Stepping Into a Map: Initial Heading Direction Influences Spatial Memory Flexibility

Transcription

1 Cognitive Science (2013) 1 28 Copyright 2013 Cognitive Science Society, Inc. All rights reserved. ISSN: print / online DOI: /cogs Stepping Into a Map: Initial Heading Direction Influences Spatial Memory Flexibility Stephanie A. Gagnon, a,b Tad T. Brunye, a,b Aaron Gardony, a,b Matthijs L. Noordzij, c Caroline R. Mahoney, a,b Holly A. Taylor b a U.S. Army Natick Soldier RDEC, Cognitive Science Team b Department of Psychology, Tufts University c Department of Cognitive Psychology and Ergonomics, University of Twente Received 20 April 2012; received in revised form 10 October 2012; accepted 18 January 2013 Abstract Learning a novel environment involves integrating first-person perceptual and motoric experiences with developing knowledge about the overall structure of the surroundings. The present experiments provide insights into the parallel development of these egocentric and allocentric memories by intentionally conflicting body- and world-centered frames of reference during learning, and measuring outcomes via online and offline measures. Results of two experiments demonstrate faster learning and increased memory flexibility following route perspective reading (Experiment 1) and virtual navigation (Experiment 2) when participants begin exploring the environment on a northward (vs. any other direction) allocentric heading. We suggest that learning advantages due to aligning body-centered (left/right/forward/back) with world-centered (NSEW) reference frames are indicative of three features of spatial memory development and representation. First, memories for egocentric and allocentric information develop in parallel during novel environment learning. Second, cognitive maps have a preferred orientation relative to world-centered coordinates. Finally, this preferred orientation corresponds to traditional orientation of physical maps (i.e., north is upward), suggesting strong associations between daily perceptual and motor experiences and the manner in which we preferentially represent spatial knowledge. Keywords: Spatial language; Virtual environments; Orientation; Navigation As we navigate through the world, we perceive our surroundings from an egocentric perspective characterized by a body-centered frame of reference; these egocentric experiences are considered primary due to their immediacy and correspondence with natural perception and locomotion (Golledge, 1992; Levelt, 1989; Shelton & McNamara, 1997). Correspondence should be sent to Tad T. Brunye, U.S. Army NSRDEC, Attn: RDNS-WSH-S, 15 Kansas St., Natick, MA thaddeus.brunye@us.army.mil

2 2 S. A. Gagnon et al. / Cognitive Science (2013) Memory for space learned from an egocentric perspective, however, is often complemented by rapidly formed and enduring mental representations that include allocentric information about the overall structure of the environment (Mou, McNamara, Valiquette, & Rump, 2004; O Keefe & Nadel, 1978). The allocentric perspective is characterized by a world-centered frame of reference such as canonical coordinates (north/south/east/west) and, analogous to a map, uniquely represents fixed landmarks relative to one another rather than relative to the dynamic position of a navigator. Despite growing consensus that humans can develop spatial memories integrating both egocentric and allocentric information, relatively little is known about how these unique perspectives come to overlap with one another in memory, especially when spatial experience is acquired from a single perspective such as during navigation. Here, we propose that ease of alignment between egocentric experiences and developing allocentric representations might prove critical in initially developing a multi-perspective mental representation, in addition to then flexibly using spatial memories of novel environments learned from a single perspective. Could there be situations in which learning an environment is more difficult due to misalignment between body- and world-centered frames of reference? If so, how might this difficulty affect the ability to flexibly use spatial memories? We examine these questions by conducting two experiments: the first uses spatial descriptions, and the second relatively realistic virtual environments. In both cases, the environments were designed to alter the alignment between egocentric experiences and a world-centered coordinate system, and measure the effect of (mis)alignment on online and offline measures of spatial memory formation and application. 1. Perspectives and spatial memory Investigations into the form and function of spatial memory have strong historical roots in cognitive psychology, extending back several decades to seminal investigations of spatial problem solving in rats (O Keefe & Nadel, 1978; Tolman, 1948) and debates over whether the cognitive map is an appropriate metaphor to describe the nature of spatial memory (i.e., Burgess, Maguire, & O Keefe, 2002; Tversky, 1991, 1993). A fundamental question involves whether, and how, spatial perspectives influence the development and application of spatial memory. On the one hand, a number of studies have demonstrated that the perspective adopted during learning is maintained in mental representations (e.g., Perrig & Kintsch, 1985; Roskos-Ewoldsen, McNamara, Shelton, & Carr, 1998; Thorndyke & Hayes-Roth, 1982). On the other hand, many have argued that spatial memories can become abstracted from the learned perspective, allowing for flexible application to complex tasks (e.g., Brunye & Taylor, 2008a; Burgess, 2006; Ferguson & Hegarty, 1994; Mou et al., 2004; Noordzij & Postma, 2005; Taylor & Tversky, 1992). This abstraction process allows for flexible inferencing from perspectives other than those learned; for instance, learning in the egocentric perspective and then showing an ability to solve survey perspective inferences at accuracy levels similar to those following survey learning. In contrast, one can learn from the survey perspective and show high accuracy when

3 S. A. Gagnon et al. / Cognitive Science (2013) 3 solving inferences from the route perspective (e.g., Brunye, Rapp, & Taylor, 2008; Taylor & Tversky, 1992). The ability to accurately generate inferences from a perspective other than that learned is thought to indicate spatial memory flexibility; this type of flexibility affords accurate real-world spatial problem solving such as finding novel short-cuts or navigating around detours (Taylor & Brunye, in press). Some suggest that both egocentric/route and allocentric/survey representational formats may develop in parallel while learning an environment from one perspective (Burgess, 2006; Ishikawa & Montello, 2006; McDonald & Pellegrino, 1993; Ruddle, Payne, & Jones, 1997). Under this theory, during first-person navigation an individual will accumulate viewpoint-specific memories of events but may also develop a map-like, survey mental representation of the environment that she can subsequently rely upon to solve complex spatial problems. Likewise, studying a map or viewing an environment from an overhead perspective (e.g., while in an airplane) allows for both the development of a survey representation and drawing inferences about how things might appear from a ground-level perspective. 2. A map in the head? Physical maps provide reliable structure for organizing complex spatial information into a single depiction that accurately reflects an environment s spatial structure. The ubiquity of maps in Western culture has prompted some to posit that humans mentally represent environments in a format analogous to a cartographer s map (Tolman, 1948); this representation is often metaphorically referred to as a cognitive map or mental map (for a review, see Kitchin, 1994). But how far might this metaphor be stretched to account for the format of human spatial memory? Consider maps, atlases, and even Global Positioning Systems; in each of these cases, it is common to view north in a physically upward orientation. Exposure to this conventional orientation leads to some intriguing phenomena, including misassociations between north and physically upward space (Brunye et al., 2012), similar misassociations between north and elevated social status (Gagnon, Brunye, Robin, Mahoney, & Taylor, 2011), and a preferential selection of south-going routes in an effort to avoid perceived physically demanding northbound locomotion (Brunye, Mahoney, & Taylor, 2010). Thus, there is some compelling evidence that humans associatively represent north as physically up and this representation influences spatial decision making (and even real estate preferences; Meier, Moller, Chen, & Riemer-Peltz, 2011). Even though the allocentric perspective is by definition non-self-referential (Levelt, 1982), people may automatically associate allocentric terms (e.g., canonical directions) with body axes, likely resulting from first-hand experience viewing maps. Thus, not only would north be associated with up but also east might be linked with right, and west with left. As a result, if humans spontaneously attempt to develop cognitive maps during egocentric navigation, and if north is preferentially oriented toward the top of cognitive maps, similar to most physical maps, then learning new environments might rely upon a

4 4 S. A. Gagnon et al. / Cognitive Science (2013) coordinate-based global reference frame as an organizational principle. If this is the case, egocentric learning may be most efficient if the body-centered reference terms (e.g., left, right) are congruent with the typical self-referential layout of canonical terms on a map (e.g., west is left). That is, if the initial heading direction is oriented northward, then the navigator s body-centered coordinates would align with those traditionally associated with map viewing. Critically, if this is true, then misalignment between an egocentrically experienced orientation and a globally defined reference system (north/south/east/west) should lead to relative difficulty during learning. 3. Alignment effects in spatial memory Very little research has investigated how alignment between egocentric orientations and the coordinate-based (NSEW) world-centered frame of reference affects either the encoding or retrieval of spatial memory. Some work, however, indicates that spatial memory retrieval is facilitated when the facing direction of the body is aligned with an environment s global reference frame (for a review, see McNamara, Sluzenski, & Rump, 2008). Under this theory, egocentric experience is dominant in spatial memory, and people assign a spatial reference system to aid in defining an environment s intrinsic structure (Mou & McNamara, 2002; Shelton & McNamara, 2001). In the absence of a coordinate system that clearly defines canonical axes, people appear to identify a salient pattern (e.g., a row of buildings), or rely on their initial heading direction (i.e., principal reference vector; Mou et al., 2004; Shelton & McNamara, 2004), and use this information to define a privileged axis (McNamara, 2003). During navigation, egocentric experiences are tracked relative to this privileged axis (Kelly, McNamara, Bodenheimer, Carr, & Rieser, 2008). Moreover, people tend to align their memory representations with regard to this axis, such that when participants are tested on their memory for an environment, they tend to show relatively poor performance when a retrieval orientation is misaligned with their global reference system. It is largely assumed that limited performance under conditions of misalignment is due to the increased mental rotation efforts required to adopt an orientation misaligned with the privileged axis. A number of studies have investigated these types of misalignment effects within the contexts of initial path segments (Shelton & McNamara, 2004), room or carpet geometry (Kelly et al., 2008; Mou, Liu, & McNamara, 2009; Valiquette, McNamara, & Labrecque, 2007), and large-scale environment features (McNamara, Rump, & Werner, 2003). Some recent work suggests that similar results might be found when people are tasked to retrieve distal spatial information from orientations misaligned with the north-south axis. Frankenstein, Mohler, B ulthoff, and Meilinger (2012) had participants point to unseen locations in their hometown environment from orientations with varied degrees of alignment with the coordinate world-centered reference system (NSEW). When participants were oriented toward the north, they produced the lowest pointing errors. To our knowledge, this is the first study to report that individuals use a canonical coordinate system to structure their knowledge of highly familiar environments primarily experienced from

5 S. A. Gagnon et al. / Cognitive Science (2013) 5 egocentric perspectives; in this case, participants appear to use a north-south vector as a preferred axis during retrieval. However, because all participants had previously viewed maps of their hometown, these findings can only be interpreted as evidence that people were relying on spatial memories learned from maps rather than navigational experience when inferring complex survey relationships. It thus remains to be seen whether misalignment between body- and world-centered reference systems affects the acquisition and development of novel spatial memories. 4. Learning new environments In the present experiments, we investigate whether alignment between an initial path segment s heading direction and the coordinates axes defining an environment s global reference system will facilitate the spontaneous learning of unfamiliar environments. Learning new environments during exploration is a naturally egocentric experience and involves monitoring the positions of landmarks relative to the self, and understanding the positions of landmarks relative to each other and relative to a global reference system. Here, we ask how these processes interact during learning to create relatively flexible or inflexible spatial memories of unfamiliar environments. To motivate our hypotheses, we propose two juxtaposed real-life examples in which the initial path segment is either aligned or misaligned with the preferred coordinate axes. Consider walking into an unfamiliar town on a path heading due north, for instance, heading north on Main Street to enter Hanover, New Hampshire. In this case, the landmarks you experience en route are arranged in a manner congruent with that of map-based coordinates; landmarks to your right (e.g., the Post Office) are also to the east, and those to the left (e.g., the Town Hall) are to the west. In accordance with prior research, while navigating this environment you will begin to develop an allocentric representation (i.e., a cognitive map) and align your experiences with a global reference system. In the absence of knowledge pertaining to canonical coordinates, you are likely to establish your path into the environment as the principal reference vector defining the environment s intrinsic layout (Shelton & McNamara, 2004). In many cases, however, you would be cognizant of your heading direction relative to the global coordinate system; indeed, your written directions may have instructed you to enter town by heading north on NH Route 10. This practical case demonstrates a potentially overlapping alignment between an egocentric experience (e.g., to the right is the Post Office) and the global coordinate system that might define a developing cognitive map (e.g., to the east is the Post Office). In other situations, however, these two systems might be misaligned. Consider now entering the same town by heading south along Main Street; in this case, landmarks experienced to your left are now to the east (and vice versa). If humans attempt to build maplike allocentric spatial memories during egocentric navigation, and these cognitive maps are spontaneously oriented with north toward the top, then entering from the north heading southward would represent a 180 degree misalignment between egocentric experience and a global coordinate system. This extent of misalignment would thus be predicted to

6 6 S. A. Gagnon et al. / Cognitive Science (2013) produce mental rotation effort during learning and/or retrieval, and correspondingly limit the flexibility of spatial memories formed during navigation. For instance, mentally rotating the environment to represent the Post Office on your left as to the right (east) in your developing cognitive map would possibly induce mental effort that is unnecessary under conditions of relative alignment (Shepard & Hurwitz, 1984; Wickens, 1999). Further, this relative mental effort might also hinder your ability to form flexible spatial memories that could be applied to complex problem solving from an allocentric perspective, a process that is critical to finding short-cuts or navigating around detours (Chapuis, Durup, & Thinus-Blanc, 1987; Maguire, Nannery, & Spiers, 2006). 5. The present experiments Despite the clear practical and theoretical relevance of these issues, to our knowledge research has not attempted to understand how learning environments might be affected by the types of alignment effects described above. In two experiments, we examined whether costs would accrue during learning and manifest in subsequent memory flexibility when participants enter novel environments on paths headed northward, southward, eastward, or westward. In the first experiment, participants read descriptions of novel environments from either the egocentric (route) or allocentric (survey) perspective. The route perspective describes an egocentrically experienced path through the environment and the landmarks encountered along the way, and the survey perspective describes the overall allocentric layout of the environment in terms of its paths and landmarks. Critically, the descriptions either detailed the environment on initial path trajectories headed to the north, south, east, or west, and we monitored participants reading times and subsequent ability to flexibly apply their newly formed memories of the described environments. In the second experiment, we used virtual environment navigation to engage participants in egocentric navigation tasks that are thought to capture some dynamics of realworld navigation. Participants actively explored realistic virtual environments of towns; critically, their initial heading when entering each environment was either depicted (on an on-screen compass) as to the north, south, east, or west. In this study, we monitored measures of learning efficiency (e.g., total path length, time spent stopping) and the ability to effectively use their newly formed memories on a task demanding the application of allocentric knowledge. 6. Experiment 1 For our first experiment, participants read a series of Route and Survey spatial descriptions that were introduced using spatial coordinates (i.e., canonical directions). The Route descriptions described the first path segment s initial heading direction as either oriented to the north, south, east, or west (see Appendix A), and the remainder of the description used solely body-referenced terms (e.g., to your left). Correspondingly, the Survey

7 descriptions described the layout of the environment by detailing areas beginning on one side of the map, heading in the opposite direction (i.e., northward, southward, eastward, westward). This Survey description manipulation allowed us to assess the effect of canonical direction orientation in a case where we would not expect to see an alignment effect between egocentric and allocentric perspectives, since the descriptions were exclusively described using world-centered, allocentric terms. We measured sentence reading times during initial comprehension of each environment, as well as performance on a subsequent memory task requiring inferences from both survey and route perspectives. In this manner, we were able to identify comprehension costs associated with varying the initial heading direction and to assess how these costs might limit flexibility of spatial memory Method S. A. Gagnon et al. / Cognitive Science (2013) Participants A total of 40 native English-speaking Tufts University students (12 males, age M = 20.4 years, SD = 1.8) participated for monetary compensation Materials Spatial descriptions: Descriptions of 16 unique environments were created from the route and survey perspective. Similar to the texts from Taylor and Tversky (1992), the survey descriptions used canonical, allocentric terms (north/south/east/west), while the route descriptions used egocentric terms (e.g., left, right, straight ahead) after the initial canonical heading direction was provided. Each description consisted of an introductory sentence (e.g., The town of Eureka was established during the Gold Rush in the 1800s) and then 19 additional sentences describing the environment itself. Half of the environments were large-scale towns, and half were small-scale enclosed locations (e.g., museum, aquarium). The large-scale environments contained two or three geographic features, three to five landmarks, and three to five street names. The small-scale environments were divided into three or four sections that contained a total of seven to nine landmarks. The streets were all straight and intersected at 90 degree angles. To promote informational equivalence, both route and survey descriptions began with configural information about the spatial features defining the environment s borders (e.g., large-scale geographic features like mountain ranges, rivers) using language consistent with the description perspective. Both description types contained locative (Survey: South of Emerson Country Club, Pirate Lane heads east and Route: Continue past Emerson Country Club, and turn left onto Pirate Lane ) and non-locative sentences ( The supermarket gets all of their produce from local farms ). Non-locative sentences were identical in route and survey descriptions, and locative sentences conveyed spatially equivalent information. All the towns were fictitious but loosely based on real locations from various regions across the United States (e.g., southwest, midwest, northeast). Further, both small- and large-scale environments contained themed landmarks and features that were semantically related to facilitate learning of the non-locative information (i.e., the beach town contained a Surf Shop, Clam Shack, Ocean, etc.). A small pilot study (n = 4) demonstrated that readers

8 8 S. A. Gagnon et al. / Cognitive Science (2013) could produce an accurate map during description reading, for route and survey versions of all 16 environments. Full sample descriptions (one route, one survey) are provided in Appendix A. We created route and survey versions for each of the 16 environments. From these 32 descriptions, we created four versions of each, aligning the environment with each coordinate direction (north/south/east/west), resulting in 128 descriptions total. Within a given perspective, the descriptions of the four different orientations were identical with the exception of spatial terms. In the route descriptions, the navigator entered the environment oriented toward one of four cardinal directions (N/S/E/W); to do so, we rotated the environment by 90 degree increments and reorganized the description accordingly. The canonical term always appeared in either sentence 1 or 2. Although the survey descriptions conveyed the overall spatial layout, landmarks and street names were presented in the same order as route descriptions (i.e., the southward route and survey descriptions both begin in the north, proceeding toward the south), and canonical terms were modified to match the corresponding orientation of the environment. Both route and survey descriptions had similar sentence lengths and Flesch Reading Ease scores Statement verification task: Six true/false statements assessed spatial knowledge indirectly (inference) from both the survey and route perspectives (three survey and three route statements for each environment), as exemplified in Appendix B. These statements were based on knowledge that was not explicitly stated in the text but could be inferred from information in the first, middle, and final segment of the description (termed locative inference statements). Across descriptions, we equated the number of inference items that oriented the reader to the north, south, east, or west direction. In addition, there were two statements assessing paraphrased non-locative knowledge for each environment (termed non-locative paraphrased statements). Across descriptions, half the statements were correctly answered as true and half false Procedure Learning: Each participant read descriptions of the 16 environments (eight route, eight survey) in random order. Of the eight descriptions in each perspective, two each used the north-going, south-going, east-going, and west-going initial heading directions. The perspective and orientation of each environment were counterbalanced across participants. Descriptions were presented on a computer monitor using SuperLab 4.0 TM (Cedrus, Inc., San Pedro, CA, USA), one sentence at a time (14-point Arial font), interleaved with a fixation cross-centered on the screen. Each sentence remained on the screen until participants pressed the spacebar to advance to the next sentence, with a maximum time of 20 s per sentence. The reading time for each sentence was recorded. Participants were told that after reading each description their memory for that location would be tested Testing: After each description participants completed the corresponding statement verification task. The true/false statements were self-paced and presented one at a time in a randomized order; participants verified statements as quickly and accurately as

9 possible by pressing keys labeled true (C) or false (M). Accuracy and response times were recorded Results S. A. Gagnon et al. / Cognitive Science (2013) 9 Analyses of participant means were performed with the R statistical software package ( Effect sizes are provided using eta-squared (g 2 ) or Cohen s d. Our analyses here primarily focus on route and survey perspectives separately; this allows us to investigate the influence on canonical orientation within each perspective. All participants had above chance accuracy across all conditions (M = 0.74, SD = 0.10) Reading time results To adjust for differences in sentence length across perspectives as well as overall differences in participants reading rates, we calculated residual reading times (Ferreira & Clifton, 1986), the difference between actual reading times and regression-based predicted reading times (based on each participant s average reading rate in msec per character). We then averaged residual reading times within each of the three sentence segments: the first segment, which always contained the first canonical term (sentences 1 6); second segment (sentences 7 12); and third segment (sentences 13 19). This segmented approach allowed us to analyze in a manner that promoted informational equivalence across the route and survey perspectives; similar information could be derived from each segment across the two perspectives. A sentence-by-sentence approach would limit comparability across perspectives. We submitted these residual reading times from both Route and Survey descriptions to separate 3(Segment: First, Second, Third) 9 4(Canonical Direction: northward, southward, eastward, westward) repeated measures analysis of variance (ANOVA) to test for comprehension effects during learning. Reading time data are detailed in Table 1. Route description analysis revealed main effects of Segment, F(2,78) = , p <.01, g 2 = 0.45, and Canonical Direction, F(3,117) = 6.025, p <.01, g 2 = 0.03, with no interaction, F(6,234) = 0.86, p >.10, g 2 < Participants read faster as the description progressed: First (M = , SE = ), Second (M = , SE = 81.28), and Third segment (M = , SE = 80.81). Critically, reading times were fastest with the northward orientation, as indicated by planned t-tests comparing the northward to southward, t(119) = 5.81, p <.01, d = 0.47, eastward, t(119) = 2.45, p <.05, d = 0.22, and westward directions, t(119) = 3.66, p <.01, d = 0.32 (see Fig. 1A). 1 Survey description analysis demonstrated a main effect of Segment, F(2,78) = , p <.01, g 2 = 0.42, but no effect of Canonical Direction, F(3,117) = 1.71, p >.1, g 2 < 0.01, and no interaction, F(6,234) = 0.86, p >.1, g 2 < Similar to the Route descriptions, participants read faster as the Survey description progressed: First (M = , SE = 84.01), Second (M = , SE = 80.67), and Third segment (M = , SE = 77.88). However, the Canonical Direction did not have a significant effect on Survey description reading times (see Fig. 1B).

10 10 S. A. Gagnon et al. / Cognitive Science (2013) Table 1 Experiment 1: mean and standard error residual reading time data as a function of Description Perspective, Canonical Direction, and Segment Segment 1 Segment 2 Segment 3 M SE M SE M SE Route perspective North , , South East , West , Survey perspective North South , East , West (A) (B) Fig. 1. Experiment 1: mean and standard error residual reading times (in msec) for the route (A) and survey (B) perspectives, each of four canonical directions (northward, southward, eastward, westward) and three description segments Statement verification results Recall that statement verification questions were either paraphrased and non-locative or inferential and locative; locative inferences were either within- or across-perspective. Statement verification response times are presented in msec/character to account for differential statement length. Only correct trials are included in response time analyses Paraphrased non-locative statements: Accuracy on paraphrased statement verification trials was overall high (M = 0.85, SE = 0.02). For each description perspective, a one-way repeated measures ANOVA was conducted to compare the effect of Canonical Direction (northward, southward, eastward, westward) on both accuracy and response

11 S. A. Gagnon et al. / Cognitive Science (2013) 11 time. Analysis of accuracy or response times did not reveal any main effects (all ps >.1) for either Route or Survey descriptions Locative inference statements: Inference statements were either congruent or incongruent with the studied perspective; the former requires an inference only, and the latter requires an inference and perspective-switch. The ability to accurately and efficiently switch perspectives to solve inference problems demonstrates the inherent flexibility of spatial memory; importantly, the ability to efficiently perform Survey inferences after reading a description in the route perspective points to the integration of egocentric and allocentric information during learning. As such, we separately analyzed within- and across-perspective inference questions in a series of ANOVAs Within-perspective inferences: Accuracy: We analyzed within-perspective inferencing ability after reading Route and Survey descriptions. A one-way within subjects ANOVA testing the effects of Canonical Direction (northward, southward, eastward, westward) on accuracy for Route inferences showed no effect of initial heading direction (p >.05, M avg = 0.74, SD avg = 0.13). However, the same ANOVA on accuracy data for Survey inferences showed an effect of Canonical Direction, F(3, 117) = 2.93, p < 0.05, g 2 = 0.07; Bonferroni corrected t-tests did not reveal any significant differences among heading directions, but there was a trend that descriptions described from east-to-west resulted in the least accurate verifications (M = 0.74, SD = 0.21), whereas descriptions described from west-to-east were most accurate (M = 0.83, SD = 0.20) Within-perspective inferences: Response times: One-way within-subject ANOVAs looking at the effects of Canonical Direction on response times did not show any effects for either Route or Survey inferences (all ps >.1) Across-perspective inferences: Accuracy: For analysis of across-perspective inferences, we calculated difference scores for each participant by subtracting the average accuracy for within-perspective inferences. Thus, we can assess performance for a Survey inference after reading a Route description (requiring across-perspective inference), relative to average performance on Survey inferences after reading the description from the Survey perspective (a within-perspective inference), and vice versa. One-way within-subject ANOVAs looking at the effects of Canonical Direction on accuracy did not show any effects for Route or Survey across-perspective inferences (all ps >.1) Across-perspective inferences: Response times: As with the analysis of acrossperspective accuracy, we again computed difference scores for each participant by subtracting the average response times for within-perspective inferences. A one-way within-subject ANOVA revealed an effect of Canonical Direction on response times for Survey inferences after reading Route descriptions, F(3, 117) = 4.07, p < 0.01, g 2 = Specifically, response times for Survey inferences were fastest after reading a Route description with an initial northward heading, as indicated by planned t-tests

12 12 S. A. Gagnon et al. / Cognitive Science (2013) comparing the northward to southward, t(39) = 3.55, p <.01, d = 0.56, eastward, t (39) = 2.38, p <.05, d = 0.38, and westward directions, t(39) = 2.80, p <.01, d = 0.44 (all other ps >.1; see Fig. 2A). In contrast, a similar ANOVA did not reveal any effect of Canonical Direction on response times for Route inferences after reading Survey descriptions (p >.1; see Fig. 2B) Experiment 1 Discussion Aligning the initial path segment s heading direction with the preferred direction (south to north) of the north-south reference axis during Route description reading facilitated the learning of an unfamiliar environment s allocentric structure, yielding the predicted advantages during comprehension and subsequent test. Specifically, Route sentence reading times increased when the initial heading orientation deviated from the northward direction. In general, readers showed faster reading times when the route began oriented northward, thus aligning body- and world-centered frames of reference. Compellingly, this northward comprehension advantage persisted numerically across all three description segments (the effect was greatest in the second segment, and north was significantly faster than south in the third segment, p <.001, uncorrected). Relative to this northward heading, comprehension costs were found in all other directions but were most pronounced with the greatest angular deviation from a northward heading (180 degrees; entering heading southward). These results provide evidence that readers appear to use an allocentric framework congruent with the standard map layout to structure their developing spatial memory; comprehension costs likely accrue due to effort involved in mentally (A) (B) Fig. 2. Experiment 1: mean and standard error response times to correctly verified across-perspective inference statements for each of the four canonical directions (northward, southward, eastward, westward). (A) depicts data from learning the route perspective and answering survey inference statements, and (B) depicts data from learning the survey perspective and answering route inference questions.

13 S. A. Gagnon et al. / Cognitive Science (2013) 13 rotating the environment to afford congruence between body- and learned world-centered frames of reference. We also found support for our prediction that aligning body- and world-centered frames of reference during Route learning facilitates the development of flexible spatial knowledge. The predicted response time results demonstrate that when Route description readers are required to switch perspectives at test a task requiring and demonstrating spatial flexibility they are especially efficient doing so when they learned from an initial northward heading. This effect was specific to across-perspective inferencing, when participants typically show the most difficulty given the required problem solving (inferencing) and perspective-switching (Brunye & Taylor, 2008a; Taylor & Tversky, 1992). A number of studies demonstrate that readers can switch perspectives to solve spatial inferences (Brunye et al., 2008; Ferguson & Hegarty, 1994; Taylor, Naylor, & Chechile, 1999; Taylor & Tversky, 1992); however, this process appears to be relatively time- and resource-consuming (Brunye & Taylor, 2008a; Deyzac, Logie, & Denis, 2006; Pazzaglia, De Beni, & Meneghetti, 2007). This work suggests that perspective-flexible mental representations may be formed during reading but are relatively restricted, at least when readers are limited to a single read-through (i.e., Brunye & Taylor, 2008b). Most critically, after reading a Route description with an initial northward heading direction, performance on subsequent Survey inferences was numerically similar to when readers had initially acquired information from a Survey description (see Fig. 2A). These data uniquely demonstrate that the extent to which Route descriptions are structured in a manner that promotes the parallel development of egocentric and allocentric mental representations during reading predicts the flexibility of spatial memory. In accordance with this hypothesis, the benefit of an initial northward heading direction also suggests that readers are structuring their memory of the environment using a principal reference vector that is defined by a route s first path segment (Shelton & McNamara, 2004). If this is the case, then the initial heading direction established in the Route descriptions defines the intrinsic structure of the environment, and as such, the axis formed by the initial heading direction serves as an anchor structuring the subsequent memory representations. Notably, our results suggest that aligning the principal reference vector (i.e., the initial heading direction) with the self-referential global coordinate reference frame (e.g., east to the right) enables the development of a mental representation that incorporates both egocentric and allocentric information, even though the learning format is strictly single-perspective. An additional, though tentative, result should also be noted. The structure of Survey descriptions involves detailing from an allocentric perspective, using solely world-centered coordinates, and is analogous to describing the visual layout of a map. Thus, while Survey descriptions detailed environmental features in the same order as the headingcomparable route perspective, Survey comprehension should not be affected by incongruence between body- and world-centered reference frames because these two types of reference frames should never conflict in a Survey description, regardless of where the description begins. Rather, it appears that comprehension was numerically deficient when the progression of spatial structure was incongruent with a mental scanning path, which

14 14 S. A. Gagnon et al. / Cognitive Science (2013) typically begins in the left. In this manner, readers showed some limited evidence of developing map-like mental imagery during Survey description reading. Supporting this contention, a number of studies have demonstrated that a participant s primary spoken language predicts attentional allocation toward particular elements of visual or verbal information. Specifically, with primary languages that read from left to right, participants tend to have visual scanning habits that move from left to right (Pollatsek, Bolozky, Well, & Rayner, 1981; Spalek & Hammad, 2005; Vaid & Singh, 1989). Further, the extent to which information is aligned with the left-to-right organization predicts comprehension and subsequent test performance (Harsel & Wales, 1987). This tendency to initially orient attention to the left results in visual scanning that begins in the upper left and then moves rightward (and downward). Thus, Survey descriptions that begin describing the westward areas of the environment are well aligned with this tendency, assuming that there is an association between world-centered canonical terms (e.g., west) and their corresponding learned self-referential terms (e.g., left). This effect provides some suggestion that Survey description reading promotes the formation of allocentric mental imagery that is organized by visual attention principles. As suggested by the visual attention literature, our results with survey descriptions should not replicate in participants whose primary language reads from right to left (e.g., Arabic; Maass, Pagani, & Berta, 2007; Spalek & Hammad, 2004; Tversky, 1991). 7. Experiment 2 The results of Experiment 1 provide strong evidence that aligning an initial egocentric experience (acquired through reading route perspective spatial descriptions) with a global coordinate reference frame oriented in a manner congruent with map-based depictions (i.e., north is up) facilitates efficient learning of an environment. Experiment 2 was conducted to assess the generality of this effect; namely, the present experiment uses virtual environment navigation to evaluate whether the effect observed in spatial descriptions can be extended to an egocentric navigation task similar to real-world navigation. The results from Experiment 1 leave open the possibility that the alignment between body-centered and world-centered frames of reference is uniquely beneficial for learning spatial information via written descriptions. To test this, in Experiment 2 participants were given a fixed amount of time to actively explore four realistic large-scale virtual environments, where their initial heading when entering the environment was briefly depicted on the screen via an on-screen compass oriented either to the north, south, east, or west. To promote comparability to Experiment 1 s reading time measures, we monitored two online measures of learning that we considered indicative of navigators processing difficulty during learning: first, we measured participants total path length during navigation, and second the amount of time they stopped virtual locomotion during learning. Indeed, situations involving disorientation or reorientation tend to elicit brief cessation of locomotion, perhaps in an effort to gain one s bearings by looking around to reassess one s

15 current location (Aretz & Wickens, 1992; Elvins, Nadeau, Schul, & Kirsh, 2001). As anecdotally described by Newcombe and Ratliff (2007), when one emerges from a subway station she stops to look around in an effort to reorient; indeed, this behavioral marker of reorientation has been experimentally observed in both rats (Cheng, 1986) and humans (Hermer-Vazquez, Spelke, & Katsnelson, 1999). Further supporting this pattern, Elvins et al. (2001) found that providing navigators with more comprehensive spatial knowledge prior to navigation led to lower rates of looking around and re-orienting. Given these earlier findings, we hypothesized that higher degrees of mental rotation effort involved in integrating body- and world-centered reference frames may elicit more frequent cessation of locomotion (i.e., time stopped) and hence shorter overall path lengths (see also, Olmos, Liang, & Wickens, 1997). Finally, as in Experiment 1, we also measured the offline ability to effectively use newly formed memories on an inference task demanding the application of allocentric knowledge Method S. A. Gagnon et al. / Cognitive Science (2013) Participants Thirty-two Tufts University undergraduate students (17 males, age M = 19.9 years, SD = 1.6) participated for monetary compensation. Each participant s prior video game experience was measured using a common video game questionnaire (Basak, Boot, Voss, & Kramer, 2008), containing a series of eight questions probing for frequency, proficiency, and type of video gaming experience. Overall, participants reported lowmoderate video gaming frequency (M = 1.20 on scale from 0 to 4); the 12 participants who self-identified as video gamers also reported low-moderate frequency of use (M = 1.79 h/week) Materials Virtual environments: We created four large-scale environments using the Unreal Tournament video game editor (Unreal Engine 2; Epic Games, Raleigh, NC, USA). Each environment measured approximately 68,376 square meters and contained a solid blue sky devoid of global spatial cues (e.g., clouds, sun) and 16 distinct and unique labeled landmarks (e.g., Bakery, Town Hall). Only open space between landmarks was navigable; participants could not enter buildings or navigate beyond the boundaries of the environment. In each environment, the navigation starting position was located in an entrance hallway comprised of generic buildings (see Appendix D). While navigating through this hallway into the main environment, the Unreal software rendered a compass rose on the center right of the screen, indicating the initial allocentric heading. Navigation through the hallway took approximately 10 s, after which the compass rose disappeared. Avatar height was 1.67 m and avatar speed was fixed at 9.14 m/s. Extraneous movement types included in the software (e.g., jumping, crouching, and weapon use) were disabled. Participants navigated the virtual environments on a 20 widescreen LCD display with 1,440 x 900 native resolution with a simulated field of view of 90 and a viewing distance of approximately 2 ft using a standard keyboard (WASD movement controls)

16 16 S. A. Gagnon et al. / Cognitive Science (2013) and mouse (to control orientation). During navigation, the Unreal software automatically measured navigator position in x- and y-coordinates, and heading orientation (yaw) at 50 ms intervals. These data were automatically output into a text file and analyzed by software written in house Spatial statement verification task: After navigating each environment for a fixed amount of time, allocentric spatial knowledge was assessed through 120 true/false statements. As in Experiment 1, the true/false statements were self-paced and presented one at a time in a randomized order using SuperLab 4.0 TM (Cedrus, Inc., San Pedro, CA, USA) participants verified statements as quickly and accurately as possible by pressing keys labeled true (C) or false (M). Accuracy and response times were recorded. For each of the four virtual environments we first calculated all possible combinations of the 16 unique landmarks, resulting in 16!/(2!(16 2)!) = 120 unique combinations (e.g., bank post office). The order of these 120 combinations was randomized, and we then selected the first 60 as north/south statements (e.g., bank north post office) and the last 60 as east/ west statements (e.g., bank east post office). Then, we split each group of 60, designating the first 30 as correct statements and the last 30 as incorrect statements. All statements were presented in the following format: Landmark 1 is (north/south/east/west) of Landmark Procedure Upon arrival, participants provided informed consent and completed questionnaires. Afterward, they completed a brief practice session, familiarizing themselves with the virtual environment controls, the experiment instructions, and the spatial statement verification task. The experimenter explained that participants would first navigate each environment for a fixed time trying to learn as much as they could about the environment including layout information and landmark names and locations. The experimenter instructed participants to record that they had visited the landmarks they encountered by virtually touching red flags adjacent to each landmark during navigation. Participants were also warned that the compass rose would disappear once upon entering the main environment. Participants then navigated the first of four environments for 7 min, which previous pilot studies had confirmed to be an adequate amount of time to reach all 16 landmarks. Following the exploratory navigation phase, participants completed the spatial statement verification test. Upon completion, participants took a break of at least 5 but no longer than 10 minutes and then repeated this process for the remaining three environments Results Online navigation performance On average, participants were able to reach the majority of the 16 total landmarks (M = 15.3, SD = 0.59) within the 7 minutes of navigation time for each environment, regardless of the initial heading direction, F(3,87) = 1.63, p >.20, g 2 = The following analyses consider the total path length traveled within the environment, measured

17 S. A. Gagnon et al. / Cognitive Science (2013) 17 in Unreal Units (UUs; 16 UUs = approximately 1 linear ft or m), and the amount of time the participant spent momentarily stopped during navigation (i.e., when speed was less than 100 UUs/s; M speed = 432.4, SD speed = 57.0). On average, total path length and the amount of time spent stopped were strongly correlated, such that a longer path length through the environment was related to less time spent momentarily stopped (r =.996, p <.001). Note that the online navigation data from two participants were excluded due to a computer error Total path length: First, initial heading direction marginally influenced the total path length traveled during navigation, F(3,87) = 2.53, p =.06, g 2 = To specifically test the hypothesis that the northward heading direction would uniquely allow the navigator to travel a greater distance within the environment, we conducted planned t-tests comparing the northward to southward, t(29) = 2.28, p <.05, d = 0.42, eastward, t (29) = 2.58, p <.05, d = 0.47, and westward directions, t(29) = 2.34, p <.05, d = 0.43 (all other ps >.1; see Fig. 3A) Amount of time stopped: Similarly, initial heading direction marginally influenced the amount of time the participant stopped moving during navigation, F (3,87) = 2.56, p =.06, g 2 = Again, planned t-tests confirmed that a northward heading direction resulted in less time stopping, relative to southward, t(29) = 2.30, p <.05, d = 0.42, eastward, t(29) = 2.50, p <.05, d = 0.46, and westward directions, t (29) = 2.43, p <.05, d = 0.44 (all other ps >.1; see Fig. 3B). (A) (B) Fig. 3. Experiment 2: mean and standard error for dependent measures of online navigation performance for each of the four canonical directions (northward, southward, eastward, westward). (A) depicts the total path length traveled through an environment, and (B) depicts the total amount of time (in seconds) the navigator was stopped within an environment.

18 18 S. A. Gagnon et al. / Cognitive Science (2013) Spatial statement verification All participants showed above chance accuracy across all spatial inferences (M = 0.74, SD = 0.14). Analyses include only accurate trials, and all trials with response times below 1,000 ms were removed. 2 Analyses were conducted on raw response time data given equated statement character length Accuracy: A one-way repeated measures (Canonical Direction: northward, southward, eastward, westward) ANOVA did not reveal any effects (p >.1) Response times: A one-way repeated measures (Canonical Direction: northward, southward, eastward, westward) ANOVA indicated an effect of initial heading direction, F (3,93) = 5.44, p <.01, g 2 = Specifically, the northward initial heading direction resulted in faster response times for Survey inferences, as revealed by planned t-tests comparing the northward to southward, t(31) = 3.25, p <.01, d = 0.57, eastward, t (31) = 2.79, p <.01, d = 0.49, and westward directions, t(31) = 2.86, p <.01, d = 0.51 (see Fig. 4). No other comparisons reached significance (all ps >.5) Experiment 2 discussion The present experiment replicated the main findings of Experiment 1, extending them beyond spatial descriptions into the domain of virtual navigation. Together, the two experiments provide convergent evidence that aligning body- and world-centered reference frames facilitates the parallel integration of egocentric and allocentric information in spatial memory when learning an unfamiliar environment from a first-person perspective. Fig. 4. Experiment 2: mean and standard error response times to correctly verified across-perspective (i.e., survey) inference statements for each of the four canonical directions (northward, southward, eastward, westward).

19 S. A. Gagnon et al. / Cognitive Science (2013) General discussion Learning a new environment often entails integrating first-person perspectives with information about the coordinate structure of the surroundings. In the present experiments, we asked whether global coordinate cues help structure mental representations formed from the egocentric perspective, and further, whether the use of these cues would affect the development of flexible spatial memories. We proposed that aligning body-centered (e.g., left, right) with self-referential but world-centered (e.g., west, east) representational formats during Route description reading or first-person virtual environment exploration would (a) minimize the effort required for mental rotation during initial learning of an environment, and (b) promote allocentric knowledge, and improve the ultimate flexibility of spatial memories formed during egocentric navigation. In two experiments, participants learned multiple environments through either spatial descriptions (Experiment 1) or naturalistic virtual environment exploration (Experiment 2), where the initial allocentric heading (i.e., north, south, east, west) was varied across environments. After each environment, participants were tested on their ability to draw spatial inferences from memory. Experiment 1 demonstrated a comprehension benefit via faster sentence reading times for initially northward-oriented Route descriptions; Experiment 2 showed facilitated navigation efficiency after entering the environment in a northward heading direction. Similarly, both experiments demonstrated improved performance on Survey inferences after learning environments from a northward-oriented egocentric perspective. Importantly, Experiment 1 demonstrated that this benefit for across-perspective Survey inferencing brought performance to the same level as if readers had originally learned the environment from the Survey perspective. Taken together, our data support and extend theoretical accounts that emphasize the importance of both egocentric and environmental cues in defining reference axes that structure spatial memory (Mou, Zhao, & McNamara, 2007; Mou et al., 2004; Shelton & McNamara, 2004). In such frameworks, navigators track a reference axis intrinsic to the environment, such as the edge of a carpet, room geometry, or the edges of buildings or bodies of water; egocentrically experienced information is continually referenced to this relatively global structure, and this referencing ultimately defines a privileged axis during recall (McNamara et al., 2008). In the present work, we provided coordinate information about participants initial heading direction as they entered environments; we hypothesized that the direction would be used to define a reference axis with which to organize their spatial memory. Results advance recent theoretical frameworks in two primary ways. First, we found strong evidence that participants will use global coordinate axes (NSEW) to define the structure of a novel environment during egocentric learning. Second, experiencing environments in a manner that aligns coordinate axes with a body-centered reference system (i.e., heading northward) facilitates the development of perspectiveabstracted mental representations congruent with typical map orientations (north-up). Why might spatial learning be facilitated by an initial northward heading direction? Mentally simulating locomotion during Route description reading (i.e., Brunye et al.,

20 20 S. A. Gagnon et al. / Cognitive Science (2013) 2010) similar to the experience of navigating through the paths and turns in a virtual or real-world environment relies strongly upon the representation of body-centered axes (see also, Noordzij, Zuidhoek, & Postma, 2006; Struiksma, Noordzij, & Postma, 2009). We have suggested that experience with maps and the typical compass rose configuration results in associations between canonical terms and body-centered axes, causing worldcentered terms to become in essence self-referential (e.g., west = left). Aligning bodycentered (egocentric) with world-centered (allocentric) frames of reference may be most efficient when these two dimensions promote angular congruence between first-hand experience and the typical structure of maps, that is, when right is east and left is west. With a northward heading, the right side of the environment corresponds to the east (to the right on a map-based mental image), and the left side of the environment corresponds to the west (to the left on a map-based mental image). With initial headings other than northward, for instance southward, the body-centered frame of reference (e.g., to the left) must be aligned with a world-centered frame of reference (e.g., eastward), involving a transformation that does not align with our customary association of cardinal directions and body axes. Our results support this notion by demonstrating greatest comprehension costs when the required mental rotation was largest (180 degrees). The extent that the correspondence between body-centered and world-centered frames of reference is incongruent, more effort is necessary toward comprehension due to increased mental rotation (Wickens, 1999). The present results are also congruent with recent literature suggesting that people preferentially orient mental maps of familiar environments with north upward (Frankenstein et al., 2012). If people indeed orient mental maps with the north in an upward orientation, then entering an environment from any direction other than from the south will lead to a misaligned cognitive map, which makes learning more difficult and memory more error prone (i.e., Boer, 1991; Rieser, 1989). Our results lend strong support to this contention and may be analogous to results found with map learning: Maps depicted in a north-up orientation (i.e., in aircraft, vehicles, or even tourist maps) tend to facilitate the learning of environmental layouts relative to those depicted in a track-up orientation (Aretz & Wickens, 1992). One possible limitation of our study, however, is not clarifying the nature of spatial memory when readers or navigators learn in the complete absence of canonical information. Of course, it would be difficult for participants to perform tests demanding the application of canonical knowledge without explicit reference to canonical direction during learning. Some evidence, however, suggests that when participants are not provided explicit canonical cues in spatial descriptions, they tend to typically position the entrances to buildings or large-scale environments in the south or at the bottom of a map (i.e., with the first path segment heading north; De Beni, Pazzaglia, Gyselinck, & Meneghetti, 2005; Meneghetti, Pazzaglia, & De Beni, 2012). Though our data cannot speak directly to this issue, given this earlier work we expect that data from a condition absent of an initial canonical heading would pattern similar to data from the northward heading condition. In the introduction, we distinguished several theoretical positions that attempt to characterize the form and function of spatial memories derived from varied-perspective

21 S. A. Gagnon et al. / Cognitive Science (2013) 21 experiences. At least three models were proposed first, one suggesting that learned perspectives are strictly maintained in memory (e.g., Perrig & Kintsch, 1985; Roskos-Ewoldsen et al., 1998; Thorndyke & Hayes-Roth, 1982); second, one suggesting that spatial memories can become (over time) abstracted from learned perspectives (e.g., Brunye & Taylor, 2008b; Mou et al., 2004; Taylor & Tversky, 1992); and third, one suggesting that both egocentric/route and allocentric/survey representational formats develop spontaneously and in parallel during learning (e.g., Burgess, 2006; Ishikawa & Montello, 2006; Ruddle et al., 1997). Converging evidence from the present experiments suggest support for the parallel model of spatial memory development. When difficulties arose aligning the body- and world-referenced systems, participants showed evidence of these difficulties not just at test but also during learning. Specifically, we found evidence of slowed reading times (Experiment 1) and more time stopped during navigation (Experiment 2) when the initial heading direction was anything other than northward. If participant were not attempting to integrate world-centered coordinates into their developing spatial memories, then online performance would not be expected to show this pattern. Further, follow-up analyses demonstrated that Experiment 2 test performance did not become more accurate over the course of the 120 survey-format test items, suggesting that some degree of world-centered knowledge was already represented in memory and was not affected by repeated testing. We note some limitations of the presently reported work, and motivation for future research. There was no compelling evidence that initial heading directions are critical for memorizing survey perspective spatial descriptions. We did, however, find some suggestion that mental images formed from survey descriptions may be organized in accordance with principles derived from the mental scanning literature (i.e., top-to-bottom, left-toright). In this case, we found numerical reading time advantages with southward and eastward progressing descriptions; future work might further investigate whether these biases are specific to canonical directions (i.e., with mental maps) or generalized to different content types (e.g., with diagrammatic reasoning). This type of work seems important to further elucidating the format of spatial mental representations derived from varied sources. Second, we note that the inclusion of a control condition without explicating an initial coordinate direction might help elucidate some of the default processing strategies people use when navigating or reading route descriptions. We expect that in the absence of an initial heading direction, people would assume a northward heading. Future work might experimentally investigate this possibility. Overall, we demonstrate that learning environments from the egocentric perspective and developing flexible spatial memories involves interactions between both egocentric and allocentric information. Route descriptions and virtual environment navigation directly convey egocentric information, but we also demonstrate that readers and navigators use allocentric cues to structure their memory for the described spaces. When the allocentric spatial structure implied by an initial heading direction conflicts with the body-centered frame of reference, readers and navigators incur comprehension costs. These costs accrue and impair the flexibility of spatial memory. We also provide evidence that Survey description readers develop mental images that preserve some of the characteristics of

22 22 S. A. Gagnon et al. / Cognitive Science (2013) attentional deployment during real-world visual tasks. These results lend support for theoretical advances in our understanding of the form and function of spatial memory (e.g., Burgess, 2006; Ishikawa & Montello, 2006; McDonald & Pellegrino, 1993), suggesting that the development of flexible spatial memories involves not just a single-perspective representation of an experienced space but a complementary and parallel development of alternate-perspective representations. When these representational formats merge, they afford the flexibility necessary for successfully solving real-world spatial problems. Acknowledgments SAG and TTB contributed equally to this manuscript; author order was determined randomly. Notes 1. Follow-up analyses demonstrated that small-scale environment descriptions tended to be read faster than large-scale ones (p <.01); this effect did not interact with description perspective (p =.47) or canonical direction (p =.17). 2. On the basis of minimum sentence length of seven words; at a reading speed of 175 ms per word, we considered the fastest reasonable response time as approximately 1,500 ms. References Aretz, A. J., & Wickens, C. D. (1992). The mental rotation of map displays. Human Performance, 5, Basak, C., Boot, W. R., Voss, M. W., & Kramer, A. F. (2008). Can training in a real-time strategy videogame attenuate cognitive decline in older adults? Psychology and Aging, 23, Boer, L. C. (1991). Spatial ability and orientation of pilots. Oxford, England: John Wiley & Sons. Brunye, T. T., Gagnon, S. A., Waller, D., Hodgson, E., Tower-Richardi, S., & Taylor, H. A. (2012). Up north and down south: Implicit associations between topography and cardinal direction. Quarterly Journal of Experimental Psychology, 65, Brunye, T. T., Mahoney, C. R., & Taylor, H. A. (2010). Moving through imagined space: Mentally simulating locomotion during spatial description reading. Acta Psychologica, 134, doi: /j. actpsy Brunye, T. T., Rapp, D. N., & Taylor, H. A. (2008). Representational flexibility and specificity following spatial descriptions of real-world environments. Cognition, 108, Brunye, T. T., & Taylor, H. A. (2008a). Working memory in developing and applying mental models from spatial descriptions. Journal of Memory and Language, 58, doi: /j.jml Brunye, T. T., & Taylor, H. A. (2008b). Extended experience benefits spatial mental model development with route but not survey descriptions. Acta Psychologica, 127(2), doi: /j.actpsy

23 S. A. Gagnon et al. / Cognitive Science (2013) 23 Brunye, T. T., Mahoney, C. R., Gardony, A. L., & Taylor, H. A. (2010). North is up(hill): Route planning heuristics in real-world environments. Memory & Cognition, 38, Burgess, N. (2006). Spatial memory: How egocentric and allocentric combine. Trends in Cognitive Sciences, 10(12), doi: /j.tics Burgess, N., Maguire, E. A., & O Keefe, J. (2002). The human hippocampus and spatial and episodic memory. Neuron, 35, doi: /s (02) Chapuis, N., Durup, M., & Thinus-Blanc, C. (1987). The role of exploratory experience in a shortcut task by golden hamsters (Mesocricetus auratus). Animal Learning & Behavior, 15, Cheng, K. (1986). A purely geometric model in rat s spatial representation. Cognition, 23, De Beni, R., Pazzaglia, F., Gyselinck, V., & Meneghetti, C. (2005). Visuospatial working memory and mental representation of spatial descriptions. European Journal of Cognitive Psychology, 17, doi: / Deyzac, E., Logie, R. H., & Denis, M. (2006). Visuospatial working memory and the processing of spatial descriptions. British Journal of Psychology, 97, Elvins, T. T., Nadeau, D. R., Schul, R., & Kirsh, D. (2001). Worldlets: 3-D thumbnails for wayfinding in large virtual worlds. Presence, 10, Ferguson, E. L., & Hegarty, M. (1994). Properties of cognitive maps constructed from texts. Memory & Cognition, 22(4), doi: /bf Ferreira, F., & Clifton, C., Jr (1986). The independence of syntactic processing. Journal of Memory and Language, 25, doi: / x(86) Frankenstein, J., Mohler, B. J., B ulthoff, H. H., & Meilinger, T. (2012). Is the map in our head oriented north? Psychological Science, 23, 2. Gagnon, S. A., Brunye, T. T., Robin, C., Mahoney, C. R., & Taylor, H. A. (2011). High and mighty: Implicit associations between physical space and social status. Frontiers in Psychololgy, 2, 259. Golledge, R. G. (1992). Place recognition and wayfinding: Making sense of space. Geoforum, 23(2), Harsel, Y., & Wales, R. (1987). Directional preference in problem solving. International Journal of Psychology, 22, doi: / Hermer-Vazquez, L., Spelke, E., & Katsnelson, A. (1999). Sources of flexibility in human cognition: Dual task studies of spatial and language. Cognitive Psychology, 39, Ishikawa, T., & Montello, D. R. (2006). Spatial knowledge acquisition from direct experience in the environment: Individual differences in the development of metric knowledge and the integration of separately learned places. Cognitive Psychology, 52, doi: /j.cogpsych Kelly, J. W., McNamara, T. P., Bodenheimer, B., Carr, T. H., & Rieser, J. J. (2008). The shape of human navigation: How environmental geometry is used in the maintenance of spatial orientation. Cognition, 109, Kitchin, R. (1994). Cognitive maps: What are they and why study them? Journal of Environmental Psychology, 14, Levelt, W. J. M. (1982). Cognitive styles in the use of spatial direction terms. In R. J. Jarvella & W. Klein (Eds.), Speech, place, and action (pp ). Chichester, UK: John Wiley & Sons. Levelt, W. J. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press. Maass, A., Pagani, D., & Berta, E. (2007). How beautiful is the goal and how violent is the fistfight? Spatial bias in the interpretation of human behavior. Social Cognition, 25, doi: /soco Maguire, E. A., Nannery, R., & Spiers, H. J. (2006). Navigation around London by a taxi driver with bilateral hippocampal lesions. Brain, 129, McDonald, T. P., & Pellegrino, J. W. (1993). Psychological perspectives on spatial cognition. In T. Garling & R. G. Golledge (Eds.), Behavior and environment: Psychological and geographical approaches (pp ). Amsterdam: Elsevier. doi: /s (08) McNamara, T. P. (2003). How are the locations of objects in the environment represented in memory? In. C. Freksa, W. Brauer, C. Habel & K. F. Wender (Eds.), Spatial cognition III: Routes and navigation,

24 24 S. A. Gagnon et al. / Cognitive Science (2013) human memory and learning, spatial representation and spatial reasoning, LNAI 2685 (pp ). 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 and Review, 10(3), McNamara, T. P., Sluzenski, J., & Rump, B. (2008). Human spatial memory and navigation. In H. L. Roediger, III (Ed.), Cognitive psychology of memory. Volume 2 in J. Byrne (Editor-in-chief), Learning and memory: A comprehensive reference (pp ). Oxford, England: Elsevier. Meier, B. P., Moller, A. C., Chen, J., & Riemer-Peltz, M. (2011). Spatial metaphor and real estate: North-south location biases housing preference. Social Psychological and Personality Science, 2, Meneghetti, C., Pazzaglia, F., & De Beni, R. (2012). The mental representation derived from spatial descriptions is north-up oriented: The role of visuo-spatial abilities. Lecture Notes in Computer Science, 7463, Mou, W., Liu, X., & McNamara, T. P. (2009). Layout geometry in encoding and retrieval of spatial memory. Journal of Experimental Psychology: Human Perception and Performance, 35, Mou, W., & McNamara, T. P. (2002). Intrinsic frames of reference in spatial memory. Journal of Experimental Psychology: Learning, Memory and Cognition, 28, Mou, W., McNamara, T. P., Valiquette, C. M., & Rump, B. (2004). Allocentric and egocentric updating of spatial memories. Journal of Experimental Psychology: Learning, Memory and Cognition, 30, doi: / Mou, W., Zhao, M., & McNamara, T. P. (2007). Layout geometry in the selection of intrinsic frames of reference from multiple viewpoints. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33, Newcombe, N.S., & Ratliff, K.R. (2007) Explaining the development of spatial reorientation: Modularityplus-language versus the emergence of adaptive combination. In J. Plumert & J. Spencer (Eds.), The emerging spatial mind (pp ). New York: Oxford University Press. Noordzij, M. L., & Postma, A. (2005). Categorical and metric distance information in mental representations derived from route and survey descriptions. Psychological Research, 69(3), doi: / s y Noordzij, M. L., Zuidhoek, S., & Postma, A. (2006). The influence of visual experience on the ability to form spatial mental models based on route and survey descriptions. Cognition, 100, doi: / j.cognition O Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map. Oxford, England: Oxford University Press. Olmos, O., Liang, C.-C., & Wickens, C. D. (1997). Electronic map evaluation in simulated visual meteorological conditions. International Journal of Aviation Psychology, 7, Pazzaglia, F., De Beni, R., & Meneghetti, C. (2007). The effects of verbal and spatial interference in the encoding and retrieval of spatial and non-spatial texts. Psychological Research, 71, Perrig, W., & Kintsch, W. (1985). Propositional and situational representations of text. Journal of Memory and Language, 24(5), doi: / x(85) Pollatsek, A., Bolozky, S., Well, A. D., & Rayner, K. (1981). Asymmetries in the perceptual span for Israeli readers. Brain & Language, 14, doi: / x(81) Rieser, J. J. (1989). Access to knowledge of spatial structure at novel points of observation. Journal of Experimental Psychology: Learning, Memory, & Cognition, 15, Roskos-Ewoldsen, B., McNamara, T. P., Shelton, A. L., & Carr, W. S. (1998). Mental representations of large and small spatial layouts are viewpoint dependent. Journal of Experimental Psychology: Learning, Memory, & Cognition, 24, doi: / Ruddle, R. A., Payne, S. J., & Jones, D. M. (1997). Navigating buildings in desk-top virtual environments: Experimental investigations using extended navigational experience. Journal of Experimental Psychology: Applied, 3, doi: / x

25 S. A. Gagnon et al. / Cognitive Science (2013) 25 Shelton, A. L., & McNamara, T. P. (1997). Multiple views of spatial memory. Psychonomic Bulletin & Review, 4, Shelton, A. L., & McNamara, T. P. (2001). Systems of spatial reference in human memory. Cognitive Psychology, 43, Shelton, A. L., & McNamara, T. P. (2004). Orientation and perspective dependence in route and survey learning. Journal of Experimental Psychology: Learning Memory and Cognition, 30(1), doi: / Shepard, R. N., & Hurwitz, S. (1984). Upward direction, mental rotation, and discrimination of left and right turns in maps. Cognition, 18, Spalek, T. M., & Hammad, S. (2004). Supporting the attentional momentum view of inhibition of return: Is attention biased to go right? Perception & Psychophysics, 66, doi: /bf Spalek, T. M., & Hammad, S. (2005). The left-to-right bias in inhibition of return is due to the direction of reading. Psychological Science, 16, doi: /j x Struiksma, M. A., Noordzij, M. L., & Postma, A. (2009). What is the link between language and spatial images? Behavioral and neural findings in blind and sighted individuals. Acta Psychologica, 132, doi: /j.actpsy Taylor, H. A., & Brunye, T. T. (in press). I go right, north and over: Processing spatial language. In D. Waller & L. Nadel (Ed.), Handbook of spatial cognition (pp ). Washington, DC: APA Publications. Taylor, H. A., Naylor, S. J., & Chechile, N. A. (1999). Goal-specific influences on the representation of spatial perspective. Memory & Cognition, 27, Taylor, H. A., & Tversky, B. (1992). Spatial mental models derived from survey and route descriptions. Journal of Memory and Language, 31(2), doi: / x(92)90014-o Thorndyke, P. W., & Hayes-Roth, B. (1982). Differences in spatial knowledge acquired from maps and navigation. Cognitive Psychology, 14(4), doi: / (82) Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55, doi: / h Tversky, B. (1991). Spatial mental models. In G. H. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory (pp ). San Diego, CA: Academic Press. Tversky, B. (1993). Cognitive maps, cognitive collages, and spatial mental models. In A.U. Frank & I. Campari (Eds.), COSIT 93, Lecture Notes in Computer Science 716 (pp ). Berlin: Springer. doi: / _2 Vaid, J., & Singh, M. (1989). Asymmetries in the perception of facial affect: Is there an influence of reading habits? Neuropsychologia, 27, doi: / (89) Valiquette, C. M., McNamara, T. P., & Labrecque, J. S. (2007). Biased representations of the spatial structure of navigable environments. Psychological Research, 71, Wickens, C. D. (1999). Frames of reference for navigation. In D. Gopher & A. Koriat (Eds.), Attention and performance XVII: Cognitive regulation of performance: Interaction of theory and application (pp ). Cambridge, MA: The MIT Press.

26 26 S. A. Gagnon et al. / Cognitive Science (2013) Appendix A Experiment 1: Sample Map (for illustrative purposes only; participants were only presented with verbal descriptions) and Description Excerpts from the route perspective, for northward versus southward. Town of Avondale is depicted Route Perspective: Northward Avondale is a typical small beach town. The town s population triples in the summer when tourists come. To reach Avondale, drive north along Atlantic Ave. to where the road crosses the Pawcatuck River. Continuing straight on Atlantic Ave. for another half mile you see the ocean straight ahead. You turn left onto Shore Rd. from Atlantic Ave., and see the Ocean on your right. (continued) Route Perspective: Southward Avondale is a typical small beach town. The town s population triples in the summer when tourists come. To reach Avondale, drive south along Atlantic Ave. to where the road crosses the Pawcatuck River. Continuing straight on Atlantic Ave. for another half mile you see the ocean straight ahead. You turn left onto Shore Rd. from Atlantic Ave., and see the Ocean on your right. (continued)

27 S. A. Gagnon et al. / Cognitive Science (2013) 27 Appendix B Experiment 1: Full sample descriptions from the route and survey perspective Route Perspective (Southward: Heading South) Avondale is a typical small beach town. The town s population triples in the summer when tourists come. To reach Avondale, drive south along Atlantic Ave. to where the road crosses the Pawcatuck River. Continuing straight on Atlantic Ave. for another half mile you see the ocean straight ahead. You turn left onto Shore Rd. from Atlantic Ave., and see the Ocean on your right. After a few blocks Shore Rd. ends. You are forced to make a left onto Ocean View Highway, heading back to the Pawcatuck River. Traveling down Atlantic Ave. toward the ocean, Shore Rd. is ahead to your left. Most of the town is along Shore Rd. by the Ocean. Shore Rd. used to be gravel but was recently paved. There are often pedestrians waiting to cross the street, so cars must drive carefully. After turning from Atlantic Ave. onto Shore Rd., the Carousel is on your left. The Carousel is the oldest one in the country. Continuing straight on Shore Rd., right before the intersection with Ocean View Highway, the Surf Shop is on your right. The Ocean is directly behind the Surf Shop. The Surf Shop rents surf boards and offers lessons in the afternoons. Across the street from the Surf Shop, on your left as you turn from Shore Rd. onto Ocean View Highway, is the Clam Shack. The Clam Shack won an award for the best clam fritters in the nation. After passing the Clam Shack on Ocean View Highway, Captain Kidd s Way is on your left. Captain Kidd s Way is a dead-end street. Survey Perspective (Southward: Beginning in North) Avondale is a typical small beach town. The town s population triples in the summer when tourists come. Avondale and the surrounding areas are bordered by four major landmarks: the Pawcatuck River, Atlantic Ave., Ocean View Highway, and the Ocean. The Pawcatuck River makes up the northern border. Running northsouth along the western border of this region is Atlantic Ave. The eastern border is made up of Ocean View Highway. The ocean forms the southern border of the town, and Shore Rd. runs parallel to the coast. Shore Rd. connects Atlantic Ave. and Ocean View Highway. Most of the town lies in the south by the ocean, along Shore Rd. Shore Rd. used to be gravel but was recently paved. There are often pedestrians waiting to cross the street, so cars must drive carefully. On the northern side of Shore Rd, just east of its intersection with Atlantic Ave, is the Carousel. The Carousel is the oldest one in the country. South of Shore Rd., near its intersection with Ocean View Highway, is the Surf Shop. The Ocean is directly south of the Surf Shop. The Surf Shop rents surfboards and offers lessons in the afternoons. North of the Surf Shop, on the other side of Shore Rd., is the Clam Shack. The Clam Shack won an award for the best clam fritters in the nation. Captain Kidd s Way is a dead-end street just north of the Clam Shack. Captain Kidd s Way comes off of Ocean View Highway.

28 28 S. A. Gagnon et al. / Cognitive Science (2013) Appendix C Experiment 1: Sample inference statements from the route and survey perspective, corresponding to a south-heading description of Avondale (Appendix A) Route Perspective Statements 1. From Atlantic Ave., take a right onto Shore Road to reach the Carousel (false) 2. If the ocean is on your left, Atlantic Ave. is in front of you (true) 3. After turning onto Ocean View Highway from Shore Rd., the clam shack is on your right (false) Survey Perspective Statements 1. Shore Rd. heads east from Atlantic Ave (true) 2. The Carousel is south of Shore Rd (false) 3. The Clam Shack is north of the Surf Shop and west of Ocean View Highway (true) Appendix D Experiment 2: Sample survey layout of one of the virtual environments, with the entrance hallway shown in white with a red arrow (left), and view from starting location in the entrance hallway, depicting a north-oriented compass (right)