«Stefan Münzer*, Hubert D. Zimmer*, Maximilian Schwalm*, Jörg Baus+ & Ilhan Aslan+ *Department of Psychology, Brain and Cognition Unit + German ...»
Computer Assisted Navigation and the Aquisition of Route and Survey Knowledge 13 In the survey knowledge test, subjects had to replace thumbnail pictures of the intersections onto a roadmap of the zoo. For evaluating the accuracy of survey knowledge, the deviations of the replacement from the correct place on the map were measured in pixels, and all deviation values were subsequently averaged. One subject had to be excluded from the data set due to missing data. Inspection of the descriptive results already showed that survey knowledge was quite poor in all experimental conditions in which participants had used navigation assistance. Deviation values of up to 200 pixels for the replacement of an intersection on a complete roadmap of the zoo that had been shown as a picture with 1024 x 768 pixels mean quite poor accuracy. In contrast, the mean deviation value in the map-based wayfinding condition was about 78 pixels which indicates a considerably better replacement performance. This impression was confirmed in an one-way between subjects analysis of variance with navigation condition as factor, which revealed a significant difference for the mean deviation in pixel, F (3, 59) = 8.31, MSE = 5462, p 0.001. A planned contrast showed that the map-based wayfinding condition differed significantly from the three navigation assistance conditions, F (1, 59) = 22.85, MSE = 5461.6, p 0.001).
These data might suggest that the difference between the map condition and the assistance conditions were larger for the survey knowledge than for the route recognition performance. A comparison of effect sizes using Cohen's d revealed that this was not true.
The size of the effect of the map condition compared with the mean of the assistance conditions was d = 1.347 for route recognition performance, and the same comparison revealed d = 1.513 for survey knowledge. Thus, the effect of the map condition was strong, and the effect did not differ considerably with respect to the dependent variable (route recognition vs. survey knowledge).
Computer Assisted Navigation and the Aquisition of Route and Survey Knowledge 14
4. Discussion The main hypothesis that motivated the present experiment was that using a navigation assistance system results in poor spatial memory, because subjects would incidentally learn only that information they have dealt with during the wayfinding activity. It was therefore expected that users of navigation assistance systems would acquire route knowledge, but their memory for spatial survey information would be poor. In contrast, it was hypothesized that map users would acquire both route and survey knowledge as a side effect of their wayfinding effort. The pattern of results supports these hypotheses. Map users acquired better survey knowledge as well as better route knowledge, compared to the knowledge acquired by assistance users. However, more specific hypotheses regarding the effects of presentation modes by a navigation assistance system on memory were not confirmed.
It has been expected that the acquisition of survey knowledge would be enhanced by a visual animation of the spatial context because an allocentric perspective is thereby presented in addition to the egocentric view. Due to results from explicit route learning, it has been furthermore assumed that visual presentation of direction information would be superior over an auditory presentation. None of these manipulations of the presentation format did influence learning in the real environment. All three assistance groups showed comparable results. Our explanation is that these manipulations were inefficient because the information provided was not “actively” processed by the pedestrians. The provided information was not needed for the primary goal of wayfinding. Since participants did not know that remembering would be relevant, they "passively" used the readily available direction information for navigation. In other words our attempts to enhance spatial information presentation did not overcome the less elaborated mode of processing of the navigation assistance users during their wayfinding activity. The observation of poor spatial orientation knowledge under these circumstances is in line with the view that spatial learning is an effortful and error-prone process (e.g.
Thorndyke & Hayes-Roth, 1982; Aginsky et al., 1997; Gillner & Mallot, 1998).
Computer Assisted Navigation and the Aquisition of Route and Survey Knowledge 15 Map users not only acquired better survey knowledge but they also acquired better route knowledge, which additionally supports an “active learning hypothesis”. We assume that this learning advantage comes from the additional active effort that map users invested in dealing with the route information during wayfinding. Unlike the assistance users the map users had to derive the direction information by relating the pictures of the intersections to the marked intersections on a fragmentary map (thereby processing mental spatial transformations) and, by walking a route segment from memory, they had to keep the to-beadopted direction for the current segment in working memory. These processes are likely to enhance memory encoding and consolidation. Studies manipulating cognitive load show that working memory is indeed involved in spatial learning tasks (e.g. Lindberg & Gärling, 1981;
Rossano & Moak, 1998). More specifically, visuo-spatial working memory (Logie, 1995) – which is viewed as a specialized working memory component within the tripartite working memory model of Baddeley (1986) – is thought to play an important role in spatial orientation (Smyth & Waller, 1998; Garden, Cornoldi & Logie, 2002; Bosco, Longoni & Vecchi, 2004;
Coluccia & Louse, 2004; Coluccia & Martello, 2004). In the map condition these working memory processes should leave 'traces' for long-term memory. Moreover, if working memory is viewed as an activated sub-set of long-term memory (e.g., Cowan, 1988, 1995), more elaborated working memory structures should lead to better (i.e. more elaborated and highly accessible) long-term memory structures.
We thus consider the active encoding explanation as the most likely reason for the superior spatial orientation knowledge of the map users. An alternative explanation would be that survey views were more familiar for the map users than for the assistance groups since they had studied map fragments. Map users might remember the allocentric views and match memory of the shape of the ways on a studied fragment with the shape of ways on the complete map during testing (however this is considered difficult given the complex map without any landmarks that was used in the test). Admittedly, this argument is a consequence Computer Assisted Navigation and the Aquisition of Route and Survey Knowledge 16 of our dependent measure of survey knowledge, which involved using a roadmap of the zoo for the relocation of places. Other, additional measures such as a pointing test or sketch maps could have helped to further substantiate the difference between assistance and map users.
However, the measure used in the present study reflects everyday spatial learning and retrieval. Both learning and retrieval of spatial orientation knowledge involves moving through and thereby studying the real environment as well as studying a map (as travelers often do) that represents it. In other words, the perception of the real environment and the cognitive processing of internal and external representations of it are intertwined both when learning and when retrieving spatial information in everyday orientation. From our point of view, the decisive point is that map users actively dealt with the information presented, and not that they matched roadmap fragments seen in the study phase. This argument is supported by the fact that assistance users also had spatial context information available during wayfinding. The visual animations that were provided in two of the assistance conditions comprised three intersections and the shape of the ways connecting them from an allocentric perspective. That is, the assistance users also saw roadmap fragments. Nevertheless survey memory of the navigation assistance users was poor. This suggests that the presence of survey information does not suffice. The information has to be used during walking to become effective for spatial learning.
Finally, one might speculate that participants reconstruct survey knowledge from route memory. They might solve the replacement task by relating the remembered view-based pictures to the bird's eye perspective shown on the roadmap at the time of test, i.e., they mentally walk the route again and try to find the appropriate intersections on the map by transforming the egocentric perspective of a particular intersection into an allocentric perspective. This strategy would have been possible since the start position was provided as a reference point on the map in the replacement task. Direction knowledge (i.e. in which direction one walked at a particular intersection) would help to successfully solve the Computer Assisted Navigation and the Aquisition of Route and Survey Knowledge 17 replacement task. However, this presupposes that the order of intersections is remembered without errors, since after a single mistake all subsequent reconstructions would be wrong.
There is no indication in the data that participants applied such a strategy, neither in the assistance nor in the map conditions. Participants misplaced intersections at the beginning of the walk in the same way as they misplaced intersections later on the route. Performance was generally poor independent of the serial position. On the contrary, when making an attempt to describe the observed replacement behavior of the participants, it resembles more a reconstruction of configurations by provisional placements and replacements than by following an explicit strategy. Finally, if the strategy would have been actually applied, then both navigation assistance users and map users should have been successful in the replacement task. Because the data show a clear advantage of map users, we therefore believe that the better performance of map-users in the replacement test is a consequence of the better memory of survey information acquired during wayfinding.
In summary, the present study puts forward an extended focus on the design of navigation assistance systems including their consequences for memory and for acquisition of spatial orientation knowledge. The design of navigation systems is usually optimized with respect to cognitive ergonomics during wayfinding. Direction information is provided in a way such that navigation is as easy as possible. This may have consequences on spatial learning, which is rarely discussed. We hypothesized that navigation assistance systems reduce spatial processing to a minimum, and consequently spatial knowledge would be poor.
Our data confirm this hypothesis. Both survey and route knowledge were considerably better in the map guided group than in the navigation assistance groups. Hence, if spatial learning would be relevant for users – e.g. for tourists, but also for elderly and/or disabled persons –, it is necessary to adopt the design in a way that the acquisition of spatial orientation knowledge is also supported. The presentation of additional allocentric information using animations and/or the manipulation of presentation modality did not work in the present real environment Computer Assisted Navigation and the Aquisition of Route and Survey Knowledge 18 study. We assume that those manipulations were inefficient because they did not initiate active encoding. It remains therefore a challenge for intelligent human-computer interaction design to find a way to both present information and to require the user to deal actively with this information. It might be suggested that for spatial learning, route instructions and working with maps with the requirement of some spatial processing could be combined. Such a specific combination – taking an individual's preferences and spatial strategies into account – may stimulate and support the acquisition of spatial orientation knowledge.
Acknowledgments This research was supported by a grant from the Deutsche Forschungsgemeinschaft in a Special Collaborative Research Group on Resource Adaptive Cognitive Processes (SFB