How do we read YAH maps? An Eye-Tracking Study Hana Kmecova 1,*, David Araci 1, Primrose Moyo 1, and Jan. M. Wiener 1 Department of Psychology, Bournemouth University Poole, BH12 5BB, UK Abstract. This paper presents an eye-tracking experiment investigating gaze behaviour when solving wayfinding tasks using You-Are-Here maps. Furthermore, we investigated the influence of the type of instructions given prior to the wayfinding task. In the experiment, participants were presented with a map consisting of three key interest areas; an overview map, a registry, and an enlarged insert containing local directional information. Participants were given instructions to find a target location that was situated in the same building as their starting location, or in a different building. In addition, the instructions given to participants either included the name of the building containing the target location or not. Results revealed that instructions did not influence the accuracy of wayfinding decisions but did affect response times and gaze behaviour. Specifically, the way participants distributed their gaze when reading the maps depended on the specific instructions. Keywords: Spatial cognition, you-are-here maps, wayfinding, eye-tracking, gaze behaviour, route formation, visual search, decision making 1. Introduction Static maps are widely used in complex architectural environments (Lynch, 1960). They allow users to acquire knowledge about their current location, increase spatial awareness and facilitate identification of route options (Klippel, Hirtle & Davies, 2010, Liben, Myers, Christensen & Bower, 2013). Research demonstrates that wayfinding maps should be devoid of unnecessary visual clutter (Schmid, Richter, & Peters, 2010) and allow users to connect structures depicted on a map with structures in the physical world (Freska, 1999; Klippel et al., 2010, Maxwell, 1976, Dent, Torguson, Hodler, 2008). One way to allow for clearer correspondence between the physical environment and the map is by explicitly indicating the position of the map reader. The presence of an indicator stating one s location on the map is commonly known as a You-Are-Here (YAH) symbol. Factors such as map-terrain correspondence, syntactic clarity, completeness, alignment and the YAH symbol itself can increase the effectiveness of a map (Klippel, Freska & Winters, 2006; Levine, 1982). The present study was informed by findings from a preceding study on the efficacy of wayfinding maps in Poole Hospital in Dorset, UK (Kmecova et al., in preparation). Three map conditions were compared: In Condition 1, the original map utilized in Poole Hospital was used. In Condition 2, the alignment and order of infor-
mation was improved, visual complexity was reduced and a YAH symbol was introduced. Finally, in Condition 3, the maps in Condition 2 were supplemented with a Local Directional Insert (LDI), which is a magnified YAH area. Accuracy increased across the conditions, reaching ceiling levels in Condition 3, with response time decreasing across conditions. Accuracy and response times also improved when participants were provided with the building name of the target destination in addition to the unit name. Figure 1. An example of one of the maps used in the present study. The six interest areas examined were; the Title (top), the Registry (left), the Overview Map (centre), the Key (top right), the No Smoking/Phone sign (bottom) and the LDI (bottom right). The current study aimed to further explore the influence of both the LDI and the type of instructions received on the efficacy of wayfinding maps in a more controlled laboratory-based experiment. Figure 1 displays one of the maps used in the current study. The LDI provides directional information relevant to spatial decision making at the current location at different levels of detail: it provides directional information to all units in the same building, units in other buildings, however, are not listed but directions to all other buildings are provided. The Registry and Overview Map provide information about the entire environment. Our work with Poole hospital also revealed inconsistencies regarding the information provided in hospital appointment letters: while they always contain the specific unit name, they may or may not contain the name of the building. Here we investigate the effects of such inconsistencies. 2. Method Participants: 27 participants (14 females, mean age 21; range 18 to 28) unfamiliar with Poole Hospital took part in the experiment. They were students of Bournemouth University and received course credit compensation for their participation. Procedure: Each participant took part in 24 trials. Participants were first presented with instructions comprising of a target location within the hospital (e.g., Go to Der-
matology), followed by a map of Poole Hospital. Participants' task was to indicate the direction in which they would proceed in order to navigate to the target location. A total of 6 maps were utilised (see Figure 1), each with a different current location. The study used a 2x2 design with the following independent variables: First, Instruction (name of the building containing the target unit present or absent) and second Target Location (in same building as current location or in different building). The resulting four experimental conditions were; with+same = target location includes building name and is in the same building as the start location, with+different = target location includes building name and is in a different building to the start location, without+same = target location does not include building name and is in the same building as the start location, without+different = target location does not include building name and is in a different building as the start location. Experimental Setup: Maps were presented at a resolution of 1024 x 768 pixels on a 20 monitor. Participants were seated in front of the monitor at a distance of ~60 cm. The resulting visual angle of the monitor was 37 degrees (horizontally) x 28 degrees (vertically). Gaze behavior was recorded using a SR Research Ltd. Eyelink 1000 eye tracker, sampling pupil position at 500Hz. Accuracy and reaction times were recorded with a response box Analysis of Gaze Behavior: For each map presented, six interest areas were defined; Title, Registry, Overview Map, Key, LDI (Local Directional Insert) and the No Smoking/Phone sign (see Figure 1). Dwelling time (time gaze remained in a given interest area) was analyzed to determine how participants read the maps. Figure 2: Left: Accuracy (correct responses); right: Mean response times (milliseconds) for correct.
3. Results 3.1. Behavioral Data Performance: On average, participants chose the correct direction in 84% of trials (see Figure 2 left). An ANOVA (factors: inclusion of building name in instructions [with, without] & target location in the same building or in a different building [same, different]) did not reveal significant main effects (instructions: p=.15; target location: p=.47) but a significant interaction (F(1,26) = 13.12, p =.001, see Figure 2). Response time analysis: Only correct trials were included in the response time analysis. Participants' mean response time was 8794msec. An ANOVA (factors: instructions [with, without], target location [same, different]) revealed a significant main effect of instructions (F(1,26) = 8.397, p =.008) and target location (F(1,26) = 27.61, p <.001) as well as a significant interaction (F(1,26) = 51.26, p <.001, see Figure 3). Post hoc analysis suggests that the main effects as well as the interaction was primarily driven by longer response times in trials in which the target location was situated in a different building and the instructions did not include the building name (condition: without+different; 14693ms as compared to 5381msec in the without+same condition). In addition to these effects, participants response times decreased over the 24 trials (F(1,23) = 37.63, p <.001) which reflects learning of the general layout of 1 2 3 4 Figure 3: Graph displaying average dwell time for each interest area across all experimental condition (1 same+without, 2 same+with, 3 different+without, 4 different+with).
the maps. Gaze Behavior Analysis of dwelling time demonstrated that participants attended to the Registry, the overview Map, or the LDI for over 99% of the time. The further analysis therefore concentrated on these three interest areas. Average dwell proportion for LDI was highest (55.7%), followed the overview map (22%), and the registry (19,2%). An ANOVA revealed a main effect of interest area (F=(2,1176)=471.5, p<.001). Post hoc tests demonstrated that all comparisons were significant at p<.03). Average dwell time for the registry was 2302msec, 3391msec for the LDI, and 1629msec for the overview map. The fixation maps in Figure 4 suggest that participants used different viewing strategies in different conditions. An ANOVA (factors: instructions [with, without], target location [same, different], interest area [LDI, Overview Map, Registry) revealed significant main effects of instructions (F(1,26) = 9.68, p =.004), target location (F(1,26) = 24.48, p <.001) and interest area (F(2,52) = 19.17, p <.001). Post hoc analysis indicated that dwell time was different for each interest area (all p <.001). Importantly, the analysis also revealed a significant three way interaction (F(2,52) = 6.16, p =.004, see Figure 4) which highlights differences in viewing strategies, particularly between the different+without conditions and the remaining conditions (this is also reflected in the different fixation patterns in Figure 3 and the dwell time patterns in Figure 4). Figure 4: Mean dwell time for each of the 3 interest areas for the all conditions.
4. Discussion In the current study we examined gaze behavior when utilizing YAH-maps. The design of the maps was informed by findings from a case study in Poole Hospital suggesting that wayfinding maps for large complex environments with several buildings should contain a registry and an aligned overview map as well as an aligned Local Directional Insert (LDI). While map inserts are widely used by cartographers to show a blow-up area where greater detail is needed (Dent, et al., 2008), the insert used in this study contains additional site specific directional information. The LDI therefore resembles an insert map with embedded wayfinding sign. One may argue that the inclusion of LDIs (see Figure 1) makes the overview map dispensable, as movement decision can be drawn on basis of LDI and registry alone. Results of the current study, however, demonstrate that participants did attend to the overview maps (~20% of the trial time). While further research is needed to understand this result in full, a possible explanation is that overview maps in contrast to LDIs allow estimating overall route shape and length and may therefore decrease uncertainty during navigation. The longest response times were found when participants were in a building different to that containing the target unit and instructions only included the unit name (condition: without + different), but not the name of the building the unit was located in. Gaze analysis shows that participants had to refer to the registry to determine the building name in order to make movement decisions. This result demonstrates that in order to benefit from the inclusion of an LDI, sufficient instructions need to be provided prior to the wayfinding task. This has implications for practice: For example, as it often cannot be predicted where visitors enter a hospital, appointment letters should provide visitors with both the unit and building names. 5. References Dent, B., Torguson, J., Hodler, T. (2008). Cartography: Thematic Map Design, McGraw-Hill. Freksa, C. (1999). Spatial aspects of task-specific wayfinding maps a representation-theoretic perspective. In J. S. Gero, & B. Tversky (Eds.) Visual and Spatial Reasoning in Design, (pp. 15 32). University of Sydney. Klippel, A., Freksa, C., & Winter, S. (2006) You-are-here maps in emergencies The danger of getting lost. Journal of Spatial Science, 51(1), 117 131. Klippel, A., Hirtle, S., & Davies, C. (2010). You-are-here maps: Creating spatial awareness through map-like representations. Spatial Cognition & Computation, 10(2-3), 83 93. Kmecova, H., Wiener, M. J., & Araci, D. (in preparation). Redefining map design to optimise wayfinding in a complex hospital environment: An empirical investigation Levine, M. (1982). You-are-here maps: Psychological considerations. Environment and Behavior, 14(2), 221 237. Liben, L. S., Myers, L. J., Christensen, A. E., & Bower, C. A. (2013). Environmental-Scale Map Use in Middle Childhood: Links to Spatial Skills, Strategies, and Gender. Child Development. Lynch, K. (1960). The image of the city. Cambridge: MIT Press. Maxwell, E., A. (1976). Geometry by transformation. Cambridge: Cambridge University Press. Schmid, F., Richter, K.-F., & Peters, D. (2010). Route Aware Maps: Multigranular Wayfinding Assistance. Spatial Cognition & Computation, 10(2-3), 184 206.