Potentials and Limitations of Mobile Eye Tracking in Visitor Studies

Abstract

The main goal of this article is to demonstrate the potentials and limitations of mobile eye tracking (MET) in visitor studies and other social science research. We provide empirical examples of MET research in the context of a comparative study of two exhibitions at two museums in Germany. The article underlines the potentials of MET in combination with other methods and in comparison with conventional forms of observation and interviewing. On the basis of our case material we provide recommendations for social scientists who consider integrating mobile eye-tracking in their field research.

Keywords

cognitive science mobile eye tracking observation visitor studies field research

Introduction: Background and Characterization of Mobile Eye-Tracking

One goal of this article is to demonstrate the potentials and limitations of mobile eye tracking (MET) in museum visitor studies. Another goal is to provide empirical examples of visitor research and draw some conclusions about the potentials and limitations for future prospects of MET in museum visitor studies and in social science field research more generally. Thereby we hope to put anthropologists and other social scientists engaged in field research into the position to make an informed decision as to whether and how they may apply MET technology in their future research. We provide background information on the available MET tools and what is required to make good use of MET in field research. We then discuss the potentials and limitations of MET for visitor studies, first in general and second in terms of our own study at two museum exhibitions in Germany. We present MET results that provide us with an inventory of visitors' scan patterns that can be used in a bottom-up fashion to construct a “script” or “schema”, in this case of museum exhibition visits but with potentially wider applications. We conclude by suggesting MET as a complementary tool for social science research on culture and cognition.

As a first step toward understanding MET technology and the kind of data that it provides, some background knowledge about visual perception in action is essential. The way we perceive things visually differs depending on the context, for instance when being seated in an armchair reading a book or when sitting in front of a screen watching videos or looking at images - or when strolling through a museum gallery. Seeing things while being involved in bodily action involves more parts of the brain than in a position of rest, and it is connected especially to the “executive” of our “working memory” (Baddeley, 2007) and to our procedural long-term memory (Land & Tatler, 2009).¹ The implicit or procedural knowledge involved in visual perception in action is stored in schemata/scripts and is in a great measure unconscious in nature so that agents themselves find it hard to verbalize and manipulate their behavior (Land & Tatler, 2009; cf. Schank & Abelson, 1977 about scripts).

The discussion of visual perception relies on three important technical terms which are also important for our assessment of MET, namely “fixations,” “saccades,” and “scan patterns”. Our vision consists of so-called eye movement patterns or scan patterns or gaze patterns that are made of combined “saccade and fixate” strategies. Fixations can be explained as very short “stops” of eye movements indicating that attention is being resided somehow. Saccades are very rapid eye movements between fixations that are done spontaneously and that can vary in duration. During saccades we can hardly perceive anything, we are basically “blind.” Hence, eye movements are not done smoothly but in an apparently zigzag fashion that seem to be done very unsystematically at first (Land & Tatler, 2009). Eye-tracking technology records these movements and it is the first task of MET data analysis to distinguish fixations from saccades and then to investigate correlations between the observed patterns, meanings and the goals of attention (cf. Holmqvist, Nyström, Andersson, Dewhurst, Jarodzka, & van de Weijer, 2011).

Whereas stationary eye trackers require participants to be seated in front of a screen, METs can be mounted on the participants’ head while they move around freely engaging in various tasks. There are a number of head-mounted eye tracking devices available. We have used the ASL MobileEye and the Locarna PT Mini (see below) but many of the following points refer to MET technology in general.

All of these devices generally work with two cameras (see Figure 1): a scene camera that records the scene and the environment from the participant’s perspective and an eye camera that records his or her eye movements using harmless infrared lights, the so-called dark pupil tracking. The two cameras are mounted on frames of a pair of eyeglasses but most devices only record one eye of the participant. These glasses sometimes are mounted on a cap or even on a helmet. As Figure 1 shows, most devices today still need to be fixed in a very stable manner on the participants' head. However, the current trend is to produce MET devices that are worn like regular glasses by the participant and that can track both eyes (so-called binocular eye-tracking) which reduces the parallax error of our two-eyed vision that is more pronounced when relying on MET technology that tracks only one eye (cf. www.eyetracking-glasses.com or http://www.tobii.com/en/analysis-and-research/global/ and cf. Holmqvist et al., 2011, p. 60 for parallax error).

Figure 1.

Calibration of a mobile eye tracker (photo: Kira Eghbal-Azar).

Since both separately recorded videos have to be brought together and processed into a single video by a computer after recording has taken place, the calibration of the cameras is essential. What processing does is to synchronize the two recordings so that we receive a single video with a marker (a fixation cross or circle) that is permanently shown while we watch the recorded scene from the participant's perspective. This video indicates eye movements in action based on the eye camera together with the picture of the three-dimensional space through which the participant has moved (cf. Mayr, Knipfer, & Wessel, 2009 and the manuals of MET devices). This process makes MET rather complex to use in field research, making some preparatory training with this technology necessary (Holmqvist et al., 2011, p. v-vi, 1).²

One of the reasons why eye tracking has not yet played a role in the social sciences is that in its stationary version it requires a typical experiment-like setting, cutting out the social context and the context of agency. With MET becoming more readily available, this limitation is less severe today.³ We therefore suggest MET as a new feasible method for social scientists conducting field research, but we are also aware that a number of problems still remain that researchers need to take account of.

Pros and Cons of Mobile Eye Tracking Technology in the Context of Visitor Studies

In their summary of visitor studies methodology Yalowitz and Bronnenkant (2009) are expressing the dominant view when they stress that the main method regarding visitor studies is still observation, no matter how observation data are collected: e.g. by the paper-and-pencil technique or through videotaping or other tracking technologies.⁴ The same probably holds more generally since many social scientists, especially anthropologists, who are conducting field research all over the world, continue to apply the paper-and-pencil technique for observation in the field. We do not expect that MET technology will completely replace these other forms of observation but that it will be complementary - in what ways remains to be seen.

Mayr et al. (2009) provide a pioneering exploratory MET study with “first insights into informal learning in museums” (p. 195.). However, they neglect that visitors are not simply “integrating information that is spatially distributed” (p. 190.) but are engaged in what could be called “appropriating” the exhibition. Appropriating exhibitions means that visitors are not only gathering information but are also actively looking for an emotional and aesthetical experience (Hein, 2000) in a particular social setting (Falk, Koran, & Dierking, 1986), a point to which we shall return below. In their summary Mayr et al. (2009) list four potentials of MET as a recording device (p. 196-197.) which can be taken as a useful point of departure:

“Data richness” through the inclusion of information about the environment (e.g., other visitors present, the visibility, or approachability of objects and media).

High external “data validity” because data are recorded “objectively” by cameras (p. 196), reducing perspective errors and providing more ecological validity due to applicability in natural settings.

Providing a “nonreactive measurement,” because eye movements can hardly be manipulated due to its unconsciousness (see above) so that it can be usefully related independently to conscious reflection (see below).

Allowing statistical analysis, depending on the sample size and research question.

Mayr et al. (2009) then go on to list eight limitations of MET (pp. 198–200):

No recording of “covert attention and mental spotlight” since eye fixations per se do not tell us anything about the goal of the visitor's attention.

Only “limited conclusions about cognitive processing” are possible because the conclusions that we arrive at through MET are typically “interpretations of eye-tracking data that are often based on assumptions and heuristics about underlying cognitive processes” (p. 198).

Some “obtrusiveness of measurement” as data might be distorted because visitors are aware of being evaluated by the MET or are simply irritated by wearing it.

“Selective sampling” in the sense that not everyone can wear an MET device since visitors with corneal dysfunctions (that prohibit reflections of the infrared lights) or who wear regular glasses are disqualified (Since then devices have become available, e.g., from Locarna, that allow wearing regular glasses in addition to the eye tracker).

“Limited temporal and spatial accuracy”: There is (currently) a time limitation due to the temporal resolution of eye trackers that has to be divided for the recording of two videos at the same time (hence 50 Hz in fact mean 25 Hz), so that very short fixations might be missed. If distances between the visitor and objects often differ from the calibrated distance the spatial accuracy is not given at all times (p. 199; the parallax error, see above). Like every single lens camera the scene camera only records one part of the world and is more limited than our vision.

“Laborious data analysis” is required because data analysis has still to be done largely manually by the researcher due to the mobile setting that allows visitors to walk individually through galleries so that the stimuli and scenes change all the time.⁵

Costs: METs are very expensive (each of the devices we used in our study costs around 25 000 Euro, binocular mobile eye-trackers are currently sold at around 40.000 Euro).

New “ethical concerns”: Because participants can hardly manipulate or hide their eye movements, a point that has to be included in procedures of consent and at the same time this is one of the important assets of this technology.

In conclusion, Mayr et al. (2009) suggest combining MET with other methods such as interviews, questionnaires, or verbal reports to gain valid interpretations of scan patterns. One of the reasons why MET needs to be combined with other methods is that the marker (based on the eye camera) that is permanently shown in the processed MET video data cannot be equalized all the time with actual attention to the object or scene captured by the scene camera. While past research assumed a strong “eye-mind hypothesis” (Just & Carpenter, 1980) as if eye movements would provide a direct insight into cognition, current research does not assume that we can draw simple conclusions from fixations and scan patterns to the goals of attention and to the underlying cognitive processes (Mayr et al., 2009; Land & Tatler, 2009). Complementary to MET there are three different kinds of verbal reportings which may be selected according to the research question or the task at hand and with respect to the relevant memory system. These are concurrent, retrospective or cued retrospective reporting (cf. van Gog, Paas, van Merrienboer, & Witte, 2005; van Someren, Barnard, & Sandberg, 1994; Ericsson & Simon, 1993):

—Concurrent reporting means verbal reporting while doing a particular task.

—Retrospective reporting means verbal reporting in retrospection after completing the task.

—Cued retrospective reporting means verbal reporting after completing a particular task with a cue, in this case typically while watching the processed recording.⁶

In our own study, described in more detail below, we have chosen for cued retrospective reporting since it promises to lead to more exact and controlled results as the given cue (a processed eye tracking video) triggers more exact memories which would otherwise be constructed without the assistance of a cue (cf. van Gog et al., 2005).

Mobile Eye-Tracking in the Field: Two Exhibitions at Two Museums as a Comparative Test Case

The potentials and limitations outlined above were arrived at through a single study that consisted of an exhibition set up solely for the purposes of exploring MET with three participants (Mayr et al., 2009). We now report on research in two settings that were natural in the sense that the two comparative case studies were carried out at museums with regular exhibitions that were set up by curators with a conventional training in their field for just plain visitors who frequent these museums. Putting the MET at work in these “wild” conditions is a more realistic test case of this tool for interested field researchers. The results that constitute our empirical examples are preliminary as they await full analysis in a forthcoming PhD thesis (Eghbal-Azar, forthcoming). The two exhibitions concerned were the temporary exhibition: “Pacific Oases: Living and Surviving in the West Pacific” (at the Linden-Museum Stuttgart, Figure 2, cf. Heermann, 2009) and the permanent exhibition: “nexus” at the Museum of Modern Literature (LiMo) in Marbach (cf. Gfrereis, 2009). The overall project covers a broad range of issues to do with the distinction of the culturally close (modern German literature in Marbach) and the culturally distant (Pacific ethnography in Stuttgart) with regard to the visitors who are resident in Germany. Other issues of comparison concern the modes of the exhibitions (temporary versus permanent exhibition) and particular presentation styles of displaying objects and modes of implementing the use of various digital and non-digital media in correspondence with the particular intentions of the curators. Furthermore the two exhibitions are placed in institutions that are very different in terms of their histories and buildings (old versus newly established) and with regard to their subject matter. Despite these differences there is one common feature that is very relevant in the context of MET studies, namely that both exhibitions do not suggest a single or main pathway for visitors.

Two different eye trackers were used in the two exhibitions: We applied the ASL MobileEye eye tracker (designed 2004; http://www.asleyetracking.com/Site/Products/MobileEye/tabid/70/Default.aspx) for the MET study at the Linden-Museum Stuttgart (see Figure 2) and the Locarna PT Mini (designed 2010; http://www.locarna.com/products.html) at the LiMo in Marbach. Our sample size was eight visitors in each of both MET studies (n = 16 in sum). These were in each case four “experts” (visitors with prior knowledge of the subject matter, e.g., students of German studies or cultural anthropology) and four “novices” (with only cursory or even no prior knowledge of the subject matter). At this stage we have not yet fully analyzed differences between “experts” and “novices”. As far as the authors know, these two comparative MET studies to date had the largest sample size for examining regular exhibitions applying MET technology. For more information about both samples see Tables 1 and 2.

Table 1.

Sample of the Mobile Eye Tracking (MET) Study at the Linden-Museum Stuttgart “Pacific Oasis” Exhibition: Visitors' Social Data.

Visitor Category	Gender	Age	Profession
Expert	Male	25	Student of cultural anthropology
Expert	Female	24	Student of cultural anthropology
Expert	Female	24	Student of cultural anthropology
Expert	Female	24	Student of cultural anthropology
Novice	Male	36	Mechanical engineer
Novice	Male	29	Student of English philology
Novice	Male	25	Student of technology management
Novice	Female	19	Student of linguistics

Table 2.

Sample of the Mobile Eye Tracking (MET) Study at the LiMo “Nexus” Exhibition in Marbach: Visitors' Social Data.

Visitor Category	Gender	Age	Profession
Expert	Male	23	Teacher trainee for German
Expert	Female	24	Teacher trainee for German
Expert	Male	27	Master degree in German studies
Expert	Male	24	Student of philology
Novice	Female	23	Student of Roman philology and history
Novice	Male	43	Technician for machine tools
Novice	Male	24	Student of software engineering
Novice	Male	32	Student of philosophy and industrial mechanic

All visitors had been first-time visitors to the exhibitions and had no prior knowledge about these exhibitions or how to navigate through them. We applied MET in both cases in an exploratory fashion, to help documenting and analyzing the scan patterns of visitors, i.e. what visitors “really viewed” as they moved freely through the exhibition. According to Land & Tatler (2009, p. 41) “free-viewing” allows the visitor “to select their own high-level approaches to looking at scenes”. Consequently, by recording free viewing as detailed as possible we were aiming to find their implicit scripts and strategies or what one may consider their natural, habitual way of appropriating exhibitions. All visitors got the following open and standardized instruction for their exhibition visit after calibrating the MET: “Please view the exhibition naturally at your own speed, following your own wishes and needs. There are no further specifications, even no time specification on how to carry out this visit. Your knowledge acquisition about the exhibition will not be tested afterwards”.

Since we wanted to find out what the visitors actually attended to during their visit we used cued retrospective reporting to elicit their goals of attention, avoiding priming effects as much as possible. After the exhibition visit all visitors were asked to watch their own processed eye-tracking video and to verbalize with this cue in retrospection. At that stage, we gave the following standardized instruction: “Now I present you the video recorded by the MET during your visit of the exhibition. While watching the video, please describe spontaneously what you viewed, perceived, thought and felt at various points and what you paid attention to.” We also conducted interviews directly after the reporting.⁷

The figures we provide give an overview about the procedure, the time effort for conducting and analyzing these MET studies and the mass of data that even such a small sample produces (see Tables 3 and 4).

Table 3.

Estimated Time Effort for Preparing and Analyzing the Mass of Data.

Tasks	Details	Total time (min)^a
Transcription of all reports	10 min transcription time for one min audio recording	8,650
Connecting all audio recording of reporting with corresponding MET videos	Approx. 30 min for one video	480
MET manual data analysis without audio (estimated)^b,e	Estimated 1 min recording = 5–20 min manual data analysis	4420–17,680
MET manual analysis with audio (estimated)^c,e	Estimated 1 min recording = 10–25 min manual data analysis	8,650–21,625
Transcription of all interviews	10 min transcription time for 1 min audio recording	3,460
Connecting all interviews with corresponding questionnaires		540
Estimated time requirement for preparing data for analysis and analyzing all 16 videos with and without audio completely on the microlevel^d	Corresponds to 11-22 weeks of full-time work	26,200–52,435

Note. ^aData analysis is not yet fully completed. Results mentioned in this article are preliminary results to provide potentials of MET.

^bCalculation based on MET recording.

^cCalculation based on audio recordings of cued retrospective reporting.

^dMicrolevel = all 10 distinct scan patterns that have been found exemplarily in both MET studies so far. Time requirement excluding needed breaks during manual data analysis and excluding analysis of Interview/questionnaires.

^eData analysis of MET data depends on visitors’ behavior, level of analysis, number of displayed exhibits, and given information in the exhibition, size of the exhibition area, kind of analysis (with or without audio), and the experience of the researcher. These time efforts are estimations because the 16 tests are not yet fully analyzed. The preliminary results mentioned in this article are based on manual video data analysis concerning systematic, repeating, and eye-catching scan patterns.

Table 4.

Time Effort Conducting the Two Comparative Mobile Eye Tracking (MET) Studies.

Tasks (chronologically)	Details	Total time (Min)
Test setups at the museums	60 min/test	960
Calibrations	Varied from 5 min to 30 min (including breaks)	240
MET recordings	Varied from 12 min 52 sec to 1 hr 16 min 30 sec (excluding breaks)	884
MET data processings	Equates appr. recording time	884
Cued retrospective reporting^a
(separate audio recording)		865
Interviews^b		346
Removals of test setups at the museum	60 min/test	960
Time requirement for the 16 tests		5,139
Time schedule for data collection^c	One day of work for each test	16 days

Note. ^aPure reporting time without preparation and instruction time. Differentiates from MET recording time due to time spent in exhibition and time spent reporting.

^bStandardized interviews based on a questionnaire with open and closed questions. These interviews have been recorded completely. Closed questions had been documented on the questionnaires as well.

^cNote that five tests did not work out due to technical issues. So in fact data collection took much longer to get the complete sample size of n = 16.

Figure 2.

Mobile eye tracking (MET) study at the Linden-Museum Stuttgart “Pacific Oases” exhibition (photo: Kira Eghbal-Azar).

Mayr et al. (2009, see above) listed twice as many limitations than potentials of MET for visitor studies. The first lesson suggested by our examples is that it is worthwhile to first discuss the actual qualities of MET potentials before starting to compare potentials and limitations by counting them. What are, in fact, the benefits of applying MET in our research? After all, the costs (time, money, energy) are considerable. Is MET really worth doing? If MET as an objective measurement cannot measure aesthetical experiences, as it has been criticized (Kaube, 2010), why bother?

Here are some details of our research that speak to these broad questions:

Detail 1:

Eye movements in action are very rapid (Land, & Tatler, 2009). For example, a sequence of one visit to the literature exhibition that lasted about 10 sec included no less than eight distinct scan patterns that the MET recorded! There is no way that applying conventional observation methods could record such a large number of rapid changing behavioral patterns in such a detailed manner over a long period. The observer is likely to miss out some patterns because observation depends on the acute awareness of the observer (with no rewind option). By contrast, exploratory MET data can be stored and then analysed repeatedly.

Detail 2:

In conventional observation, head and body movements can be treated as a proxy for scan patterns, but there are eye movements that are done without moving the head or the rest of the body which do show up in the MET record but which otherwise would remain undetected (cf. Mayr et al., 2009). For example, we observed an “alternating gaze” between two tiny exhibits that are positioned very close to each other so that the visitor does not even have to move his or her head but only the eyes.⁸ Without MET these gazes would be undetected because there are no indicators from which an observer could conclude the visitors’ gaze.

Detail 3:

Conditions in the museum gallery can make observation difficult or even impossible due to an obstructed view caused by the particular style of the presentation or simply by too many other visitors being in the way, by light reflections or by relative darkness. At the LiMo, this forced us to drop conventional observations at the level of objects in close proximity within the showcases in favour of the larger scale only, while the MET allowed us to include the participants’ perspective at all levels. The often unpredictable and uncontrollable obstructions that make conventional observation difficult were bypassed applying MET.

Detail 4:

Even when eye movements can be observed, fixations never can due to their very nature (cf. Mayr et al. 2009). In the example given earlier in Detail 1, eight fixations were made within 10 sec. Two of these eight fixations were directed at a pistol that was exhibited. While these fixations indicate that attention is being arrested, we still do not know the goal of attention and the reasons for paying attention until we combine MET data with other methods. MET therefore also serves a function as an exploratory tool and not just as a tool for quantifying observations and for testing assumptions as might be expected. It does not exclude other approaches to aesthetical experiences but it may complement them in interesting ways.

Implication of Mobile Eye Tracking Data for Current Controversies in Visitor Studies

From these details we may draw some intermediate conclusions, the most important one being the need for combining MET with other methods. These combinations may be similar to what Mayr et al. (2009) suggest but other, additional ideas for further possible combinations emerge from our case material. On the basis of these empirical examples, we may revisit the issue of MET potentials.⁹

Firstly, equipped with MET data we may reconsider some of the unresolved controversies in visitor studies. For instance, can we equate stopping or the time spend looking at an exhibit with attending to this exhibit (Serell, 1997; Yalowitz & Bronnenkant, 2009)? The visitor who had given attention to the pistol (note it is the only pistol displayed in the literature exhibition) not only spent time viewing it and fixating it but also reported on it afterwards. However, there were many other exhibits that this visitor viewed much longer but without reporting on them. By applying MET we might be able to conclude that more quantity (length of time and number of stops) that we spend looking at exhibits, or more generally at exhibitions for that matter, do not necessarily translate into a qualitative difference - as Serell (1997) and other visitor researchers seem to assume. Serell (1997) found out that most visitors only spend about 20 minutes at an exhibition and only view one third of the displayed objects. There is a considerable amount of empirical research about visitors' movement patterns in exhibitions that has been carried out since the beginning of tracking studies (without applying MET). The results are partially contradictory. One attempt to deal with these contradictory results in terms of underlying universal principles is the general value principle that assumes that visitors are “saving steps” (Bitgood, 2006). In other words, the assumption is that visitors walk as little as necessary in order to minimize their costs for getting more value from their visit. Since they do not have control over the benefits offered in exhibitions (i.e. the design of the exhibition and the level of information provided) they concentrate on their “costs” in order to arrive at an optimal visit (Bitgood, 2006 and Rounds, 2004). Unfortunately, `saving steps' as a principle cannot explain every phenomenon encountered in exhibitions. Moreover, it does not account for the most important and distinguishing feature that characterizes a museum exhibition visit in comparison to other forms of information gathering: The classical way of appropriating exhibitions compared to watching a movie or reading a book is “strolling and viewing” (“gehen und sehen” Korff & Thiemeyer, 2008, p. 137; cf. Korff, 2003), i. e. the ability to go off and to explore in multiple possible ways and from a number of perspectives. This is what makes museum exhibition visits such a unique experience and arguably this is the main attraction that museums hold. The agendas that we have set for ourselves during a visit and that influence our attention (Falk, Moussouri, & Coulson, 1998) may override the time that we devote to a particular object and the “cost” factor more generally. Discussing alternative explanations of visitor behaviour has a much better basis once we have good data on both, the time spent on particular objects, the fixations that go with an exhibition visit and other data from other sources - allowing for a triangulation of methods as George E. Hein (1998, p. 75.) proposed it long ago.

Secondly, with MET data we may also contribute more specifically to the debate as to whether there is an underlying universal principle that explains the behavior of visitors and whether it consists of the type of behavioural rules that have been suggested (cf. Bitgood, 2006 and Rounds, 2004). In other words, we can discuss whether there is an “exhibition visit script” - as one may want to call it - for appropriating and experiencing exhibitions. On the basis of MET data we suggest that it is possible to reconstruct this script “bottom-up” by looking at distinct scan patterns that may form a repertoire of subscripts from which visitors draw as they combine these subscripts for their complex navigation strategies that help them to successfully meet the goals and agendas at the superordinate level (cf. Holmquist et al., 2011 and cf. Falk et al., 1998). MET data thereby allows us to flesh out the presumed script and add substance to the assumption that exhibition visits are guided by scripts.

On the basis of some preliminary, manual analysis of MET videos recorded in the “Pacific Oasis” exhibition, we were able to compile a list of systematic and repetitive behavioral patterns that may be considered distinctive key elements of an “exhibition visit script”. So far, this list consists of 14 scan patterns (scanning the objects and scenes displayed) that we have observed, described, labeled and classified (cf. Holmquist et al. 2011, p. 253-285 on the theory of scanpaths). However, this list is probably incomplete and it is likely that in further analysis other scan patterns will emerge or that closely associated scan patterns will be bundled together. Twelve of these 14 scan patterns were operationalized for quantitative analysis so that we were able to successfully double check them with systematic conventional observation at the level of the nine selected sections within the “Pacific Oases” exhibition at Stuttgart. This in turn allows us to claim that there is some robustness of this list across different methods of observation. Ten out of these 12 patterns were also evident in the MET data recorded in the “nexus” exhibition at Marbach. This is remarkable because these two exhibitions are very different in many respects so that it is not trivial to ask whether there is a similar script at work in visits to these two exhibitions.

On the basis of our preliminary data, we can now safely assume that at least these 10 distinctive scan patterns are very frequent in the way that visitors appropriate and experience exhibitions. They are good candidates for being part of the larger, general “exhibition visit script” that researchers have long assumed to be at work without having the detailed record to show what it consists of.

Some of the above mentioned scan patterns appear to be particularly representative for the practices of “looking and acting” (Land & Tatler, 2009) or “gehen und sehen” (strolling and viewing; Korff & Thiemeyer, 2008) in a three-dimensional gallery space. For the purposes of our discussion, here a few examples will suffice:

Changing perspective: We define this pattern as sequentially and systematically viewing an exhibit from a number of different vantage points or perspectives, a characteristic feature that the three-dimensional space (right/left, up/down, front side/back side) of an exhibition allows. For this scan pattern, visitors also have to move their head and sometimes even their body. This provides us with indicators that are more readily observable. There are typical indicators from the cued retrospective response that go together with the visitor doing a “change of perspective”, e.g. visitor statements such as wanting to know “how this mask was put up” on the wall.

Insight: We define this as looking inside an object (not necessarily from above). To realize this “insight” visitors have to move closer to the exhibits first with their body and second with their head before they can gain a look inside. For example “insights” were often recorded when visitors approached the men’s house model from Palau that was part of the “Pacific Oases” exhibition. One participant reported about the specific characteristics of its floor structure that was only visible when taking the “insight”, providing matching verbal evidence that an “insight” took place.

Backward gaze: Following Mayr et al. (2009) we define this pattern, typically part of an initial (“selection of information and visual research”, p. 195) orientation in a gallery space, as one that visitors employ to orientate themselves again after viewing a gallery as they gaze backwards before they leave a gallery. This includes at least a combined eye and head movement and sometimes even an eye-head-body movement (cf. Holmqvist et al., 2011, pp. 264-265 for a comparable event also called look-back, return or refixation).

Social gaze: We define this pattern as a gaze directed toward other persons in the environment. Sometimes a “social gaze” only consists of an eye-movement but mostly it is a combined eye and head movement and sometimes even an eye-head-body movement. When visitors reported that they felt affected by staff or foreign visitors like school classes (their closeness or noise) they often changed their direction and moved away. Sometimes this report was accompanied by a “social gaze” which underlines that social factors, and other senses such as hearing, are also influencing visitors’ movement patterns and not only the visual affordances of the displayed objects or the presentation style.

Thirdly, we note that an “exhibition visit script”, whatever its particular form, is not made up of a collection of gazes and scan patterns alone. Rather, we suggest that this may be an example of what Hutchins (1995) has called “distributed cognition” since it involves aspects of the materialized exhibition, above all the “affordances” of objects that allow insights or of the exhibition structures that channelize “social gazes” in a certain way (cf. Gibson, 1986, pp. 127-143). Building on the earlier work by Mayr et al. (2009), Eghbal-Azar (forthcoming) aims to provide a first comprehensive generic classification of museum visitors' eye movement patterns connected to affordances in exhibitions based on mobile eye-tracking that is lacking so far. The observed scan patterns are of course not limited to exhibitions but form a subsection of a larger repertoire of scan patterns used in everyday life. The “exhibition visit script” would denote a particular bundle, sequence and frequency of scan patterns. Similarly the scan patterns are related to expectations (some personal, others publicly debated) of what there is to see in an exhibition (including the “must-sees”). Although cognitive visitor research defines an exhibition visit as an “open-ended task” (Mayr et al., 2009, p. 191.), the motivations attributed to visitors are usually those of informal learning and receiving information. These appear to be rather passive strategies given that, in most cases, visitors are not allowed to act upon objects but only to view them. By contrast, the MET data supports the view that visitors are much more active than they otherwise may appear to be. Visitor research by social scientists tries to accommodate motivations and goals that go beyond effective information gathering, considering visitors as appropriating and experiencing exhibitions in an active and embodied fashion (MacDonald, 2002, p. 219).

We can therefore organize our MET observations not only in terms of a list of gazes employed by visitors looking for information. Rather, the observed scan patterns are also employed as navigation strategies looking for emotional and aesthetical experiences and with reference for social aspects to do with the presence of other persons. On the basis of our results we think it is plausible (and likely) that an “exhibition visit script” typically includes a number of navigation strategies and can flexibly accommodate a spectrum of agendas (see above). With this conclusion we can also deal with raised eye-brows towards MET that criticizes the inclusion of an “objective” measuring device as being reductionist. Applying MET does not entail that exhibition visits are reduced to information gathering strategies guided by principles of optimality and rational choice (cf. Bitgood, 2006 and Rounds, 2004) since the observed scan patterns may result from different - and at times conflicting - motivations and contextual conditions. Milekic (2010) has noted that MET in visitor studies provokes controversies between the disciplines concerned because “the major problem in adopting these technologies is the divide that exists between traditional notions of Art and Science.” We suggest that including the technology into different interdisciplinary approaches may help to design more integrated research questions and to arrive at more satisfying answers.

Conclusion: The Potentials and Limitations of Mobile Eye Tracking and Prospects for its Use in Field Research

MET in combination with other methods can provide us with new insights into very individual experiences, appropriation strategies, and goals of visitors. It can get us a step closer toward “strolling and viewing” an exhibition from the visitor's perspective. It may also help us to detect and outline unconscious “exhibition visit script(s)” that usually can hardly be verbalized by the agents themselves. These potential benefits matter to social scientists because MET opens the door toward an investigation that links unconscious aspects with socially and culturally constituted forms of embodied knowledge. It adds to our knowledge about what people look at, what they look for and why (cf. Land & Tatler, 2009, p. 222). Cultural schemata and scripts, whether they apply to museum exhibition visits or to other practices are notoriously difficult to investigate through external observations or through videotaping alone. The strength of MET is that it can help to break down fundamental questions of cognition and practice into very precise queries such as “What do persons look at while they are interacting with other persons and why?”, “What do persons communicate with their words and what do they communicate with their eyes?” The answers to such precise questions can then be aggregated into more general ones such as “How robust is the scan pattern that an individual uses to view a particular scene?” and “How similar is this scan pattern between individuals?.”

MET not only generates very detailed and precise data outside the laboratory but it also allows us to store that data so that it can be analysed again and again quantitatively or qualitatively with many possible variations depending on the research question. A single MET recording thereby potentially provides much more data than conventional observation. As with any other method, MET, too, is limited in what it can provide. For instance, there are technical limitations to do with the fact that our visual perception is only partially captured by a camera and that applying MET presupposes the availability of electronic gadget also under adverse field conditions. Future technical developments might reduce some of these technical limitations.

In the social sciences, we will want to combine MET with other methods such as field diary notes, interviews, questionnaires and cued retrospective reporting that provide help when trying to link our observations to the agents' goals of attention. MET does not replace the investigator's sense of understanding the particulars of a research setting, rather it presupposes such an underlying understanding because only on this basis can we tune its application to the requirements of the research context and the particular research question at hand. Applying MET is often laborious, certainly at this stage as we are lacking a software tool for comprehensive automatic analysis. However, the real challenge is the integration of this technique and the data that it produces into a social science research agenda that by its very nature will always rely on a number of complementary methods.

Footnotes

Authors’ Note

We would like to thank the German Federal Ministry of Education and Research (BMBF) for funding this project. We thank two anonymous reviewers who commented on an earlier version of this article, sole responsibility for the content of the article rests with the authors. We also thank Prof. Dr. Stephan Schwan, Dr. Daniel Wessel, and our student assistants, Linda Greci and Marie-Luise Saile, as well as the Media Technology Department of the KMRC for their support. This research would not have been possible without the cooperation of the Linden-Museum Stuttgart, especially Dr. Ingrid Heermann, and the German Literature Archive (DLA), and its Museum of Modern Literature, especially Dr. Heike Gfrereis and the visitors who agreed to participate in the study. We would also like to thank Prof. Dr. Thomas Thiemeyer, Yvonne Schweizer, and Felicitas Hartmann.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article has been generated within the research project “knowledge&museum: archive – exhibit – evidence” that is funded by the German Federal Ministry of Education and Research (BMBF) license number 01UB0909.

Notes

References

Baddeley

(2007). Working memory, thought and action. Oxford, England: Oxford University Press.

Bitgood

(2006). An analysis of visitor circulation: Movement patterns and the general value principle. Curator, 49, 463–475.

Eghbal-Azar

(Forthcoming). Affordances, Appropriation and Experience at Museum Exhibitions: From (Eye) Movement Patterns to Exhibition Visit Scripts. PhD thesis to be submitted.

Ericsson

K. A.

Simon

H. A.

(1993). Protocol analysis. Verbal reports as data (rev. ed.). Cambridge: MIT Press.

Falk

J. H.

Koran

J. J. R.

Dierking

L. D.

(1986). The things of science: Assessing the learning potential of science museums. Science Education, 70, 503–508.

Falk

J. H.

Moussouri

Coulson

(1998). The effect of visitors ‘ Agendas on Museum Learning. Curator, 41, 107–120.

Gfrereis

(2009). Didaktik des Schweigens: Das Literaturmuseum der Moderne des Deutschen Literaturarchivs Marbach. Der Deutschunterricht, 2009, 20–29.

Gibson

J. J.

(1986). The Ecological Approach to Visual Perception. Hillsdale: Lawrence Erlbaum.

Heermann

(2009). Südsee-Oasen. Leben und Überleben im Westpazifik Begleitbuch zur Sonderausstellung im Linden-Museum Stuttgart (5. Dezember 2009 – 6. Juni 2010). Linden-Museum Stuttgart/Neustadt/Wstr: Nino Druck.

10.

Hein

G. E.

(1998). Learning in the museum. London, England: Routledge.

11.

Hein

H. S.

(2000). The museum in transition: A philosophical perspective. London, England: Smithsonian Institution Press.

12.

Holmqvist

Nyström

Andersson

Dewhurst

Halszka

van de Weijer

(2011). Eye tracking: A comprehensive guide to methods and measures. Oxford University Press.

13.

Hutchins

(1995). Cognition in the wild. Cambridge: MIT Press.

14.

Just

M. A.

Carpenter

P. A.

(1980). A theory of reading: From eye fixation to comprehension. Psychological Review, 87, 329–354.

15.

Kaube

(2010). Richtige Betrachtung: Marbach blinzelt. Frankfurter Allgemeine Zeitung 26th August 2010. Retrieved from http://www.faz.net/artikel/C30108/richtige-betrachtung-marbach-blinzelt-30302784.html

16.

Korff

(2003). Staging science. Museumskunde, 68, 67–72.

17.

Korff

Thiemeyer

(2008, October). Authentizität versus Szenografie. Wie das Museum das Bild des Krieges verändert. Paper presented at the 47. Deutschen Historikertag in Dresden 2008. Proceeding retrieved from http://books.google.de/books?id=k9Hj1Uyt5r4C&pg=PA137&lpg=PA137&dq="gehen+und+sehen"+Thiemeyer&source=bl&ots=FIzLsqjv5L&sig=gbjWJoR2ku4n3uVWmla9e-M77Bw&hl=de&ei=HIkG

18.

Land

M. F.

Tatler

B. W.

(2009). Looking and acting: Vision and eye movement in natural behaviour. New York, NY: Oxford University Press.

19.

MacDonald

(2002). Behind the scenes at the science museum. Oxford, England: Berg.

20.

Mayr

Knipfer

Wessel

(2009). In-sights into mobile learning: An exploration of mobile eye tracking methodology for learning in museums. In Vavoula

Pachler

Kukulska-Hulme

(Eds.), Researching mobile learning: Frameworks, tools and research designs (pp. 189–204). Oxford, England: Peter Lang.

21.

Milekic

(2010). Inside your head: Use of eye-tracking for museum installations and websites. Retrieved from http://www.uarts-eyetracking.org/wp-content/uploads/2010/11milekicNodemPaper.pdf">http://www.uarts-eyetracking.org/wp-">http://www.uarts-eyetracking.org/wp-content/uploads/2010/11milekicNodemPaper.pdf

22.

Rounds

(2004). Strategies for the curiosity-driven museum visitor. Curator, 47,389–412.

23.

Schank

R. C.

Abelson

R. P.

(1977). Scripts, plans, goals and understanding: An inquiry into human knowledge structures. Hillsdale, NJ: Lawrence Erlbaum.

24.

Scheper-Hughes

(1994). Embodied Knowledge. In Borofsky

(Ed.), Assessing Cultural Anthropology. New York: Graw-Hill, 229–242.

25.

Serell

(1997). Paying attention: The duration and allocation of visitors‘ time in museum exhibitions. Curator, 40, 108–125.

26.

van Gog

Paas

van Merriënboer

J. J. G.

Witte

(2005). Uncovering the problem solving process: Cued retrospective reporting versus concurrent and retrospective reporting. Journal of Experimental Psychology: Applied, 11, 237–244.

27.

van Someren

M. W.

Barnard

Y. F.

Sandberg

J. A. C.

(1994). The think aloud method: A practical guide to modeling cognitive processes. London, England: Academic Press.

28.

Yalowitz

S. S.

Bronnenkant

(2009). Timing and tracking: Unlocking visitor behavior. Visitor Studies, 12, 47–64.