J Police Crim Psych DOI 10.1007/s11896-016-9213-x
Using an Eye Tracking Device to Assess Vulnerabilities to Burglary Thomas Zawisza 1
&
Ray Garza 1
# Society for Police and Criminal Psychology 2016
Abstract This research examines the extent to which visual cues influence a person’s decision to burglarize. Participants in this study (n = 65) viewed ten houses through an eye tracking device and were asked whether or not they thought each house was vulnerable to burglary. The eye tracking device recorded where a person looked and for how long they looked (in milliseconds). Our findings showed that windows and doors were two of the most important visual stimuli. Results from our follow-up questionnaire revealed that stimuli such as fencing, beware of pet signs, cars in driveways, and alarm systems are also considered. There are a number of implications for future research and policy. Keywords Burglary . Vulnerability to burglary . Decision-making . Eye tracking . Visual attention
Introduction Eye tracking devices (ETDs) capture data that otherwise could not be collected. One advantage of using eye tracking devices is that it provides objective measurements as to where and what a person is looking at and for how long (in milliseconds; Rayner 1998). Such technology has been used in a number of fields, including medical research (Krupinski and Nishikawa 1997; Kundel et al. 2007), research in psychology (Dixson et al. 2014; Rayner 1998; Suschinsky et al. * Thomas Zawisza
[email protected] Ray Garza
[email protected] 1
Texas A&M International University, Laredo, TX, USA
2007), marketing (e.g., Clement 2007), and sports research (Abernethy and Russell 1987), to name a few. By using such technology as a primary method of data collection, insight can be gained as to what effect visual stimuli and eye movements have on the decision-making process. The use of this technology, however, has been almost nonexistent in the field of criminology and criminal justice. To our knowledge, only three such studies exist that have used such technology, and only two of those directly test the influence of eye movements and visual stimuli on decision-making (Busey et al. 2011; Jacques et al. 2015; Neveu et al. 2011). One area of research in criminology and criminal justice that would benefit from the use of ETDs is burglary research. Over the past four decades, there has been an impressive body of research dedicated to understanding where burglaries occur (Bernasco and Luykx 2003; Bernasco and Nieuwbeerta 2005; Johnson and Bowers 2010; Wiles and Costello 2000; Wright et al. 1995) and how burglaries are committed (see Cromwell and Olson 2004; Rengert and Wasilchick 1985; Wright and Decker 1996). Though this body of research has provided valuable information on how certain cues increase or decrease the likelihood of burglary, we do not know the extent to which these offenders focused on those cues. The scope of this research is twofold. First, our aim was to better understand the effect of visual stimuli and eye movements on the immediate decision to burglarize. This provided us precise recordings on where individuals focus their visual attention while making a decision to burglarize. The importance of this methodology is that it allowed us to investigate the intuitive responses when deciding to burglarize and the post-intuitive responses once participants were asked for reasons to their responses, an aspect that has yet to be studied in burglary research. The second aim of this study was to spark a discussion on the use of ETDs in criminology and criminal justice research. Burglary is just one area of research that
J Police Crim Psych
would benefit from the use of this technology. The use of ETD technology seems to have implications in other related areas of research such as theft, eyewitness testimony, shoot-noshoot scenarios, and crime scene investigation. Burglary Literature Indeed, for a burglary to occur, motivation is a necessary requirement, but it is not sufficient (Cromwell and Olson 2004; Rengert and Wasilchick 1985); a suitable target must also be available. Evaluation of targets occurs in three stages: first, at the neighborhood level, then at the street level, and, lastly, at the site-specific level (house; Bernasco and Nieuwbeerta 2005; Cromwell and Olson 2004). At the neighborhood level, offenders first look for neighborhoods in which they are not easily identifiable. Such places are often located within an offender’s awareness space (areas where he/she is familiar with) and action space (areas where he/she conducts daily business in; Brantingham and Brantingham 1981; Rengert and Wasilchick 1985). Not only are most burglaries within offenders’ awareness space, but offenders and the place also tend to be ethnically and culturally similar to their place of residence (Rengert and Wasilchick 1985). Offending in areas that are demographically similar reduces the likelihood of being reported by a passerby since there is a sense of belongingness. If an offender finds a place where he/she would stand out, extra measures are taken to reduce detectability. For instance, offenders may dress as maintenance or city workers to create an illusion that they belong in the neighborhood to reduce suspicion (Wright and Decker 1996). With respect to street design and location, offenders tend to shy away from places that are complicated to navigate (for example, see Beavon et al. 1994; Johnson and Bowers 2010). Beavon et al. (1994) showed that neighborhoods that are not easily accessible from major roadways are less attractive than those in which offenders can easily access. Neighborhoods that have a large number of turns are also viewed as less advantageous than those that have direct routes to major roadways. Vernon and Lasley (1992) also demonstrated how dead-end streets could impact the occurrence of crime. In South Las Angeles, they created pseudocommunities by blocking off ten city blocks. At the end of the trail, when compared to control areas, dead-end streets had a 20 % reduction in Part I index crimes. The most important stimuli for target selection at the site level are residence occupancy and visibility (Logie et al. 1992; Nee and Taylor 2000; Wright and Decker 1996). Cars in driveways, open windows, blinds that are drawn, and houses with lights on are some of the most obvious signs of residence occupancy. Experienced burglars look for more obscure cues when evaluating occupancy, such as running air conditioners (Rengert and Wasilchick 1985). As one offender noted, when a house is locked up, with no air conditioning on, during
summertime, it is not likely that a person is home (Rengert and Wasilchick 1985). Other stimuli such as fences, trees, and shrubs also influence the decision to burglarize at the site level. Shrubbery and the presence of trees provide offenders with cover, reducing detectability (Clarke and Hope 1984; Cromwell and Olson 2004). Shrubbery in proximity to a house is especially problematic since it provides camouflage close to the target. Further, houses with opaque fencing around a perimeter are viewed as favorable to burglary since the view to the offender from passersby and neighbors is obstructed (Clarke and Hope 1984; Logie et al. 1992). At first, the risk for burglars is high since they may have to climb the privacy fence to enter the property; however, once over the fence, the likelihood of an offender being observed by neighbors and passersby decreases. At the point of entry, dogs, security systems, and locks are three of the most common stimuli that affect the decision to burglarize (Clarke and Hope 1984; Cromwell and Olson 2004; Rengert and Wasilchick 1985). Most offenders will avoid houses where dogs are present. Dogs are problematic because they are capable of attracting attention and can be a real threat to violence. At first, offenders in Cromwell and Olson’s (2004) sample overlooked houses with dogs or had contingency plans for dealing with dogs. Yet, even those with contingency plans eventually agreed that houses with dogs are not viable targets. Houses with security systems have less risk for burglary than those without (Buck et al. 1993). Buck et al. (1993) reported that houses with alarms were 2.71 times less likely to be burglarized than houses without alarms. Most offenders are not equipped to deal with alarms; they lack the tools or knowledge to disarm them (Cromwell and Olson 2004). As one burglar told Cromwell and Olson (2004), BSometimes I pick a house to do and when I get up close I can see the wires taped to the window and I know they got an alarm. I just move on^ (p. 29). Of course, there are those offenders who circumvent alarms. For instance, one burglar described how he drilled a hole in the garage wall to avoid tripping an alarm. Some offenders (see Rengert and Wasilchick 1985) view a locked door as a problematic obstacle, which requires time and skill to overcome. These offenders would rather search for a house requiring less effort to enter. Others (see Bennett and Wright 1984; Logie et al. 1992; Wright and Decker 1996) are not deterred by locks and deadbolts since they can be overcome with skill and ingenuity. It appears as though the effect of locks on the decision to burglarize is moderated by other factors, particularly time. When questioned about the presence of a lock, one burglar said, BI gotta be in and out in 2, 3 min. I ain’t got time to mess with no tough lock^ (Cromwell and Olson 2004, p. 31). It is unlikely, then, that locks are insignificant stimuli.
J Police Crim Psych
ETDs in Social Scientific Research The use of eye tracking technology in investigating decisionmaking has primarily been used in psychological research (see Rayner 1998). Research in psychology has shown that differences in eye movements indicate an initial stage in processing (Yantis 2005) and level of difficulty in reading (Rayner 1998; Rayner et al. 2010) and demonstrate interest when viewing a specific area. Linguistic research has been a proponent of using eye tracking devices because they provide insight into the perceptual and reading span of readers. Rayner et al. (2010) found that slow, young, dyslexic, and old readers demonstrate a reduction in perceptual span when reading text. More pertinent to our study, ETDs have shown that the amount of time one spends viewing a specific region of interest (ROI) highlights a level of attractiveness toward that particular region. Recently, a collection of research has shown that time spent on a particular image indicates interest and attraction that can be quantitatively measured in milliseconds. Research in the field of attraction and visual processing has shown that infants stare longer at female faces and familiar adults, indicating a level of interest (Langlois et al. 1987). Others have shown that individuals will particularly look at images longer, such as pictures of the opposite sex that contain erotic or appealing content (Lykins et al. 2006). However, recent research has used more objective techniques when recording visual measurements, such as tracking an individual’s first fixations, gaze duration, and total time. These measurements allow researchers to investigate the level of interest or attractability an individual has toward a particular area that cannot be obtained through qualitative measurements. Although most of these studies have been done in the field of physical attraction (Suschinsky et al. 2007; Dural et al. 2008; Hewig et al. 2008; Dixson et al. 2010), they lend to this study the importance of viewing time when looking at images, and decision-making. That is, ETD provides researchers with a behavioral measurement in recording interest to specific characteristics. Most, if not all, of ETD studies used first fixations, gaze duration, fixation count, and total time as their common method of visual analysis to indicate visual preference, attraction, or interest.
opposite was true when viewing pictures of adults; the control group had greater penile blood flow. No significant difference was found between the two groups when viewing a neutral image. As noted by the authors, virtual immersion made it more difficult for the participants to modify penile response to the images since they were fully immersed and eye movement was tracked. To study the decision-making of shoplifters, Jacques et al. (2015) used a purposive sample of active shoplifters and recorded their eye movements during simulated shoplifting events. To track eye movements, participants wore glasses that recorded their line of sight and the movement of their right pupil. Software combined these two images to create a frameby-frame, in milliseconds, video of the encounter. After the shoplifting simulation, the authors asked the participants a number of open-ended questions to better understand their decision to shoplift or not. Their findings indicate that (1) passively seeing security cameras may increase the awareness of risks involved in shoplifting; (2) shoplifters often avoid looking directly at security cameras to appear more natural; and (3) ETDs enhance the validity of qualitative responses by offsetting memory decay and limiting participant embellishment. In the field of forensic science, Busey et al. (2011) used ETDs to test whether or not there was a difference in the methods used to correctly identify pairs of latent prints, one lifted and one known, between novice latent print examiners and expert latent print examiners. They used a mounted ETD that showed participants a pair of prints, and they were asked to determine whether the prints were a match or non-match. Not unexpectedly, expert examiners performed better in both experiments than did the novice examiners. They found that novice examiners tended to view prints quicker and focused on details that were easily readable. Expert examiners tended to spend more time viewing prints and focused on minute details of both sets of prints. When expert examiners were given an unrestricted amount of time to view prints, they spent more time looking at the prints, first looking for areas with high clarity. This is much different from what has been reported in the medical field, where experts in their respective areas were found to make quick eye movements to identify problem areas quicker.
ETDs in Criminology Current Study To our knowledge, there are only a few studies that have used ETD technology to investigate criminological phenomenon. Neveu et al. (2011) used virtual immersion and penile plethysmography to evaluate sexual arousal toward children. Level of sexual arousal was compared between males who had sexually abused children (n = 29) and males with no prior sexual deviance (n = 27). The results showed that males who had sexually abused children had greater penile blood flow than the control group when viewing images of children. The
The current study focuses on visual attention in the decisionmaking process when assessing vulnerability to burglary. Previous research in the eye tracking criminology literature has primarily focused on shoplifting and the areas where shoplifters look at when inside a place of business (Jacques et al. 2015). However, eye tracking in decision-making research in assessing vulnerability to burglary is nonexistent. By analyzing visual responses to specific characteristics in houses, we
J Police Crim Psych
can address whether characteristics in a person’s house are more salient when looking at vulnerabilities to burglary. Although the current study does not use experienced burglars, we attempt to utilize a new methodology that can quantify in milliseconds the decision-making process and visual attention of areas that individuals may consider important in houses that are susceptible to or at risk of being burglarized. In this study, photos of front-facing houses were used and inputted into an eye tracking device. While previous research has shown that decisions about target selection are based on signs of occupancy, nearby activity, and neighborhood characteristics, we primarily use front-facing photos of houses only to capture an ecological image that can be seen clearly in an eye tracking device. Therefore, in this study, we seek to answer the following: (1) What characteristics are salient when deciding to burglarize a house? (2) Does visual attention relate to burglary vulnerability? (3) What is the relationship between visual attention and the socioeconomic status of homes? This study will also look at participants’ responses as to why they decided to burglarize a home. Research in criminology and related fields has shown that the amount of time spent on specific areas demonstrates visual interest to that particular area. In this study, a battery of visual measurements will be used to identify early stages of visual processing (i.e., first fixation duration) and later-stage processing (i.e., gaze duration, total time, and fixation count). The study lends itself to the following hypotheses: (1) characteristics that suggest a home’s probable value, such as the windows, will receive the most visual attention; (2) houses that are vulnerable to burglary will result in increased viewing time; and (3) houses of high socioeconomic status will result in increased viewing time.
Method Participants Participants were 65 (men = 13 and women = 52) undergraduate students from Texas A&M International University (TAMIU) in Laredo, TX, ranging in age from 18 to 47 years (M = 22.75, SD = 6.30). Participants signed up for the study through the university SONA system and received course credit for participation. Given the nature in which the study was announced, and the method in which the participants signed up for the study, the sample of participants is considered a convenience sample. Furthermore, only one participant in this sample indicated that he/she had committed a burglary. For all intents and purposes, then, we are considering this sample of participants to be novice burglars, with little to no experience of burglary prior to this study.
Materials Ten color photographs of front-facing houses were taken in Laredo, TX. The houses were located from different social economic areas around the city to represent homes that were considered of low, medium, and high economic status. Although the exact income per house was not reported, the houses were located in areas that are known to be characteristic of income disparities. The houses were classified as having specific ROIs to categorize which areas would be the recording point when entered into the eye tracking device. The ROIs used were the windows and doors. The images were input into the SR-Experiment Builder and altered to meet the size requirements for clear viewing. The same process was done for all images where the windows and doors were classified as ROIs. Procedure Participants signed up for the study using the online participant sign-up software SONA. Participants chose a specific time and day to participate and were given course credit using the online software as well. Upon arriving, participants were given consent forms and instructed on how to proceed using the eye tracker. Participants were instructed to sit comfortably and rest their chin on the chin rest to maintain stability. Once seated, visual calibrations and validations were done to ensure that the eye tracking device was accurately recording the correct visual movement. This allowed us to see whether there were any discrepancies between the visual instruments and where the participants were actually looking. The distance between the chin rest and the eye tracker was kept standardized (55 cm), as has been done in studies examining visual perception and attention. In all eye tracking trials, the recording was monocular, meaning that the right eye was being recorded. The recording was done on the Eye-Link 1000 Tower Mount Head Supported System (SR Research Ltd., Ontario, Canada) running Windows OS 7. A fixation was defined as lasting longer than 50 ms. Instructions were presented both orally and visually. Participants were instructed to read the instructions presented on the computer screen; however, the research assistant would also repeat the instructions to ensure that participants comprehended the procedure and know how to use the remote control to make ratings. A Microsoft Sidewinder Plug and Play Gamepad was used to rate the vulnerability to burglary. In each trial, participants were shown a picture of a house in Laredo and were asked to rate whether the house was vulnerable to being burglarized, where the left trigger indicated Bno^ and the right trigger indicated Byes.^ There were a total of ten images of various houses in Laredo, TX, and all of the trials were randomized for each participant. After each house was shown, a fixation point
would appear in the center of the screen so that participants could fixate their eyes on the fixation dot before proceeding to the next trial. The purpose of providing a fixation point was to ensure that there were no errors and that fixations were being recorded accurately. If a participant did not focus their visual attention on the fixation point, the eye tracker would not show the next image. Once participants were done with viewing images in the eye tracker, they were given a follow-up questionnaire on why or why not the houses seemed most vulnerable to burglary. Aside from demographic information, participants were also asked whether they had ever burglarized another person’s home. The eye tracker recorded the following visual measurements: first fixation duration, gaze duration, fixation count, and total time. Each visual measurement was dependent on the image having a specified ROI, with the exception of total time. First fixation duration was the amount of time spent on the first ROI (i.e., windows and doors) that participants viewed. Gaze duration was defined as the total amount of time spent on a specific ROI for each trial. Total time was the amount of time spent on the entire image. Therefore, for total time, ROIs were not specifically analyzed individually as total time records the amount of time spent on the entire image. Lastly, fixation count was defined as the number of times a ROI was fixated upon.
Results Data for the eye tracker were submitted to a linear mixed effects analysis using IBS SPSS (version 22), with items (images) and subjects (participants) as random effects and ROIs and burglary decision as fixed effects.
First Fixation Duration (ms)
J Police Crim Psych 350 300 250 200
Windows
150
Doors
100 50 0
Fig. 1 Mean first fixation duration (in milliseconds) as a function of regions of interest
most visual attention. Gaze duration should be considered a post-automatic visual response since participants are still viewing other parts of the houses, as opposed to a first fixation, which is more automatic. There was a main effect for ROIs [F(1, 817.88) = 224.09, p < .001]. Post hoc analysis revealed that participants focused most of their visual attention toward the windows than the doors [t(817.88) = 14.97, p < .001] (see Fig. 2). Fixation Count Fixation count was measured by recording the total amount of fixations that entered and reentered a ROI. There was a significant main effect for ROI [F(1, 816.76) = 225.48, p < .001]. Participants made more fixations toward the windows (M = 5.39) than the doors (M = 1.38) in all of the houses used [t(816.76) = 15.42, p < .001] (see Fig. 3). Total Time
First Fixation Duration
Gaze Duration For gaze duration, the average time that participants focused on a particular area throughout the entire trial was used. Regions of interest were used to determine which characteristics of the house (e.g., doors and windows) would receive the
Total time was measured by recording the amount of time participants viewed each photo. There was a significant main effect for house viewed [F(9, 810.56) = 9.15, p < .001]. In order to see whether viewing time was consistent with the 3500 Gaze Duration (ms)
First fixation duration was measured by the amount of time participants spent on their first fixation. In many eye tracking studies, a fixation is defined as lasting longer than 50 ms (Dixson et al. 2014). We examined the ROIs to determine where participants automatically view the houses and how long their fixations were. The results revealed a significant main effect for ROIs [F(1, 820.89) = 7.10, p = .008]. In this analysis, participants focused most of their first fixation time on the doors of the houses [t(820.89) = 2.67, p = .008] than on the windows (see Fig. 1).
3000 2500 2000 1500 1000
500
Windows Doors
0
Fig. 2 Mean gaze duration time (in milliseconds) for each house as a function of region of interest
J Police Crim Psych 6000
12 10 8
Windows
6
Doors
4 2 0
Mean Total Time
Mean Fixation Count
14
5000 4000 3000
Total Time
2000 1000
0
Low SES
Medium SES
Social Economic Status
High SES
Fig. 3 Mean fixation count for each house as a function of region of interest
Fig. 5 Mean viewing time (in milliseconds) for each house categorized by socioeconomic status
choice to burglarize a home, participants’ decisions were entered as a binary variable (i.e., NO/YES) in predicting the amount of time they viewed each house. There was an interaction between the house shown and the decision made [F(9, 820.48) = 4.62, p < .001]. Post hoc analysis showed that for most of the homes used, participants’ decision to burglarize resulted in less time spent viewing the home. Figure 4 shows the overall distribution of total time and decision to burglarize. The houses were then divided into three levels, each representing the socioeconomic status of the neighborhoods. The three levels of socioeconomic status (SES) were low, medium, and high. Socioeconomic status was significant when measuring total time [F(2, 896) = 11.42, p < .001]. Houses that were classified as of high socioeconomic status were viewed more than medium economic status homes [t(631) = 3.69, p < .001] and low economic status homes [t(634) = 4.32, p < .001] (as shown in Fig. 5).
consistently shown throughout each trial, the doors and windows of each of the houses were looked at the longest. The fixation duration for each of these areas ranges from 480 to 843 ms. As shown, these areas have a deep red color, indicating a long duration of fixation. Other characteristics that were looked at to a lesser extent were cars, fences, garage, doors, mailboxes, and, for one house, beware of dog sign.
Fixation Maps
Mean Total Time (ms)
Fixation map analysis was conducted as a means to present a visual representation of where a person was looking at and for how long. Figures 6, 7, 8, 9, 10, 11, 12, 13, 14, and 15 (see Appendix) present the fixation maps for each of the trials. As 8000 7000 6000 5000 4000 3000 2000 1000
Not Burglarize Burglarize
0
Fig. 4 Mean viewing time (in milliseconds) for each house as a function of their burglary decision
Post-study Questionnaire After participants finished the eye tracking portion of the study, we asked them to complete a post-study questionnaire. Of particular interest were two follow-up questions: (1) BOf the houses you viewed, did you observe any that seemed vulnerable to burglary? Why or why not?^ (2) BOf the houses you viewed, did you observe any that seemed not vulnerable to burglary? Why or why not?^ The major factor that influenced participants’ decision-making was housing characteristics. Fifty-four of the 65 participants indicated that the aesthetics of a house influenced their decisions. Generally, participants belonged to one of two groups: those who preferred Bfancy, big^ houses (n = 30) and those who preferred houses that looked Brun-down^ (n = 24). The first group of participants believed that aesthetically pleasing houses were best suited for burglary because Ba house that big and that nice, should have a big family with lots of stuff inside.^ They imagined these houses having nice furniture, electronics, jewelry, and Bfancy dishes.^ For the other group of participants, houses that looked aesthetically pleasing were Blikely to have some type of security system^ and burglarizing Bthose houses would be risky.^ Instead, these participants favored houses that Blooked like no one really cared about them.^ Another participant said, Bthe houses in the poor neighborhoods seemed easy to penetrate. The doors seem really easy to force, and the windows seemed really easy to open or break.^ In almost all of the follow-up questionnaires, participants noted that windows were important because they allowed
J Police Crim Psych
them to Bsee if anyone was inside^ or Bif the house had anything worth stealing.^ This reaffirms our findings of the ETD portion of this paper that windows are important factors in the decision-making process. Surprisingly, the presence of a car was only a mild to moderate deterrent. Some participants used cars as an indicator of wealth; the nicer the car, the more likely the house is to have valuable items. It was not clear, however, whether or not they would burglarize that house in that moment or, as we hypothesize, that these participants would return at a later time when there was no resident occupying the house. The majority of participants also indicated that the house with the Bbeware of dog^ sign is not likely one they would burglarize. As one participant stated, BI’d rather not burglarize that house because I would have to deal with the dog.^ It seems that the participants in this sample and the experienced burglars in prior research view dogs as threats of detection. We also found that participants were more likely to say that houses with fences were less vulnerable to burglary than houses without. One participant indicated that fences limited access to the house. The participant said, Bit would be hard to get over and then get away.^
Discussion The purpose of the present study was to examine visual attention when making decisions in burglarizing a house. We addressed whether specific areas in a house were more salient when searching for vulnerabilities to burglary. A variety of visual measurements were used to record the early-stage processing (i.e., first fixation duration and gaze duration) and late-stage processing (i.e., total time and fixation count) in decision-making. The current study showed that in the early stage of visual processing, the doors received the longest first fixation duration. That is, when the eye first fixates on a particular region of interest (ROI), it fixates longer when viewing the door. Gaze duration showed that participants focus most of their time on the windows before viewing other features and when returning to previously viewed areas. To our knowledge, measuring the early stages of viewing time has not been done in the field of criminology. This suggests that since the door is mainly the center point of a house, viewing time tends to gravitate to the door when first viewing a house, but at first glance, the windows are more salient. However, one of the main focal points of this study was to address whether certain characteristics were more salient when deciding to burglarize a house. To accurately measure this, fixation count was used to record the overall visual interest when viewing houses. The windows received the most visual attention. For fixation count, participants spent a significantly more amount of time on the windows than the doors. This
finding suggests that windows are more salient when deciding to burglarize a house regardless of type (i.e., SES). Another focus of this study was to determine whether the time spent on each image would reflect the decision to burglarize the house. In most of the houses that were rated as burglarable, participants spent less time viewing the house than the houses not vulnerable to burglary. When separated by socioeconomic status, homes of high socioeconomic status received the most viewing time. The findings of the current study are mostly consistent with previous literature on burglary in criminology (see Logie et al. 1992), eye tracking research in criminology (see Jacques et al. 2015), and psychology (see Rayner 1998). Research in criminology has shown that burglars are less likely to focus on locks and doors due to the timeconsuming effort to unlock them and because of time constraints (Cromwell and Olson 2004, p. 31). The participants in our study did not find the doors particularly salient since most of their visual attention was focused on the windows and not the doors. Although at first view the doors did receive the most visual attention, it was only in reference to the first fixation and not the entire trial length visually inspecting the house. The present study supports previous research on eye tracking research suggesting that time spent is an indication of interest. Studies on facial attraction and physical attraction have revealed that time spent on particular regions of interest is a measurement of attractability and interest (Langlois et al. 1987). Therefore, in our study, we can conclude that participants did find the windows as attractive regions when making a decision to burglarize a house. Our second hypothesis suggesting that time would reflect the decision to burglarize was not supported. We found that participants who chose to burglarize a house spent less time viewing the house. Although research by Busey et al. (2011) had demonstrated that experts in the field of latent print examination spend less time than novice examiners, in this study, experts were not used. We do not know whether less time in deciding to burglarize reflects expertise in burglary as our participants were predominantly undergraduate students, who, except for one participant, had no expereince prior to this research. There were three novel findings that contribute to the field of criminology examining burglary. First, this study is the first to examine burglary using eye tracking technology. Eye tracking technology provides an immense amount of information that can address early- and latestage processing when deciding to burglarize a house. Decision time may have significant implications in home security, as we know that deciding to burglarize a house is not a time-consuming process. In this study, we find that there are differences when first viewing the house (e.g.,
J Police Crim Psych
first fixation duration and gaze duration) and the overall viewing time (e.g., total time and fixation count). At first, participants gravitate toward the doors, but then spend a considerable amount of time on the windows. This suggests that doors are the automatic area of visual importance and windows are the post-automatic area of interest. Second, there are specific characteristics that individuals focus most of their attention to when assessing vulnerability to burglary. The windows attract the most visual attention, even when using images that had other surrounding characteristics, such as fences, signs, and vehicles (see Fig. 10). Though previous research has collected self-report data and qualitative measurements to suggest that windows may be attractive to burglary, in this research, we quantitatively find that windows are visually appealing. Lastly, visual attention is significant when considering the socioeconomic status of the neighborhood. In this study, we can conclude that houses of high socioeconomic status do receive considerable amount of visual attention. A house of high socioeconomic status suggests that there are highly valuable items, and visual attention is allocated to those houses that are of high economic value. Limitations This study is not without limitations. One of the first limitations we noticed was that peripheral vision could not be accounted for. For instance, some of the houses had fencing around the perimeter, and the fixation map shows that individuals looked at least for a moment at these fences; however, they were not the most salient areas within the fixation heat maps. Yet, in the followup questionnaire, several participants noted that the presence of fences influenced their decision-making. It could be that the relative size of the fencing, or other structures on the premises of the house, does not require an individual to fixate on that specific area or that the peripheral regions of the house are not that important. Rather, one could look at a window and still notice fencing or other characteristics within the perimeter of that house. As such, it may be that participants were indeed focused on a fence, but not in a way that could have been empirically measured. Another limitation of this study is the sample in which the data came from. The participants in this study were students at a university, over 90 % Hispanic, who were, for the most part, inexperienced burglars. It is unknown how these results compare to individuals with prior experience. Because of this unique population, caution must be taken when interpreting the results; they may not be generalizable to other populations. Finally, ethnographic investigations of active burglars indicate that a large portion of offenders prefer to enter a
house through the backyard because they are less likely to be detected. In fact, Bennet and Wright’s (Clarke and Hope 1984) sample of offenders indicated that they would prefer to enter through a backyard because it reduces the chances of a neighbor observing them. Our investigation could not account for alternate entrances to a house. It is possible that cues in the front yard negatively impacted a person’s decision, while cues in the backyard may entice participants to respond positively toward burglarizing a house. Despite the noted limitations, our findings have a number of implications for future research in assessing burglary using eye tracking technology. Eye tracking technology is a robust method for quantifying the visual interest in assessing vulnerability to burglary because it provides a quantitative analysis that indicates user interest. That is, ETDs are behavioral measurements in detecting interest. Researchers investigating interest in burglary should consider this methodology because it further strengthens selfreport and qualitative responses in assessing vulnerability. In using eye tracking technology, there are a couple of suggestions researchers may consider when embarking on using this methodology. First, the specific measurements used indicate different levels of visual interest, either indicating automatic processing or post-automatic conscious responses. Using first fixation duration and number of first fixations (not used in this study) is a strong method in addressing the early stages or automatic visual interest to particular stimuli. Gaze duration, total time, fixation count, and visual regressions (not used in this study) are a strong method in addressing overall visual attention, or post-automatic visual interest. Although visual regressions were not used, they can be used to discern where is the last focal point a person looks at when making the final decision to burglarize. Second, we recognize that what is seen in one’s periphery could not be accounted for in this study; however, our study was not the first to mention such a limitation. Several participants in the shoplifting study of Jacques et al. (2015) noted that one does not have to look directly at a security camera to know it is there. Unlike those of Jacques et al. (2015), however, participants’ answers in our sample to the post-study questionnaire were corroborated by the heat maps. In other words, there were no fixation points on the heat maps that did not match with the answers participants gave. The differences in our findings from those of Jacques et al. (2015) could be related to the settings in which the studies were conducted in. For instance, our study was in a controlled setting, with relatively low stakes; we did not ask participants to commit a burglary. On the other hand, the participants in the Jacques et al. (2015) study could have stolen an item from the store
J Police Crim Psych
and, as a result, were more likely to be aware of the stimulus around them. Nevertheless, we can only offer suggestions as to how to combat this problem. First, researchers might limit pictures viewed at a stationary machine so that peripheral vision would not matter. Of course, this would require researchers to focus on one or two main stimuli, restricting the number and types of research questions that could potentially be asked. Perhaps more feasible is to explore technologies that mitigate the impact of peripheral vision. The use of this technology is relatively new in criminal justice and criminology. Our hardware/software package does not have the capabilities to control for peripheral vision, so exploration of other hardware/software combinations is needed. Third, researchers could simply ask about peripheral vision. Future studies may consider manipulating stimuli (i.e., including security cameras and signs) in one’s periphery to see whether participants are able to detect changes in images. Finally, as a methodological suggestion, researchers should consider using manipulations that are consistent with previous research on burglary, such as manipulating signs, fences, or vehicles. That way, different focal points or ROIs can examine whether other characteristics that are in the peripheral vision are more salient. Since burglars are less likely to burglarize a house with noticeable occupancy, occupancy could be manipulated. Lastly, to increase the ecological validity of the study, using actual burglars to visually inspect houses and record their decision to burglarize would further support whether expertise translates to (a) less time making a decision, as seen in Busey et al. 2011, and (b) difference in the types of cues that were not seen from this sample of novice burglars.
always moved from the front door to a window; windows were much more important than doors on target vulnerability. We hypothesize that windows provide a person with important information, such as residence occupancy. Finally, we found that fences, cars in driveways, and beware of dogs signs were discussed to great length during the post-study questionnaire, but these stimuli did not receive as much visual attention as the windows or doors. It is possible that participants actually looked at these stimuli longer, but they were in their periphery and could not have been measured by our technology. It is also possible that these stimuli have such great influence on the decision to burglarize that not much visual attention is required. These findings do not deviate from what was found in prior research; doors and windows are the two most important stimuli when assessing a target. This research is unique in that it showed how much time a person spends looking at different stimuli. Though our study had notable findings, more research in this area is needed. It is still unknown as to how experienced burglars would perform compared to those with no experience as it is related to the eye tracking device. We also know that the target evaluation is not made in a vacuum; stimuli from the street and neighborhood also influence decision-making. Unfortunately, the apparatus that was used in this study would not allow us to use pictures of entire neighborhoods. The characteristics of houses would become too small and participants would not have been able to see them. Future studies could also study the potency of stimuli by manipulating certain features of houses, such as burglar bars, to investigate how it influences decision-making.
Conclusion
Appendix
Prior research has relied on the use of pictures or interviews to understand how cues influence a person’s decision to burglarize (Cromwell and Olson 2004; Logie et al. 1992; Nee 2015; Nee and Taylor 2000; Rengert and Wasilchick 1985; Wright and Decker 1996). Our study sought to further understand how cues influence a person’s decision to burglarize by using an eye tracking device to measure where and for how long a person looks at different stimuli. This type of technology has yet to be used for investigating burglary. Our study had three notable findings. First, participants’ initial focus was on the front door. This was not surprising since the front door is the main entry point into a house. Second, participants almost
Fig. 6 Heat map trial 1
J Police Crim Psych
Fig. 7 Heat map trial 2
Fig. 11 Heat map trial 6
Fig. 8 Heat map trial 3
Fig. 12 Heat map trial 7
Fig. 9 Heat map trial 4
Fig. 13 Heat map trial 8
Fig. 10 Heat map trial 5
Fig. 14 Heat map trial 9
J Police Crim Psych
Fig. 15 Heat map trial 10
References Abernethy B, Russell DG (1987) The relationship between expertise and visual search strategy in a racquet sport. Hum Mov Sci 6:283–319 Beavon DJ, Brantingham PL, Brantingham PJ (1994) The influence of street networks on the patterning of property offenses. Crime Prevention Studies 2:115–148 Bennett T, Wright R (1984) Burglars on burglary: Prevention and the offender. Gower, Aldershot Bernasco W, Luykx F (2003) Effects of attractiveness, opportunity and accessibility to burglars on residential burglary rates of urban neighborhoods. Criminology 41:981–1002 Bernasco W, Nieuwbeerta P (2005) How do residential burglars select target areas? A new approach to the analysis of criminal location choice. Br J Criminol 45:296–315 Brantingham PJ, Brantingham PL (eds) (1981) Environmental criminology. Sage, Beverly Hills Buck AJ, Hakim S, Rengert GF (1993) Burglar alarms and the choice behavior of burglars: a suburban phenomenon. J Crim Just 21:497–507 Busey TA, Yu C, Wyatte D, Vanderkolk J, Parada F, Akavipat R (2011) Consistency and variability among latent print examiners as revealed by eye tracking methodologies. JFI 61:60–91 Clarke R, Hope T (eds) (1984) Coping with burglary-research perspectives on policy. Kluwer-Nijhoff, Boston Clement J (2007) Visual influence on in-store buying decisions: an eyetrack experiment on the visual influence of packaging design. J Mark Manag 23(9–10):917–928 Cromwell PF, Olson JN (2004) Breaking and entering: burglars on burglary. Thomson/Wadsworth, Belmont Dixson BJ, Grimshaw GM, Linklater WL, Dixson AF (2010) Watching the hourglass. Hum Nat 21:355–370 Dixson BJ, Grimshaw GM, Ormsby DK, Dixson AF (2014) Eyetracking women’s preferences for men’s somatotypes. Evol Hum Behav 35:73–79
Dural S, Cetinkaya H, Gulbetekin E (2008) The role of the waist to hip ratio in evaluation of female physical attractiveness: eye-tracker data. Turk J Psychol 23:75–91 Hewig J, Trippe RH, Hecht H, Straube T, Miltner WH (2008) Gender differences for specific body regions when looking at men and women. J Nonverbal Behav 32:67–78 Jacques S, Lasky N, Fisher BS (2015) Seeing the offenders’ perspective through the eye-tracking device: methodological insights from a study of shoplifters. J Contemp Crim Justice 31:449–467 Johnson SD, Bowers KJ (2010) Permeability and burglary risk: are culde-sacs safer? J Quant Criminol 26:89–111 Krupinski EA, Nishikawa RM (1997) Comparison of eye position versus computer identified microcalcification clusters on mammograms. Med Phys 24:17–23 Kundel HL, Nodine CF, Conant EF, Weinstein, SP (2007) Holistic component of image perception in mammogram interpretation: gazetracking study 1. Radiology 242:396–402 Langlois JH, Roggman LA, Casey RJ, Ritter JM, Rieser-Danner LA, Jenkins VY (1987) Infant preferences for attractive faces: rudiments of a stereotype? Dev Psychol 23:363–369 Logie R, Wright R, Decker S (1992) Recognition memory performance and residential burglary. Appl Cogn Psychol 6:109–123 Lykins A, Meana M, Kambe G (2006) Detection of differential viewing patterns to erotic and non-erotic stimuli using eye-tracking methodology. Arch Sex Behav 35:569–575 Nee C (2015) Understanding expertise in burglars: from pre-conscious scanning to action and beyond. Aggress Violent Behav 20:53–61 Nee C, Taylor M (2000) Examining burglars’ target selection: interview, experiment or ethnomethodology? Psychol Crime Law 6:45–59 Neveu SM, Trottier D, Goyette M, Rouleau JL, Renaud P (2011) Identifying men who sexually abused children with virtual immersion and penile plethysmography. J Cyber Ther Rehabil 4:274–276 Rayner K (1998) Eye movements in reading and information processing: 20 years of research. Psychol Bull 124:372 Rayner K, Slattery TJ, Belanger NN (2010) Eye movements, the perceptual span, and reading speed. Psychon Bull Rev 17:834–839 Rengert G, Wasilchick J (1985) Suburban burglary: a time and a place for everything. CC Thomas, Springfield Suschinsky KD, Elias LJ, Krupp DB (2007) Looking for Ms. Right: allocating attention to facilitate mate choice decisions. Evol Psychol 5:428–441 Vernon RL, Lasley JR (1992) Police/citizen partnerships in the inner city. FBI Law Enforcement Bull 61:18–22 Wiles P, Costello A (2000) The’ road to nowhere’: the evidence for travelling criminals (Vol. 207). Home Office Research, Development and Statistics Directorate, London Wright RT, Decker SH (1996) Burglars on the job: streetlife and residential break-ins. UPNE, Boston Wright R, Logie RH, Decker SH (1995) Criminal expertise and offender decision making: an experimental study of the target selection process in residential burglary. J Res Crime Delinq 32:39–53 Yantis S (2005) How visual salience wins the battle for awareness. Nat Neurosci 8:975–977
