isolating observer-based reference directions in human spatial memory: head, body, and the...

Available online at www.sciencedirect.com

www.elsevier.com/locate/COGNIT

Cognition 106 (2008) 157–183

Isolating observer-based reference directionsin human spatial memory: Head, body,

and the self-to-array axis q

David Waller *, Yvonne Lippa, Adam Richardson

Department of Psychology, Miami University, Oxford, OH 45056, USA

Received 31 May 2006; revised 5 January 2007; accepted 12 January 2007

Abstract

Several lines of research have suggested the importance of egocentric reference systems fordetermining how the spatial properties of one’s environment are mentally organized. Yet rel-atively little is known about the bases for egocentric reference systems in human spatial mem-ory. In three experiments, we examine the relative importance of observer-based referencedirections in human memory by controlling the orientation of head and body during acquisi-tion. Experiment 1 suggests that spatial memory is organized by a head-aligned referencedirection; however, Experiment 2 shows that a body-aligned reference direction can be moreinfluential than a head-aligned direction when the axis defined by the relative positions of theobserver and the learned environment (the ‘‘self-to-array’’ axis) is properly controlled. A thirdexperiment shows that the self-to-array axis is distinct from – and can dominate – retina, head,and body-based egocentric reference systems.� 2007 Elsevier B.V. All rights reserved.

Keywords: Spatial memory; Reference frames; Reference system; Egocentric representations

0010-0277/$ - see front matter � 2007 Elsevier B.V. All rights reserved.

doi:10.1016/j.cognition.2007.01.002

q This research was supported by NIMH Grant MH068245 to David Waller. We thank Parker Huston,Ambika Gautam, and Shaina Rood for assistance with conducting the experiments and NathanGreenauer, Eric Hodgson, and two anonymous reviewers for helpful comments on previous drafts of thisarticle. We also thank Frances Wang for informal comments that inspired us to conduct Experiment 2.

* Corresponding author. Fax: +1 513 529 2420.E-mail address: [email protected] (D. Waller).

mailto:[email protected]

158 D. Waller et al. / Cognition 106 (2008) 157–183

1. Introduction

Perceiving and mentally representing the locations of objects and places inone’s environment is a fundamental prerequisite to most of human behavior.Because locations can be specified only in and through a reference system (orreference frame), the scientific investigation of the reference systems used inperception and memory has become an increasingly important area of contempo-rary research in spatial cognition (Behrmann, 2000; Carlson-Radvansky & Irwin,1993; Hinton & Parsons, 1981; Levinson, 1996; Pani & Dupree, 1994; Rock,1973; Shelton & McNamara, 2001b). One aspect of cognitive reference systemsthat has received a great deal of research attention is the reference direction thatis used to code spatial information. (Another aspect, the origin of the referencesystem, has received considerably less research attention.) When applied to thememorial coding of spatial information, a reference direction can be thought ofas conceptual ‘‘north’’ (Shelton & McNamara, 2001b), a privileged or preferreddirection that is used to organize and define angular relationships. Broadlyspeaking, reference directions can be defined with respect to: (a) environmentalfeatures such as salient landmarks or the North Pole, yielding an environmental

reference direction (McNamara, Rump, & Werner, 2003; Werner & Schmidt,1999) (b) the inherent structure of a stimulus or stimulus array, yielding an intrin-

sic reference direction (Mou & McNamara, 2002), or (c) the orientation of partof an observer’s body, yielding an egocentric reference direction (Shelton &McNamara, 1997).

One critical issue in the field of spatial cognition is to determine which refer-ence direction is typically used in a given situation. In the last decade, severallines of investigation have converged to suggest the importance of egocentric ref-erence directions as a particularly influential means of mentally organizing thespatial properties of one’s environment. For example, investigators have shownthat when asked to imagine a familiar environment, patients with unilateral sen-sory neglect fail to report objects that would appear on the neglected side of theirbodies’ midline (Bisiach & Luzzatti, 1978), suggesting an egocentric coding ofthese scenes (Easton & Sholl, 1995; Werner & Schmidt, 1999). Work in normalpopulations has shown that pointing to targets from an imagined viewpoint istypically influenced by the targets’ location in body-centered coordinates (Sholl,1987; Franklin & Tversky, 1990). Similarly, patterns of errors in pointing to tar-gets after disorientation suggest that navigation in one’s immediate environmentis largely governed by transient egocentric mental representations of space,instead of enduring allocentric representations (Waller & Hodgson, 2006; Wang,1999; Wang & Spelke, 2000, 2002).

Some of the most compelling evidence for the importance of egocentric refer-ence directions in memory for spatial layouts has come from the work of McNa-mara and his colleagues who have shown repeatedly that people are relativelyaccurate at imagining or recognizing orientations within a layout of objects whenthose orientations were experienced during learning. Likewise, people are relative-ly error-prone when imagining or recognizing orientations that were not experi-

D. Waller et al. / Cognition 106 (2008) 157–183 159

enced during learning (Diwadkar & McNamara, 1997; Mou & McNamara, 2002;Roskos-Ewoldsen, McNamara, Shelton, & Carr, 1998; Shelton & McNamara,1997, 2001b; Valiquette, McNamara, & Smith, 2003). These effects can be inter-preted as indicating the importance of egocentric experience to spatial memory.Although variables such as cultural norms (Levinson, 1996), the structure ofthe stimuli (Mou & McNamara, 2002), the structure of the environment (McNa-mara et al., 2003; Shelton & McNamara, 2001b; Werner, 2001), and the instruc-tions given during encoding (Attneave & Reid, 1968; Fery & Magnac, 2000) canaffect the way in which spatial layouts are mentally represented, the literature isconsistent with the idea that there is a strong tendency to store information aboutspatial layout by means of an egocentric reference system, and that it requiresparticularly salient information from other sources (often induced by experimentalmanipulations) to overcome this tendency. Indeed, even when other reference sys-tems are used to organize spatial information in memory, these reference systemsare likely to be chosen based on egocentric cues (McNamara, 2003; Valiquetteet al., 2003).

Given the predominance of egocentric reference directions in spatial memory, wethought that it was important to understand more precisely the nature of this codingsystem. We approached this issue by considering three possible bases for an egocen-tric reference direction, each of which uses the inherent structure of a body part todefine conceptual ‘‘north.’’ First, an egocentric reference direction may be based sim-ply on the orientation of the retina, organizing spatial information in terms of theretinal images – the views – that were experienced during learning. Second, it isknown that relatively early processes in the central nervous system integrate retino-centric information with information about the eyes’ orientation relative to the head(see for example Colby, 1998). Thus, information about the positions of objects rel-ative to the facing direction of the head is available to memory, and provides a sec-ond possible basis for an egocentric reference direction. Finally, a somewhat moresophisticated egocentric coding system can integrate head-based coordinates withproprioceptive and efferent information about the orientation of the head relativeto the body (i.e., the trunk). The result is a reference direction based on the orienta-tion of the body (Kopinska & Harris, 2003).

Although there is solid evidence for people’s tendency to use egocentric refer-ence systems to code information about their environment, the degree to whichretina-, head-, or body-based reference directions dominate memorial coding ofspatial layout is largely unknown. In general, past experiments have either con-founded these reference directions, by asking participants to learn with headand body aligned (Diwadkar & McNamara, 1997; Presson, DeLange, & Hazel-rigg, 1989; Roskos-Ewoldsen et al., 1998; Shelton & McNamara, 2001b; Sholl& Bartels, 2002; Sholl & Nolin, 1997; Waller, Montello, Richardson, & Hegarty,2002), or have not controlled them, by asking participants to maintain a constantbody orientation, while freely altering their gaze direction (Hintzman, O’Dell, &Arndt, 1981; Presson & Montello, 1994; Rieser, 1989; Valiquette et al., 2003;Experiment 1; Werner, 2001). In perhaps the only prior research that specificallyaddressed the precise basis for egocentric reference frames in memory for spatial


layout, Sholl (1999) asked people to point to unseen targets based on an imaginedlocation in a campus environment that was learned either from a map or fromdirect experience. Sholl used an established pattern of response asymmetries(see Franklin, Henkel, & Zangas, 1995; Franklin & Tversky, 1990) to inferwhether information about the layout was coded with respect to either retina-based or body-based reference direction. She concluded that map-acquired spatialinformation uses primarily a retina-based reference system (because it was notassociated with the front–back response asymmetry), whereas spatial informationacquired from self-motion and navigation is primarily coded in terms of a body-based reference system (because it did show such an asymmetry). Sholl did notdraw conclusions about the relative influences of head- and body-based referencedirections.

In the present experiments, we examine the relative importance of head- andbody-based reference directions in memory by controlling the orientation of thesereference systems during acquisition. Consider, for example, the situation illustrated

Glue

Stapler

Eraser

Mouse

Calculator

Batteries

Scissors

Tape

Ruler

Pencil

Disk

Glue

Stapler

Eraser

Mouse

Calculator

Batteries

Scissors

Tape

Ruler

Pencil

Disk

Fig. 1. Depiction of the stimulus layout and learning conditions in Experiment 1. Participants in the StayMisaligned group learned by standing in front of the array (depicted with open circles) with their headfacing the array. The direction of their head (dashed arrow) and body (dotted line) were misaligned by 72�.Participants in the Walk Misaligned group (not shown) also learned with head and body misaligned, but inaddition, walked in the direction that their body faced, starting from the solid arrow and repeating thecircuit indicated by the dotted line. A third group (Stay Aligned) is not shown.


in Fig. 1. An observer stands with her body facing Northeast (to the upper right inthe figure) and views an array of objects by turning her head 72� to her left. Theevidence we have reviewed suggests that, in many situations, this observer willremember the layout in terms of an egocentric reference system; however, it isunknown whether the layout is more likely to be remembered in terms of theobserver’s viewing direction or in terms of her body’s facing direction. In Experi-ment 1, we used this situation to examine the relative influence of head- andbody-based reference directions in memory for spatial layout. The results suggestedto us a novel possible basis for a reference direction – one defined by the relation-ship between an observer and the spatial array that he or she observes. In Exper-iment 2, we again examined the relative effects of head- and body-based referenceframes while controlling for this additional reference direction. Experiment 3showed that this additional reference direction is distinct from other possibleobserver-based reference systems.

We assessed the degree to which a given reference direction was used to rep-resent spatial location in memory by examining performance on two tasks. First,participants were asked to imagine different orientations within a rememberedarray, and to judge directions relative to these imagined headings. Several inves-tigators have used such a task to infer that the imagined orientations from whichrelations are most accurately and quickly retrieved represent those that are ‘‘pre-ferred’’ in memory (Roskos-Ewoldsen et al., 1998; Shelton & McNamara, 1997;Sholl & Bartels, 2002; Waller et al., 2002). Preferred orientations are typicallyassumed to be those that are directly encoded, and thus presumably serve toorganize other (nonpreferred) spatial relationships (Mou, McNamara, Valiquette,& Rump, 2004; Shelton & McNamara, 2001b; Valiquette et al., 2003). Second, atthe end of each experiment, we asked participants to construct a map of the lay-out that they had learned. We then used the orientation of their constructed mapto assess the orientation (and thus the reference direction) that was preferred inparticipants’ memory.

2. Experiment 1

In Experiment 1, we contrasted performance on tests of spatial memory amongthree groups of participants that differed in the disparity between the orientationof their head and the orientation (and/or movement) of their body during acquisi-tion. One group (Stay Aligned) viewed an array of objects from a stationary posi-tion, with their head and body aligned, facing the array. A second group (StayMisaligned) learned the array while maintaining the posture depicted in Fig. 1 – withtheir head facing the array, and their body misaligned by 72�. We then examined thedegree to which participants in this group showed facilitation at test with imaginingorientations that had been aligned with their body. Facilitation at body-aligned ori-entations was examined relative to both the head-aligned orientations in the StayMisaligned group and to the performance of the Stay Aligned group at the sameimagined orientations.


The results of several pilot experiments had suggested that, in general, partici-pants in the Stay Misaligned condition would tend to code spatial information withrespect to a head-aligned reference direction (Waller, Richardson, & Lippa, 2004). Inanticipation of a relatively weak tendency to remember the array with respect to abody-based reference direction, we included a third condition in Experiment 1 thatattempted to strengthen the degree to which spatial memory would be encoded ina body-based reference frame. In this condition (Walk Misaligned), participants alsoviewed an array with a 72� disparity between their head and body orientation.Unlike the Stay groups, however, participants in the Walk Misaligned group walkedduring learning in the direction that their body faced. Thus, the orientation of theparticipant’s body coincided with the direction that their body moved during acqui-sition (i.e., their course).

Very little prior research has examined the degree to which the direction ofone’s movement through an environment can help to organize spatial memory;however, several related investigations have shown that, in general, bodily move-ment during acquisition can influence spatial memory. Yamamoto and Shelton(2005) showed that the spatial memories formed solely on the basis of one’smovement through an environment (i.e., without vision) share many of the samequalities as visually acquired spatial information. Richardson, Montello, andHegarty (1999) suggested that movement through an environment can enhancethe flexibility of one’s memory of it. Similarly, Valiquette et al. (2003) haveshown that moving through an environment while experiencing multiple body ori-entations can improve the flexibility of spatial memory, relative to a situation inwhich the body maintains a single orientation at learning (see also Presson,DeLange, & Hazelrigg, 1987). Although many of these experiments have con-trolled the orientation of the participant’s body during learning, they generallydo not enable claims about the role of the direction of the participant’s move-ment. We used the Walk Misaligned condition to address the issue of whetherthe direction of one’s movement during acquisition can affect the strength of abody-based reference direction used to encode spatial memory.

To enable conclusions about egocentric reference frames in these experiments, itwas important to minimize the degree to which nonegocentric reference frames couldbe used by participants to encode the spatial layout. We attempted to minimize thedegree to which an environmental reference frame could be used by placing the arrayin the center of a round room. Similarly, we minimized the effectiveness of intrinsicreference frames by using a circular array of objects. The symmetry of the environ-ment and the spatial array maximized the chances that participants would use anegocentric frame of reference when committing the layout to memory.

2.1. Method

2.1.1. Participants

Forty-nine University students participated in the experiment in return for creditin their introductory psychology course. One of these participants voluntarily with-drew from the experiment after completing fewer than half of the testing trials. The


remaining sample consisted of 48 participants, composed of 24 men and 24 women.Three gender-balanced groups, each with 16 participants, were randomly assigned tothe three experimental conditions (Stay Aligned, Stay Misaligned, and WalkMisaligned).

2.1.2. Materials

Participants learned an array of 11 common objects arranged on a 1.07 m diam-eter round table (see Fig. 1). Ten of the objects were located near the perimeter of thetable, spaced at equal intervals of 36�. The 11th object was placed at the center of thetable. This table was located at the center of a 3.00 m diameter round enclosure inwhich the participant stood (or walked) to learn the array. This enclosure was con-structed from black opaque fabric. Its ceiling was also black, with a round hole in thecenter through which a round light shone. When standing in the enclosure, it was notpossible to see the walls of the larger room in which it resided.

After learning the layout, participants were tested in another lab room, arounda corner and down a hall from the room in which they learned the array. Testquestions were administered in a V8 head-mounted display (HMD) from VirtualResearch which provided 640 · 480 resolution and a 60� diagonal field of view.These questions were composed of two stimuli: an orienting stimulus (e.g., ‘‘Atthe glue facing the tape’’) and a target stimulus (e.g., ‘‘Point to the pencil’’). Par-ticipants controlled the onset of each stimulus by pressing a hand-held triggerbutton. Participants responded to the test questions by turning their head to indi-cate a direction and pressing the button to record their response. The partici-pants’ head direction (yaw) was continually tracked by means of an IntersenseInertiaCube2 inertial tracker that was mounted to the participant’s HMD andwhich provided online (72 Hz) directional data within 1� RMS of error. Datafrom the tracker were also used to update the viewpoint depicted in the HMDas participants responded to the target stimulus (providing optic flow of a tex-tured ground plane and sky during participants’ responses.) Randomization andpresentation of the stimuli, as well as the collection of direction estimationsand latencies, were controlled through a scripting facility in the Python program-ming language, supplemented with a utility module written by Andrew Beall spe-cifically for virtual environment applications.

2.1.3. Procedures

Participants were met at the laboratory in which they were to be tested and werebriefed about the nature of the experiment. They were then randomly assigned toone of three learning conditions that differed in their body posture and/or movementduring learning. The experimenter told participants in the Stay Aligned conditionthat they would learn the array by standing in front of it, with their head and bodyfacing forward. Participants in the Stay Misaligned condition were instructed thatthey would face the center of the array during learning, but that they would standwith their body facing 72� clockwise from their head direction. Participants in theWalk Misaligned condition were instructed that they would learn the array whilewalking and while continually maintaining a 72� disparity between their course


and heading (see Fig. 1). These participants practiced walking in this manner beforebeing exposed to the array.

The experimenter then escorted the participant to the learning room. Before enter-ing, the participant was asked to close his or her eyes, and was then led into the cir-cular enclosure in the room. Correct body alignment during learning was achievedby means of a large wooden template, placed on the floor at the learning location,that depicted two footprints facing in the direction to which the participant’s bodyshould be oriented. Before viewing the object array, participants were instructedto look down and use the template to assume the correct body posture. Correct headalignment was ensured by giving participants a visual target toward which they wereinstructed to look initially and to use to fix their head direction. For participants inthe two Stay conditions, this visual target was the object in the center of the array.For participants in the Walk Misaligned group, the visual target was a small coloredmarker placed on the wall of the enclosure. Participants in the two Stay groupslearned the array from the same location. Participants in the Walk Misaligned groupstarted their walk from a different location (see Fig. 1), and passed through the Staygroups’ learning location halfway through their course. Once the Walk Misalignedparticipants had walked past the array, they turned away from it, and walked backto their starting position, while still maintaining their learning posture. They repeat-ed walking this course with the correct posture until they indicated that they wereready to be tested. Before leaving the room for testing, all participants were askedto point to each object in a random order with their eyes closed. If the experimenterdetected a pointing error, the participant was asked to continue learning for anotherminute.

For testing, participants made 44 judgments of relative direction (JRDs), eachof which consisted of an imagined heading (e.g., ‘‘At the glue, facing the disk’’)and a target (e.g., ‘‘Point to the stapler’’). The first four judgments were practiceitems. The remaining 40 judgments were randomized separately for each partici-pant, and were composed of eight replications at each of five imagined headings.These imagined headings were arbitrarily defined with 0� as the direction in whichparticipants’ heads faced during learning (e.g., from the center of the table [theglue] to its top [the stapler]). Positive and negative headings corresponded toclockwise and counterclockwise rotations of the viewpoint, respectively. The fiveimagined headings were 0�, ±72�, and ±144�. For example, an item asking par-ticipants to imagine being at the glue, facing the mouse represented an imaginedheading of �144�. Note that, by this convention, the body orientation of partic-ipants in the two misaligned conditions was always at 72�. Half of the test itemsasked participants to imagine being at the center of the table (‘‘at the glue stick’’),with the other half asking them to imagine being on its perimeter. At each levelof imagined heading, the direction and mean magnitude of the correct responseswere counterbalanced. Across all items, the number of times each object was usedas a target was equal.

At the beginning of each testing trial, participants stood with head and bodyaligned, and then pressed a button to trigger the appearance in the HMD of an ori-enting stimulus supplying an imagined heading. After viewing the orienting stimulus,


they pressed the button again to display the target stimulus. At this point, a texturedground plane and sky became visible in the HMD and provided optic flow to partic-ipants as they indicated the direction to the target. Participants then turned theirhead (and pressed the button) to indicate the target direction, relative to the imag-ined heading. The participants’ head orientation before and after presentation ofthe target stimulus were recorded and subsequently used to compute signed estima-tion errors. Mean errors were calculated from the absolute values of these errors. Inaddition, response latencies on each item were recorded as the time between theonset of the orienting stimulus and the time at which participants had turned theirhead more than 10� to respond. Thus, the longer movement times that are requiredfor greater rotations were not included in our latency measure.

After testing, participants were seated at a table and asked to construct a mapof the environment. These maps were constructed by placing small cardboardpieces representing the objects in the learned array on a sheet of grid paper onwhich a large circle was drawn. Participants were instructed to consider this circleas the table on which the learned objects had been placed, and to construct theirmap accordingly.

2.2. Results

For all participants, data from the four practice trials were not analyzed. Anadditional 21 trials from 18 participants (representing 1.09% of the data) wereremoved because the presentation time of the orienting stimulus was too short(less than 0.25 s) to be read by the participant. These trials were invariably theresult of the participant accidentally pressing the trigger button twice at the

Table 1Absolute pointing error and latency for imagined headings in Experiment 1

Condition Imagined heading Absolute error(degrees)

Latency (s)

Mean SD Mean SD

Stay Aligned �144 40.24 24.19 11.12 4.10�72 37.10 15.39 10.18 4.03

0 25.98 11.38 7.29 1.5772 35.48 12.03 10.08 2.12

144 44.22 21.75 11.06 4.07

Stay Misaligned �144 37.24 13.83 14.34 5.74�72 31.61 10.49 11.64 4.46

0 24.81 10.65 8.59 3.2872 35.66 14.07 12.61 4.96

144 44.63 25.56 13.34 6.30

Walk Misaligned �144 38.29 20.19 10.88 2.72�72 41.60 12.86 9.64 2.14

0 34.09 15.11 9.28 2.4772 44.14 17.96 10.20 2.62

144 45.75 22.89 11.15 3.29

Imagined Heading

-144 -72 0 72 144

Mea

n D

iffic

ulty

-0.6

-0.4

-0.2

0.0

0.2

0.4

Stay AlignedStay MisalignedWalk Misaligned

Imagined Heading

-144 -72 0 72 144

Mea

n D

iffic

ulty

-0.6

-0.4

-0.2

0.0

0.2

0.4

Stay AlignedStay MisalignedWalk Misaligned

Fig. 2. Mean difficulty (defined in the text) of imagining various headings in the learned array among thethree groups in Experiment 1. Error bars represent standard errors.


beginning of the trial. Preliminary analyses confirmed that neither gender norimagined location (i.e., at the center versus the perimeter of the array) exertedan effect or interacted with the independent variables of interest, so data were col-lapsed across these variables.

Table 1 presents mean absolute pointing error and mean latency for eachgroup across the five imagined headings. In general, there was no evidence of aspeed accuracy tradeoff, and the effects of experimental condition and imaginedheading were similar for both pointing accuracy and latency. To facilitate analysisand depiction of the results, we thus combined errors and latencies into onedependent variable called Difficulty, following procedures adapted from Walleret al. (2002).1 More specifically, we defined Difficulty on each trial as the averagez-score for absolute pointing error and response time, with both variables beingnormalized individually (based on each participant’s distribution of errors ortimes).

Mean difficulty for each group across the five imagined headings is illustratedin Fig. 2. For each of the three experimental groups, the easiest headings to imag-ine were those that were head aligned during learning. Difficulty increased steadi-ly as imagined headings deviated from this head-aligned perspective. The effectsof imagined heading and learning condition on item difficulty were examined ina 5 (imagined heading: �144�, �72�, 0�, 72�, 144�) · 3 (group: Stay Aligned, StayMisaligned, Walk Misaligned) mixed factor ANOVA with imagined heading as awithin-subjects factor. This analysis revealed a significant main effect of imaginedheading (F(4, 180) = 35.76, p < .01). This effect was largely accounted for by itsquadratic trend, which accounted for 68.43% of the variance associated withthe effect of imagined heading. The particularly strong quadratic component of

1 Separate analyses of errors and latencies yielded consistent results and identical conclusions as thosebased on the combined measure of difficulty.

Table 2Number of participants for each condition of Experiment 1 constructing their map in a given orientation

Condition Map orientation (object at top)

�36 (scissors) 0 (stapler) 36 (eraser) 108 (ruler) 180 (disk)

Stay Aligned 16Stay Misaligned 1 14 1Walk Misaligned 4 8 3 1


this effect indicates that difficulty increased as imagined headings differed fromhead aligned.

In general, the effect of imagined heading was not as pronounced for partici-pants in the Walk Misaligned group as it was for the other participants. Fig. 2shows that group differences were most notable at the 0� imagined heading.The omnibus interaction between imagined heading and condition was significant(F(8, 180) = 2.57, p = .016), and was further analyzed with tests of simple maineffects at the two imagined headings (0� and 72�) that corresponded to partici-pants’ head and body alignment during learning. For head-aligned items (imag-ined heading of zero), difficulty was significantly different among the threegroups (F(2, 45) = 5.69, p = .006). Orthogonal contrasts at imagined headings ofzero revealed that this effect was primarily associated with differences betweenthe Walk group and the two Stay groups (t(45) = 6.45, p < .01), which did notdiffer from each other (t(45) = 0.12, p = .906). For body-aligned items (imaginedheadings of 72�), difficulty was significantly higher relative to head-aligned items(F(1, 45) = 67.12, p < .01) but not significantly different among the three groups(F(2, 45) = 0.51, p = .95). The contrasts testing for specific differences betweengroups were also not significant for body-aligned imagined headings (botht’s < 1). We also analyzed these effects with an interaction contrast that examinedthe change in Difficulty between 0� and 72� imagined headings for the Walk Mis-aligned group, as compared with the two Stay groups. This contrast was signif-icant (t(45) = 2.53, p = .015).

Finally, we examined the maps that participants constructed at the end of theexperiment. For this analysis, we coded the orientation during learning of the objectthat participants placed at the top of their map. For example, referring to Fig. 1, if aparticipant placed the tape at the top of her map, this was coded as a 72� (body-aligned) orientation. Table 2 presents the cross-classification of participants’ exper-imental group and the alignment of their map. The percentage of participants whoconstructed a map that was aligned with their head orientation during learningwas 100%, 87.5%, and 50% for the Stay Aligned, Stay Misaligned, and Walk Misa-ligned conditions, respectively. Among the eight participants in the Walk Misalignedgroup who did not construct a head-aligned map, map alignment was varied. Nota-bly, four of these participants constructed their map at a �36� alignment (corre-sponding to the scissors being placed at the top), and none of them constructed amap with a body aligned (72�) orientation.


2.3. Discussion

The results of Experiment 1 suggest that, when one’s head and body are misa-ligned during learning, memory of a spatial layout is coded with respect to ahead-aligned – not a body-aligned – reference direction. Evidence for the dominantuse of a head-aligned reference direction comes from the observation that, for allgroups, head-aligned orientations were significantly easier to imagine than body-aligned orientations. Indeed, the difficulty of imagining orientations within thelearned array increased steadily as these imagined orientations deviated from beinghead aligned. Table 1 illustrates that these effects were present in both errors andlatencies. In the two groups for which there was a disparity between the orientationof the head and body during learning, there was no indication that imagining body-aligned orientations were facilitated at all. Additionally, an overwhelming majorityof participants in the Stay Aligned and Stay Misaligned groups (more than 93%)constructed a map of the layout in an orientation that had been aligned with theirhead during learning.

Interestingly, however, the facilitation in imagining head-aligned directions wassignificantly attenuated in the Walk Misaligned group. It is interesting to note that,while the difficulty of imagining body-aligned orientations was comparable across allthree experimental groups, participants in the Walk Misaligned group had signifi-cantly more difficulty than those in the other groups with imagining head-alignedorientations. This increased difficulty was the result of differences in both accuracyand latency (see Table 1). Moreover, the orientation of the maps constructed by par-ticipants in the Walk Misaligned group was not as frequently head-aligned, andshowed greater variability than that of the other groups. The act of walking in a fixeddirection thus appears to decrease the strength to which memories are encoded bymeans of a head-aligned reference system; however, it does not clearly enhance theuse of a body-based reference system.

3. Experiment 2

In Experiment 1, we observed that performance based on imagining head-alignedorientations was worse for participants who had walked during learning than forparticipants in the stationary groups. One possible reason for this difference is that,for participants in the Walk Misaligned group, the spatial relationships betweenthemselves and the object array continually changed during learning. In particular,the direction between the participant and the center of the object array varied as par-ticipants in the Walk Misaligned group moved. On the other hand, for participantsin the stationary groups, the direction between themselves and the object array wasstable. Moreover, for stationary participants, this direction coincided with their headalignment during learning. If people use the direction between themselves and a loca-tion in the environment to help define a dominant reference axis in memory, then it ispossible that the performance advantage of the stay conditions over the walking con-dition for head-aligned imagined headings in Experiment 1 was at least partially due


to an additional reference direction defined by the relative positions of the partici-pant and the center of the object array. It is thus possible that the conclusions fromExperiment 1 about the dominance of a head-based reference system may need to betempered, in order to account for this additional reference axis.

We called the direction between the participant and the center of the object arraythe self-to-array axis and designed Experiment 2 to determine the relative influenceof head- and body-based reference systems when this axis does not coincide withone’s head direction. Participants were asked to learn an array while standing sta-tionary in front of it (see Fig. 3). The relative positions of the participant and thearray defined an self-to-array axis that we arbitrarily labeled as 0�. Participantslearned the array with their heads facing to the left (or right) by 36�, and their trunkfacing to the right (or left) by 36�. Thus, as in Experiment 1, participants viewed anarray with a 72� disparity between the orientation of their head and body. In Exper-iment 2, however, this disparity was bisected by the self-to-array axis. After learning,we tested participants on imagined headings that coincided with their head orienta-tion, their body orientation, and the orientation of the self-to-array axis. If the per-formance advantage in Experiment 1 for imagining head-aligned orientations wasmainly driven by the self-to-array axis, we should find a performance advantagefor items requiring participants to imagine being aligned with the self-to-array axis,

Stapler

Eraser

Mouse

Calculator

Batteries

Scissors

Tape

Ruler

Pencil

Disk

Stapler

Eraser

Mouse

Calculator

Batteries

Scissors

Tape

Ruler

Pencil

Disk

Fig. 3. Depiction of the stimulus layout and learning condition in Experiment 2. The facing directions ofparticipants’ body (dotted arrow) and head (solid arrow) were each separated by 36� from the self-to-arrayaxis (dashed arrow).


relative to those that require imagining being aligned with the head. By the sametoken, if the performance advantage in the previous experiments for head-aligned tri-als was mainly driven by the fact that these were aligned with participants’ heads, weshould find a significant performance advantage for imagining head-aligned orienta-tions, relative to orientations that are aligned with the self-to-array axis. Finally,Experiment 2 enabled us to assess the relative influence of a head- and body-basedreference system when the former does not coincide with the self-to-array axis.

3.1. Method

3.1.1. Participants

Forty-one University students participated in the experiment in return for creditin their introductory psychology course. One of these participants voluntarily with-drew from the experiment, reporting symptoms of illness during testing. Her datawere not analyzed. The remaining sample consisted of 40 participants, including20 men and 20 women.

3.1.2. Materials and procedure

Materials for Experiment 2 were identical to those of Experiment 1 with theexception that the target array consisted of only 10 objects that were placed aroundthe perimeter of the table; there was no central object.

Procedurally, Experiment 2 was similar to the Stay Misaligned condition ofExperiment 1 with the following exceptions. In the learning room, approximatelyhalf (n = 21) of the participants were instructed to use the footprint template to aligntheir body at +36� relative to the self-to-array axis. The remaining participants wereasked to align their body at �36�. They were then directed to turn their head toeither �36� or +36� before removing their blindfold. Thus, all participants learnedthe array with a 72� difference between their head and body orientation, with theself-to-array axis bisecting this difference.

For testing, participants responded to 64 questions, composed of eight replica-tions of eight imagined headings: �144�, �72�, �36�, 0�, 36�, 72�, 144�, and 180�.At each level of imagined heading, the direction and mean magnitude of the correctresponse were counterbalanced. As in Experiment 1, test items were questions of theform ‘‘Imagine you are at X, facing Y. Point to Z.’’ Across all items for each partic-ipant, the number of times each object was used as an imagined location (X), an ori-enting object (Y), and a target (Z) was approximately equal.

3.2. Results

For all participants, data from the four practice trials were not analyzed. Addi-tionally, 16 trials from 10 participants (representing 0.63% of the data) were removedbecause they involved a very short (less than 0.25 s) presentation time of the orient-ing stimulus and likely indicated trials on which the participant accidentally pressedthe trigger button twice at the beginning of the trial. For all analyses we report, thefacing direction of the body (left versus right) was entered as a between-subjects fac-

Table 3Absolute pointing error and latency for imagined headings in Experiment 2

Imagined heading Absolute error (degrees) Latency (s)

Mean SD Mean SD

�144 39.45 19.91 11.94 4.45�72 36.82 18.06 11.20 3.83�36a 32.98 13.59 10.01 3.11

0b 32.72 14.36 10.23 3.4236c 32.55 13.25 10.92 3.5672 35.54 17.75 10.89 3.57

144 39.68 19.96 11.96 4.18180 37.78 21.72 12.15 4.22

a Corresponds to body alignment at learning.b Corresponds to self-to-array axis at learning.c Corresponds to head alignment at learning.


tor, but did not exhibit a main effect or interaction. As a result, the analyses wereport collapse over this factor. Collapsing over this factor involved recoding theimagined headings for half of the participants, such that, for all participants, animagined heading of +36� corresponded to a head-aligned item, an imagined head-ing of �36� corresponded to a body-aligned item, and an imagined heading of 0� cor-responded to a self-to-array-aligned item.

Table 3 presents mean absolute pointing error and mean latency across the eightimagined headings. In general, there was no evidence of a speed accuracy tradeoff,and, as with Experiment 1, analyses were conducted on the composite variable calledDifficulty.2

Fig. 4 illustrates that, in general, body-aligned and self-to-array-aligned imaginedheadings were answered most easily, with difficulty increasing steadily as imaginedheadings became more disparate from these perspectives. The effect of imaginedheading on Difficulty was examined in a one way repeated measures ANOVA andwas significant (F(7, 273) = 11.76, p < .01). As in Experiment 1, a majority(62.80%) of the main effect of imagined heading was accounted for by its quadratictrend.

We further examined the relative difficulty of head-aligned, body-aligned, andself-to-array-aligned imagined headings by means of three planned pairwise con-trasts. Body-aligned imagined headings were significantly easier than head-aligned(t(39) = 2.47, p = .018). Similarly, self-to-array-aligned headings were significantlyeasier than head aligned (t(39) = 2.12, p = .04). Performance on body-aligned itemswas not significantly different from that on self-to-array-aligned items.

Finally, we examined which object participants placed at the top of the maps thatthey constructed. All but two participants placed the stapler at the top of their map,indicating a map that was aligned with the self-to-array axis. One participant alignedher map with the body axis at �36�, and another aligned his map to �72�.

2 As in Experiment 1, separate analyses of errors and latencies yielded consistent results and identicalconclusions as those based on the combined measure of difficulty.

Imagined Heading

-144 -72 -36 0 36 72 144 180

Mea

n D

iffic

ulty

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

B H

Imagined Heading

-144 -72 -36 0 36 72 144 180

Mea

n D

iffic

ulty

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

B H

Fig. 4. Mean difficulty (defined in the text) of imaging various headings in the learned array in Experiment2. Imagined headings that were aligned with the body and head during learning are indicated with a B andH, respectively. Error bars represent standard errors.


3.3. Discussion

In Experiment 2, participants were more efficient at imagining orientations thathad been aligned with the self-to-array axis or with their body than they were withimagining orientations that had been head aligned. Table 3 shows that differences inlatencies were the strongest indicator of this effect, whereas error levels were fairlyconsistent at approximately 30� across head-, body-, and self-to-array-alignments.Given the magnitude of participants’ errors at their most accurate imagined head-ings, it is possible that error measures did not provide a sufficiently sensitive measureof differences among imagined headings that differed by only 36�. The results of themap-placement task provided additional support for the idea that participants’memory may have been largely organized in terms of the self-to-array axis; nearlyall participants’ maps were aligned with this axis.

It is particularly notable that the significant ease with imagining body-aligned ori-entations (relative to head aligned) in Experiment 2 was the reverse of the significanteffect found in Experiment 1. In conjunction with the findings from Experiment 1,the present results suggest that a head-based reference system is not necessarily adominant means of coding spatial layout, unless the axes of a head-based systemcoincide with an additional organizational structure, such as the self-to-array axis.More specifically, the present results suggest that the strength of the head-alignedreference system found in Experiment 1 was probably not driven by head alignmentper se, but rather by its coincidence with the self-to-array axis.

4. Experiment 3

The conclusion that bringing the self-to-array axis under experimental controlwas necessary to uncover the relative strengths of head- and body-based reference


frames in spatial memory called for a clear understanding of the nature of theself-to-array axis. As we have conceptualized it, the self-to-array axis representsa heretofore unrecognized basis for a reference direction in spatial memory.Clearly, the reference direction defined by the self-to-array axis is neither intrinsicnor environmental, because it is not invariant over changes in the observer’s loca-tion. Additionally, unlike a purely egocentric reference direction (e.g., a body-based reference direction) the self-to-array axis does not use the intrinsic structureof an observer’s body part (e.g., the facing direction of the trunk) to define a ref-erence direction. Although the self-to-array axis is observer-based, its direction isdefined by the relationship between the observer and his or her environment –not by the observer or the environment exclusively. This relationship was notphysically instantiated by any particular object, but rather had to be inferredand abstracted by participants from their interpretation of the global structureof the environment. For participants to have apprehended this somewhat abstractrelationship between themselves and their environment and to have used it toorganize spatial memory may represent a noteworthy feat of high-level spatialcognition.

However, there is an alternative account of the self-to-array axis in theseexperiments that is consistent with the idea that participants did not interpretabstract relationships in their environment, but rather encoded only the spatialinformation that was immediately available to them in the learning situation. Thisis because in Experiments 1 and 2, the self-to-array axis may have been largelycoincident with the alignment of the participant’s retina. For instance, in Exper-iment 2, it seems likely that in order to apprehend the array of objects, partici-pants would have generally needed to align their retinas with the self-to-arrayaxis. To the extent that retinal orientation in these experiments was coincidentwith the self-to-array axis, our conceptualization of and conclusions about theself-to-array axis may apply only to retina-based egocentric reference frames.Before making conclusions about the higher-level cognitive processes required

A

BC

A

B

C

A

BC

A

B

C

Fig. 5. Schematic plan view of the stimulus layout and learning condition in Experiment 3. Participantslooked into a video camera while learning the positions of three objects on a table next to them. Therelative directions among objects were aligned with either the self-to-array axis (e.g., A and B) or theegocentric axes (e.g., A and C). Participants learned both an acute (left) and obtuse (right) arrangement ofobjects.


to apprehend and encode spatial information in terms of a self-to-array axis, wewanted to determine the degree to which relatively low-level retinal encodingcould account for the findings of Experiments 1 and 2.

We conducted Experiment 3 to disambiguate the self-to-array axis from theretinal axis. As depicted in Fig. 5, participants learned an array with their head,body, and retina aligned with a viewing axis that differed by 40� from the self-to-array axis. This set-up thus required participants to learn an array that was pre-sented in their peripheral vision. The 40� disparity between the retina and self-to-array axis was necessarily smaller than the disparity we used in Experiments 1and 2 because of the limited spatial resolution and acuity of participants’ periph-eral vision. The relatively small misalignment between reference frames, in con-junction with the limited spatial resolution of peripheral vision resulted in ourusing a relatively simple stimulus array that was composed of only three objects.Because such a small number of objects cannot generate a suitable number ofinterobject angles, we did not ask participants at test to make judgments of rel-ative direction, but rather focused our measures exclusively on the map construc-tion task. Because Experiments 1 and 2 had confirmed a close correspondencebetween the conclusions enabled by participants’ judgments of relative directionand those enabled by the map construction task, we regarded the latter task asan adequate means of assessing which reference direction is preferred in spatialmemory (see also Shelton & McNamara, 2001a).

4.1. Method

4.1.1. Participants

Twenty-three University students participated in the experiment in return forcredit in their introductory psychology course. Data from seven of these participantswere subsequently eliminated because the participants were deemed by two indepen-dent raters as not having complied with the learning instructions. These eliminationprocedures are described in greater detail below. The remaining sample consisted of16 participants, composed of 8 men and 8 women.

4.1.2. Materials

Participants learned two arrays of objects while seated on a round stool in thesame round room that was used in Experiments 1 and 2. The two arrays wereeach composed of three common objects (array A: glue stick, stapler, and eraser;array B: glue stick, tennis ball, and water bottle) that were arranged on a 1.07 mdiameter round table (see Fig. 5). In each array, the glue stick was located at thecenter of the table and served, in conjunction with the participant’s seating posi-tion, to define an self-to-array axis. The participants were asked to fix their gazeon a video camera that was placed approximately at eyeheight, 0.74 m away fromthe participant, and separated by 40� from the self-to-array axis (see Fig. 5). Thedirection between the participant and the video camera thus defined the partici-pants’ retinal axis.


Two objects in each array were located near the table’s perimeter – one alignedwith the self-to-array axis and the other aligned with the retinal axis. We calledthese objects the self-to-array object and the retinal object, respectively. For eachparticipant, one of the arrays comprised an acute angle, with the glue stick at thevertex and the self-to-array object on the far side of the table relative to the par-ticipant (see Fig. 5, left). The second array comprised an obtuse angle, with theglue stick at the vertex and the self-to-array object on the near side of the tablerelative to the participant. The orders of the arrays’ angles (acute or obtuse) andthe order of their composition (array A or array B) were counterbalanced acrossparticipants.

Testing consisted of the same map construction task that was used in previous exper-iments with the exception that the glue stick was already marked on the center of themap. This task was conducted in the same lab room as was used in Experiments 1and 2. Note that, to the extent that the self-to-array axis is preferred in spatial memory,we would expect the self-to-array object to be placed at the top of participants’ maps forthe acute layout, and at the bottom of their maps for the obtuse layout. Likewise, to theextent that a retina-based axis is preferred, we would expect the retinal object to beplaced at the top of the participants’ maps for both layouts.

4.1.3. Procedures

Each participant was met individually at the laboratory and was told that theinvestigators were examining the spatial precision of peripheral vision. As such,the participant would be required to learn two layouts of objects without directlylooking at them. Participants were told that, during learning, they must look directlyinto a video camera and that if, subsequently, the experimenters determined thattheir gaze had deviated from the camera, their data would be eliminated. Partici-pants were then shown each of the objects that would compose the arrays, and toldwhich objects would be grouped together in each array. Pilot experiments had shownthat this familiarization period significantly enhanced participants’ ability to recog-nize the objects during learning.

The experimenter then escorted the participant to the learning room. As with pre-vious experiments, the participant was asked to close his or her eyes before enteringthe room, and was then led by the experimenter into the circular enclosure in theroom. Participants were seated with head and body aligned toward the video camera.Participants kept their eyes closed while the experimenter started the video record-ing, constructed the first array, and then stood behind the participant. The partici-pant was then instructed to open his or her eyes and to maintain their gazedirectly into the camera. The experimenter asked the participant to identify eachobject and to commit the layout of objects to memory. After indicating that theyhad adequately learned the array, participants were asked to close their eyes whilethe experimenter set-up the second array. Learning for the second array then pro-ceeded as it had for the first.

For testing, participants returned to the other lab room, and constructed maps ofeach of the two arrays. Finally, the participants were debriefed about the true natureof the experiment.


4.2. Results

To ensure that all participants had complied with the instructions to look awayfrom the object array, we asked two independent raters to view the participants’ vid-eo tapes and to identify participants that they deemed to have shifted their gazedirection toward the object array. The raters agreed that six of the participants inap-propriately shifted their direction of gaze. One rater additionally identified a seventhsuch participant. All seven participants were subsequently replaced with new partic-ipants. The video tapes of these seven additional participants were reviewed by theraters who deemed these new participants to have adhered to the instructions.

We then analyzed the orientation of the participants’ maps. Instead of categoriz-ing each map according to the object placed at the top, we pursued a more fine-grained analysis by measuring the degrees from vertical that participants had placedthe retinal object and the self-to-array object on their maps. For these analyses, wedefined ‘‘vertical’’ as the direction between the bottom and the top of each partici-pant’s map and measured the signed deviation from vertical for each object and eachmap. These deviations were then analyzed with circular statistics (see, for example,Batschelet, 1981) and are depicted in Fig. 6. For the acute arrays, participants onaverage placed the self-to-array object 3.19� ± 10.73 (for a 95% confidence interval)from vertical (i.e., the top of their map), and the retinal object 43.89� ± 10.62 (for a95% confidence interval) from vertical. For the obtuse arrays, participants on aver-age placed the self-to-array object 2.30� ± 14.75 (for a 95% confidence interval) from

m = 45.42o

r = 0.92

m = 177.70o

r = 0.87

m = 3.19o

r = 0.93m = 43.89o

r = 0.93

Self-to-array objectRetinal object

m = 45.42o

r = 0.92

m = 177.70o

r = 0.87

m = 3.19o

r = 0.93m = 43.89o

r = 0.93

Self-to-array objectRetinal objectSelf-to-array objectRetinal object

Fig. 6. Circular histograms of the directions of the objects in the acute (left) and obtuse (right) arraysconstructed by participants in Experiment 3. In general, the object aligned with the self-to-array axisduring learning (represented for each participant by the solid circles) was more closely aligned with thevertical axis of participants’ maps than the object that was aligned with the retinal axis (represented by theopen circles.) Also depicted are the direction (m) and length (r) of the mean vector (see Batschelet, 1981) aswell as its 95% confidence interval.


vertical (i.e., the bottom of their map), and the retinal object 45.42� ± 11.23 (for a95% confidence interval) from vertical. Of the 32 maps constructed by the partici-pants, 27 of them (approximately 84%) depicted the self-to-array object as beingmore aligned with vertical than the retina-aligned object. The five maps for whichthe self-to-array object was less aligned with vertical than the retina object were com-posed by only three of the participants.

4.3. Discussion

Experiment 3 dissociated the self-to-array axis from the retinal axis duringlearning and examined which of these two axes was subsequently preferred inspatial memory. The results clearly supported the dominance of the self-to-arrayaxis over the retinal axis. This dominance is especially impressive when one notesthat both the head and the body were aligned with the retina in this experiment.Thus, despite the redundant reference directions offered by the head, body, andretina, participants still tended to conceptualize the array as being aligned withthe self-to-array axis. These findings further temper the conclusions from Exper-iment 1 about the strength of a head-based reference direction. Indeed, the resultsof Experiments 2 and 3 strongly suggest that the effect of head alignment foundin Experiment 1 was primarily driven by the head’s coincidence with the self-to-array axis in that experiment.

Despite its relative strength in memory, the self-to-array axis may have been some-what influenced by egocentric reference frames in the present experiment. For example,in Experiment 3, the average deviation from vertical of the self-to-array object wasbiased a few degrees in the direction of the retinal object. Moreover, for only five(31%) of the participants was the average deviation of the self-to-array object less than10�. Thus, the self-to-array object was rarely depicted exactly at the top of participants’maps. Instead, its orientation was typically pulled in the direction of the retina object.This finding indicates that, although the self-to-array axis likely dominated the organi-zation of spatial memory, this memory may have also been partly affected by otheravailable egocentric reference frames. An alternative (but not mutually exclusive) pos-sibility is that the placement of the self-to-array object in Experiment 3 was biased by aspatial prototype such as the center of a quadrant (see, for example, Huttenlocher,Hedges, & Duncan, 1991).

Finally, it should be noted that, unlike Experiments 1 and 2, the present con-clusions are based solely on participants’ performance on the map-placementtask. Given the physical constraints of dissociating a retinal axis from otherpotential reference frames, this limitation was necessary for the current experi-ment. However, it is possible that the self-to-array axis is especially sensitive tothe map task (perhaps, for example, constructing a map emphasizes self-to-arrayrelations more than JRDs do) and that the present results thus exaggerate itsinfluence. For example, in Experiment 2, participants’ performance on the map-placement task indicated a more extreme effect of the self-to-array axis thandid their performance on judgments of relative direction. Nonetheless, thesetwo tasks have led to largely consistent conclusions both in Experiments 1 and


2, as well as in the existing literature on spatial memory (Shelton & McNamara,2001a). Additionally, in our opinion, both tasks have high face validity. Becausethe influence of the self-to-array axis in this experiment was large and unambig-uous, it seems likely that it indicates a real effect on spatial memory, even if ourmeasuring technique possibly amplified it.

5. General discussion

These experiments explored the relative strength of head- and body-based referencedirections in spatial memory when the cues from both intrinsic and environmental ref-erence systems were minimal. Experiment 1 provided initial evidence in such situationsfor the dominance of a head-aligned reference direction over a body-aligned referencedirection. However, this evidence was qualified by the results of Experiments 2 and 3,which showed that much (perhaps all) of the dominance of a head-aligned referencedirection inExperiment1derived from its superposition withanother referenceaxis thatis defined by the direction between the person and a location in the environment – adirection that we term the ‘‘self-to-array axis.’’ Experiment 3 further showed that thisself-to-array axis is distinct from egocentric axes and that it too can exert a significantinfluence on the organization of spatial memory. These findings have implications forunderstanding the roles of head- and body-based reference systems in spatial memory,as well as illuminating the importance of the spatial relationships between an observerand his or her environment. We address each of these topics in the following sections.

5.1. Head-based or body-based coding?

The totality of our data is most consistent with the idea that, when nonegocentricreference frames are properly controlled, memory for spatial location is more likelyto be coded with respect to a body-based reference direction than with respect to ahead-based one. In many cases, coding the locations of objects in one’s immediate envi-ronment in terms of a body-based reference system is probably quite functional, inas-much as body-based coding facilitates the types of interactions that one can expect tohave with those objects. For example, it is common to interact with an object on a near-by table by moving toward the table and grasping the object. Because the body is nec-essarily involved with actions such as walking and grasping, these actions must bemediated by a body-based reference system. Spatial memory may best facilitate theseactions by coding information directly in terms of a body-based system. More gener-ally, it is possible that the reference system(s) used to code spatial information dependon how one is subsequently able (or intends) to use that information. This hypothesiscan provide motivation for future experiments that examine whether the selection of adominant reference frame in memory depends on the degree to which one can interactwith the remembered environment. For example, this hypothesis would predict thatlearning about an array of objects from a pictorial depiction of the array (see, for exam-ple, Chua & Chun, 2003; Waller, 2006) would lead to a less dominant role of a body-based reference system than what we observed in our experiments, because one’s body


would presumably not be used to interact with pictorially depicted objects. It is inter-esting to note that Sholl’s (1999) conclusion that maps are coded with respect to a ret-ina-based reference system while navigable spaces are coded with respect to a body-based system is also fully consistent with this hypothesis.

In the introduction, we noted that several previous studies examining the refer-ence directions used in spatial memory have either not controlled (Hintzmanet al., 1981; Presson & Montello, 1994; Rieser, 1989; Valiquette et al., 2003; Exper-iment 1; Werner, 2001) or have confounded (Diwadkar & McNamara, 1997; Pressonet al., 1989; Roskos-Ewoldsen et al., 1998; Shelton & McNamara, 2001b; Sholl &Nolin, 1997; Sholl & Bartels, 2002; Waller et al., 2002) the relative orientations ofthe head and body during learning. For example, Waller et al. (2002) asked partic-ipants to sit in a chair and remember several arrays of objects that were laid-out infront of them. The size of these arrays, in conjunction with their proximity to theparticipant, meant that participants needed to turn their heads approximately 30�to the left and right in order to fixate on each object. Waller et al. subsequently con-trasted the ease with which participants imagined orientations in the arrays that were‘‘aligned’’ during learning with those that were not. Because head alignment was notstrictly fixed at learning, these investigators apparently implicitly defined alignmentwith respect to the body’s orientation at learning (see also Wang, in press; Werner,2001). The results of the present experiments provide some support for this assump-tion that a body-based reference direction is used to organize spatial memory in sucha situation. However, these experiments also make clear that people opportunistical-ly use multiple cues for reference frames, and that investigators must be mindful ofadditional possible reference axes that can increase the possibility that a head-basedreference direction is used instead of a body-based direction.

5.2. The role and implications of the self-to-array axis

In order to examine the relative effects of head- and body-based reference direc-tions, we attempted to create a learning situation in which nonegocentric referenceframes, such as those introduced by walls of a room or a geometric arrangementof objects, were minimized. Despite these efforts, it became clear in these experimentsthat the mere presence of the observer created an asymmetry to the environment thatwas apparently leveraged as a basis for a preferred orientation in spatial memory.More specifically, our conjecture that people used an axis defined by the relativelocations of themselves and the object array received strong support in Experiment3. A noteworthy part of these experiments thus involves the discovery and descrip-tion of a basis for spatial organization in memory that involves neither egocentricnor environmentally based reference frames exclusively, but rather involves a refer-ence system based on the relationship between an observer and the environment. Inall three of the present experiments, this self-to-array axis likely accounted for muchof the organization of participants’ spatial memory. This axis was particularly influ-ential when it coincided with the orientation of the head, as it did in Experiment 1.The coincidence of the head- and self-to-array axes in Experiment 1 resulted in amore prominently represented reference direction than the body axis.


The conclusion that participants’ relative ease at imagining head-aligned orienta-tions in Experiment 1 was primarily due to the additional support of the self-to-arrayaxis illustrates the opportunistic nature of the selection of reference frames in spatialmemory. McNamara (2003) describes the selection of a reference frame in spatialmemory as a process of choosing among various egocentric and environmental cues.From this description, it is easy to speculate that as the number of cues supporting agiven reference frame increases, the probability of selecting that reference frame toorganize spatial memory also increases. Thus, when two potential reference axes(e.g., a head-axis and an self-to-array axis) coincide, they may be able to providea strong dominant axis by mutually reinforcing each other (see also Millar & Al-At-tar, 2004; Yoon & Shelton, 2005). The added benefit of using such mutually reinforc-ing reference axes can be further illustrated in the present experiments by examiningthe relative magnitudes of participants’ pointing errors and latencies between Exper-iments 1 and 2 (see Tables 1 and 3). For example, the easiest headings to imagine inExperiment 2 (body-aligned) were significantly more difficult than the easiest head-ings to imagine in Experiment 1 (head-aligned). This difference may be well-ex-plained as a facilitation for head alignment in Experiment 1 that arose as theresult of coding space in terms of multiple coincident reference frames that mutuallysupported each other.

The impetus for this research involved discovering the unique contributions ofhead- and body-based reference frames in spatial memory. However, we found thatwe could not address this issue without introducing the concept of the self-to-arrayaxis. The finding that this strong and dominant reference direction is distinct from(and preferred to) head-, body-, and retina-aligned axes is an important finding fromboth a methodological and a theoretical point of view. Methodologically, we notethat it can be extremely difficult to construct a learning situation that fully eliminatesthe self-to-array relationship as a potential reference direction for spatial memory.Future work examining the reference frames involved in spatial memory must there-fore account for this possible reference axis.

Our finding about the dominance of the self-to-array axis helps both to confirm andto augment existing theories of spatial representation. Recent work by McNamara andcolleagues (McNamara, 2003; Mou & McNamara, 2002; Mou et al., 2004; Valiquetteet al., 2003) has illustrated that, rather than coding a mere ‘‘snapshot’’ of one’s egocen-tric experience, memory for spatial arrays may depend on additional (likely effortful)mental processing at encoding that involves assessing the global structure of the envi-ronment and choosing a reference system that maximizes coherence and efficiency.Because people’s determination and use of the self-to-array axis requires an assessmentof the overall structure of one’s environment, our conclusions are highly consistentwith this account. On the other hand, according to the model proposed by McNamaraand colleagues, people generally select a reference axis that is based on the intrinsicgeometry of the layout. In the present experiments, participants learned layouts thatgenerally had no salient intrinsic axes. It is thus probably more parsimonious to con-sider the reference axes selected in the current experiments as being based not on theintrinsic properties of the layout, but rather on other directions, such as those definedby egocentric and self-to-array relationships. Thus, although the concept of the self-to-


array axis is in keeping with the idea that effortful processes can be engaged whenencoding spatial information, it also represents an alternative to the dominant roleof an intrinsic axis offered by McNamara’s theory.

Finally, it is worthwhile to consider how and whether our findings about the dom-inant role of the self-to-array axis may generalize to everyday spatial cognition in nat-ural environments. The use of spatial memory in everyday situations can be morecomplicated than the situations examined in the present experiments for at least tworeasons. First, it is common for object locations in a naturalistic environment to belearned only as a result of movement through the environment. Real-world arrays thatrequire movement to apprehend clearly complicate the definition of the observer’sposition on which the concept of the self-to-array axis critically relies. The relativelypoor performance at imagining head-aligned orientations among participants in theWalk condition of Experiment 1 (compared to participants in the two stationary con-ditions) suggests that, in situations when the user moves during learning, the self-to-ar-ray axis may be less dominantly represented in memory than when the user learns froma stationary position. This does not mean, however, that the use of a self-to-array axis isnecessarily irrelevant for moving observers. It is quite conceivable, for example, that amoving observer samples a momentary self-to-array axis, either before or during move-ment, and relies on this single axis to organize spatial memory. The limited range ofmap orientations for participants in the Walk group of Experiment 1 (see Table 2) isconsistent with the hypothesis that different observers selected different self-to-arrayaxes within the range of possible axes that were available during their walk.

A second complication with generalizing the present results involves the factthat arrays of objects in naturalistic environments are often unbounded and ill-de-fined. In the present experiments, the objects that participants were asked to learnwere clearly and unambiguously set-off from a relatively sparse environment;however, in daily life spatial arrays are rarely so well-defined. Most theories ofspatial representation (even theories that differ radically in their description ofhow spatial relations are mentally represented) are consistent with the notion thatreal world places are apprehended by keeping track of a relatively small numberof functionally important objects in one’s environment (O’Keefe & Nadel, 1978;Wang & Brockmole, 2003; Wang & Spelke, 2000). It is natural to speculate thatthese small sets of objects constitute the spatial arrays that are remembered andused to represent places in memory. Many of these naturally formed spatialarrays likely do not have sufficient inherent structure to define a salient referenceaxis, and thus it seems likely to us that a remembered observation point could behelpful in anchoring a reference direction in spatial memory. In the end, thestrength and ease of use of the self-to-array axis in the present experiments makesit difficult to believe that it would only be used in laboratory experiments.

References

Attneave, F., & Reid, K. W. (1968). Voluntary control of frame of reference and slope equivalence underhead rotation. Journal of Experimental Psychology, 78, 153–159.


Batschelet, E. (1981). Circular statistics in biology. London: Academic Press.Behrmann, M. (2000). Spatial reference frames and hemispatial neglect. In M. Gazzaniga (Ed.), The

cognitive neurosciences (2nd ed., pp. 651–666). Cambridge: MIT Press.Bisiach, E., & Luzzatti, C. (1978). Unilateral neglect of representational space. Cortex, 14, 129–133.Carlson-Radvansky, L. A., & Irwin, D. E. (1993). Frames of reference in vision and language: where is

above? Cognition, 46, 223–244.Colby, C. (1998). Action-oriented spatial reference frames in cortex. Neuron, 20, 15–24.Chua, K., & Chun, M. M. (2003). Implicit scene learning is viewpoint dependent. Perception &

Psychophysics, 65, 72–80.Diwadkar, V. A., & McNamara, T. P. (1997). Viewpoint dependence in scene recognition. Psychological

Science, 8, 302–307.Easton, R. D., & Sholl, M. J. (1995). Object-array structure, frames of reference, and retrieval of spatial

knowledge. Journal of Experimental Psychology: Learning, Memory, & Cognition, 21, 483–500.Fery, Y. A., & Magnac, R. (2000). Attenuation of alignment effect with exocentric encoding of location.

Perception, 29, 789–799.Franklin, N., Henkel, L. A., & Zangas, T. (1995). Parsing surrounding space into regions. Memory &

Cognition, 23, 397–407.Franklin, N., & Tversky, B. (1990). Searching imagined environments. Journal of Experimental

Psychology: General, 119, 63–76.Hinton, G. E., & Parsons, L. M. (1981). Frames of reference and mental imagery. In A. D. Baddeley & J.

Long (Eds.). Attention and Performance (Vol. 9, pp. 261–277). Hillsdale, NJ: Erlbaum.Hintzman, D. L., O’Dell, C. S., & Arndt, D. R. (1981). Orientation in cognitive maps. Cognitive

Psychology, 13, 149–206.Huttenlocher, J., Hedges, L. V., & Duncan, S. (1991). Categories and particulars: prototype effects in

estimating spatial location. Psychological Review, 98, 352–376.Kopinska, A., & Harris, L. R. (2003). Spatial representation in body coordinates: evidence from errors in

remembering positions of visual and auditory targets after active eye, head, and body movements.Canadian Journal of Experimental Psychology, 57, 23–37.

Levinson, S. C. (1996). Frames of reference and Molyneux’s question: Crosslinguistic evidence. In P.Bloom, M. A. Peterson, L. Nadel, & M. F. Garrett (Eds.), Language and space (pp. 109–169).Cambridge MA: The MIT Press.

McNamara, T. P. (2003). How are the locations of objects in the environment represented in memory? In C.Freksa, W. Brauer, C. Habel, & K. Wender (Eds.), Spatial cognition III: Routes and navigation, human

memory and learning, spatial representation and spatial reasoning (pp. 174–191). Berlin: Springer-Verlag.McNamara, T. P., Rump, B., & Werner, S. (2003). Egocentric and geocentric frames of reference in

memory of large-scale space. Psychonomic Bulletin & Review, 10, 589–595.Millar, S., & Al-Attar, Z. (2004). External and body-centered frames of reference in spatial memory:

evidence from touch. Perception & Psychophysics, 66, 51–59.Mou, W., & McNamara, T. P. (2002). Intrinsic frames of reference in spatial memory. Journal of

Experimental Psychology: Learning, Memory, & Cognition, 28, 162–170.Mou, W., McNamara, T. P., Valiquette, C. M., & Rump, B. (2004). Allocentric and egocentric updating of

spatial memories. Journal of Experimental Psychology: Learning, Memory, & Cognition, 30, 142–157.O’Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map. Oxford: Clarendon.Pani, J. R., & Dupree, D. (1994). Spatial reference systems in the comprehension of rotational motion.

Perception, 23, 929–946.Presson, C. C., DeLange, N., & Hazelrigg, M. D. (1987). Orientation specificity in kinesthetic spatial

learning: the role of multiple orientations. Memory & Cognition, 15, 225–229.Presson, C. C., DeLange, N., & Hazelrigg, M. D. (1989). Orientation specificity in spatial memory: what

makes a path different from a map of a path? Journal of Experimental Psychology: Learning, Memory,

& Cognition, 15, 887–897.Presson, C. C., & Montello, D. R. (1994). Updating after rotational and translational body movements:

coordinate structure of perspective space. Perception, 23, 1447–1455.Rock, I. (1973). Orientation and form. New York: Academic Press.


Richardson, A. E., Montello, D. R., & Hegarty, M. (1999). Spatial knowledge acquisition from maps andfrom navigation in real and virtual environments. Memory & Cognition, 27, 741–750.

Rieser, J. J. (1989). Access to knowledge of spatial structure at novel points of observation. Journal of

Experimental Psychology: Learning, Memory, & Cognition, 15, 1157–1165.Roskos-Ewoldsen, B., McNamara, T. P., Shelton, A. L., & Carr, W. (1998). Mental representations of

large and small spatial layouts are orientation dependent. Journal of Experimental Psychology:

Learning, Memory, & Cognition, 24, 215–226.Shelton, A. L., & McNamara, T. P. (1997). Multiple views of spatial memory. Psychonomic Bulletin &

Review, 4, 102–106.Shelton, A. L., & McNamara, T. P. (2001a). Visual memories from nonvisual experiences. Psychological

Science, 12, 343–347.Shelton, A. L., & McNamara, T. P. (2001b). Systems of spatial reference in human memory. Cognitive

Psychology, 43, 274–310.Sholl, M. J. (1987). Cognitive maps as orienting schemata. Journal of Experimental Psychology: Learning,

Memory, & Cognition, 13, 615–628.Sholl, M. J. (1999). Egocentric frames of reference used for the retrieval of survey knowledge learned by

map and navigation. Spatial Cognition & Computation, 1, 475–494.Sholl, M. J., & Bartels, G. P. (2002). The role of self-to-object updating in orientation-free performance on

spatial-memory tasks. Journal of Experimental Psychology: Learning, Memory, & Cognition, 28, 422–436.Sholl, M. J., & Nolin, T. L. (1997). Orientation specificity in representations of place. Journal of

Experimental Psychology: Learning, Memory, & Cognition, 23, 1494–1507.Valiquette, C. M., McNamara, T. P., & Smith, K. (2003). Locomotion, incidental learning, and the

selection of spatial reference systems. Memory & Cognition, 31, 479–489.Waller, D. (2006). Egocentric and nonegocentric coding in memory for spatial layout: evidence from scene

recognition. Memory & Cognition, 34, 491–504.Waller, D., & Hodgson, E. (2006). Transient and enduring spatial representations under disorientation

and self-rotation. Journal of Experimental Psychology: Learning, Memory, & Cognition, 32, 867–882.Waller, D., Montello, D., Richardson, A. E., & Hegarty, M. (2002). Orientation specificity and spatial

updating of memories for layouts. Journal of Experimental Psychology: Learning, Memory, &

Cognition, 28, 1051–1063.Waller, D., Richardson, A., & Lippa, Y. (2004, November). Egocentric reference frames in spatial

memory: Head versus body. Poster session presented at the 45th Annual Meeting of the PsychonomicsSociety, Minneapolis, MN.

Wang, R. F. (1999). Representing a stable environment by egocentric updating and invariantrepresentations. Spatial Cognition & Computation, 1, 431–445.

Wang, R. F. (in press). Spatial processing and view-dependent representations. In F. Mast & L. Jancke(Eds.), Spatial Processing in Navigation, Imagery, and Perception. Berlin: Springer.

Wang, R. F., & Brockmole, J. R. (2003). Human navigation in nested environments. Journal of

Experimental Psychology: Learning, Memory, & Cognition, 29, 398–404.Wang, R. F., & Spelke, E. S. (2000). Updating egocentric representations in human navigation. Cognition,

77, 215–250.Wang, R. F., & Spelke, E. S. (2002). Human spatial representation: insights from animals. Trends in

Cognitive Science, 6, 376–382.Werner, S. (2001). Role of environmental reference systems in human spatial memory. Poster presentation

at the 42nd annual meeting of the Psychonomics Society, Orlando, FL.Werner, S., & Schmidt, K. (1999). Environmental reference systems for large-scale spaces. Spatial

Cognition & Computation, 1, 447–473.Yamamoto, N., & Shelton, A. L. (2005). Visual and proprioceptive representations in spatial memory.

Memory & Cognition, 33, 140–150.Yoon, B. C., & Shelton, A. L. (2005). What’s up? Reference frames and spatial transformations. Poster

presentation at the 46th annual meeting of the Psychonomics Society, Toronto, ON.

isolating observer-based reference directions in human spatial memory: head, body, and the...

Documents