number-
report measure of individual differences examined here did not based instructions compared to visual instructions, since with
show an inuence. Whether individual differences as measured visual instructions several of these units have to be combined
by standardized psychometric ability tests systematically interact mentally in order to solve the task.
with the role of environmental features for path choice strategies As was discussed in the previous section, the variation of visual
would need to be claried by future studies but goes beyond the vs. number-based task instruction had the most prominent
scope of the current work. In addition, the sample size of 32 modulating effect on path choice strategies in this study. Together
participants in the present study would be considered small for with the differences in using the external navigation aid of maps,
a study focusing on individual differences. Even for tasks with less this highlights the adaptive nature of waynding behaviors on
precise requirements a substantial subset of participants chose the different scales. The nding illustrates that one must be very careful
non-dominant strategy, indicating an inuence of individual path
choice. So the strong inuence of environment on strategy choice is
no contradiction to the existence of individual differences as 6
But this later fact alone is almost certainly not responsible for the substantial
reported in other studies. difference in task duration observed here.
218 lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho

when generalizing research ndings in waynding studies, as these study the map and orient themselves in the building. This would
task characteristics inuence the strategy choice and may modulate clearly not be the case in an emergency where people are under
effects of geometrical properties (environmental features) as well stress and tend to panic. This as well as other research (Klippel
as inter-individual differences. It also implies that building users et al., 2006) shows that the deployment and design of emergency
who have a general idea about the location of a room, but do not maps have to be re-considered and be subject to empirical
know the specic room number (or look for a named function like testing.
a radiology department or departmental library), may choose Naturally, these results do not imply that maps are inadequate
rather different paths than people who have been given the more for spatial learning or navigation per se. Thorndyke and Hayes-Roth
precise information about the room number. (1982), like many others, have impressively shown that maps can
This has important consequences for the design of the circula- be an excellent means of acquiring spatial knowledge, especially
tion system in complex multi-level buildings to make sure that both survey representations which normally require a rather long time
strategies are equally well supported. Especially, designers should to acquire via route navigation alone. The present ndings imply
take care to fully connect all oors both vertically and horizontally, that the standard wall-mounted maps, but not maps in general, are
unlike the discontinuities in the present building. Brown, Wright, inadequate for navigating complex multi-level settings like the
and Brown (1997) describe similar navigation decits in a more building in our study. It is an important topic for our subsequent
qualitative Post-Occupancy Evaluation (POE) of a hospital complex studies to identify how wall-mounted maps should be improved in
with separated inpatient and outpatient sections. In their study order to support these tasks in a better way. We suspect that it will
participants who initially confused the building sections, used the be especially important to make sure that all parts of the building
wrong entrance and then employed a oor strategy (possibly trig- ensemble and all vertical connections are properly included in such
gered by knowing a room number) got stuck at closed corridors in improved maps. Also, it should be helpful if at least those wall-
the upper oors. The present study provides a controlled experi- mounted maps positioned near vertical decision points (staircases)
mental conrmation for this qualitative observation. would be supplemented with a means to illustrate the topology of
the vertical connections, e.g., with a cross-sectional diagram. The
4.3. Maps adequate level of schematization for this support is also subject to
future research.
When given the opportunity, unfamiliar participants tried to
compensate for their lack of prior knowledge by making use of the 4.4. Future research
wall-mounted maps. They did so both longer and more often than
familiar users. However, surprisingly it turned out that the oor Together with a group of information designers the authors
maps did not have any positive impact on waynding performance, are currently investigating opportunities for improving the
neither for rst time nor for regular visitors. guidance system in this building ensemble. Maps have been re-
In fact, reading the maps required additional time and did not designed together with a consistent set of signage to rooms in
receive any payback in performance enhancement (Butler et al., the setting. A follow-up study involves the same tasks in order to
1993). One might suspect that the map decits relate to a lack of allow for a comparative analysis of the current decits and pos-
alignment, but we were able to show that this is not a decisive sible improvements. Preliminary results (Holscher, Buchner,
factor here (see Section 3.1). Why did the oor maps not help in Brosamle, Meilinger, & Strube, 2007) indicate that even large,
waynding? In a study on outdoor navigation Soh and Smith- salient maps which were designed particularly for the purpose of
Jackson (2004) demonstrated the impact of cultural differences on easy navigation may not allow rst-time visitors to reliably
successful map usage, likely due to differences in level of experi- compensate for their lack of knowledge in such a complex real-
ence with map conventions and experience with the particular type life building setting. Instead signs appear to be the main attrac-
of environment. For the present study, we can safely assume that tors of attention and people tended to follow them instead of
the participants were familiar both to the general type of map and intensively studying the maps. This aspect of our research has
the type of environmental setting (public university buildings), and a focus on real-world application and further studies are needed
simply lacked knowledge about the particular building. Despite to get a fuller understanding of the role of external navigation
violating several design principles (Klippel, Freksa, & Winter, 2006), aids for successful waynding and for incidental as well as de-
the re plans do provide detailed and accurate survey information liberate spatial learning.
about single oors within each building section. However, the With respect to more basic research into navigation strategies
waynding tasks require across-level as well as across-section and cognitive representation of the environment, we will continue
navigation. Participants may have had trouble to integrate our efforts to extend the indoor waynding research to further
information from single oor maps into a coherent multi-level types of public buildings with different spatial layouts and
representation of the whole building or even parts of it. In this properties. Both the settings in Holscher et al. (2006) and in the
particular building ensemble, the oor numbering is inconsistent present study owe their complexity largely to vertical structure,
between the main parts of the building (KG I vs. KG III) and the oor while within a single level the path choices were relatively
numbering is not transparent with respect to the mezzanine levels. limited. It remains to be investigated how the hierarchically- based
Nonetheless, the main decit of the re plans clearly generalizes to waynding strategies that we have identied extend to layouts that
other multi-level multi-building environments: In most real-world provide mostly within-oor horizontal extend and complexity.
cases, the re plans of complex buildings do not cover the area Hospital or airport terminal buildings would be prime examples of
beyond a single building section or oor and thus fail to supply the such settings to validate the scope of applicability as well as the
necessary information for route planning across oors or sections. adaptive value of human waynding strategies. We start by looking
The re plans were not only designed to help re ghters at a building with a less complex internal structure to further
entering the building to quickly assess its topography in the case untangle the relationship of hierarchical route choice elements
of a re, but also to help visitors to quickly nd a way out in the (predominantly oors) and least-angle based directional cues. Here
case of an emergency. In our study participants were instructed initial results also indicate that these two general choice elements
to nd the target room as quickly as possible. At the same time may be combined adaptively, partly based on which spatial
they were instructed to solve the task in a calm manner without dimension is dominant in the mental representation of an
any hurry, so they actually had as much time as they wanted to individual user.
lscher et al. / Journal of Environmental Psychology 29 (2009) 208219
C. Ho 219

Acknowledgements D. S. McNamara, & J. G. Trafton (Eds.), Proceedings of the 29th annual cognitive
science society (pp. 377382). Austin, TX: Cognitive Science Society.
Holscher, C., Meilinger, T., Vrachliotis, G., Brosamle, M., & Knauff, M. (2006). Up the
The research is supported by the German Research Foundation down staircase: waynding strategies and multi-level buildings. Journal of
(DFG) in the Transregional Collaborative Research Center Spatial Environmental Psychology, 26(4), 284299.
Cognition (SFB/TR-8). We are grateful to Martin Brosamle and Klippel, A., Freksa, C., & Winter, S. (2006). The danger of getting lost YAH maps in
emergencies. Journal of Spatial Sciences, 51(1), 117131.
Leonard Reinecke for support with data collection and analysis, and Lawton, C. A. (1996). Strategies for indoor waynding: the role of orientation.
to Lara Webber for proof-reading and comments. Parts of this study Journal of Environmental Psychology, 16(2), 137145.
were originally reported at the International Spatial Cognition 2006 Lee, P. U., & Tversky, B. (2001). Costs of switching perspectives in routed and survey
descriptions. Proceedings of the 23rd annual conference of the Cognitive Science
conference, Bremen, Germany, September 2006. Society. Edinburgh: Scotland.
Levine, M., Jankovic, I., & Palij, M. (1982). Principles of spatial problem solving.
References Journal of Experimental Psychology: General, 11, 157175.
Meilinger, T., Holscher, C., Buchner, S., & Brosamle, M. (2007). How much
information do you need? Schematic maps in waynding and self localisation.
Brosamle, M., & Holscher, C. (2007). Multi-level complexity in terms of space
In T. Barkowsky (Ed.), Spatial cognition V Reasoning, action, interaction (pp.
syntax: a case study. Proceedings of 6th International Space Syntax Symposium
381400). Berlin: Springer.
Istanbul, I, 44.0144.12.
Meilinger, T., Riecke, B.E. & Bulthoff, H.H. (2007). Orientation specicity in
Brown, B., Wright, H., & Brown, C. (1997). A post-occupancy evaluation of
long-term-memory for environmental spaces Proceedings of the 29th
waynding in a pediatric hospital: research ndings and implications for
annual conference of the Cognitive Science Society, Nashville, USA, pp.
instruction. Journal of Architectural and Planning Research, 14(1), 3551.
479484.
Butler, D. L., Acquino, A. L., Hissong, A. A., & Scott, P. A. (1993). Waynding by
Moeser, S. D. (1988). Cognitive mapping in a complex building. Environment and
newcomers in a complex building. Human Factors, 35(1), 159173.
Behavior, 20(1), 2149.
Coluccia, E., & Louse, G. (2004). Gender differences in spatial orientation: a review.
Montello, D. R. (1991). Spatial orientation and the angularity of urban routes: a eld
Journal of Environmental Psychology, 24, 329340.
study. Environment and Behavior, 23, 4769.
Conroy Dalton, R. (2003). The secret is to follow your nose: route path selection an
Montello, D. R., & Pick, H. L. (1993). Integrating knowledge of vertically aligned
angularity. Environment and Behavior, 35(1), 107131.
large-scale spaces. Environment and Behavior, 25(4), 457484.
Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Verbal reports as data.
Passini, R. (1992). Waynding in architecture (2nd ed.). New York: Van Nostrand
Cambridge, MA: MIT Press.
Reinhold Company.
Fontaine, S. (2001). Spatial cognition and the processing of verticality in
Richardson, A. R., Montello, D. R., & Hegarty, M. (1999). Spatial knowledge acqui-
underground environments. In D. R. Montello (Ed.), COSIT 2001 (pp. 387399).
sition from maps, and from navigation in real and virtual environments.
Berlin: Springer.
Memory and Cognition, 27, 741750.
Garling, T., Lindberg, E., & Mantyla, T. (1983). Orientation in buildings: effects of
Shelton, A. L., & McNamara, T. P. (2004). Orientation and perspective dependence in
familiarity, visual access, and orientation aids. Journal of Applied Psychology,
route and survey learning. Journal of Experimental Psychology: Learning, Memory
68(1), 177186.
and Cognition, 30, 158170.
Gray, W. D., Sims, C. R., Fu, W.-T., & Schoelles, M. J. (2006). The soft constraints
Soeda, M., Kushiyama, N., & Ohno, R. (1997). Waynding in cases with vertical
hypothesis: a rational analysis approach to resource allocation for interactive
motion. Proceedings of MERA 97: Intern. conference on environmentbehavior
behavior. Psychological Review, 113(3), 461482.
studies 559564.
Hegarty, M., Richardson, A. E., Montello, D. R., Lovelace, K., & Subbiah, I. (2002).
Soh, B. K., & Smith-Jackson, T. L. (2004). Inuence of map design, individual
Development of a self-report measure of environmental spatial ability.
differences, and environmental cues on way nding performance. Spatial
Intelligence, 30, 425447.
Cognition and Computation, 4, 137166.
Hegarty, M., & Waller, D. (2005). Individual differences in spatial abilities. In P. Shah,
Thorndyke, P. W., & Hayes-Roth, B. (1982). Differences in spatial knowledge
& A. Miyake (Eds.), The Cambridge handbook of visuospatial thinking (pp. 121169).
acquired from maps and navigation. Cognitive Psychology, 14, 560589.
Cambridge: Cambridge University Press.
Wang, R. F., & Brockmole, J. R. (2003). Simultaneous spatial updating in nested
Hirtle, S. C., & Jonides, J. (1985). Evidence of hierarchies in cognitive maps. Memory
environments. Psychonomic Bulletin & Review, 10, 981986.
& Cognition, 13(3), 208217.
Weisman, J. (1981). Evaluating architectural legibility: way-nding in the built
Hochmair, H., & Frank, A. U. (2002). Inuence of estimation errors on waynding
environment. Environment and Behavior, 13(2), 189204.
decisions in unknown street networks analyzing the least-angle strategy.
Wiener, J. M., Schnee, A., & Mallot, H. A. (2004). Use and interaction of navigation
Spatial Cognition and Computation, 2(4), 283313.
strategies in regionalized environments. Journal of Environmental Psychology,
Holscher, C., Buchner, S., Brosamle, M., Meilinger, T., & Strube, G. (2007). Signs and
24(4), 475493.
maps Cognitive economy in the use of external aids for indoor navigation. In

Você também pode gostar