Hex Maze: A new virtual maze able to track acquisition and usage of three
navigation strategies
⁎
Meg J. Spriggsa,b, , Ian J. Kirka,b, Ronald W. Skeltonc
a
School of Psychology, University of Auckland, New Zealand
b
Brain Research New Zealand, New Zealand
c
Department of Psychology, University of Victoria, Canada
A R T I C L E I N F O A B S T R A C T
Keywords: Spatial navigation is a complex and multi-faceted skill that, in humans, is understood to encompass two distinct
Allocentric navigational strategies, namely allocentric and egocentric navigation. These differ in the frame of reference used
Egocentric and the brain networks activated. However, egocentric navigation can be further divided into two, equally
Spatial navigation distinct strategies depending on whether the navigator is using subject-to-object relations (egocentric-cue) or
Virtual maze
direction of body turns (egocentric-response) to navigate. To date, there are no experimental paradigms able to
Sex differences
distinguish between participants’ employment of allocentric, egocentric-cue and egocentric-response strategies,
and to track their usage over time. The current study presents the Hex Maze: a novel virtual environment that
can not only distinguish between the three navigational strategies, but can also be used to index aspects of
strategy use such as preference, acquisition, stability and competence. To illustrate this, 32 male and 32 female
participants were presented with the Hex Maze and sex differences in strategy usage were explored. While the
results offer some support for previously identified sex differences in strategy preference, there were no sig-
nificant sex differences in the novel measures of strategy acquisition, stability, or multi-strategy competence.
Additionally, our results suggest that strategy preference does not preclude learning to competently navigate
using other strategies. Importantly, the current study offers validation for the Hex Maze as an unbiased method
of exploring spatial navigation, and it is anticipated that this easy-to-use tool will be valuable across research and
clinical settings.
⁎
Corresponding author at: The University of Auckland, Private Bag 92019, Auckland 1142, New Zealand.
E-mail address: mspr827@aucklanduni.ac.nz (M.J. Spriggs).
https://doi.org/10.1016/j.bbr.2017.11.041
Received 24 June 2017; Received in revised form 17 November 2017; Accepted 30 November 2017
0166-4328/ © 2017 Elsevier B.V. All rights reserved.
Please cite this article as: Spriggs, M.J., Behavioural Brain Research (2017), https://doi.org/10.1016/j.bbr.2017.11.041
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
navigation can be referred to as ‘cue-based’, as it relies of the location of As can be seen from the above discussion, virtual versions of the
cues in the environment. However, egocentric navigation also en- RAM and the MWM only distinguish between allocentric navigation,
comapsses navigating based on the direction of body-turns, which can and one of the types of egocentric navigation (egocentric-response and
be referred to as ‘non-spatial’ or ‘response-based’ as it does not involve egocentric-cue respectively). Additionally, measures of performance are
encoding of environmental information [11,12,7]. Accordingly, ego- largely limited to accuracy, latency and performance on a single probe
centric navigation is frequently associated with activation across both trial. While such measures indicate strategy choice and task proficiency,
the parietal lobe (posterior cingulate, precuneus, inferior parietal lo- they do not provide more detailed information about strategy usage,
bule) [21,16,17,22] and striatum [11,23,12]. The parietal lobe is un- such as acquisition, stability, and competence. These limitations may
derstood to be involved in “body referencing” [1,24] and may therefore have implications for how we interpret individual, or group based
be better suited for cue-based navigation. Conversely, the striatal differences in spatial navigation, for example, the widely cited sex
system is associated with stimulus-response driven behaviour [25,11] differences in strategy preference and navigational efficiency. Both
as well as the encoding of spatial layout [1] and may thus be better subjective (self-report) and objective (providing navigational informa-
suited for response-based egocentric navigation. Similarly to allocentric tion) measures have previously indicated that males are more likely to
navigation, a wider network of regions has frequently been identified in use allocentric strategies while females are more likely to use egocentric
neuroimaging studies of egocentric navigation including frontal and strategies [43–46]. Additionally, while males tend to perform better in
occipital regions, some of which showing some overlap with those ac- tasks requiring allocentric navigation, this male-advantage is reduced
tivated in allocentric tasks [20]. or eliminated when egocentric strategies can be employed
In recent years, human spatial navigation has increasingly been [32,37,47–49]. However, whether these sex differences are due to
studied in virtual environments. The use of a virtual environment has strategy preference (the strategy a participant will chosen most often
several advantages over real world navigation including easy control over all others) and/or competence (accuracy and efficiency using a
and manipulation of experimental factors, and allowing participants particular strategy), and whether there are sex differences in strategy
active exploration, repetitive practice, and feedback [26,27]. There are acquisition, or stability remains unclear. Furthermore, it is uncertain
however some limitations to virtual environments, such as a narrower whether these sex differences would persevere in an environment in
point of view and limited resolution. Regardless, virtual reality, and the which navigators have the choice of using all three navigation strategies
cognitive maps built within, show good translation to real world spatial (allocentric, egocentric-cue, and egocentric-response).
navigation and are easy to transfer between clinical and research set- Here we present a new paradigm for studying spatial navigation; the
tings [28,27]. Hex Maze. The Hex Maze is a 6 arm RAM, but follows the paradigmatic
The virtual environments used to study spatial navigation in hu- procedures of a MWM. This means that while it has a countable number
mans are often derivatives of well established rodent tasks, such as the of pathways, and thus a limited number of choices as to where the goal
radial arm maze (RAM) and the Morris water maze [29]. The RAM has a might be, there is only one goal in a fixed location, and trials begin from
central hub with radiating arms, ranging in number from just four to 17 variable start locations (always at the end of an arm, rather than in the
[30]. The rodent RAM is usually constructed so that distal cues are centre of the maze). Participants are presented with alternating
visible, and a subset of the arms are baited with food. In general, testing learning and probe trials. In learning trials (called “Find-It” trials),
takes two forms: i) working memory, in which within-trial arm re-visits participants must locate an invisible platform. On probe trials (called
are considered errors and; ii) reference memory, in which maximally “Show-Me” trials), they must then return to the location of the platform
successful performance requires the animal to only visit the arms that on the previous Find-It trial. Importantly, a shift in maze configuration
have been baited on previous trials. Virtual versions of the RAM also between Find-It and Show-Me trials means that participants will give
require human participants to navigate to ‘goals’ hidden at the end of a different responses on probe trials depending on which of the three
selection of the arms, and consequently remember which arms they navigational strategies they use (allocentric, egocentric-cue and ego-
have and have not visited [31–34,10,12]. Performance is measured in centric-response). This not only allows an examination of strategy
number of errors (wrong arms visited), and latency. Navigational preference, but also provides more detailed information on strategy use,
strategy use is determined using a small number (1 or 2) of probe trials. such as acquisition, stability, and competence. The aims of the current
Hidden goals can be located allocentrically using the extra-maze en- study were to 1) establish the validity and virtues of the paradigm, 2)
vironment, or egocentrically using the spatial arrangement of arms of provide a baseline from young, healthy university students and 3) ex-
the maze (e.g., “the first goal is in the second arm to the left of the start plore sex differences in the various aspects of strategy use elucidated by
position”). This egocentric navigation corresponds to ‘response-based’ the paradigm.
navigation.
One of the limitations of the rodent RAM is the potential use 2. Materials and methods
proximal visual or olfactory cues [29]. The rodent MWM [35,36] was
designed specifically to address this issue. The MWM consists of a cir- 2.1. Participants
cular pool of opaque water, featureless on the inside, and surrounded
by curtains so that distal cues can be manipulated (or removed). A ty- Participants were 64 undergraduates recruited primarily from a first
pical paradigm might involve learning the position of a hidden platform year Psychology class (32 Men and 32 Women, M Age = 21.0,
over repeated trials with varying start position (i.e., in the N,S,E, or W SD = 3.8) who received partial course credit for participating. The
quadrants of the maze). Subsequent to learning the position of the study was approved by the University of Victoria’s Human Research
platform (a day later than the last hidden platform trial for example), Ethics Board. After consent was obtained, participants were given a
probe trials are run in which the platform is removed altogether. Spatial brief survey to collect demographic information (age, handedness, etc.),
memory can be assessed by the amount of time spent in, or distance to check exclusion criteria (English not first language, brain injury,
swam within the previous target quadrant. Virtual versions of the MWM psychiatric and neurological disorders) and some potential influences
also require human participants to locate a single hidden goal or plat- (exhaustion, dizziness, and recent food, alcohol or tobacco).
form and performance is measured via accuracy, latency, and path
length [32,37,11,38,39,40–42]. A single probe trial is used to de- 2.2. Apparatus
termine whether the platform was located using an allocentric (extra-
maze environmental cues) or egocentric-cue (an intramaze landmark) All navigational testing was completed in the Hex Maze. The maze
strategy. Because the start location is varied, an egocentric-response was built using the UnReal Development Kit (UDK® Epic Megagames),
strategy cannot be used. run on a Window 7 desktop computer (Dell, 2.83 GHz) using a 23”
2
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
Fig. 1. Hex maze design. a) A view of the Hex Maze from above (Note that participants do not see the maze from this angle). b) Four views of Hex Maze looking north (i), south (ii), east
(iii), and west (iv) from the centre of the maze.
1680 × 1050 screen resolution placed on a desktop ∼80 cm from the the south with a large island (Fig. 1b). The room also had two sets of
participant. narrower but equally high windows to the east and west which afforded
Participants moved using an Xbox 360-type controller with buttons views of the mountain ranges coming down to meet the large body of
programmed to allow only left, right and forwards movements (both water almost exactly at the east and west. At the end of each arm,
joysticks were disabled). floating above the low wall bounding the arena, was a whitish/grey
All statistical analyses were conducted in Microsoft Excel. sphere.
Fig. 1a provides an aerial view of the maze environment. The maze
was contained in a large circular arena with an apparent diameter 2.3. Navigation training and testing
∼60m. It was bounded by a low stone wall and had a tiled floor. The
floor of the maze was trisected by three recessed alleys each 1/10th the 2.3.1. Training
diameter of the arena, at 60° angles to each other with one alley or- Throughout training and testing, participants were instructed as to
iented East-West, and a hexagonal plate in the centre so that the floor the nature of the upcoming trial by the experimenter reading from a
tiling was not a clue to orientation. The entire arena was enclosed in a script. Training consisted of three phases: 1) exploration, 2) visible
round room that was twice the diameter of the arena. Two large win- platform trials, and 3) an explicit probe.
dows to the arbitrarily labelled north and south afforded views to a Firstly, the exploration phase was used to familiarize participants
mountain range that peaked at the north, and a large body of water to with the environment and how to move using the controller. At the start
3
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
of the single “explore” trial, participants were teleported to the eastern windows (but is no longer marked by the dark grey sphere and is no
extreme of the room outside of the area (grassy area visible in Fig. 1a), longer the second arm to the left). If they went to the end of the west
facing out the window. They were instructed to explore the room, look arm, they were scored as navigating by an egocentric “cue” strategy as
out the windows, and to let the experimenter know when they were this is now marked by the dark grey sphere. If they took the second left
satisfied with their knowledge of the environment and the use of the turn and went to the southeast arm, they were scored as navigating by
controller. an egocentric “response” strategy, as this corresponds to the original
The exploration phase was followed by four visible platform trials to turning direction. If they went to either of the other two arms, or re-
familiarize the participant with the platform-finding procedure. turned to the original start arm, they were scored as not having learned
Participants were teleported to the end of one arm of the area and were the goal location or any strategy.
told to move to a visible platform within the arena as quickly and di- On the second set of 5 trial pairs, the “Find-It” trials started in the
rectly as possible, and step on it. The visible platform locations were, in southwest, and the platform was at the end of the east arm (the second
order: 1) the centre of the arena, 2) the far end of the alley in line with right) and was marked by a blue sphere. “Show-Me” trials started at the
the starting alley, 3) the second alley on the right and 4) the first alley end of the west arm and the blue sphere was moved to the end of the
on the left. Once participants reached the platform, it made an audible northeast arm. Once again, participants were scored by their arm
sound and they were instructed that they could look around as much as choice: to the same place (east), to the same cue (blue-sphere, north-
they liked, but would be prevented from moving off the platform by an east) or by the same response pattern (taking the second right, south-
invisible barrier. For these four trials, an inner wall was raised to mask east).
the outside landscape and the spheres marking the end of each arm In sum, on learning trials, the three navigational strategies were
were removed. convergent, while on the probe trials they were divergent and revealed
Finally, participants were presented with an explicit probe. strategy selection. At no point were participants informed of this di-
Participants were told that this was the first “Show-Me” trial and that vergence on the Show-Me trials or that the start position would be
the experimenter wanted to see how good their “intuition” was by different. The most important measure of performance was arm choice
having them go to where they thought the invisible platform might be on the probe (“Show-Me”) trials, however distance and latency were
hidden on the next trial. The trial started with the participant in the also recorded. The use of two different maze configurations (with dif-
centre of the maze, facing east. One of the six white spheres was re- ferent places, cues and responses) reduces the effects of overtraining
placed with a dark grey sphere. The purpose of this trial was to see how and allows additional measurements such the effects of experience on
many participants could guess that the platform would be hidden in the acquisition of a new goal location and the stability of strategy selection.
arm marked by the coloured sphere, or whether participants were un-
intentionally biased by any other aspects of the maze environment. 2.3.3. Ancillary tasks
Participants were told to go to the end of the arm where they thought Three tasks were administered to assess the extent of each partici-
the platform would be hidden. Once the participant reached the end of pant’s navigational knowledge. First, three “forced-strategy” probes
an arm, there was an audible beep and they were trapped by an in- were administered (Fig. 2b). Participants were instructed to go to the
visible barrier. location of the platform on the last set of trial pairs (configuration 2).
“Place” probe trials started in the centre of the Hex maze (to obviate
2.3.2. Testing response navigation) and there were no cue spheres. The landscape was
In order to measure acquisition of navigational knowledge and visible through the windows. The “Cue” probe also started in the centre
strategy preference, participants were given 10 pairs of trials in sets of of the maze, but the windows were blocked by a blank wall (to obviate
five across two maze configurations (Fig. 2a). Each pair of trials con- allocentric navigation). Each arm was marked with one of three white
sisted of a learning trial followed by an explicit probe trial, labelled spheres and three coloured spheres corresponding to the cues of the
“Find-It” trials and “Show-Me” trials respectively. Prior to the first first set (grey), the second set (blue), and an unused option (green). The
learning trial, participants were informed that they would be alter- Response probe began at the end of one arm. There were no spheres
nating between two types of trials across the next two blocks of trials, present and the windows were blocked by a blank wall. In all trials,
and the two types of trials were described. On “Find-It” trials, their task participants were instructed to go to the end of an arm to make their
was find an invisible platform at the end of one arm, which would choice and responses were counted as to whether they were correct
visibly rise-up below them and make a sound when it was found. Like relative to the first set or second set of trial pairs.
on the visible platform trials, they were not able to get back off the The second ancillary task was a “Sense of Direction” task [50,40,41]
platform once on it, but could look around as much as they liked and to test for the ability to point to landscape features that were not in
were asked to inform the experimenter when ready to move on. “Find- sight (and sense of immersion or presence) and cognitive map. Parti-
It” trials were always followed by “Show-Me” trials, in which there was cipants were teleported to the north edge of the room, looking north-
no platform. Instead, participants were tasked with returning to the wards (towards the mountains). They were asked to use their hands in
location of the platform on the previous “Find-It” trial. Once they their air above them to point to where they thought the water was.
reached the end of their chosen arm, they would hear a beep and an Their responses were recorded on paper and later scored as follows: 0-at
invisible barrier would keep them from leaving the arm. Participants the display screen, 1-to one side of the screen, 2-straight to left or right,
were also told that for each configuration, the platform would remain in 3-behind the midline, to the side, 4-straight behind.
the same place, and that they would be informed on each trial whether The third ancillary task was a Room Reconstruction task
it was a “Find-It” or “Show-Me” trial. [38,50,40,41]. Participants were presented with a template for placing
In the first set of five trial pairs, the “Find-It” trials started at the end pictures representing the landscape, the walls, and the radial maze
of the east arm and the goal (platform) was at the end of the southwest floor. First they were asked to pick 4 cardinal landscapes from a random
arm (see Fig. 2a), which was the second arm to the left and was marked array of 8 cardinal and semi-cardinal views and were awarded a point
by a dark grey sphere. The “Show-Me” trials started at the end of the for each cardinal landscape in the correct position relative to the first
northwest arm, and the grey sphere was move to the end of the west one placed (maximum − 4). Then they were asked to place the four
arm. Shifting the configuration of the maze on the “Show-Me” trial walls, and given 1 point if all were correct relative to each other and
enables a distinction between three navigational strategies based on another point if at least one wall was correct relative to a cardinal
participants’ chosen arm. If participants went straight to the southwest landscape. Then they were asked to place a representation of the floor
arm, they were scored as navigating by an allocentric “place” strategy (with 6 arms) with a blue circle representing the cue sphere of the
as this is the original location of the platform as indicated by the arena second set, and were given 1 point if the blue sphere was oriented
4
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
towards the east landscape picture. Finally, they were asked to place on This thus provided the groundwork to use the Show-Me trials to probe
the “floor” a small red circle representing the platform and were given 1 strategy prevalence, rate of learning and strategy acquisition.
point if it was correct relative to the first cardinal landscape placed. The It was also important to examine participants previous navigational
maximum score on this task was 8. experience so as to assess the impact of this on measures of Hex Maze
At the conclusion of the test session, participants were asked a series performance. The composite gaming experience score revealed that
of 10 questions about the strategies they used in the Hex maze and a most participants had limited previous gaming experience (M = 5.08,
series of 7 questions about their experience with videogames, maps and SD = 3.22) and was used to group participants into low/no gaming
compasses and other experiences that might have affected how they experience (score < 3, N = 21), moderate gaming experience (score
navigated in the Hex maze. Three separate measures of gaming ex- 3–6, N = 21) and high gaming experience (score > 6 N = 22). The
perience (experience with videogames, 3D games, and 2D games), each controller experience score also revealed a relatively low level of pre-
measured on a Likert scale from 0 (never) to 5 (daily) were combined vious experience (M = 1.97, SD = 1.36) and was used to split partici-
into a composite “gaming experience” score (out of 15). Controller pants into groups of little/no controller experience (score < 2 N = 34)
experience was also measured on Likert scales from 0 (no experience) to and high control experience (score > 1 N = 30). Interestingly, less
5 (daily). Finally, participants were asked whether they had previous than half of the participants (N = 30) reported any map/compass ex-
map/compass experience and if so, how frequently (1 = occasionally to perience. As such, participants were grouped by whether or not they
5 = daily). had previous map/compass experience, rather than by frequency.
First, it was important to validate the Show-Me trials as a means of A substantial proportion of participants acquired and used each of
evaluating navigation strategy. As part of the post-maze questionnaire, the three possible navigational strategies (allocentric, egocentric-cue,
participants were asked a direct questions regarding strategy con- and egocentric-response) regardless of sex (Fig. 3). One-sample t-tests
sistency across Find-It and Show-Me trials. 82.3% whose responses against zero indicated that all three strategies were used a significant
could be evaluated (25/30 males, 26/32 females) used the same proportion of the time (t(63) = 6.60, p < 0.0001). One sample t-tests
strategy on Show-Me trials as they had on the preceding Find-It trial. against chance indicated that the allocentric (place) strategy was
5
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
chosen more often than chance (1 in 5) (t(63) = 4.43, p < 0.0001, p > 0.05, ηp2 = 0.035) and no interaction (F(3.56) = 1.177,
d = 0.554), and when combined, the two egocentric strategies (cue and p > 0.05). Additionally, a 3(game experience) x 2(control experience)
response) were chosen more often than chance (2 in 5) (t(63) = 2.30, x 3(map/compass experience) ANOVA on average latencies revealed no
p =0.024, d = 0.287). Pairwise comparisons indicated that the allo- significant effect of previous game experience (F(2.58) = 0.13,
centric strategy was chosen on more trials that the egocentric-cue p > 0.05, ƞp2 = 0.005), previous controller experience (F(1.58)
strategy (t(30) = 2.4, p = 0.02, d = 0.3). No other pairwise compar- = 1.55, p > 0.05, d = 0.028) or map/compass experience (F(2.58)
isons of strategies were significant (p > 0.05) and there were no sex = 0.07, p > 0.05, ƞp2 = 0.001) and no significant interactions (all
differences within strategies (pairwise t-tests p > 0.05). p > 0.05). While men were, on average, faster regardless of strategy,
planned post-hoc t-tests showed that they were significantly so only
when their strategy preference was for the egocentric-cue strategy (t(9)
3.2. Rate of learning
= 2.36, p = 0.04, d = −1.2) (Fig. 4b).
Participants navigated the maze quickly. All latencies on the visible
platform trials were low, reaching an asymptote of between 10 and
3.3. Strategy use
15 seconds (M = 10s, SD = 3.91s). After the first Find-It trial, 80% of
the latencies were below 15 seconds (Trials 2–10 M = 14s, SD = 6s)
The ability of the Hex Maze to track the use of the three strategies
and 80% of both men and women reached this level of performance by
over trials made it possible to examine four novel aspects of strategy use
the fifth trial of the first set, and the second trial of the second set.
and sex differences within these areas: 1) distribution of preferred
Strategy preference did not appear to be a factor in how quickly
strategies, 2) acquisition of preferred strategies, 3) stability and origin
participants learned the maze nor how quickly they could navigate to
of strategy preference, and 4) strategy preference vs strategy compe-
the hidden platform on Find-It trials (Fig. 4a). A 3 (strategy) x 2 (sex)
tence. These results are discussed below.
ANOVA on average latencies (trials 2–10) revealed no effect of strategy
(F(3.56) = 0.076, p > 0.05, ηp2 = 0.004) or sex (F(1.56) = 2.021,
6
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
7
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
8
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
capable of identifying the relative positions of the windows and plat- presented once, or presented only a small number of times over large
form position relative to the landscape outside the windows (Fig. 7b). delays. The Hex Maze is novel in that ‘Show-Me’ trials are presented
Two 2 × 4 ANOVAs, revealed no significant effects of sex or after every learning trial throughout testing. This not only improves
strategy preference on cognitive map development (all p > 0.05). The reliability and applicability to studies requiring averaging over a large
only significant effect present was that male participants who preferred number of trials (i.e., neuroimaging studies), but also allows for
allocentric or cue-based egocentric navigation were better than the strategy usage to be tracked throughout testing. van Gerven et al. [42]
female participants with the same strategy preference (post-hoc pair- presented participants with alternating standard trials (similar to Hex
wise t-test; allocentric t(29) = 2.637, p = 0.01 d = 0.98; egocentric- Maze Find-It trials) and inter-trial probes (similar to Hex Maze Show-
cue t(9) = 2.657, p = 0.03, d = 1.85). Me trials), but were limited to distinguishing between egocentric and
allocentric navigation. The Hex Maze is the first virtual maze to include
repeated ‘probe’ trials that distinguish between three navigational
4. Discussion
strategies.
The current results also demonstrate that the Hex Maze is easy to
The current paper presents a novel method for assessing three dis-
learn and does not bias strategy choice. All strategies were used on a
tinct spatial navigation strategies in a virtual environment. The Hex
considerable proportion of trials. For both men and women, the
Maze is configured as a 6 arm RAM, and is run procedurally as a MWM
strategy used most often was the allocentric strategy, with the ego-
with one platform location, and varied start locations. The paradigm
centric-cue and egocentric-response strategies used for a similar
involves participants repeatedly alternating between locating an in-
number of trials. When grouped together, the egocentric strategies were
visible platform (learning, or ‘Find-It’ trials), and returning to the pre-
used on an equivalent number of trials as the allocentric strategy. This
vious location of that platform (probe, or ‘Show-Me’ trials).
indicates that the maze is not heavily biased towards or against a
Importantly, the Hex Maze not only distinguishes between three navi-
particular strategy (or sex). Additionally, trial latencies indicate that
gational strategies (allocentric, egocentric-cue, and egocentric-re-
the task was easy to learn, with male and female participants reaching
sponse) but also allows further examination of aspects of strategy usage,
asymptote (15 s) after an average of 2 and 4 trials respectively.
such as acquisition, stability, and competence.
Latencies were not significantly affected by strategy choice or the in-
The Hex Maze is easily accessible and easy to use. The maze was
teraction between sex and strategy choice. This is contrary to the notion
constructed in the Unreal Development Kit virtual environment (www.
that males are more efficient allocentric navigators
unrealengine.com) and has been designed so that different key strokes
[31,32,37,48,49,40,41], and suggests that females who prefer to navi-
call different trials, meaning that trial order and several room features
gate allocentrically are equally as efficient.
can be modified without programming. Additionally, the maze creates a
The design of the Hex Maze also allows for an exploration of a
log file with each run, which not only records user locations over time,
number of novel aspects of strategy usage including strategy preference,
but also records the outcome of each trial (e.g., when the platform is
acquisition, stability and competence. Firstly, strategy ‘preference’ was
reached on Find-It trials and which arm is chosen on Show-Me trials).
defined as the strategy used for the most trials by a participant. The
The results of the current study offer validation for the use of the
majority of participants (male and female) demonstrated a preference
Hex Maze as a tool for assessing various features of spatial navigation.
for one strategy, with an approximate 50/50 split between those who
Firstly, the results demonstrate the novel use of “Show-Me” trials to
preferred the allocentric strategy, and those who preferred one of the
assess the strategies participants employ. Show-Me trials require par-
egocentric strategies. Interestingly, previous studies using the RAM
ticipants to move to the position where the invisible platform was
have generally found that a much larger proportion of their participants
hidden on the preceding Find-It trial. Due to a shift in the configuration
use an egocentric-response strategy than an allocentric strategy
of the maze, their response will differ depending on the navigation
[33,10,12]. As previous studies employing alternative maze designs
strategy they used. In the current study, 82.3% of participants reported
have identified similarly high proportions of allocentric navigators
using the same strategy on Find-It and Show-Me trials, indicating that
[56,57,40,41], this suggests a potential bias in the standard RAM. The
Show-Me trials are a credible index of strategy employment.
Hex Maze does not appear to be affected by bias for any one strategy.
The use of a probe trial to determine strategy usage is common
There was a trend for more males to prefer the allocentric strategy,
practice in ‘dual-solution’ rodent mazes [51–54]. Additionally, previous
and more females to prefer the egocentric strategies (primarily the ego-
virtual MWM studies conducted by our group have used similar ‘drop-
response strategy) which is consistent with the premise that males and
the-seed’ trials to distinguish between egocentric and allocentric navi-
females have a tendency to use allocentric and egocentric strategies
gation [55,40,41]. However, such probe trials are usually only
9
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
10
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
configurations. Additionally, participants performed better on the preference and competence are not analogous, and that multiple stra-
forced choice probe for their preferred strategy than either of the other tegies can be learnt concurrently and with proficiency. It is envisaged
strategies. If the responses of participants were due to errors, their that this paradigm will be widely applicable and will be beneficial in
performance would have been poor on the corresponding probe. As furthering our understanding on the cognitive and neural basis of
such, the argument of errors in strategy uses seems non-parsimonious. human navigation, and the factors that may influence it.
The primary contribution of the Hex Maze to future research is that
it is able to index and characterize three different spatial navigation Funding
strategies. Previous research has only focused on the distinction be-
tween allocentric and egocentric navigation, however these results MJS is supported by Brain Research New Zealand PHD scholarship.
demonstrate cue-based egocentric navigation and response-based ego-
centric navigation can (and should) be differentiated. As these strate- Conflicts of interest
gies are understood to have different neural correlates, the ability of the
Hex Maze to distinguish between the strategies will be important when There are no conflicts of interest to report.
studying factors that may influence these brain areas, such as stress
[58,54,42], genetics [33,69,70] brain injury [71–73], ageing [74,75] Acknowledgements
and neurodegenerative disease [76,28,77,78]. It will also be of interest
to future studies to explore the relationship between the measures of The authors would like to thank Maureen Krulak for data collection.
strategy usage afforded by the Hex Maze and navigational ability, using Additionally, the authors would like to thank Thomas Ferguson and
a measure such as the Santa Barbara Sense of Direction Scale [79]. As Dustin Van Gerven for assisting in the development of the virtual en-
many aspects of the task can be modified with simple key strokes (e.g., vironment.
trial order, configuration, available cues), the Hex Maze is well suited
for use across different populations. Additionally, the repetitive nature References
of the ‘probe’ (Show-Me) trials renders this task favorable for neuroi-
maging and neurophysiology studies that require averaging over a large [1] N. Burgess, Spatial cognition and the brain, Ann. N. Y. Acad. Sci. 1124 (2008)
77–97.
number of trials.
[2] T. Madl, K. Chen, D. Montaldi, R. Trappl, Computational cognitive models of spatial
It is important to note a couple of limitations of the current study. memory in navigation space: a review, Neural Netw. 65 (2015) 18–43.
Firstly, the sample population was comprised of English first-language [3] N. Burgess, Spatial memory: how egocentric and allocentric combine, Trends Cogn.
Sci. 10 (2006) 551–557.
university students in Canada, which limits the generalizability of the [4] T. Meilinger, G. Vosgerau, Putting egocentric and allocentric into perspective,
results. Secondly, it could be argued that the sample size was small for a International Conference on Spatial Cognition. Springer (2010) 207–221.
methodological study, and that the lack of significant sex effects could [5] R. Klatzky, Allocentric and egocentric spatial representations: definitions, distinc-
tions, and interconnections, in: C. Freksa, C. Habel, K. Wender (Eds.), Spatial
be due to insufficient power. However, previous research from our lab Cognition, Lecture Notes in Computer Science, Springer, Berlin/Heidelberg, 1998,
has established a sample size of 15–20 per group is sufficient for de- pp. 1–17.
[6] J. O’Keefe, An allocentric spatial model for the hippocampal cognitive map,
tecting sex differences [80,40,41]. Additionally, with the potential for Hippocampus 1 (1991) 230–235, http://dx.doi.org/10.1002/hipo.450010303.
clinical applications, it was important to demonstrate the sensitivity of [7] J. O’Keefe, L. Nadel, The Hippocampus as a Cognitive Map, Oxford University Press,
the Hex Maze with a sample size close to what would be seen in such Oxford, 1978.
[8] J. O’Keefe, Place units in the hippocampus of the freely moving rat, Exp. Neurol. 51
studies. This sensitivity is demonstrated by the included effect sizes for (1976) 78–109.
significant effects. Nevertheless, it is important to acknowledge that, [9] E. Antonova, D. Parslow, M. Brammer, A. Simmons, S. Williams, G.R. Dawson,
based on the current results, we cannot claim an absence of sex dif- R. Morris, Scopolamine disrupts hippocampal activity during allocentric spatial
memory in humans: an fMRI study using a virtual reality analogue of the Morris
ference (i.e., one cannot claim support for the null hypothesis) and it Water Maze, J. Psychopharmacol. (Oxf.) 25 (2011) 1256–1265.
will be important for future research to substantiate the current study [10] L. Dahmani, V.D. Bohbot, Dissociable contributions of the prefrontal cortex to
hippocampus- and caudate nucleus-dependent virtual navigation strategies,
with a larger cohort. Third, we did not include a measure of naviga- Neurobiol. Learn. Mem. Memory Decis Making 117 (2015) 42–50, http://dx.doi.
tional ability in the current study. Previous studies using our virtual org/10.1016/j.nlm.2014.07.002.
MWM have found no relationship between maze performance and [11] C.F. Doeller, J.A. King, N. Burgess, Parallel striatal and hippocampal systems for
landmarks and boundaries in spatial memory, PNAS 105 (2008) 5915–5920.
questionnaires of navigational ability [50,81]. As the relationship be- [12] G. Iaria, M. Petrides, A. Dagher, B. Pike, V.D. Bohbot, Cognitive strategies depen-
tween navigational ability and strategy usage remains elusive [82], it dent on the hippocampus and caudate nucleus in human navigation: variability and
change with practice, J. Neurosci. 23 (2003) 5945–5952.
was not a pertinent to the main goals of the current study to include
[13] K. Iglói, C.F. Doeller, A. Berthoz, L. Rondi-Reig, N. Burgess, Lateralized human
such a measure. However, as there may be some influence of naviga- hippocampal activity predicts navigation based on sequence or place memory, Proc.
tional ability on strategy use, this is an important avenue for future Natl. Acad. Sci. 107 (2010) 14466–14471.
[14] K. Jordan, J. Schadow, T. Wuestenberg, H.J. Heinze, L. Jancke, Different cortical
studies. Finally, the Hex Maze did not appear to encourage develop- activations for subjects using allocentric or egocentric strategies in a virtual navi-
ment of a cognitive map of the environment outside the arena. This gation task, Neuroreport 15 (2004) 135–140.
could suggest that the allocentric cues provided by the Hex Maze are [15] O. Baumann, J.B. Mattingley, Medial parietal cortex encodes perceived heading
direction in humans, J. Neurosci. 30 (2010) 12897–12901, http://dx.doi.org/10.
not actually being used within an allocentric cognitive map. However, 1523/JNEUROSCI.3077-10.2010.
this seems unlikely as the allocentric environment is similar to that used [16] K. Gramman, J.E. Sharkawy, H. Deubel, Eye-movements during navigation in a
virtual tunnel, Int. J. Neurosci. 119 (2009) 1755–1778.
in previous MWM studies where participants have demonstrated good [17] C.-T. Lin, T.-C. Chiu, K. Gramann, EEG correlates of spatial orientation in the human
cognitive map development [38,50,40,41]. It is unclear why there is retrosplenial complex, Neuroimage 120 (2015) 123–132.
such a discrepancy between the Hex Maze and the MWM in this regard, [18] S.D. Vann, J.P. Aggleton, E.A. Maguire, What does the retrosplenial cortex do? Nat.
Rev. Neurosci. 10 (2009) 792.
and it will be important for future studies to examine this further. [19] S.D. Vann, J.P. Aggleton, Extensive cytotoxic lesions of the rat retrosplenial cortex
Here, we present a novel paradigm for exploring human spatial reveal consistent deficits on tasks that tax allocentric spatial memory, Behav.
navigation; the Hex Maze. In the current study we were able to char- Neurosci. 116 (2002) 85.
[20] M. Boccia, F. Nemmi, C. Guariglia, Neuropsychol. Rev. 24 (2014) 236–251, http://
acterize not only strategy preference and rates of learning, but also the dx.doi.org/10.1007/s11065-014-9247-8.
stability and acquisition of strategy use. Together our results show that [21] R.J. Blanch, D. Brennan, B. Condon, C. Santosh, D. Hadley, Are there gender-spe-
cific neural substrates of route learning from different perspectives? Cereb. Cortex
across all three strategies, acquisition was quick and relatively stable, 14 (2004) 1207.
and while there was some indication of sex differences in strategy [22] M. Plank, H. Müller, J. Onton, S. Makeig, K. Gramann, Human eeg correlates of
preference, there appear to be no evidence for significant sex differ- spatial navigation within egocentric and allocentric reference frames, Spat. Cogn.
VII (2010) 191–206.
ences in any of our novel measures of acquisition, stability or multi [23] N. Etchamendy, K. Konishi, G.B. Pike, A. Marighetto, V.D. Bohbot, Evidence for a
strategy competence. Additionally, we have demonstrated that strategy virtual human analog of a rodent relational memory task: a study of aging and fMRI
11
M.J. Spriggs et al. Behavioural Brain Research xxx (xxxx) xxx–xxx
12