Science Briefs

What They See is What You Get: Eye Tracking of Attention in the Anxiety Disorders

Individuals with anxiety disorders differ in how they think about, remember, and attend to threatening aspects of their environment.

By Thomas Armstrong and Bunmi O. Olatunji, PhD

Thomas ArmstrongThomas Armstrong is a graduate student in the Emotion and Anxiety Research Laboratory at Vanderbilt University in Nashville, Tennessee. He completed undergraduate training at Lewis & Clark College in Portland, Oregon, working with Brian Detweiler-Bedell and Mark Becker on research investigating emotional influences on visual search. Prior to his graduate studies, he also coordinated research on the role of disgust in social cognition in David Pizarro’s laboratory at Cornell University. At Vanderbilt, Thomas employs eye tracking and facial electromyography to study emotion processing abnormalities associated with obsessive-compulsive disorder (OCD). He is currently working on a project investigating the relationship between disgust sensitivity and attentional biases in contamination-based OCD.  

Bunmi OlatunjiBunmi Olatunji
is an assistant professor in the Department of Psychology at Vanderbilt University in Nashville, Tennessee. He currently serves on the editorial boards of the journals International Journal of Cognitive Therapy, Journal of Anxiety Disorders, and Behavior Therapy. He has published more than 70 journal articles and book chapters and has participated in several conference presentations. As director of the Emotion and Anxiety Research Laboratory at Vanderbilt University, he is currently examining the role of basic emotions as they relate to the assessment, etiology, and maintenance of anxiety-related disorders. His research has been funded by the National Institute of Mental Health and the Anxiety Disorders Association of America.

Ever since psychology’s “cognitive revolution,” information-processing biases have been thought to play an important role in the etiology and maintenance of anxiety disorders. Both clinical observation and experimental research suggest that individuals with anxiety disorders differ in how they think about (Butler & Matthews, 1983), remember (Amir, Foa, & Coles, 1998), and attend to (MacLeod & Matthews, 1988) threatening aspects of their environment. While early theorists (Beck, 1976; Eysenck, 1992) posited an “anxiety schema,” assumed to create threat-related biases at all stages of stimulus processing, recent research has looked more closely at specific components of attention thought to play a role in anxiety disorders. By inserting emotional stimuli into classic attention tasks (e.g. Stroop task; Stroop, 1935; cueing task; Posner, 1980; visual search task; Schneider & Shiffrin, 1977), clinical scientists have identified patterns in the emotional modulation of attention that may characterize anxiety disorders. An aggregate of studies suggest a “vigilant-avoidant” pattern of attention, as individuals with anxiety disorders initially attend to threatening stimuli more than healthy controls, yet subsequently attend less (Mogg & Bradley, 1998). However, controversy remains over which components of attention underlie these phenomena. Although reaction time measures of attention, such as the Stroop and modified dot probe, have provided crucial insights, the adoption of eye tracking technology, which provides a more direct and continuous measure of attention, promises significant advances in our understanding of information-processing biases in affective disorders.

Reaction time measures of emotional modulation of attention and their limitations

To explain the initial increase of attention to threat in anxiety disorders, some researchers (Mogg & Bradley, 1988) stress “vigilance” for threat, a phenomenon involving facilitated detection and subsequent biases in the orientation of attention. However, others (Derryberry & Reed, 2002; Fox, Russo, Bowles, & Dutton, 2001) argue that biases begin after detection, through the maintenance of attention on identified threats. Primary evidence for the latter view comes from the Emotion Stroop task (MacLeod, 1991), in which quick and accurate responding to a word’s color requires ignoring the word’s meaning. Individuals with anxiety disorders show delayed color responses when words are threatening (more so than healthy controls), suggesting that anxiety disorders are characterized by increased attentional engagement by threat, as well as difficulty disengaging attention from threat (Williams, Mathews, & MacLeod, 1996). However, the Emotional Stroop does not allow the demonstration of an orientation bias, as there are not multiple stimuli in different locations competing for attention. In addition, the response interference caused by emotional word content may not be the result of attention, per se, but instead some other aspect of one’s emotional reaction to the word’s meaning (MacLeod, Mathews, & Tata, 1986).

The modified dot probe (MacLeod et al., 1986) in which participants respond to a neutral “probe” placed behind one of two simultaneously presented pictures, improves on the Emotional Stroop by allowing multiple stimuli to compete for the engagement, as well as the orientation of attention. Despite this improvement, reaction times in the dot probe task may fail to discriminate components of attention, as faster responses to probes at the location of a threatening stimulus could be the result of orienting to that location first, or of maintaining attention to that location, once fixated (Koster, Crombez, Verschuere, & De Houwer, 2004). As Weierich, Treat, and Hollingworth (2008) note, the 500 ms asynchrony between stimulus onset and probe presentation used in most dot probe studies allows for multiple fixations during the stimulus presentation, making it unclear which component(s) of attention are responsible for decreased response latencies. In addition, the modified dot probe has limited ability to register biases in later stages of processing. The inhibition of detailed, elaborative processing of threat described in multiple theories of anxiety (Foa & Kozak, 1986; Mogg, Bradley, De Bono, & Painter, 1997) can, in theory, be measured by longer probe onset asynchronies; however, a continuous measure of attention, as opposed to the snapshot provided by reaction times, would address the question with more efficiency.

Measuring eye movements through eye tracking technology

Although improved reaction time measures have been developed to better differentiate components of attention (e.g. Posner’s (1980) cuing paradigm), many researchers have turned to eye tracking technology to overcome the limitations inherent in manual reaction time measures. The measurement of eye movements has been pursued for over a century; however, only in the last thirty years have accurate, non-invasive, methods been developed (Duchowski, 2003). Today, the most popular methods involve directing a camera and infrared light source at the participant’s eye(s). By recording the surface of the eye with a video camera, the pupil can be detected by its lack of reflectance (dark pupil tracking; Richardson & Spivey, 2004); alternatively, with a bright light aimed at the eye, the pupil can be identified by the light reflecting through the pupil, off of the retina (light pupil tracking; Richardson & Spivey, 2004). While locating the pupilary reflection provides the primary indicator of eye movement, a second corneal reflection is often used to control for head movement (Duchowski, 2003).  An alternative method of eye tracking involves electro-oculography, the use of electrodes placed near the eye to record changes in electrical potentials produced by eye movements (e.g., Rohner, 2002).

The eye movements measured by these methods are more closely linked to attention than key press behavior, which occurs downstream of intervening response selection and skeletal muscle movement (Weierich et al., 2008).  Indeed, eye movements are a direct indicator of overt attention, that is, the selection of stimuli for fine-grained, foveal perception. In addition to providing a highly direct measure of visual attention, eye tracking systems also allow continuous measurement of eye movements, with gaze location typically sampled at rates of once per 16.7 ms (60 Hz) or faster. By directly and continuously measuring eye movements (see Figure 1 for a sample scan path), eye tracking devices greatly enable researchers to parse the orientation and engagement of attention, as the locations of initial fixations indicate orientation (i.e. where one looks first), while the duration of these fixations indicate the engagement of attention (i.e. how long one looks). Eye tracking also provides richer data for the analysis of later attentional processes. Whereas extended dot probe tasks can reveal the probability of attending to one location at a single point in time, eye tracking allows the comparison of fixation durations at multiple locations across the time course of the trial. In other words, eye tracking devices liberate the measurement of attention from the parameters of the task. The dot probe requires varying the presentation time of the stimulus in order to register either early or late attentional processes. Whereas additional conditions, or even studies are required to assess different components of attention or stages of processing in the dot probe paradigm, eye tracking can assess a multitude of attentional processes within the same trial.

Extant research on eye movements in anxiety disorders

Although small in number, preliminary eye tracking studies have begun to clarify the nature of increased attention to threat in anxiety disorders. One emerging conclusion is that initial increases in attention to threat cannot be fully accounted for by engagement alone. The strongest evidence thus far comes from Mogg, Millar, and Bradley’s (2000) research in patients with generalized anxiety disorder (GAD). These authors measured eye movements during a dot probe task, in which happy, sad, or angry faces were presented alongside a neutral face. While GAD patients did not differ from controls or patients with major depressive disorder (MDD) in their responding to sad or happy faces, they showed reduced latencies to first fixation on the angry face, and exhibited an orientation bias on trials with angry faces, looking at the angry face first more often than the neutral face. This pattern of speeded detection, coupled with biased orientation, fits squarely with Mogg and Bradley’s (1998) conceptualization of vigilance in anxiety disorders.

Interestingly, Mogg and colleagues found no attentional bias in reaction times with the dot probe task in which they measured eye movements. This finding is inconsistent with the hypothesis that attentional biases involve increased maintenance of attention (i.e. disability disengaging attention) to threat. However, Mogg and colleagues dot probe task involved a relatively long (1000 ms) asynchrony between picture onset and probe onset, leaving open the possibility that difficulty disengaging attention occurred on a shorter timescale. Unfortunately, Mogg and colleagues did not report the duration of initial fixations, which would shed light on this possibility. However, similar studies with social anxiety (Garner, Mogg, and Bradley, 2006) and specific phobia (Rinck & Becker, 2006) have reported the duration of initial fixation to threat stimuli, and have not found evidence of increased maintenance of attention in viewing paradigms reminiscent of the dot probe task. Together, these studies suggest that reduced reaction times in the dot probe task may not reflect difficulties disengaging attention.

However, these findings do not rule out the possibility that threat holds attention in other contexts. Indeed, other eye tracking studies, in which the threat stimulus serves as a distractor, have found difficulty disengaging attention in anxious or phobic participants. For example, Miltner, Krieschel, Hecht, Trippe, and Weiss (2004) found that a single spider, placed as a target in a 4 x 4 array with 15 other flowers, did not hold attention longer in the initial fixations of spider phobics, compared to those of controls. However, when a mushroom was added as the target, and the spider left in the array as a distractor, spider phobics begin to show delayed reaction times to the target reminiscent of engagement effects in the Stroop task. When eye movements were analyzed, it was found that delayed manual reactions appeared to be the result of difficulty disengaging attention from the spider, when it was fixated prior to the mushroom target (which appeared to occur by chance, in the absence of an orientation effect). These findings have been replicated in multiple experiments with spider phobics (Gerdes, Alpers, & Pauli, 2008; Rinck, Reinecke, Ellwart, Heuer, & Becker, 2005).

It appears that vigilance for threat, as well as difficulty disengaging threat, both contribute to information-processing abnormalities in anxiety disorders. However, each bias may be expressed in a different context. Vigilance-based biases seem to predominate when participants freely view a simple scene, while engagement-based biases arise during more complex search tasks with conflicting task demands. How these biases may differentially contribute to anxiety disorders is unclear. Vigilance is believed to contribute to anxiety disorders by promoting an exaggerated sense of danger and vulnerability to threat (Mogg & Bradley, 1998).  Difficulty disengaging attention, on the other hand, may interfere more with daily functioning, by distracting attention from ongoing tasks when threats are identified (Bishop, 2007).

Perhaps the most consistent finding in the eye tracking and anxiety literature relates to the phenomenon of avoidance. Continuous measurement of eye movements has revealed avoidance of disorder-specific threat on multiple timescales, ranging from 3 to 60 seconds (Garner et al., 2006; Hermans, Vansteenwegen, & Eelen, 1999; Pflugshaupt et al., 2005, 2007; Rinck & Becker, 2006; Rohner, 2002). In these studies, avoidance has been operationalized most often as a decrease or sustained reduction in the time spent fixating a threat stimulus, relative to a neutral stimulus (e.g. Rinck & Becker, 2006). However, Pflugshaupt and colleagues (2005) operationalized avoidance as the distance of fixations away from spiders embedded in a naturalistic scene, and found a similar pattern of avoidance in phobic individuals. Although a few studies have found avoidance beginning as early as the first fixation on a threat stimulus (e.g. Garner et al., 2006), a more typical finding is avoidance beginning somewhere between 1500-2000 ms following exposure to threat stimuli (Hermans, Vansteenwegen , & Eelen, 1999; Rinck & Becker, 2006; Rohner, 2002). Attentional avoidance is believed to maintain anxiety disorders by preventing elaborative processing of feared stimuli, which in turn prevents reappraisal, and maintains learned associations with harm (Mogg & Bradley, 1998). Given the apparent robustness of attentional avoidance in anxiety disorders, it would seem that recent attempts to retrain attention (e.g. Amir, Weber, Beard, Bomyea, & Taylor, 2008) might benefit by adding tasks that address this later bias, instead of focusing solely on earlier biases in the opposite direction.

Limitations of extant research

One difficulty in comparing the findings of these eye tracking studies is inconsistency in how researchers operationalize attentional biases. For example, “vigilance for threat” has been used to describe diverse, and sometimes unrelated findings. As mentioned above, Mogg and colleagues (2000) use the term “vigilance” to describe aspects of anxious individuals’ first fixation only, namely, its latency and target. However, the same construct is used elsewhere (Mogg & Bradley, 2006) to describe processes involving multiple fixations (Rinck et al., 2005). In two experiments, Rinck and colleagues found that phobics detected spiders faster than other insects when featured as targets in an array of 19 identical distractor animals (dragonflies, dogs, cats, butterflies). While similar to Mogg et al.’s findings at first glance, analysis of eye movements revealed that, in the phobic group, detection of spiders was hastened by reduced fixation times on neutral distractors. In other words, speeded detection of spiders was not related to orientation (i.e. finding the spider in fewer fixations) but instead to quick disengagement from neutral targets until the threat was found. This example highlights the need to decompose broad phenomena such as “vigilance,” “maintenance,” or “avoidance” into more specific processes, which may represent different routes to the same outcome (e.g. delayed or speeded detection of a target stimulus).

Future Directions

Eye tracking is a promising methodology for future research on anxiety disorders, providing crucial improvements over manual reaction time measures of attention. However, eye tracking measures need not replace reaction time measures. Indeed, research thus far suggests that eye tracking can actually improve the utility of a dot probe (Mogg et al., 2000) or visual search task (Gerdes et al., 2008; Rinck et al., 2005) by clarifying the determinants of reaction time. Future research should continue to combine eye tracking with previously used reaction time measures, in order to enhance the interpretation of reaction time data.

In addition, given eye tracking’s increased sensitivity to detect attentional biases, it would be an ideal tool for assessing attentional biases in anxiety disorders such as obsessive-compulsive disorder where such biases have proven to be somewhat elusive (Summerfeldt & Endler, 1998).  Eye tracking may also be a useful supplement for research on treatment outcome. If attentional biases play a causal role in anxiety disorders, as hypothesized elsewhere (e.g. Mogg & Bradley, 1998), they should be attenuated by successful treatment. Eye tracking could also be used to measure cognitive vulnerability for the development of anxiety-related psychopathology. In this sense, eye tracking may provide a means of detecting an endophenotype marked by attentional biases.


Amir, N., Foa, E. B., & Coles, M. E. (1998). Automatic activation and strategic avoidance of threat-relevant information in social phobia. Journal of Abnormal Psychology, 107, 285–    290.

Amir, N., Weber, G., Beard, C., Bomyea, J., & Taylor, C. (2008). The effect of a single session attention modification program on response to a public speaking challenge in socially anxious individuals. Journal of Abnormal Psychology, 117, 860-868.

Beck, A. T. (1976). Cognitive therapy and the emotional disorders. New York: International Universities Press.

Bishop, S. J. (2007). Neurocognitive mechanisms of anxiety: An integrative account. Trends in Cognitive Science, 11, 307-16.

Butler, G., & Matthews, A. (1983). Cognitive processes in anxiety. Advances in Behaviour Research and Therapy, 5, 51-62.

Derryberry, D., & Reed, M.A. (2002). Anxiety-related attentional biases and their regulation by attentional control. Journal of Abnormal Psychology, 111, 225-236.

Duchowski, A. T. (2003). Eye Tracking Methodology: Theory & Practice. London: Springer-Verlag.

Eysenck, M. W. (1992). Anxiety: The cognitive perspective. Hove, England: Psychology Press.

Foa, E. B., & Kozak, M. J. (1986). Emotional processing of fear: Exposure to corrective information. Psychological Bulletin, 99, 20–35.

Fox, E., Russo,R., Bowles, R., & Dutton, K. (2001). Do threatening stimuli draw or hold visual attention in subclinical anxiety? Journal of Experimental Psychology: General, 130, 681-700.

Garner, M., Mogg, K. & Bradley, B. P. (2006). Orienting and maintenance of gaze to facial expressions in social anxiety. Journal of Abnormal Psychology, 115, 760-770.

Gerdes, A. B. M., Alpers, G. W. & Pauli, P. (2008). When spiders appear suddenly: Spider phobic patients are distracted by task-irrelevant spiders. Behaviour Research and Therapy, 46, 174-187.

Hermans, D., Vansteenwegen, D., & Eelen, P. (1999). Eye movement registration as a continuous index of attention deployment: Data from a group of spider anxious students. Cognition and Emotion, 13, 419-434.

Koster, E. H. W., Crombez, G., Verschuere, B., & De Houwer, J. (2004). Selective attention to threat in the dot probe paradigm: Differentiating vigilance and difficulty to disengage. Behaviour Research and Therapy, 42, 1183–1192.

MacLeod, C. M. (1991). Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109, 163–203.

MacLeod, C., & Mathews, A. (1988). Anxiety and the allocation of attention to threat. Visual Attention and Anxiety. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 40, 653-670.

MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional bias in emotional disorders. Journal of Abnormal Psychology, 95, 15-20.

Miltner, W.H.R., Krieschel, S., Hecht, H., Trippe, R., & Weiss, T. (2004). Eye movements and behavioral responses to threatening and nonthreatening stimuli during visual search in phobic and nonphobic subjects. Emotion 4, 323-339.

Mogg, K., & Bradley, B. (1998). A cognitive-motivational analysis of anxiety. Behaviour Research and Therapy, 36, 809-848.

Mogg, K., & Bradley, B. P. (2006).Time course of attentional bias for fear-relevant pictures in spider-fearful individuals. Behaviour Research and Therapy, 44, 1241-50.

Mogg, K., Bradley, B., de Bono, J., & Painter, M (1997). Time course of attentional bias for threat information in non-clinical anxiety. Behaviour Research and Therapy, 35, 297- 303.

Mogg, K., Mathews, A., & Weinman, J. (1987). Memory bias in anxiety. Journal of Abnormal Psychology, 96, 94–98.

Mogg, K., Millar, N., & Bradley, B. P. (2000). Biases in EMs to threatening facial expressions in generalized anxiety disorder and depressive disorder. Journal of Abnormal Psychology, 109, 695–704.

Pflugshaupt, T., Mosimann, U. P., von Wartburg, R., Schmitt, W., Nyffeler, T., & Mu¨ri, R. M. (2005). Hypervigilanceavoidance pattern in spider phobia. Journal of Anxiety Disorders, 19, 105–116.

Pflugshaupt, T., Mosimann, U. P., Schmitt, W. J., von Wartburg, R., Wurtz, P., Lüthi, M.,  et al. (2007). To look or not to look at threat? Scanpath differences within a group of spider phobics. Journal of Anxiety Disorders, 21, 353-366.

Posner, M.I. (1980). Orienting of attention. Quarterly Journal of Experimental Psychology, 32, 3-25.

Richardson, D. C., & Spivey, M. J. (2004). Eye Tracking: Characteristics and Methods. In Wnek. G.& Bowlin, G. (Eds.) Encyclopedia of Biomaterials and Biomedical Engineering, (pp.           568-572), Marcel Dekker, Inc.

Rinck, M., & Becker, E.S. (2006). Spider fearful individuals attend to threat, then quickly avoid it: Evidence from eye movements. Journal of Abnormal Psychology, 115, 231-238.

Rinck, M., Reinecke, A., Ellwart, T., Heuer, K., & Becker, E.S. (2005). Speeded detection and increased distraction in fear of spiders: Evidence from eye movements. Journal of Abnormal Psychology, 114, 235-248.

Rohner, J.C. (2002). The time-course of visual threat processing: High trait anxious individuals eventually avert their gaze from angry faces. Cognition and Emotion, 16, 837–844.

Schneider, W. & Shiffrin, R. M. (1977). Controlled and automatic human information processing: 1. Detection, search, and attention. Psychological Review, 84, 1-66.

Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643-662.

Summerfeldt, L. J., & Endler, N. S. (1998). Examining the evidence for anxiety-related cognitive biases in obsessive– compulsive disorder. Journal of Anxiety Disorders, 12, 579–598.

Weierich, M. R., Treat, T. A., & Hollingworth, A. (2008). Theories and measurement of visual attentional processing in anxiety. Cognition & Emotion, 22, 985-1018.

Williams, J. M., Mathews, A., & MacLeod, C. (1996). The emotional Stroop task and psychopathology. Psychological Bulletin, 120, 3–24.