Route Repetition and Route Reversal: Effects of Age and Encoding Method

Abstract

Previous research indicates age-related impairments in learning routes from a start location to a target destination. There is less research on age effects on the ability to reverse a learned path. The method used to learn routes may also influence performance. This study examined how encoding methods influence the ability of younger and older adults to recreate a route in a virtual reality environment in forward and reverse directions. Younger (n=50) and older (n=50) adults learned a route by either self-navigation through the virtual environment or through studying a map. At test, participants recreated the route in the forward and reverse directions. Older adults in the map study condition had greater difficulty learning the route in the forward direction compared to younger adults. Older adults who learned the route by self-navigation were less accurate in traversing the route in the reverse compared to forward direction after a delay. In contrast, for older adults who learned via map study there were no significant differences between forward and reverse directions. Results suggest that older adults may not as readily develop and retain a sufficiently flexible representation of the environment during self-navigation to support accurate route reversal. Thus, initially learning a route from a map may be more difficult for older adults, but may ultimately be beneficial in terms of better supporting the ability to return to a start location.

Age-related differences in multiple aspects of spatial navigation have long been documented in the research literature (for reviews, see: Lithfous, Dufour, & Després, 2013; Moffat, 2009). For example, the development and use of a flexible, allocentric representation, which involves encoding environmental information in terms of the spatial relationships amongst features (i.e., a cognitive map; O’Keefe & Nadel, 1978; Tolman, 1948), is negatively affected by advancing age (e.g., Head & Isom, 2010; Iaria, Palermo, Committeri, & Barton, 2009; Moffat, Elkins, & Resnick, 2006). Deficits are also observed in learning a specific route in the environment, which involves a series of stimulus-response associations as well as environmental information encoded from the perspective of the navigator (i.e., egocentric representation) (e.g., Cushman, Stein, & Duffy, 2008; Head & Isom, 2010; Moffat, Zonderman, & Resnick, 2001; Wilkniss, Jones, Korol, Gold, & Manning, 1997). Furthermore, older adults demonstrate difficulties in the acquisition and retrieval of critical environmental features that are components of these representations, including recognizing scenes, recalling landmarks, ordering landmarks along a route, and associating directional information with landmarks (e.g., Cushman et al., 2008; Head & Isom, 2010; Jansen, Schmelter, & Heil, 2010; Wilkniss et al., 1997). However, important aspects of route learning have not been sufficiently examined.

Most of the existing research on route learning has examined the ability of older adults to learn a specified route in one direction from a start location to a target location. However, it is equally important to understand the ability to return to the start location of a recently traveled route in a novel environment. This route reversal ability has rarely been examined in older adults. Reversing a path through the environment likely requires transformation of the forward route knowledge so that critical intersections can be correctly navigated from the opposite direction. Thus, it has been argued that route reversal may require an allocentric representation (Golledge, 1999; Wiener, Kmecova, & de Condappa, 2012). Wiener and colleagues (2012) observed greater age effects on aspects of route knowledge for the reverse direction than for the initial forward direction, which is consistent with observations of potentially greater age-related differences in allocentric compared to egocentric navigation (Begega et al., 2001; Moffat, 2009). However, this study did not involve actual self-navigation of the route, which may place greater demands on the use of route knowledge.

The method by which individuals encode new environments may also impact learning and retrieving a route in the forward and reverse directions. Direct navigation from a first-person perspective (i.e., self-navigation) and studying a map are two ways that individuals frequently learn a route through a new environment. Previous research indicates that these two methods may produce qualitatively different representations. For example, learning through self-navigation tends to result in greater access to orientation-specific knowledge, whereas studying a map appears to provide greater access to configural (i.e., survey-like) knowledge of the environment (e.g., Taylor, Naylor, & Chechile, 1999; Thorndyke & Hayes-Rothe, 1982; Zhang, Zherdeva, & Ekstrom, 2014). Furthermore, prior studies suggest that individuals who learn through self-navigation are better able to estimate route distances, whereas individuals who study maps perform better on tasks of Euclidean distances (Taylor et al., 1999; Thorndyke & Hayes-Rothe, 1982). These findings suggest that self-navigation may be sufficient to support learning a route in one direction. However, considering that a survey-like representation may afford greater flexibility in navigating, map study of a route may be particularly beneficial for route reversal.

Prior research demonstrates age-related impairments in the acquisition of route knowledge for both learning methods. More specifically, older adults acquire less environmental knowledge than younger adults when they learn a route in one direction from a first-person perspective (e.g., Cushman et al., 2008; Head & Isom, 2010; Jansen et al., 2010; Wilkniss et al., 1997). Similarly, older adults evidence deficits on tasks requiring them to traverse routes in one direction after studying a map (e.g., Carelli et al., 2011; Wilkniss et al., 1997). A recent study directly compared the influence of a first-person perspective to a survey perspective for learning a spatial layout (Yamamoto & DeGirolamo, 2012). Greater age-related differences were observed for indicating landmark locations on a 2D representation of the environment after encoding from a first-person perspective than after map study. As the authors note, this finding suggests that older adults may be less likely to derive an allocentric representation when experiencing an environment from a first-person perspective and/or have more difficulty translating from a first-person perspective to a survey perspective. Thus, when required to reverse a route, older adults may actually benefit from a survey view during encoding, as they might not derive an appropriately flexible representation to the same degree as younger adults from first-person, self-navigation.

The current study examined how two different encoding methods (i.e., self-navigation and map study) influence the ability of both younger and older adults to recreate routes in both the forward (i.e., route repetition) and reverse directions (i.e., route reversal). Performance on tasks of route learning in a virtual reality (VR) environment was examined. Participants initially learned a route from a designated start location to a target location by either map study or by self-navigation through the VR environment across multiple study-test trials. Participants were later tested on their ability to traverse the route in the forward and reverse directions. Consistent with previous findings (e.g., Carelli et al., 2011; Head & Isom, 2010; Wiener, et al., 2012; Wilkniss et al., 1997), we hypothesized age-related impairments in the ability to learn and remember routes in the forward direction. As an extension of recent research (Wiener et al., 2012), we expected that actual traversal of the route in reverse would evidence greater age-related impairments than traversal of the route in the forward direction.

In terms of encoding method, we hypothesized that there might be greater age-related impairments for learning the route in the forward direction via map study as this condition requires not only development of an allocentric representation, but also translation from a survey perspective to first-person navigation in the VR environment. The self-navigation learning condition should instead facilitate the development of a series of stimulus-response associations from an egocentric perspective relevant for traversing the route in the forward direction. Conversely, we hypothesized greater age-related deficits in route reversal for the self-navigation condition than the map study condition. Viewing the map during encoding may increase the likelihood of developing a more flexible representation of the intersections of the environment (Yamamoto & DeGirolamo, 2012), which should particularly facilitate route reversal. Lastly, we also examined landmark knowledge, including the temporal ordering of the landmarks as well as the directional information associated with each landmark. We expected that performance would be higher when ordering and providing directional information for the forward route than the reverse route, and that age effects would be greater for reversing this information. This would be consistent with similar findings using related tasks in a recent investigation (Wiener et al., 2012). We also explored the influence of encoding method on landmark knowledge. It is conceivable that the survey view of the layout of the environment afforded by map study better supports the ability to later reverse the order and directional information for landmarks.

Method

Participants

Younger adults (n=50) were undergraduates at Washington University in St. Louis, and received course credit for participation in this study. Older adults (n=50) were recruited from the Older Adults Volunteer Pool in the Psychology Department at Washington University and the Research Participant Registry at Washington University School of Medicine, and received monetary compensation for participation in this study. All participants were screened for major medical conditions (i.e., neurological conditions such as Parkinson’s or Huntington’s disease, stroke/TIA, seizure disorder, myocardial infarction, encephalitis, meningitis, uncontrolled hypertension or diabetes, head injury, and history of a mood disorder). Older adult participants were also screened for gross cognitive status using a cut-off score of <5 on the Short Blessed Test (Katzman et al., 1983). Participants provided written consent in accordance with the guidelines of the Washington University Human Research Protection Office.

Sample characteristics are presented in Table 1. Older adults had more years of education (p<.001) and higher scores on a composite measure of health conditions (p<.001), which was the sum of the presence of a heart problem, open heart surgery, treated hypertension, hypercholesterolemia, and treated diabetes (total: 0–5). Participants completed a measure of map experience prior to the experimental tasks, as well as measures of computer experience and vocabulary during a 15- minute delay (see below for details). The map experience questionnaire was administered prior to completing the experimental tasks. Participants indicated on a Likert scale (1 to 7) their experience with using a map to find their way around a new place (e.g., a new neighborhood) or a new building (e.g., a new museum), their ability to use a map to get around a new place or new building without major difficulty, and their level of comfort with using a map to find their way around a new place or new building. A map experience composite was created from the average of these responses (range: 1–7). The computer experience questionnaire was administered after the immediate route reversal phase (see below for details). Participants indicated on a Likert scale (0 to 7) experience with computers, computer games, and VR games. A computer experience composite was created from the average of these experiences (range: 0–7). The Shipley Institute of Living Verbal Meaning Test (Shipley, 1940) consists of 40 multiple-choice items in which participants must choose one of four listed words that is a synonym of the specified target word (range: 0–40). Older adults had significantly lower scores on the computer experience composite (p<.001), higher scores on the map experience composite (p=.023) and higher scores on the vocabulary test (p<.001). In addition, there was a non-significant trend for older adults to take longer to complete a visuomotor expertise test (p=.051), which was designed to assess the ability to efficiently use a joystick to maneuver through a virtual environment (see below for detailed task description). Finally, there were no significant differences across encoding method groups (all ps>.105), nor any significant age group by encoding method interactions (all ps>.124) for any of these variables.

Table 1.

Sample characteristics.

	Total Sample		Older Adults		Younger Adults
	Older Adults	Younger Adults	Self-Navigation	Map Study	Self-Navigation	Map Study
N	50	50	25	25	25	25
Gender (m/f)	19/31	18/32	12/13	7/18	10/15	8/17
Age (mean (SD))	71 (8)	20 (1)	71 (9)	70 (7)	19 (1)	20 (2)
Age range (yrs)	59–89	18–23	59–89	59–87	18–21	18–23
Education (mean (SD))*	16.7 (2.6)	13.3 (1.4)	17.0 (2.7)	16.5 (2.6)	12.9 (1.0)	13.7 (1.6)
Education range (yrs)	12–24	12–18	12–24	12–22	12–15	12–18
Health conditions composite*	.62 (1.0)	.02 (.1)	.72 (1.2)	.52 (.7)	.04 (.2)	.00 (.0)
Shipley Vocabulary (mean (SD))*	34.9 (3.5)	32.4 (3.1)	34.7 (3.6)	35.1 (3.4)	31.7 (2.8)	33.0 (3.3)
Computer experience composite (mean (SD))*	2.4 (1.2)	4.1 (1.3)	2.1 (1.1)	2.7 (1.2)	4.1 (1.4)	4.1 (1.2)
Map experience composite (mean (SD))*	5.5 (1.2)	5.0 (.8)	5.4 (1.4)	5.5 (1.1)	5.0 (.9)	5.0 (.8)
Visuomotor expertise time (mean (SD))*	55.5 (3.6)	54.0 (3.3)	56.0 (3.4)	54.9 (3.8)	54.7 (2.8)	53.3 (3.7)

^*Indicates a significant difference between younger and older adults at p<.05. There were no significant differences between any encoding groups.

Experimental procedures

The experimental session lasted 1.5–2 hours. Environments and tasks were adapted from past work (Allison, Fagan, Morris, & Head, 2016; Hartley, Maguire, Spiers, & Burgess, 2003; Head & Isom, 2010). Participants were randomly assigned to 1 of 2 conditions for learning a specific route: self-navigation (i.e., following arrows along a route several times in a VR environment) or map study (i.e., studying a 2D printed map of a designated route). WorldViz Vizard and Autodesk 3ds Max software were used to create a non-immersive desktop virtual maze environment that was presented on an Alienware laptop with a 17-inch monitor. The maze consisted of a series of interconnected hallways with 14 landmarks and four colored wallpaper patterns. The same maze environment and route were used across both learning conditions (see Supplemental Figure 1). The route through the environment consisted of 11 turns and was approximately 4700 virtual units long. The route took about 90 seconds to complete without errors. A joystick was used to maneuver through the environment. The computer recorded distance in virtual units and time.

Practice and visuomotor expertise

Participants completed a practice session and a visuomotor expertise test in a separate VR environment, which consisted of a long hallway with several bends. The experimenter provided instructions on joystick use and maneuvering through the environment during practice. For the visuomotor expertise test, participants traversed to a blue area at the end of the hallway within 65 seconds. Participants were given three attempts to complete the visuomotor expertise test within 65 seconds. All participants fulfilled this requirement. Following the visuomotor expertise task, participants did not receive any feedback unless they failed to reach the end location of the route within the time limit.

Learning phase

For the learning phase, participants completed multiple study-test trials. For study, participants either followed a specific route marked by arrows (i.e., self-navigation condition) in the VR environment for 5 minutes, or participants studied a 2D map with the route designated by a dashed line for 5 minutes (i.e., map study condition). The map was in color and provided the names of landmarks (see Supplemental Figure 1).

During test, participants in both conditions traversed the studied path one time without arrows in the VR environment. Participants were given a maximum of 5 minutes to complete each test trial. Participants were required to complete two consecutive study-test trials without errors prior to proceeding to the route reversal phase. A correct trial was defined as recreating the exact same route presented during study without making an incorrect turn. The number of correct test trials (0–4) was the dependent variable used in analyses. Participants who met the criterion of two consecutive correct trials in the first two trials were given a score of four. Participants who met criterion in the first three trials were given a score of three, and participants who met criterion by the fourth trial were given a score of two. Finally, participants who only had one correct trial had a score of one. Participants who were unable to meet the criterion by the end of the fourth study-test trial were assigned a score of zero, and still completed the remainder of the experimental session. During this learning phase, participants were unaware that they would be required to later reverse the route.

Immediate route reversal phase

Following the learning phase, participants were asked to traverse the route in the reverse direction from the end location to the start location in the VR environment across two consecutive trials. Participants were given a maximum of 5 minutes to complete each route reversal trial. There were two dependent variables for this phase. The first dependent variable was the average time to complete the two trials. The second dependent variable was the number of correct trials (i.e., 0, 1, or 2). A correct trial was defined as recreating the exact same route presented during study in the reverse direction without making an incorrect turn.

Delay phase

The delay phase occurred after a 15-minute delay. During the delay, participants completed the Shipley Institute of Living Verbal Meaning Test and the computer experience questionnaire. Participants traversed the learned path without arrows across two trials in the forward direction (i.e., route repetition). Next, participants traversed the path in the reverse direction without arrows across two trials (i.e., route reversal). Participants were given a maximum of 5 minutes to complete each trial during the delay phase. There were two dependent variables for this phase, which were calculated separately for the forward and reverse conditions: average time to complete the two trials and number of correct trials (i.e., 0, 1, or 2).

Supplementary tasks

Following a 5-minute delay, participants recalled as many landmarks as possible (Landmark Free Recall; total correct: 0–14). The experimenter clarified any responses that were unclear or ambiguous. For example, if a participant listed “gray bin” as an object from the maze, the experimenter asked the participant, “Can you describe this item in more detail?” to determine whether the object was from the maze (e.g., trash can) or if the object was a false alarm. Participants then completed the Temporal Order Memory task. Fourteen landmarks were presented at the top of the computer monitor with empty boxes on the bottom. Participants used the mouse to move landmarks into the empty boxes in the order in which the landmarks were encountered along the route during the study-test trials of the immediate phase. The index of performance was the Spearman rank-order correlation between the actual order and the participant’s order (range: 0–1). Next, participants completed the Landmark Direction Knowledge task in which they indicated whether each landmark (n=14) was associated with a right, left, or no turn (total correct: 0–14). Participants completed both the Temporal Order Memory and the Landmark Direction tasks separately for the forward and reverse directions. Order of completion was counterbalanced between the forward and reverse directions for the Temporal Order Memory and the Landmark Direction tasks across participants and separated by a 5-minute delay.

Data analysis

Covariates

Gender, visuomotor maze time, the computer experience composite, the map experience composite, and health conditions composite were all examined as potential covariates. Younger adults had significantly better scores on the computer experience composite and the health conditions composite, and there was a strong trend for visuomotor maze time. As a result, these variables were included as covariates in all between-age group analyses.

Outliers

Univariate outliers were defined as values ≥3 standard deviations from the mean for each age group. Results were the same when outliers were removed except when noted in the results section.

Excluded and imputed data points

Four older adult participants in the self-navigation condition were unable to complete one of the two trials during the immediate route reversal phase within 5 minutes, and 1 older adult in the self-navigation condition was unable to complete either of the trials within 5 minutes. Two of these older adult participants in the self-navigation condition were also unable to complete one of the delay route reversal trials within 5 minutes. As a result, these participants were excluded from the primary immediate route reversal and delay phase analyses, respectively. However, as exclusion of the data points may lead to an over-estimation of performance, exploratory analyses were also conducted including these data points with the time data set at 5 minutes. Results were the same with the data included except when noted in the results section.

Statistical analyses

All analyses were conducted using SPSS 21. Encoding method (map study; self-navigation) and age group were categorical predictor variables in all analyses. An ANCOVA was conducted to examine performance during the learning phase. For the study-test trials, the number of correct test trials in the forward direction was the dependent variable. For the immediate route reversal phase, separate ANCOVAs were conducted with average time to complete the two trials and the number of correct trials as dependent variables. For the delay phase, separate repeated measures ANCOVAs were conducted with average time to complete the two trials and the number of correct trials in the forward and reverse directions as dependent variables. Free recall was examined using ANCOVA with the number of correctly recalled landmarks as the dependent variable. Temporal Order Memory and Landmark Direction were both separately examined using repeated measures ANCOVAs. Performance during the forward and reverse trials was the dependent variables in both analyses.

Although all younger adults were able to recreate the route two consecutive times without any errors during the learning phase, four older adults from the self-navigation condition and 12 older adults from the map study condition did not recreate the route two consecutive times without any errors. Furthermore, both age-related and encoding-related differences were present during the learning phase (see Results section). Therefore, performance during the immediate reversal and delay phases was also examined with learning phase performance added as a covariate. Results were the same with learning phase as an additional covariate except when noted in the results section.

Results

Learning phase performance

There was a significant effect of age for the learning phase, such that older adults took a greater number of trials to learn the route (F(1,93)=13.181, p<.001; partial η²=.124, 95%CI: .026–.252). There was also a significant effect of encoding method, such that individuals who learned the route with a map had fewer correct trials (F(1,93)=6.582, p=.012; partial η²=.066, 95%CI: .003–.179). Importantly, these main effects were qualified by a significant age by encoding method interaction (F(1,93)=11.569, p=.001; partial η²=.111, 95%CI: .020–.236; see Figure 1A), such that age-related differences were greater for the map condition (F(1,45)=18.428, p<.001; partial η²=.291, 95%CI: .088–.468) than for the self-navigation condition (F<1).

Figure 1.

Age effects on learning and immediate route reversal performance. A) route forward performance; B) route reversal performance - time; C) route reversal performance - accuracy; Data represent estimated marginal means (controlling for covariates; see text for details), and error bars are standard error of the mean.

Immediate route reversal phase performance

Time

There was a significant effect of age on immediate route reversal performance (F(1,88)=14.085, p<.001; partial η²=.138, 95%CI: .031–.270; see Figure 1B), such that older adults took longer to recreate the route. There was not a significant effect of encoding method (F<1), nor a significant age by encoding method interaction (F(1,88)=1.298, p=.258; partial η²=.015, 95%CI: .000–.096).

Accuracy¹

There was a significant effect of age on immediate route reversal performance (F(1,88)=14.073, p<.001; partial η²=.138, 95%CI: .031–.271; see Figure 1C), such that older adults had fewer correct trials. There was not a significant effect of encoding method (F<1), nor a significant age by encoding method interaction (F<1).

Overall, older adults were slower and less accurate at immediately reversing the route. However, performance was not influenced by whether the route was learned through self-navigation or map study.

Delay phase performance

Time

There was a significant effect of age on delay phase performance (F(1,92)=10.438, p=.002; partial η²=.102, 95%CI: .015–.226), such that older adults took longer when recreating routes across the forward and reverse directions (see Figure 2A). There was also a significant effect of direction (F(1,92)=5.914, p=.017; partial η²=.060, 95%CI: .002–.172), such that individuals took longer to recreate the route in the reverse direction. There was not a significant effect of encoding method (F<1). Notably, there was a non-significant trend for an age by direction interaction (F(1,92)=3.822, p=.054; partial η²=.040, 95%CI: .000–.141), which did reach significance when controlling for learning performance (F(1,91)=4.049, p=.047; partial η²=.043, 95%CI: .000–.146). Thus, age-related differences were greater for the reverse direction (F(1,91)=8.684, p=.004; partial η²=.087, 95%CI: .009–.208) than for the forward direction (F(1,91)=1.715, p=.194; partial η²=.018, 95%CI: .000–.103). There was not a significant direction by encoding method interaction (F(1,92)=1.662, p=.201; partial η²=.018, 95%CI: .000–.101), nor a significant age by encoding method interaction (F(1,92)=1.036, p=.311; partial η²=.011, 95%CI: .000–.086). Finally, the three-way age by direction by encoding method interaction was not significant (F(1,92)=2.074, p=.153; partial η²=.022, 95%CI: .000–.110).

Figure 2.

Age effects on delay phase performance. A) route reversal performance - time; B) route reversal performance - accuracy. Data represent estimated marginal means (controlling for covariates; see text for details), and error bars are standard error of the mean. Younger adults were not included in statistical analyses due to ceiling effects (see text for details).

In the exploratory analyses that included imputed data for the participants who failed to reach the end location within five minutes, the three-way age by direction by encoding method interaction was significant with outliers removed (F(1,90)=4.073, p=.047; partial η²=.043, 95%CI: .000–.148). For younger adults, the direction by encoding method interaction was not significant (F(1,47)=1.165, p=.286; partial η²=.024, 95%CI: .000–.159). For older adults, there was a significant direction by encoding method interaction (F(1,46)=4.351, p=.043; partial η²=.086, 95%CI: .000–.258). Older adults who learned the route through self-navigation were slower in the reverse direction compared to the forward direction (F(1,22)=13.843, p=.001; partial η²=.386, 95%CI: .080–.595). In contrast, older adults who learned the route by studying a map did not significantly differ in their traversal time for the reverse and forward directions (F(1,24)=1.009, p=.325; partial η²=.040, 95%CI: .000–.254).

The three-way interaction was reduced to a non-significant trend when learning was additionally controlled (F(1,89)=3.632, p=.060; partial η²=.039, 95%CI: .000–.142). However, the same pattern of results held such that there was not a significant encoding by direction interaction for younger adults (F<1), but there was a significant encoding by direction interaction for the older adults (F(1,43)=4.382, p=.042; partial η²=.089, 95%CI: .000–.262). Again, this interaction for older adults reflected an effect of direction for the self-navigation condition (F(1,21)=13.341, p=.001; partial η²=.389, 95%CI: .076–.599), but not the map condition (F<1).

Collectively, the time data indicate that older adults were especially slower at reversing a route compared to repeating the route after a delay. In addition, the exploratory analyses provide tentative evidence that older adults were particularly slower at reversing a route after learning via self-navigation compared to learning the route through map study.

Accuracy²

Younger adults were not included in this analysis of accuracy because there was insufficient variability in their performance (i.e., n=49 with a perfect score for the forward direction and n=48 for the reverse direction). Among the older adults, there was not a significant effect of encoding method (F<1; see Figure 2B). However, the main effect of encoding method reached significance when controlling for learning performance (F(1,46)=6.127, p=.017; partial η²=.118, 95%CI: .003–.295), with lower accuracy for the self-navigation compared to the map study condition.

There was a significant effect of direction (F(1,47)=25.722, p<.001; partial η²=.354, 95%CI: .141–.518), such that older adults were less accurate for the reverse direction than for the forward direction. Importantly, there was a significant direction by encoding method interaction (F(1,47)=8.536, p=.005; partial η²=.154, 95%CI: .015–.334). Older adults who learned the route through self-navigation were less accurate in the reverse direction compared to the forward direction (F(1,23)=33.475, p<.001; partial η²=.593, 95%CI: .288–.735). In contrast, older adults who learned the route by studying a map did not significantly differ in their accuracy for the reverse and forward directions (F(1,24)=2.374, p=.136; partial η²=.090, 95%CI: .000–.324).

In summary, the accuracy results indicate that older adults were particularly less accurate at reversing a route after learning via self-navigation compared to learning the route through map study.

Supplementary tasks

Landmark Free Recall

There were not significant effects of age (F<1) or encoding method (F<1) on Landmark Free Recall performance. In addition, the age by encoding method interaction was not significant (F<1; see Figure 3A).

Figure 3.

Age effects on supplementary tasks. A) Landmark Free Recall; B) Temporal Order Memory; C) Landmark Direction. Data represent estimated marginal means (controlling for covariates; see text for details), and error bars are standard error of the mean.

Temporal Order Memory

There was a significant effect of age on Temporal Order Memory performance (F(1,93)=6.717, p=.011; partial η²=.067, 95%CI: .003–.181; see Figure 3B), such that older adults had greater difficulty sequentially ordering the landmarks. There was also a significant effect of encoding method (F(1,93)=6.060, p=.016; partial η²=.061, 95%CI: .002–.172), such that individuals who learned via map study had better performance on this task. There was not a significant effect of direction (F<1).

There was not a significant age by direction interaction (F<1), nor was there a significant age by encoding method interaction (F<1). There were non-significant trends for an encoding method by direction interaction (F(1,93)=3.224, p=.076; partial η²=.034, 95%CI: .000–.130) and a three-way age by direction by encoding method interaction (F(1,93)=3.439, p=.067; partial η²=.036, 95%CI: .000–.134). The three-way interaction became significant (F(1,92)=4.500, p=.037; partial η²=.047, 95%CI: .000–.152) when additionally controlling for learning phase performance. For younger adults, there was not a significant direction by encoding method interaction (F<1). For older adults, the direction by encoding method interaction was significant (F(1,47)=5.513, p=.023; partial η²=.105, 95%CI: .001–.279). For the forward direction, there was not a significant difference between those who initially learned the route through self-navigation and those who learned through map study (F(1,47)=1.022, p=.317; partial η²=.021, 95%CI: .000–.153). For the reverse direction, older adults who learned the route through self-navigation performed more poorly than those who learned through map study (F(1,47)=6.382, p=.015; partial η²=.120, 95%CI: .004–.296).

Landmark Direction

There was a significant effect of age on Landmark Direction performance (F(1,93)=8.157, p=.005; partial η²=.081, 95%CI: .007–.199; see Figure 3C), such that older adults had greater difficulty recalling the direction of travel associated with each landmark. There was also a significant effect of direction (F(1,93)=25.735, p<.001; partial η²=.217, 95%CI: .085–.350), such that individuals performed more poorly on this task when required to indicate the direction of travel for reversing the route. There was not a significant effect of encoding method (F(1,93)=1.711, p=.194; partial η²=.018, 95%CI: .000–.101).The age by direction (F<1), encoding method by direction (F(1,93)=1.255, p=.266; partial η²=.013, 95%CI: .000–.091), age by encoding method (F<1), and age by encoding method by direction (F<1) interactions were not significant.

In summary, older adults were able to freely recall a similar number of landmarks as younger adults. However, older adults were less able to retrieve the association between landmarks and the direction of travel. In addition, both age groups were significantly worse at indicating the travel direction for the reverse route than for the forward route, and this was not modulated by encoding method. In terms of landmark sequencing, there was a larger effect of reversal on sequencing landmarks for older adults who learned via self-navigation than for older adults in the map study condition, which is similar to the route traversal accuracy findings for the delay phase.

Post-Hoc Analyses

For the delay phase, there was evidence that older adults who learned through self-navigation had particular difficulty reversing the route. However, we did not observe a significant main effect of encoding method nor a significant age by encoding method interaction for immediate reversal. In order to better understand the discrepancy between the immediate reversal and delay phase results, we conducted a post-hoc analysis comparing accuracy of the immediate route reversal phase to that of the first two trials of the study phase (i.e., the forward direction), which all participants completed. Of particular interest was whether this analysis provided evidence consistent with the differential effects observed for the delay phase. In fact, there was a significant three-way age by direction by encoding method interaction (F(1,92)=4.678, p=.033; partial η²=.048, 95%CI: .000–.154; see Figure 4). For younger adults, the direction by encoding method interaction was not significant (F<1). For older adults, there was a significant direction by encoding method interaction (F(1,47)=6.212, p=.016; partial η²=.117, 95%CI: .003–.292). Older adults in the self-navigation condition were less accurate in immediately reversing the route compared to initially traversing the route in the forward direction (F(1,23)=7.881, p=.010; partial η²=.255, 95%CI: .017–.490). In contrast, older adults in the map study condition did not significantly differ in their accuracy in immediately reversing the route compared to initially traversing the route in the forward direction (F<1).

Figure 4.

Comparison between learning and immediate route reversal performance. Data represent estimated marginal means (controlling for covariates; see text for details), and error bars are standard error of the mean.

Examination of Figures 2B and 4 suggests that older adults’ performance in repeating the route in the forward direction improved from the initial two trials of the learning phase to the delay phase for the map study condition. To confirm this observation, we conducted a repeated measures ANOVA in the older adult group with accuracy for each phase as the dependent variables and encoding condition as a predictor variable. There was overall a significant increase in accuracy across phases (F(1,47)=18.713, p<.001; partial η²=.285, 95%CI: .087–.459) with a significant increase in accuracy for the map study condition (F(1,24)=20.928, p<.001; partial η²=.466, 95%CI: .157–.646) and a non-significant trend for the self-navigation condition (F(1,23)=3.286, p=.083; partial η²=.125, 95%CI: .000–.367), though the magnitude of improvement did not differ significantly between the encoding groups (F(1,47)=2.123, p=.152; partial η²=.043, 95%CI: .000–.193).

Lastly, examination of Figures 2B and 4 also suggests that older adults’ performance in reversing the route improved from the immediate route reversal to the delay phase for the map condition more so than for the self-navigation condition. To assess for this, we conducted another repeated measures ANOVA in the older adult group with accuracy for each of these phases as the dependent variables and encoding condition as a predictor variable. There was a non-significant trend for the phase by encoding method interaction (F(1,46)=3.930, p=.053; partial η²=.079, 95%CI: .000–.248) when learning performance was controlled, with a significant increase in accuracy for the map study group (F(1,23)=5.282, p=.031; partial η²=.187, 95%CI: .000–.430), but not the self-navigation group (F<1). The findings of these post-hoc analyses suggest that older adults in the self-navigation condition did have particular difficulty reversing the route during the immediate phase, and continued to have trouble with reversal during the delay phase. In contrast, older adults in the map study condition evidenced increasing accuracy for both the forward and reverse directions during both the immediate and delayed phases.

Discussion

The present study examined the influence of type of encoding method on traversing a novel route in forward and reverse directions in younger and older adults. Participants learned a route in one direction in a virtual maze by either studying a map or following arrows within the VR environment across multiple study-test trials. Next, participants recreated the route in the reverse direction. After a delay, participants again recreated the route in the forward and reverse directions. In addition, aspects of landmark knowledge were examined.

The observation of greater difficulty in learning a novel route for older adults in the map study condition may be related to requirements to acquire the route from a survey perspective, retrieve that route representation, and then translate that to a first-person perspective is particularly difficult for older adults. Although the current study was not designed to disambiguate the relative contributions of development and/or use of an allocentric representation versus mental spatial transformations in terms of age effects on learning from a map, previous research does suggest age-related impairments in these processes (Devlin & Wilson, 2010; Iaria et al., 2009; Joanisse et al., 2008; Yamamoto & DeGirolama, 2012).

The higher learning performance for the older adults in the self-navigation condition suggests that the development of a series of stimulus-response associations from an egocentric perspective is less problematic for older adults compared to the processes involved in the map study condition. However, the lack of any significant difference between age groups in learning the route via self-navigation is in contrast to past work indicating age effects when learning a route from a first-person perspective (e.g., Cushman et al., 2008; Head & Isom, 2010; Moffat et al., 2001; Wilkniss et al., 1997). The current study-test design, which has not typically been used in past work, may have particularly facilitated older adults’ performance. More specifically, the additional study trials subsequent to the initial study-test trial provided corrective feedback to participants, which would not have been present in studies employing one study of the environment followed by one test of route learning. However, this speculation requires further investigation.

The greater age-related difficulty observed in reversing the learned route after a delay may relate to a need for a flexible representation that is independent of the viewpoint in which the route was initially encoded. That is, the ability to successfully reverse a route is conceptualized to require such an allocentric representation (Golledge, 1999; Wiener et al., 2012). In contrast, repeating the route in the forward direction involves an egocentric representation and a fixed sequence of body turns in relation to environmental features. In fact, previous research does suggest greater age-related impairments for performance on tasks that assess the development and/or later use of a flexible, allocentric representation compared to egocentric tasks (e.g., Begega et al., 2001; Rodgers, Sindone, & Moffat, 2012; see Moffat, 2009 for a review). In addition, these behavioral findings are consistent with prior studies suggesting greater age-related effects on the hippocampus, which is associated with an allocentric representation, when compared to the caudate, which is more associated with an egocentric representation (Iaria, Petrides, Dagher, Pike, & Bohbot, 2003; McDonald & White, 1994; Morris, Garrud, Rawlins, & O’Keefe, 1982; Packard & McGaugh, 1996; Raz, Ghisletta, Rodrigue, Kennedy, & Lindenberger, 2010; Raz & Rodrigue, 2006; Walhovd et al., 2011). Additional neuroimaging research aimed at examining brain regions that may subserve the ability to successfully reverse a route are needed to better ascertain the neural substrates of relatively larger age-related deficits in route reversal ability.

Importantly, the lower reversal performance for older adults in the self-navigation condition compared to the map study condition is also consistent with the conceptualizations that route reversal benefits from an allocentric representation, and that older adults have more difficulty developing such a representation from a first-person perspective. That is, the differential effect of encoding conditions may relate to differences in the environmental knowledge that older adults acquired during learning. Prior investigations indicate that self-navigation and map study produce somewhat qualitatively distinct representations of the environment, with some differences in the type of environmental information acquired (e.g., Taylor et al., 1999; Thorndyke & Hayes-Rothe, 1982; Zhang et al., 2014). Self-navigation tends to lead to a more egocentric, viewpoint-dependent representation. Thus, older adults in the self-navigation condition may have had particular difficulty generating a more flexible representation of intersections of the environment and/or mentally transforming the fixed sequence of body turns that is relatively orientation specific. In contrast, map study tends to produce a more allocentric representation of the environment. Thus, older adults’ performance in route reversal after map study may have been particularly facilitated by the forced exposure to the spatial layout of the environment during encoding. This would be consistent with recent findings suggesting that map study benefits older adults’ retrieval of landmark locations in an environment relative to first-person encoding (Yamamoto & DeGirolamo, 2012). In fact, the post-hoc analyses comparing performance across the three phases of the study were consistent with the conceptualization that older adults in the map study condition were able to capitalize on their representation of the spatial layout of the environment to improve their performance in reversing the route. Instead, the less flexible, viewpoint-dependent representation of a fixed sequence of body turns in the self-navigation group could have continued to impede performance at reversing the route throughout.

Collectively, the route traversal phases of the current study provide information on how different encoding methods may influence older adults’ development of environmental representations and their resulting performance in learning and retrieving routes in forward and reverse directions. Self-navigation, and the associated egocentric, viewpoint-dependent representation, appears to be adequate for older adults to learn the route in one direction. However, older adults who initially learn a route via self-navigation are less able to later generate a sufficiently flexible, viewpoint-independent representation from the first-person perspective. Conversely, the multiple demands of map study may hinder initial learning of the route for older adults. Nonetheless, the exposure to the map may facilitate the development of a flexible, allocentric representation for older adults and thus their ability to traverse the route in reverse.

In addition to facilitating the ability of older adults to recreate the route in the reverse direction after a delay, there was also a benefit for older adults in the map study condition with regard to reverse landmark sequencing. The relative difficulty in the self-navigation condition is consistent with a recent finding of a non-significant trend for a greater age-related deficit in reverse landmark sequence knowledge with learning from a first-person perspective (Wiener et al., 2012). First-person self-navigation involves a sequential presentation in one direction during which there is a bias towards learning landmark order in the direction of the travel (Moar & Carleton, 1982; Janzen 2006; Schweizer, Herman, Janzen & Katz (1998). Notably, the lower performance in reverse sequencing for the self-navigation group may be associated with the greater difficulty in traversing the route in reverse for older adults in this condition. That is, lack of knowledge of the upcoming landmark could impede traversal of the route in reverse. In contrast, the relative benefit observed in reverse landmark sequencing for older adults in the map study condition could relate to the simultaneous presentation of the order of landmarks along the route on the map. Map study provided sequence information not only for Landmark A to Landmark B, but also more readily conveyed the reverse information about Landmark B to Landmark A. Thus, map study may not only facilitate acquisition of the spatial layout of the route, but also the temporal organization.

In contrast to findings for route traversal and landmark sequencing, older adults were not differentially impaired at indicating the reverse direction of travel at landmarks and performance was not modulated by encoding method. The lack of a differential age effect is in contrast to recent findings (Wiener et al., 2012). In that study, participants were required to indicate the correct direction to continue the route to either the start location or the end location at intersections, and older adults were particularly impaired on reversal trials. Moreover, all of the intersections involved three choices for continued travel, whereas in the current study some of the landmarks were at single direction turns or in the middle of hallways. Considering that older adults may have greater difficulty indicating directions associated with landmarks at decision points compared to non-decision points (Zhong & Moffat, 2016), the differences in the environments may have contributed to the discrepant findings across studies. In fact, older adults’ performance was generally higher for the current landmark direction task (68% for forward; 61% for reversal) compared to the intersection direction task in the recent report where reversal performance was noted to remain close to chance level (see Figure 3 in Weiner et al., 2012). Thus, future work could systematically vary the number of decision point landmarks to clarify how this influences age effects on the ability to associate directional information with landmarks in the forward and reverse route directions. The lack of an effect of encoding method for the landmark direction task could indicate that a survey perspective does not significantly enhance apprehension of the appropriate egocentrically-based body turns necessary at landmarks.

Within the context of these findings, it is important to note that the current study had several limitations. First, navigation performance was assessed using VR, which allowed for greater experimental control compared to natural environments. However, the extent to which the current findings may generalize to real world environments is currently unknown. Some prior research does indicate that navigation performance in a VR environment is similar to performance in the real world (e.g., Richardson, Montello, & Hegarty, 1999). Furthermore, although several studies have found higher performance in real world environments when compared to corresponding VR environments, these same studies demonstrated that the magnitude of age- and Alzheimer’s disease-related differences was similar across VR and real world environments (e.g., Kalova, Vlcek, Jarolimova, & Bures, 2005; Cushman et al., 2008; Taillade, N’Kaoua, & Sauzeon, 2015). Thus, there is evidence that deficits observed in VR environments correspond to deficits in the real world.

Relatedly, we observed age group differences in our computer experience composite, which included items on experience with video games, VR games and computers. Considering findings that greater video game experience is associated with enhanced performance on VR tasks (e.g., Murias, Kwok, Gil Castillejo, Liu, & Iaria, 2016), age-related differences in route learning performance in this study may in part be due to differences in computer/gaming experience. While the computer experience composite was included as covariate in all analyses, this may not have been sufficient to truly address age group differences (see Westfall & Yarkoni, 2016 for discussion of limitations of regression-based covariate control). However, the lack of an age group by encoding method interaction for the computer experience composite (p=.283) suggests that computer/gaming experience differences do not fully account for the observed pattern of age effects (e.g., greater age effects for self-navigation compared to map study during delay reversal). Overall, although differences in gaming/computer experience may have contributed to the findings in the current study and there may be limits on the degree of generalization to natural environments, there is evidence that we might expect to see similar age-related differences in real world environments. Nonetheless, future research examining the impact of age and encoding method on route repetition and reversal in a natural environment is needed.

Another potential limitation is that we did not assess the relative contributions of more basic cognitive processes, such as attentional capacity, processing speed or working memory, to age-related impairments in the ability to recreate routes in the forward and reverse directions. Prior work has demonstrated associations of processing and working memory with spatial navigation skill (e.g., Moffat et al., 2007). Furthermore, processing speed and working memory may account for a significant portion of age-related variance in other cognitive domains (e.g., Salthouse, 1990; Salthouse, 1996). Consequently, future studies would benefit from the incorporation of such measures to enhance understanding of the pattern of findings observed here.

Future investigations could also assess for the specific behavioral mechanisms of the age deficits in route reversal during self-navigation observed here, as well as explore ways for facilitating route reversal during this process. For example, older adults may fail to turn to view intersections from another perspective during the initial outward route as young adults do (Cornell et al., 1992; Heth et al., 2002), and thus may benefit from instructions to do so. In addition, repeated outward and then back journeys may be more typical of real-world navigation, and this process may increase the likelihood that older adults more readily retain a sufficient representation to increase accuracy in route reversal. Another avenue for future research would be to examine methods for further enhancing the benefits of map study. For example, older adults may benefit from having the map available throughout learning and retrieval phases, despite the additional translations between survey and first-person views that may contribute to their initial difficulty with map learning. Overall, more research is needed to better understand the cognitive factors that contribute to age-related difficulties in traversing routes as well as methods to improve learning routes in both the forward and reverse directions for older adults.

In summary, the current study extends an initial report on route reversal (Weiner et al., 2012) by demonstrating that older adults also have particularly difficulty traversing a learned route in reverse compared to repeating the route in the same direction. Furthermore, the comparison across encoding methods provides support for potential roles of temporal and allocentric processing in this age-related deficit in route reversal. Older adults may not as readily develop and retain a sufficiently flexible representation of the environment during self-navigation to support accurate route reversal. In comparison, learning a route using a map may be initially more effortful for older adults, but ultimately may be beneficial for retrieving landmark order information and finding the way back to the start location. Thus, map study appears to assist in apprehension and retention of the overall temporospatial organization of an environment for older adults, which may be particularly relevant for their route reversal. The present work serves to highlight that it is not possible to fully understand age effects on route learning without incorporating the return journey and considering the ways in which environmental information is acquired.

Supplementary Material

1

Supplemental Figure 1. Aerial perspective of the maze and route layout. The dashed line represents the route used for the route learning task.