Eye-tracking research is a behavioral compliment to interview and survey techniques. This method is being applied to many of our studies, and in collaboration with colleagues in linguistics, sociology and geography.
Note: Additions to this page are ongoing
An opportunity to rent an eye-tracking unit (Mobile Eye) allowed us to collect data from 31 subjects in Aug. 2009. These data related to novice through expert geoscientist engagement with a plate tectonics image and three temperature maps (created by Robert Simmon at NASA Earth Observatory). Eye tracks were digitized into ArcGIS by undergraduate students in Fall 2009 and we have successfully developed a methodology (with assistance from colleagues in GIS) for analysis of eye-tracking data using ArcGIS functionality.
Preliminary Findings from Eye Tracking Study of novice and expert interactions with Global Temperature Maps
*Libarkin, J.C.1, Clark, S.K.2, Simmon, R.3
1 Department of Geological Sciences, Michigan State University, East Lansing MI 48824
2 Department of Geology, University of Wisconsin-Eau Claire, Eau Claire, WI 54702
3 Sigma Space Corporation, Code 613.2, NASA Goddard Space Flight Center, Greenbelt, MD 20771
* To whom correspondence should be addressed. E-mail: firstname.lastname@example.org
Urgent calls for better communication of climate science are being made in light of the public’s mistrust of climate change messages (1). We suggest that climate illiteracy stems in part from the ways in which climate messengers convey visual information. Images are powerful conveyers of complex Earth systems science (2), but misinterpretations can arise from seemingly simple depictions (3). We report on an eye tracking and interview-based investigation of expert-novice interactions with authentic climate-related data, revealing striking differences in gaze behavior and discourse as a function of color palette and expertise level.
Participants (n=28) with color-normal vision were shown maps in one spectral (rainbow) and two hue-based (gray and purple) design conditions depicting global monthly mean surface temperatures for March 2009. Participants were interviewed while sequentially viewing each of the three maps on a 24” monitor. Questions were preceded by an initial ten seconds of spontaneous looking (SL). The SL condition was chosen to replicate the type of interaction with images expected during newspaper reading, internet searching, or television viewing.
Analysis of the initial 10 seconds of SL gaze data indicates that both map palette and expertise level influence viewing behavior. Participants’ gaze on the rainbow map was attracted most to red and blue portions of the display (warm and cold continental temperatures) and less to the green areas (moderate marine temperatures; Fig. 1a, b). Similar disparity of view was not observed for the hue-based maps. Similarly, all participants avoided the yellow-green area of the rainbow map’s legend (Fig. 1a, b).
Patterns of gaze behavior differed across expert-novice cohorts for all color treatments. Experts gaze at specific areas of the map was interspersed with gaze at the legend (Fig. 1c), suggestive of underlying cognitive behavior to couple a meaning to specific colors. Novices engaged primarily in a spiraling gaze pattern, suggesting an attempt to view as much of the image as possible (Fig. 1d). This aligns with findings in other studies of efficiency in visual search (4). In general, while experts systematically swept their eyes across key features of the maps and legend during the 10 seconds of SL, novices focused on dark and bright spots and did not attend to map legends. During a questioning period intended to encourage subjects to make observations, novices spent significantly more time looking at irrelevant features, while experts more consistently attended to the legend and areas of personal or geologic interest.
Coupled with verbal responses, these eye-tracking results suggest that: a) Expert gaze is driven by underlying cognition while novice gaze is driven by efficiency in image coverage; and b) Color can influence where and for how long specific features are observed. The implications of this work for climate literacy are significant. Experts and novices clearly attend to different aspects of maps, and novices are more likely to generate conclusions that are influenced by the color palette used, rather than the data itself. Surprisingly, both novices and experts preferred the rainbow map for depiction of temperature data. This preference for the rainbow palette was maintained even when participants were explicitly questioned about the lack of visible details in yellow-green transitions as depicted in Fig. 1a. We suggest that color palette significantly affects observation pathways as well as overall map interpretations, in alignment with results from other domains (5), and that images generated by scientists may not be effective for communication with novices. This work indicates that visualizations related to climate data must be carefully designed to convey intended messages and to target specific audiences.
References and Notes
1.T.E. Bowman et al., Science 330, 1044 (2010).
2.L. Whitmarsh, Publ. Und. Sci. 18, 401 (2009).
3.B.R. Newell, A.J. Pittman, Bull. Am. Met. Soc. 91, 1003 (2010).
4.C. Haimson, D. Bothell, S.A. Douglass, J.R. Anderson, Hum. Factors 46, 551 (2004).
5.L.A. Breslow, R.M. Ratwani, J.G. Trafton, Hum. Factors 51, 321 (2009).
6.This work was partially funded by grant DRL-0815930. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Figure 1. Kernel density (heat map) and gaze plots for 10 seconds of SL, plotted in ArcGIS 9.3. Mean surface temperature data are from the Atmospheric Infrared Sounder . Kernel density plots (a, b) depict concentration of gaze time, with darker purple reflecting longer gaze, for a) those subjects who viewed the rainbow map first (n=9) and b) those subjects who viewed the gray map first (n=9). Gaze plots (c, d) illustrate patterns of eye movement across an image for c) an expert (red) and d) a novice (blue). These data suggest that experience within a domain as well as color palette impact the way in which people perceive, visually inspect and attend to figures.