Individual Differences in the Encoding Processes of Egocentric and Allocentric Survey Knowledge

Authors


should be sent to Wen Wen, Department of Psychology, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan. E-mail: wen@l.u-tokyo.ac.jp

Abstract

This study examined how different components of working memory are involved in the acquisition of egocentric and allocentric survey knowledge by people with a good and poor sense of direction (SOD). We employed a dual-task method and asked participants to learn routes from videos with verbal, visual, and spatial interference tasks and without any interference. Results showed that people with a good SOD encoded and integrated knowledge about landmarks and routes into egocentric survey knowledge in verbal and spatial working memory, which is then transformed into allocentric survey knowledge with the support of all three components, distances being processed in verbal and spatial working memory and directions in visual and spatial working memory. In contrast, people with a poor SOD relied on verbal working memory and lacked spatial processing, thus failing to acquire accurate survey knowledge. Based on the results, a possible model for explaining individual differences in spatial knowledge acquisition is proposed.

1. Introduction

Spatial knowledge acquired from the environment is fundamental to human lives and has been extensively studied from a variety of viewpoints. One of the issues that has consistently attracted researchers’ interest concerns the types of knowledge about large-scale spaces (Siegel & White, 1975). Landmark knowledge is knowledge about discrete objects or scenes; route knowledge consists of sequences of landmarks and actions; and survey knowledge is configurational, map-like knowledge. Importantly, in survey knowledge, the layout of the environment is represented, with separate landmarks or routes being integrated with each other. Thus, the acquisition of survey knowledge is considered an elaborate step in the microgenesis of spatial knowledge (Montello, 1998).

In discussing the structure of mentally represented environments, the issue of spatial frames of reference bears particular importance, because the acquisition of survey knowledge requires separately acquired information being integrated in a common frame of reference. Egocentric frames of reference specify locations with respect to the body (e.g., front-back-left-right), and allocentric frames of reference define spatial relations with respect to external objects or cardinal directions (Klatzky, 1998). The important point here is whether the representations are tied to the viewer’s position and orientation. That is, although survey knowledge can be represented either egocentrically or allocentrically, the latter is more flexible or sophisticated. Sholl (1996) described this transition as the knowledge development from self-to-object relations to object-to-object relations. Recognizing the importance of these issues for human spatial cognition, this research aims to examine how people with varied spatial aptitudes differ in the understanding and processing of the two types of survey relations from the horizontal, traveler’s perspective.

We note that this type of “horizontal” learning differs from “vertical” learning in important ways. For example, in map learning, the layout of places can be seen from above and allocentric relations may be acquired first, and then transformed into egocentric relations when necessary (e.g., Thorndyke & Hayes-Roth, 1982). Although not the focus of this research, the relationship between map learning and working memory is an important research topic (e.g., Coluccia, 2008; Coluccia, Bosco, & Brandimonte, 2007; Meilinger & Knauff, 2008).

Past research has shown that people have difficulties in estimating allocentric directions and distances (Avraamides, Loomis, Klatzky, & Golledge, 2004; McNamara, Rump, & Werner, 2003; Ruggiero & Iachini, 2006) and that there are large individual differences in the structure and processes of mental representations of the environment in relation to the environmental-learning aptitude called sense of direction (SOD; Kozlowski & Bryant, 1977). People with a good SOD tend to do better on tasks that require configurational understanding of environments (Hegarty, Richardson, Montello, Lovelace, & Subbiah, 2002; Ishikawa & Montello, 2006) and to use allocentric frames of reference to describe environments (e.g., Hund & Padgitt, 2010; MacFadden, Elias, & Saucier, 2003).

To examine the reason for these differences, it should be insightful to look into the encoding processes of information from the environment, on which this research is focused. People with a good and poor SOD may differ in the processing of information in working memory (e.g., Baddeley & Hitch, 1974; Baeyens & Bruyer, 1999; Courtney, Ungerleider, Keil, & Haxby, 1996), which in turn may affect the nature of acquired mental representations. Baddeley (2003) argued that although visuospatial working memory seems to be important for spatial orientation and geographic knowledge, there is still little work on this topic.

Concerning the relationships between environmental learning and working memory, Garden, Cornoldi, and Logie (2002) and Meilinger, Knauff, and Bülthoff (2008) found that verbal and spatial working memory are involved in the acquisition of landmark and route knowledge. Because these studies lacked the consideration of survey knowledge and visual working memory, Wen, Ishikawa, and Sato (2011) examined the involvement of three different components of working memory in the acquisition of three types of spatial knowledge, in relation to participants’ SOD. They showed that people with a good SOD encoded landmarks and routes verbally and spatially, and integrated knowledge about them into survey knowledge with the support of all three components of working memory. In contrast, people with a poor SOD encoded landmarks only verbally and tended to rely on the visual component of working memory in the processing of route knowledge, failing to acquire survey knowledge.

In the Wen et al. (2011) study, however, survey knowledge was not examined through detailed configurational measures (only sequential or “topological” accuracy of sketch maps was examined), nor was it examined through separate analyses of egocentric and allocentric survey relations. Furthermore, the role of the visual component of working memory remains to be clarified. These issues led us to conduct the present research. Also, in the past studies of the effect of verbal interference (Pouliot & Gagnon, 2005) and visual impairments (e.g., Rieser, Hill, Talor, Bradfield, & Rosen, 1992; Thinus-Blanc & Gaunet, 1997), the roles of the three components of working memory have not been examined separately, nor has an understanding of egocentric and allocentric survey relations been studied from an individual difference perspective in detail.

On the basis of these research backgrounds, we examined how the verbal, visual, and spatial components of working memory are involved in the acquisition of egocentric and allocentric survey knowledge by people with a good and poor SOD. Here, egocentric survey knowledge refers to the understanding of self-to-object relations tied to a specific viewpoint, and allocentric survey knowledge refers to the understanding of object-to-object relations. We used a dual-task method, in which participants learned routes with and without a concurrent task and then were tested on their understanding of egocentric and allocentric survey relations, so that we could examine which components of working memory are involved in survey learning.

We constructed several hypotheses to be tested in this research. On the basis of past studies of spatial microgenesis, reference frames, and SOD, we hypothesized (a) that people would do better on egocentric than on allocentric survey tasks; (b) that people with a better SOD would do better on both egocentric and allocentric survey tasks; and (c) that people with a good SOD would acquire at least a certain level of allocentric survey knowledge, whereas people with a poor SOD would fail to acquire accurate survey knowledge, either egocentric or allocentric.

On the basis of past studies about spatial learning and working memory, we hypothesized (d) that people with a poor SOD would have difficulties with spatial processing, and thus rely on other components of working memory; (e) that people with a good SOD, in contrast, would process information about the environment spatially; and (f) that the involvement of visual working memory would be greater for the acquisition of allocentric than egocentric survey knowledge, as the former might be facilitated by mentally “envisioning” the whole layout.

2. Method

2.1. Participants

Thirty-two college students (18 men and 14 women) participated in the experiment. They were Chinese students at the University of Tokyo and received monetary compensation in return for their participation. We confirmed that they had no prior experience with the study area.

2.2. Materials

2.2.1. Route videos

Views along four routes in northeastern wards of Tokyo, videotaped from a car, were shown to participants as experimental stimuli. A video camera with a wide conversion lens was set on a tripod fixed onto the passenger seat, with the horizontal visual angle 65° and the vertical 46°. The videos were 4 min long each, and scenes for temporary stops were edited out. The four routes were 1,270, 1,301, 1,370, and 1,286 m in length, respectively, and contained five turns each (Figure 1).

Figure 1.

 Maps of the study routes. “S” on the maps denotes the starting point and “E” the endpoint for each route. Solid circles indicate the five landmarks pointed out to participants on each route.

For each route, five landmarks were indicated to participants; at each landmark, the video paused for 3 s and an arrow pointing to the landmark and its label were shown at the top of the screen (Figure 2). The buildings or places used for these landmarks were noticeable but very common in typical cities (e.g., a park, post office, or chain store), and the videotaped routes were in infrequently visited areas. Thus, participants could not identify or guess the areas from the landmarks.

Figure 2.

 A snapshot view at a landmark (gas station) from one route. Such snapshots were used in the direction- and distance-estimation tasks to specify the participant’s position and orientation.

2.2.2. Concurrent interference tasks

Participants learned three routes with a concurrent interference task and one route without any interference. We used three types of interference tasks (verbal, visual, and spatial tasks), by modifying the tasks used in the Meilinger et al. (2008) study. Interstimulus intervals of these interference tasks were set a priori based on results from the practice phase (details below).

In the verbal interference condition, participants responded orally whether a combination of two syllables is a Chinese word. We used 310 frequent Chinese words and 310 non-words, which were read by a 24-year-old female Chinese student speaking native Mandarin. Mean length was 701.1 ms (SD = 2.4) for words and 700.6 ms (SD = 1.8) for non-words, with no significant difference between the two.

In the visual interference condition, participants were given a time (e.g., 10:35) orally (read by the same student who read the verbal stimuli) and asked to imagine a clock with watch hands and answer orally whether the long and short watch hands pointed to the same upper or lower half of the clock face. The times were randomly generated from combinations of 1–12 h and 5–55 min, except for 3 and 9 h and 15 and 45 min (which are on the dividing line of upper and lower halves). Meilinger et al. (2008) discussed that although this task might not load on an isolated system, it placed much more load on the visual than on the spatial subsystem of visuospatial working memory. We also confirmed in a preliminary experiment that this task, when conducted concurrently, disrupted visual figure detection, but not spatial updating (or path integration).

In the spatial interference condition, participants heard a beep sound, transmitted randomly from one of three loudspeakers located 90 cm away to the left, to the right, and in front of the participant’s seat. Participants orally indicated the direction from which the sound came.

2.2.3. Observed variables

We observed participants’ performance on direction-estimation, distance-estimation, and map-sketching tasks. No time limit was imposed on any task. The major objective of our analysis is to examine whether participants’ performance on these tasks would decrease when verbal, visual, or spatial concurrent tasks are conducted, to see which of the three components of working memory are involved in the processing of these types of knowledge.

For the direction-estimation task, participants were shown snapshot views at locations where they learned landmarks in the videos (Figure 2). They were asked to imagine standing at the position and orientation in which the snapshots were taken, and to estimate the direction to another unseen landmark (egocentric direction) and the direction between two other unseen landmarks (allocentric direction). To answer egocentric directions, participants drew a line on a circle (12 cm in diameter) printed on A4-size paper, with the center representing the current position and an upward line from the center aligned with the current orientation. To answer allocentric directions, participants drew on the same page a line parallel to the interlandmark direction (a method adapted from Avraamides et al., 2004). For each route, five egocentric directions and five allocentric directions were estimated, each in random order. These 10 pairs of landmarks were selected randomly, with the restriction that the landmarks used for allocentric directions did not serve as the origin for egocentric directions.

After estimating the directions, participants were asked to estimate the straight-line distance to another unseen landmark (egocentric distance) and the straight-line distance between two other unseen landmarks (allocentrtic distance), both in meters. In the map-sketching task, participants drew maps of the learned routes in as much detail as possible, by specifying the locations of the five landmarks, on blank A4-size paper.

2.2.4. Attentional-demand control

Before the learning phase, we included a control phase to adjust attentional demands of the three interference tasks (Fernandes & Moscovitch, 2002). In this phase, we asked participants to conduct the verbal, visual, and spatial interference tasks with a continuous-reaction task and measured reaction times on the continuous-reaction task. In the continuous-reaction task, participants indicated in which of the four boxes presented horizontally on screen a small square appeared, by pressing one of the keys correspondingly arranged on a keyboard, as quickly and accurately as possible. This phase was repeated several times (three on average), and the interstimulus intervals of the three interference tasks were adjusted so that the reaction times on the continuous-reaction task reached the same level.

2.3. Procedure

Participants were tested individually, seated on a chair positioned 120 cm away from a 55-inch liquid crystal television. First, participants practiced the three interference tasks and the continuous-reaction task, and then began the attentional-demand control phase. After the control phase, participants viewed a practice video and the experimental tasks were explained. They were instructed to study the routes by paying as much attention as possible, so that they could answer questions about the routes to be asked later. Participants practiced estimating egocentric and allocentric directions and distances, using objects in the experimental room for several times, so that we could make sure that they understood the tasks fully. After practice, participants learned the first video with or without a concurrent task, and then conducted the experimental tasks with no concurrent tasks (the direction-estimation, distance-estimation, and map-sketching tasks in this order). After finishing the tasks for the first route, participants took a short break (varying in length from 1 to 5 min, according to participants’ requests) and proceeded to the second route. They repeated this process for a total of four routes (they viewed the video once for each route). The allocation of the four conditions of the learning phase (the verbal, visual, and spatial interference conditions and the control condition) to the four routes and the order of viewing the four routes were randomized across participants.

After finishing all tasks, participants filled out the Santa Barbara Sense-of-Direction (SBSOD) scale (Hegarty et al., 2002). It consists of fifteen 7-point Likert-type items about spatial and navigational abilities, preferences, and experiences, and is scored so that a higher score indicates a better SOD. One additional item, “I usually envision a two-dimensional map in mind when learning a novel route,” was also asked on a 7-point scale, so that the tendency to use imagery could be assessed (this additional item was not considered when computing participants’ SOD scores). The experiment took 90 min on an average.

3. Results

3.1. Attentional-demand control

There was no significant difference in participants’ response times on the continuous-reaction task for the three interference conditions, showing that the attentional demands of the three interference tasks were controlled to an equivalent level (Table 1). Also, the percentages of correct responses in the three interference tasks did not differ between the attentional-demand control phase and the learning phase, indicating comparable divided attention during the two phases.

Table 1. Performance on the continuous-reaction task and the interference tasks
 Interference Task
VerbalVisualSpatial
Interstimulus interval (ms)1,3075,3311,328
Response time on the continuous-reaction task (ms)670692674
% Correct on the interference task in the control phase918899
% Correct on the interference task in the learning phase918998

In the face of comparatively high performance on the spatial interference task among the three, we examined whether there was a trade-off between conducting the spatial interference task and learning the routes. The correlations between participants’ performance on the spatial interference task and their memory for the routes were not significant, indicating no trade-off. There was no indication of a trade-off for the verbal and visual interference tasks either.

3.2. Sense of direction

On the basis of their scores on the SBSOD scale, we divided participants into good- and poor-SOD groups through a median split (Mdn = 4.3). The good-SOD group (= 16) had a mean score of 5.2, and the poor-SOD group (= 16) had a mean of 3.0.1

3.3. Direction estimates

We examined absolute angular errors between estimated and actual directions, and compared performance in the three interference conditions to that in the control condition (Table 2). An alpha level of 0.05 was used for all statistical tests.

Table 2. Mean directional errors (and standard deviations) for the good- and poor-SOD groups
GroupLearning Condition
ControlVerbalVisualSpatial
  1. Note: Means for the three interference conditions were compared with those for the control condition through paired t tests. Performances at chance level are underlined.

  2. SOD, sense of direction.

  3. *< .05. **< .01.

Egocentric
 Good-SOD47.8 (25.0)76.3 (30.0)**64.8 (33.4)70.0 (28.0)**
 Poor-SOD66.7 (29.4)74.0 (28.7)72.8 (22.8)76.3 (23.3)
Allocentric
 Good-SOD32.8 (11.0) 42.2 (11.7) 42.8 (13.3)* 46.2 (11.5)**
 Poor-SOD 42.7 (14.4) 42.3 (11.3) 43.5 (12.2) 43.8 (10.5)

For the good-SOD group, performance decreased significantly in the verbal and spatial interference conditions for egocentric directions, ts(15) = 3.40 and 3.18, respectively, ps < .01; and in the visual and spatial interference conditions for allocentric directions, t(15) = 2.22, < .05, and t(15) = 3.36, < .01. For the poor-SOD group, there was no significant decrease in performance in any interference condition, for either egocentric or allocentric directions.

In the control condition, the good-SOD group did better than the poor-SOD group for allocentric directions, t(30) = −2.24, < .05. And for both the good- and poor-SOD groups, performance was better for egocentric directions than for allocentric directions in the control condition, ts(15) = −2.66 and −2.69, respectively, ps < .05.

For egocentric directions, participants’ performance was above chance (90° for egocentric directions and 45° for allocentric directions) in all interference conditions and the control condition for both the good- and poor-SOD groups. But for allocentric directions, performances were not significantly different from chance, except the good-SOD group’s performance in the control condition.

3.4. Distance estimates

We examined the correlation between estimated and actual distances, and compared performance in the three interference conditions to that in the control condition, with Fisher’s r-to-z transformation (Table 3).

Table 3. Mean distance correlations (and standard deviations) for the good- and poor-SOD groups
GroupLearning Condition
ControlVerbalVisualSpatial
  1. Note: Means for the three interference conditions were compared with those for the control condition. Performances at chance level are underlined.

  2. SOD, sense of direction.

  3. *< .05. **< .01.

Egocentric
 Good-SOD0.66 (0.21) 0.16 (0.49)*0.55 (0.45) 0.13 (0.59)**
 Poor-SOD0.29 (0.37) 0.02 (0.51)0.31 (0.51)0.29 (0.41)
Allocentric
 Good-SOD0.43 (0.37) 0.04 (0.50)*0.38 (0.43) 0.05 (0.54)*
 Poor-SOD 0.14 (0.41) −0.04 (0.48)0.37 (0.54) 0.23 (0.53)

For the good-SOD group, performance decreased significantly in the verbal and spatial interference conditions for both egocentric distances, t(15) = −2.63, < .05, and t(15) = −3.86, < .01, respectively, and allocentric distances, ts(15) = −2.43 and −2.26, ps < .05. For the poor-SOD group, there was no significant decrease in performance in any interference condition, for either egocentric or allocentric distances.

In the control condition, the good-SOD group did better than the poor-SOD group for both egocentric and allocentric distances, t(30) = 3.24, < .01, and t(30) = 2.25, < .05, respectively. For the good-SOD group, performance was better for egocentric distances than for allocentric distances in the control condition, t(15) = 2.67, < .05.

For egocentric distances, participants’ performance was at chance level (= .00) in the verbal and spatial interference conditions for the good-SOD group and in the verbal interference condition for the poor-SOD group. For allocentric distances, participants’ performance was at chance level in the verbal and spatial interference conditions for the good-SOD group and in the control and the verbal and spatial interference conditions for the poor-SOD group.

3.5. Sketch maps

We examined the configurational accuracy of sketch maps through bidimensional regression (Tobler, 1965), by identifying for each map the start and end points and the five landmarks as “anchors.” We compared performance in the three interference conditions to that in the control condition, with Fisher’s r-to-z transformation (Table 4).

Table 4. Mean bidimensional correlations (and standard deviations) for the good- and poor-SOD groups
GroupLearning Condition
ControlVerbalVisualSpatial
  1. Note: Means for the three interference conditions were compared with those for the control condition.

  2. SOD, sense of direction.

  3. < .10. **< .01.

Good-SOD0.90 (0.07)0.65 (0.24)**0.81 (0.16)†0.71 (0.26)**
Poor-SOD0.78 (0.22)0.53 (0.21)**0.77 (0.18)0.72 (0.22)

For the good-SOD group, performance decreased significantly in the verbal and spatial interference conditions, ts(15) = −3.91 and −3.07, respectively, ps < .01 (and a tendency of decrease was found in the visual interference condition, t(15) = −1.85, < .10). For the poor-SOD group, performance decreased significantly in the verbal interference condition, t(15) = −4.91, < .01. In the control condition, the good-SOD group tended to do better than the poor-SOD group, t(30) = 1.74, < .10.

We also examined the proportion of correct turns drawn on each participant’s sketch map (Table 5), as another measure of sketch-map accuracy (an angular transformation was applied for analysis). Similar to the results from bidimensional regression, for the good-SOD group, performance decreased significantly in all three interference conditions, t(15) = −4.41, < .01, t(15) = −2.77, < .05, and t(15) = −2.24, < .05, respectively. For the poor-SOD group, performance decreased significantly in the verbal interference condition, t(15) = −2.57, < .05 (and a tendency of decrease was found in the spatial interference condition, t(15) = −2.07, < .10). In the control condition, the good-SOD group did significantly better than the poor-SOD group, t(30) = 2.35, < .05.

Table 5. Proportions of correct turns on sketch maps for the good- and poor-SOD groups
GroupLearning Condition
ControlVerbalVisualSpatial
  1. Note: Means for the three interference conditions were compared with those for the control condition.

  2. SOD, sense of direction.

  3. < .10. *< .05. **< .01.

Good-SOD.86 (.23).45 (.36)**.55 (.46)*.55 (.39)*
Poor-SOD.63 (.35).36 (.32)*.63 (.35).43 (.37)†

For bidimensional correlation, we conducted comparisons with chance level using a Monte Carlo simulation, in which random numbers for the x- and y-coordinates for the seven “anchors” were generated 1,000 times and a mean bidimensional correlation was computed for each route (rs = .29, .37, .44, and .27 for the four routes in Figure 1; see Ishikawa & Montello, 2006, for details). For both good- and poor-SOD groups, performance on map sketching was better than chance in all conditions.

4. Discussion

This study examined how people with different levels of SOD differ in the use of separate components of working memory to encode survey knowledge, and the results (summarized in Table 6) show important differences between people with a good and poor SOD in the acquisition of egocentric and allocentric survey knowledge.

Table 6. Summary of the results concerning the involvement of different components of working memory for survey knowledge acquisition
GroupEgocentricAllocentric
DirectionDistanceDirectionDistance
  1. Note: SOD, sense of direction; Verb, processed in verbal working memory; Vis, visual; Sp, spatial.

Good-SODVerb, SpVerb, SpVis, SpVerb, Sp
Poor-SOD Verb(chance)(chance)

4.1. Egocentric and allocentric survey knowledge in a novel large-scale space

For both people with a good and poor SOD, estimation of egocentric directions and distances without interference was better than chance, showing at least a certain degree of sensitivity to survey properties of the environment in egocentric terms (although the performance by people with a poor SOD was worse than that by people with a good SOD, with direction errors of 66.8° vs. 47.8°, respectively, and distance correlations of 0.29 vs. 0.66). In contrast, for allocentric directions and distances, unlike people with a good SOD, performance by people with a poor SOD was at chance level without interference, showing a failure to acquire allocentric survey knowledge.

When egocentric and allocentric estimation by people with a good SOD was compared, the former was better for both directions and distances. This indicates that egocentric survey relations are easier to understand than allocentric ones even for people with a good SOD, in line with our hypothesis that environmental information is first encoded egocentrically and then transformed into allocentric representations.

As in prior studies (e.g., Ishikawa & Montello, 2006; Sholl, Kenny, & DellaPorta, 2006), the present study observed large individual differences in the acquisition of survey knowledge, in both egocentric and allocentric terms. For all tasks, people with a good SOD did better than people with a poor SOD without interference, showing that the ability to acquire survey knowledge, particularly in allocentric terms, is closely related to SOD. Our hypothesis was that this may stem from differences in the processing of egocentric and allocentric survey knowledge by people with a good and poor SOD, which we turn to in the next section.

4.2. Working memory in the acquisition of egocentric survey knowledge

For egocentric directions and distances, people with a good SOD did poorly when verbally and spatially disrupted, showing the involvement of verbal and spatial working memory in the acquisition of egocentric survey knowledge. And the deterioration of distance estimation to chance level with verbal and spatial interference implies a particularly strong involvement of these two components in distance estimation. Verbal processing may help abstracting the characteristics of landmarks and routes analytically, and spatial processing may help encoding their metric or relational properties. The past finding that verbal and spatial working memory were involved in the acquisition of landmark and route knowledge (Garden et al., 2002; Meilinger et al., 2008; Wen et al., 2011) seems to reflect the fact that these two types of knowledge are basically tied to egocentric views.

For people with a poor SOD, their performance on egocentric direction and distance estimation did not deteriorate in any interference condition. Without interference, however, their performance was as low as that by people with a good SOD when disrupted, suggesting a possible floor effect. And for egocentric distance estimation, performance by people with a poor SOD decreased to chance level with verbal interference, indicating a strong reliance on verbal working memory. As we hypothesized, due to the difficulty with spatial tasks by people with a poor SOD, they are prone to process knowledge of the environment in verbal, rather than spatial, working memory.

4.3. Working memory in the acquisition of allocentric survey knowledge

For people with a good SOD, allocentric distance estimation showed similar patterns to egocentric distance estimation: They did worse when verbally and spatially disrupted, showing the involvement of verbal and spatial working memory in the learning of allocentric distances. But for directions, their allocentric and egocentric direction estimation showed different patterns: Their performance decreased significantly when visually and spatially disrupted, showing the involvement of visual and spatial working memory in the learning of allocentric directions (in contrast with verbal and spatial working memory for egocentric directions).

These results show that all three components of working memory are involved in the understanding of allocentric survey relations, with some differences between directions and distances. The involvement of visuospatial (in particular, visual) working memory for allocentric directions is noteworthy—directions are first encoded egocentrically in verbal and spatial working memory, and then transformed into allocentric representations with the support of visual and spatial working memory. Visual working memory may play important roles in the transformation from egocentric to allocentric survey knowledge, as a working platform for integrating spatial relations as if the person mentally viewed the “map in the head,” as suggested by Wen et al. (2011).

This argument is supported by the results from map sketching. To draw accurate sketch maps, one needs to combine directions and distances and comprehend object-to-object relations beyond self-to-object relations, and thus the measure taps into allocentric survey knowledge, at least to a greater degree than egocentric survey knowledge. People with a good SOD did worse in all three interference conditions, suggesting that verbal, visual, and spatial working memory are involved in the understanding of allocentric survey relations. Contrasted to the aforementioned finding that egocentric survey knowledge was processed in verbal and spatial working memory, the involvement of visual working memory is particularly notable. In fact, people with a good SOD tended to respond positively to the additional SOD item “I usually envision a two-dimensional map in mind when learning a novel route” (participants’ responses to this item correlated 0.68 (< .05) with the mean SBSOD score).

Visual working memory, however, did not play a role in distance estimation, either egocentric or allocentric. Similar to the past finding that directions and distances are processed independently in working memory (Chieffi & Allport, 1997), allocentric directions and distances may be processed differently. A possible explanation for the lack of visual processing for allocentric distances is that distances might be recognized rather qualitatively (or categorically, such as “near-far”); but it needs further investigation as directions may also be recognized qualitatively (e.g., Frank, 1996).

For people with a poor SOD, their estimation of allocentric directions and distances was at chance level, showing that they failed to acquire allocentric spatial knowledge. As discussed above for egocentric distances, there is a strong involvement of verbal working memory for people with a poor SOD, in contrast with spatial processing by people with a good SOD. Similarly, Wen et al. (2011) showed that people with a poor SOD encoded landmark knowledge only verbally and had difficulties in integrating landmarks and routes into survey knowledge. These findings indicate that the lack of spatial encoding can be one of the reasons for the failure of acquiring survey knowledge by people with a poor SOD.

4.4. Model for the processing of spatial knowledge

On the basis of these results, we propose a model for the processing of egocentric and allocentric survey knowledge for people with a good and poor SOD (Figure 3), which extends the model of landmark, route, and survey knowledge acquisition proposed by Wen et al. (2011). We note as a caveat in interpretation that spatial knowledge, particularly survey representations, could be stored in long-term memory or constructed from separately stored pieces of information in working memory when necessary. It is rather difficult to disentangle between the two processes or possibilities from behavioral data only (e.g., Liben, 1981; Montello, 1992).

Figure 3.

 Model of spatial knowledge acquisition. Black arrows indicate the encoding processes for three types of spatial knowledge; white arrows indicate the integration of landmarks and routes and the transformation from egocentric to allocentric survey knowledge. For people with a good sense of direction (SOD), landmarks and routes are processed in verbal and spatial working memory, and then integrated into egocentric survey knowledge in these two components. Egocentric survey representations are transformed into allocentric survey representations with the support of all three components of working memory, especially the visual component. People with a poor SOD, albeit with a certain degree of sensitivity to egocentric survey relations, fail to acquire accurate survey knowledge. (Adapted from Wen et al., 2011.)

Wen et al.’s (2011) model posits that people with a good SOD process information about landmarks and routes in verbal and spatial working memory; in contrast, people with a poor SOD process information about landmarks only in verbal working memory and process information about routes in visual, as well as verbal and spatial, working memory. To this model, our results add specifications about the encoding of survey knowledge, by distinguishing between direction and distance knowledge and between egocentric and allocentric survey relations.

For people with a good SOD, egocentric survey knowledge is processed in verbal and spatial working memory, for both directions and distances, which accords with the verbal and spatial processing of landmarks and routes. Egocentric survey knowledge is then transformed into allocentric survey knowledge with the support of all three components of working memory, with distances processed in verbal and spatial working memory and directions in visual and spatial working memory.

In Baddeley’s (2000) new model of working memory, the episodic buffer is assumed to be capable of integrating information from the subsidiary systems. In relation to our model, the episodic buffer seems to contribute to integrating different formats of spatial knowledge (e.g., verbally and spatially encoded knowledge about landmarks). This could be an important issue for further research.

For people with a poor SOD, they tend to rely on the verbal component of working memory compared with the visuospatial components; and because of the lack of spatial processing of landmarks and routes, they have difficulties in acquiring survey knowledge, especially understanding allocentric survey relations. Although they show a certain degree of sensitivity to egocentric survey relations, their understanding is not particularly accurate.

This model also accounts for the results from past studies. Garden et al. (2002) found that the visuospatial sketchpad was involved in route learning by survey-type participants; our model explains it more specifically by the involvement of the spatial subcomponent for people with a good SOD. Meilinger et al. (2008) found that spatial and verbal working memory were involved in route learning; in our model, it may be explained that their participants had a relatively good SOD and their route-following task tapped into landmark and route knowledge, not survey knowledge. Concerning the separate but important issue of map learning, visuospatial processing (Coluccia, 2008; Coluccia et al., 2007) and verbal processing (Meilinger & Knauff, 2008) were suggested, which deserves further investigation.

Finally, this model potentially stimulates further interdisciplinary research. For example, from an instructional perspective, it could foster a discussion about the possibility of training people to process spatial information effectively, particularly leading people with a poor SOD to employ spatial processing. From a computational perspective, it could offer methods for simulating spatial learning by people with different spatial capabilities. Albeit with questions for further research such as the effects of different sensory modalities (e.g., Klatzky, Loomis, Beall, Chance, & Golledge, 1998), the present model provides a theoretical discussion of individual differences in cognitive mapping and implications for helping people with difficulties in wayfinding and spatial orientation.

Footnotes

  • 1

    Hegarty, Montello, Richardson, Ishikawa, and Lovelace (2006) reported a mean score of 3.6 and a standard deviation of 1.0. Similar values for our participants (= 4.1 and SD = 1.3) indicate that the classification and labeling of good- and poor-SOD groups validly reflect their SOD. At the same time, the good-SOD group had four women and the poor-SOD group 10 women, suggesting a possible effect of gender, which deserves further investigation. Male-female comparisons yielded the same results as those for SOD comparisons, except that for allocentric distance estimation, men’s performance decreased only when verbally disrupted and women’s performance was above chance in the control condition.

Ancillary