Authors: Han Zhang, Tessa R. Abagis, Clara J. Steeby, John Jonides
Categories: Article, ADHD, attentional capture, reactive control, visual search, eye movements
Source: Visual cognition
Authors: Han Zhang, Tessa R. Abagis, Clara J. Steeby, John Jonides
Effective visual search relies on reactively disengaging from distractors when the features of the distractors are unpredictable. Does this ability differ between adults with and without Attention-Deficit/Hyperactivity Disorder (ADHD)? Participants (36 with ADHD, 46 non-ADHD) completed the additional-singleton task, in which they searched for a unique shape while a uniquely coloured distractor unpredictably appeared on half of the trials. The distractor delayed manual response times in both groups, with no significant group difference. Both groups also demonstrated similar oculomotor capture effects, as indicated by the landing position of initial fixations. However, when initial fixations did land on the distractor, participants with ADHD tended to “linger” on the distractor with additional fixations and longer duration before disengaging from it, compared to those without ADHD. These results suggest that ADHD is associated with deficits in reactively disengaging from distractions rather than deficits in avoiding being captured in the first place.
While Attention Deficit Hyperactivity Disorder (ADHD) is typically associated with children, it appears in an estimated 2.5% of adults, or over 6 million adults in the United States alone (Simon et al., 2009). ADHD is diagnosed by inattentive, hyperactive-impulsive, or combined inattentive and hyperactive-impulsive. Among adults with ADHD, the inattentive subtype is the most prominent. One of the hallmarks of inattention is increased susceptibility to distraction (American Psychiatric Association, 2013). Heightened distractibility in ADHD is further associated with real-world difficulties such as poor performance in school, work, and day-to-day tasks such as driving (Gjervan et al., 2016; Nadeau, 2005; Randell et al., 2020). Examining this heightened distractibility in ADHD is thus critical to understanding one of the main sources of debilitation in these patients.
An important source of distraction arises from salient items in the external environment that are irrelevant to the imperative task. A prime example of this type of distraction is modeled by the “additional-singleton task” (Theeuwes, 1992). In this task, participants search for a uniquely shaped item (e.g., a diamond among circles) while a uniquely colored item may appear on some trials. Despite its irrelevance, the color-singleton has a robust effect on capturing visual attention (Gaspelin et al., 2017; Theeuwes, 1992; Theeuwes et al., 2003). Importantly, analyses of initial eye movements on distractor-present trials in this task have consistently shown an oculomotor capture effect, with the first fixation being more likely to land on the color-singleton distractor compared to a non-singleton item (an item that is neither the target nor the color singleton distractor).
Why does the color-singleton capture attention? Despite ongoing debates about this issue, there is growing consensus that a key factor for attentional capture lies in the unpredictable nature of task features (Awh et al., 2012; Luck et al., 2020; Theeuwes, 2018). In one version of the task, participants cannot anticipate whether the distractor will appear, where it will appear, or what color and shape it will have, as all these factors are randomized across trials. This task environment prevents participants from learning the features of the distractor over time and using such information to ignore the distractor. Indeed, past research has documented a range of conditions in which capture is reduced when certain task features become more predictable (e.g., Bacon & Egeth, 1994; Gaspelin et al., 2017; Stilwell et al., 2019; B. Wang & Theeuwes, 2018; Won et al., 2019). In some of these conditions, it was claimed that participants could even proactively suppress the distractor under the baseline (for a review, see Gaspelin & Luck, 2018).
In real-world situations, however, not all distractions can be anticipated. In such cases, successful visual search hinges on the ability to get back on track, that is, to disengage and reorient visual attention after initially being captured by a distractor. This becomes particularly relevant when the distractor elicits a strong capture effect, as seen in the additional-singleton task. A growing body of research has started to investigate conditions in which reactive rejection of the distractor can occur rapidly (Geng, 2014). However, to date, little attention has been given to reactive rejection of the distractor among those who have difficulty with distraction – that is, those with ADHD.
As its name suggests, ADHD is commonly believed to involve deficits in selective attention, making individuals more susceptible to visual distraction. Surprisingly, there is mixed evidence supporting this claim. Results remain inconclusive regarding whether individuals with ADHD truly show deficits in various aspects of selective visual attention, including resisting distractor interference (Huang-Pollock et al., 2005; Mason et al., 2004; Soutschek et al., 2013; Van Der Stigchel et al., 2007), visuospatial orienting (Huang-Pollock & Nigg, 2003; Mullane et al., 2011), and serial search (Mullane & Klein, 2008). The most consistent difference between individuals with ADHD and those without is found in the anti-saccade task, in which individuals with ADHD are more likely to commit directional errors (Chamorro et al., 2022). Nonetheless, the anti-saccade task does not involve searching for a target among distractors; instead, it requires inhibiting a prepotent orienting response to a sudden onset cue. Thus, it is debatable whether the worse performance seen in the anti-saccade task indicates deficits in selective visual attention per se or in other cognitive abilities, such as inhibitory control.
A few studies have used variations of the additional-singleton task to investigate visual search in those with ADHD. These studies generally did not find that individuals with ADHD were more susceptible to capture than those without ADHD. For example, Mason et al. (2004) reported that children with and without ADHD were equally affected by a color-singleton distractor during visual search. Van der Stigchel et al. (2007) found that boys with ADHD were no more captured by an abrupt-onset distractor than healthy controls. Finally, Sali et al. (2018) found that children with ADHD exhibited less capture by a distractor previously associated with reward compared to those without ADHD. Note that the majority of studies on visual search and ADHD have been conducted with children. If anything, these null results found in children suggest that adults with ADHD would not be more susceptible to capture than those without ADHD.
More importantly, one critical aspect that remains unclear is what happens after the capture of attention. When capture occurs, successful visual search relies on the ability to quickly disengage from the distractor so that search may continue. Individuals with and without ADHD may not differ in how likely they are to be captured by a salient distractor but they may differ in how quickly they can disengage from the capture. Eye-tracking may provide insight into this question, as it offers a moment-to-moment registration of overt attention. However, eye-tracking is rarely employed in the study of visual search in ADHD. Common dependent measures such as response times and accuracy, often used in past research, may not be fine-grained enough to offer insight into this reactive control process. Even in studies that did employ eye-tracking, the analysis of eye movements has typically been limited to initial eye movements to assess the initial allocation of visual attention; subsequent eye movements have not often been examined.
Several studies have examined reactive disengagement from the distractor by analyzing subsequent eye movements after capture in the general population. One line of research employed the oculomotor disengagement paradigm, in which participants must disengage from a distractor presented at the central location to locate a peripheral target (e.g., Born & Kerzel, 2009; Brockmole & Boot, 2009). Because the centrally presented distractor initially occupies the focus of attention, the time it takes for the eyes to move away from the distractor is often taken as a measure of disengagement time. This line of work has consistently shown that the disengagement from the distractor is delayed if it is similar to the target (Born & Kerzel, 2009; Wright et al., 2015). Using peripheral abrupt-onset distractors, Born et al. (2011) further demonstrated that the initiation of saccades toward the distractor and subsequent dwell times on the distractor are influenced by different factors – only the latter was affected by target-distractor similarity. These results suggest that the initial capture of attention is salience-driven, whereas disengagement from capture is goal-driven (Theeuwes, 2010; but see Luck et al., 2020). In a similar vein, Geng and DiQuattro (2010) found that a salient distractor strongly captured initial eye movements but at the same time facilitated rapid rejection of the distractor by shortening its dwell times. Together, this body of work suggests a dissociation between the control of capture and disengagement. It is, therefore, possible that ADHD is not associated with an impaired ability to resist initial capture but with an impaired ability to disengage from the capture.
The current study aims to compare adults with and without ADHD in their abilities to resist the initial capture by a salient distractor, as well as their ability to disengage from the distractor once captured. To accomplish this, we recruited adults with and without ADHD to complete the additional singleton task with their eye movements recorded. As in previous studies, the ability to resist initial capture was measured by the landing positions of the first fixation on each trial, that is, its probability to land on the distractor compared to an average non-singleton item. For the ability to disengage from the distractor, we examined several measures that incorporate subsequent eye movements. These measures will be introduced in detail in the Results section. Finally, we also examined response times and accuracy as canonical behavioral indices of distractor interference.
For the sake of transparency, we disclose that the reported study consists of a combination of two separate samples both of whom were tested on highly similar versions of the same task. Sample 1 is from one author’s (T.A.) project and Sample 2 is from another author’s (C.S.) project. The two studies were conducted at the same site, with the same apparatus, at approximately the same time in the semester, and followed nearly identical study protocols. Therefore, we combined them to increase statistical power. The study procedures reported below were identical between the two versions of the study unless otherwise noted.
Participants were recruited through the University of Michigan’s Introductory Psychology Subject Pool and through flyers posted on and around the campus. Participants recruited from the Subject Pool received 1.5 hours of subject pool credit for completion, and participants recruited through flyers received $15 for completion. Participants in the ADHD group must have had a previous diagnosis of ADHD by a clinician, per self-report. To substantiate the self-reported clinical diagnosis, we administered the Conners’ Adult ADHD Rating Scale (CAARS; Conners et al., 1999) and a semi-structured clinical interview (more details below).
A total of 99 participants (44 ADHD, 55 non-ADHD) were recruited. Data from 10 participants were excluded due to early termination and/or experimental errors. In addition, data from the following participants were 2 due to abnormal vision (color-blindness), 1 due to failing to provide demographic and medical information, 3 due to mean RT falling outside of two standard deviations of the overall mean, 1 due to response accuracy below 70%. The final sample consists of 36 in the ADHD group (Age: 19.50 ± 2.14, 50% Female) and 46 in the non-ADHD group (Age: 19.00 ± 1.21, 54.3% Female).
Participants completed the study in a quiet environment with dimmed lighting. The visual search task was presented on a 20.1-inch computer screen positioned at approximately 80 centimeters to the participant. The task was implemented using OpenSesame software (Mathôt et al., 2012) with functions from the PyGaze package (Dalmaijer et al., 2014). Eye movements were recorded by an EyeLink 1000 System at a sampling rate of 500 Hz.
The search array consisted of 10 unfilled shapes distributed equally on an imaginary circle with a 9.15° radius to the screen center. Items consisted of diamonds (1.71° diagonal) and circles (1.50° diameter), with the outline drawn in either red (CIE xyY color .65, .33, 24.90 cd/m^2^) or green (.30, .60, 71.35 cd/m^2^). The nine similar shapes contained a grey line (1.5° * 0.2°; .33, .33, 25.40 cd/m^2^) randomly oriented 22.5° from either the vertical or horizontal plane. The unique shape (target) contained a grey line randomly oriented vertically or horizontally. Stimuli were presented on a black (.28, .28, .20 cd/m^2^) background. Luminance was measured with PsyCalibrator (Lin et al., 2023) using a SpyderX photometer that touched the monitor screen.
The following sections are presented in the order in which the actual experiment was conducted.
Participants completed a questionnaire designed to obtain their demographic information and medical history. They were also tested for color blindness using the Ishihara color deficiency test (Ishihara, 2010).
The visual search task was adapted from Theeuwes (1992) and is illustrated in Figure 1. The object of the task was to find the uniquely shaped item in a display and respond to the orientation of the line contained within it. Each trial started with a fixation cross at the center of the screen. To ensure that the search started from the screen center, participants were required to maintain their gaze within a 2° radius of the fixation cross for 1000 milliseconds to trigger the search display. Then, the fixation cross disappeared, and the search array remained visible until response. On distractor-absent trials, all ten items were in the same color (all red or all green). On distractor-present trials, one of the non-target stimuli was in the opposite color (distractor) as the other items, thus becoming a salient distractor. Distractor-absent and distractor-present trials were randomly intermixed, each presented on 50% of the trials. The majority color, majority shape, and item locations were also randomly chosen on each trial. Participants identified the orientation of the line inside of the target shape by pressing the “Z” or “/” keys. The key assignments were counterbalanced across participants. Participants were explicitly told that color information was irrelevant and were asked to respond as quickly and as accurately as possible.
Participants first completed a practice block of 20 trials, repeated until they reached over 90% accuracy. Then, participants from Sample 1 completed 5 experimental blocks of 160 trials each, and participants from Sample 2 completed 5 experimental blocks of 80 trials each. In addition, participants from Sample 1 responded to thought probes that were presented after 10% of the trials. These thought probe responses were not analyzed in the current study.
After the visual search task, participants completed the CAARS (Conners et al., 1999) and the Short Stress State Questionnaire (SSSQ; Helton, 2004).
To substantiate participants’ self-reported clinical diagnosis, we administered the CAARS, which measures symptoms of ADHD in adults, namely inattention, hyperactivity, and impulsivity. Participants completed the long self-report version, which is composed of 66 items, each of which targets a symptom of ADHD with a distinct subscale (e.g., “Things I hear or see distract me from what I’m doing.”). Participants responded to each item with four possible frequency responses (Not at all, never; Just a little, once in a while; Pretty much, often; Very much, very frequently). Responses to each subscale were summed and final subscale T-scores were calculated through a formula considering both gender and age. A higher T-score indicates a greater level of ADHD symptoms on that subscale.
The SSSQ was not analyzed in the current study but we nonetheless report it for transparency. The SSSQ is a retrospective measure of participants’ distress, engagement, and worry states during the task. Participants responded to each item with one of five Likert-scale items to describe how much each item applied to their task mood and thoughts (Not at all, A little bit, Somewhat, Very much, and Extremely). Average scores were calculated for each subscale (distress, engagement, and worry), with a higher score indicating a greater level for each state.
As an additional check for the current presence and severity of ADHD symptoms, trained research staff conducted a 20-minute formal semi-structured clinical interview with each participant regarding ADHD. A psychiatrist trained in the diagnosis of ADHD provided training for this interview procedure. For each ADHD symptom according to DSM-5 (American Psychiatric Association, 2013), interviewers inquired about its presence, the time of its onset, its specific impact on the participant’s educational history, occupational history (if applicable), and interpersonal relationships. Participants were also asked about their pursuit of treatment (medication) and their family’s ADHD history. Participants in the ADHD group were required to demonstrate a significant presence of inattention symptoms that interfere with normal functioning or development, while participants in the control group were expected not to demonstrate these symptoms. No quantitative cut-off point was used in the interview; interviewers made a comprehensive assessment of participants’ eligibility based on their responses. The interview template can be found in the supplemental material.
In addition to the exclusion of participants as specified in the “Participants” section, we excluded trials in which participants did not pass the gaze-contingent check (i.e., trials in which participants failed to maintain their gaze within a 2° radius of the fixation cross for 1000 milliseconds to trigger the search display), did not produce eye movements, or exhibited extreme saccade latencies (< 50 milliseconds or > 1500 milliseconds). This led to the exclusion of 5.02% of trials in Sample 1 and 3.21% of trials in Sample 2.
To compute eye movement measures, we defined an annulus-shaped area of interest with an inner radius of 4.28° and an outer radius of 11.38°. Fixations and saccades that landed inside the annulus were assigned to the closest item. There was a high consistency (> 95%) between the classification results based on fixation positions and saccade positions. Therefore, except for saccade latencies, we computed eye movement measures based on fixations. We will outline the specific eye movement measures in the Results section as they are introduced.
Statistical analyses were performed using the R programming environment and relevant packages. ANOVAs were performed using type III sum of squares with partial eta-squared (ηP2) as effect size. In all ANOVAs, we included the sample of the study as a between-participant factor. However, in all analyses conducted, study sample did not significantly interact with the key effect of interest. Therefore, in the main text, we opted to omit reporting tests associated with study samples, and all comparisons of marginal means were averaged over the levels of study samples. The full ANOVA results can be found in the supplemental material.
To gauge support for the null hypothesis, we computed Bayes Factors (BF01) for ANOVAs using the anovaBF() function from the BayesFactor package (Morey et al., 2023). Following the default settings of the package, we assumed the JZS priors for the fixed effects with prior scales set to medium (i.e., rscaleFixed = 0.5). For each effect, we computed BF01 by comparing a model with the critical effect omitted against the full model (i.e., whichModels = “top”). Following convention, we interpret Bayes Factors (BF01) as 1 < BF <= 3 suggests anecdotal evidence supporting the null, 3 < BF <= 10 suggests moderate evidence supporting the null, 10 < BF <= 30 suggests strong evidence supporting the null, 30 < BF <= 100 suggests very strong evidence supporting the null, and BF > 100 suggests extreme evidence supporting the null.
Before proceeding to the main analyses, we present comparisons between ADHD and non-ADHD groups on the CAARS subscales as a check of current ADHD symptomatology. Each measure was analyzed using a two-by-two ANOVA with ADHD group (ADHD vs. non-ADHD) and study sample (sample 1 vs. sample 2) as between-subject factors. For each measure, Table 1 presents the means and standard deviations of each group (aggregated over the levels of study sample), along with the F-test for the effect, its associated effect size, and the Bayes factor. The results showed significant differences between ADHD and non-ADHD groups on all three subscales.
The analysis of RTs was based on correct trials only (97% of all trials). Figure 2A shows the means and 95% confidence intervals (CIs) for RT. We analyzed RTs using a mixed-effects ANOVA with distractor presence as a within-participant factor, ADHD group as a between-participant factor, and study sample as another between-participant factor. We found a significant main effect of distractor presence, F (1, 78) = 596.52, p < .001, ηP2=.884, BF01 < 1/100, which indicates extreme evidence against the null hypothesis. The main effect of ADHD group was not significant, F (1, 78) = 1.22, p = .273, ηP2=.015, BF01 = 1.25, indicating anecdotal evidence in favor of the null hypothesis. Importantly, the interaction between distractor presence and ADHD group was not significant, F (1, 78) = .57, p = .454, ηP2=.007, BF01 = 3.40, indicating moderate evidence in favor of the null hypothesis.
Comparisons of the marginal means further showed that the distractor-presence cost on RT in the non-ADHD group was 158 milliseconds, t (78) = 17.84, p < .001. For the ADHD group, there was a similar effect of 168 milliseconds, t (78) = 16.82, p < .001. This 10-millisecond group difference did not reach statistical significance. Overall, there was moderate evidence indicating that participants with and without ADHD were similarly affected by the presence of a salient distractor, as measured by RTs.
The analysis of response accuracy was based on all trials. Figure 2B shows the means and 95% confidence intervals for response accuracy. The data were analyzed using the same mixed-effects ANOVA as in the analysis of RTs. There was again a significant main effect of distractor presence, F (1, 78) = 7.40, p = .008, ηP2=.087, BF01 = 1/5.17, indicating moderate evidence against the null hypothesis. The main effect of ADHD group was not significant, F (1, 78) = .17, p = .682, ηP2=.002, BF01 = 4.01, indicating moderate evidence in favor of the null hypothesis. Importantly, the interaction between distractor presence and ADHD group was also not significant, F (1, 78) = .21, p = .646, ηP2=.003, BF01 = 4.07, indicating moderate evidence in favor of the null hypothesis.
Comparisons of the marginal means show that there was a 0.8% decrease in response accuracy in the non-ADHD group when the distractor was present, t (78) = 2.40, p = .019. For the ADHD group, there was a similar 0.5% decrease in response accuracy, t (78) = 1.51, p = .135. Overall, the results did not show a meaningful group difference in response accuracy between participants with and without ADHD.
Next, we examined the magnitude of capture as measured by the landing position of the first fixation on each trial. For each participant, we computed the proportion of trials in which the first fixation landed on the target, on the color-singleton distractor, and on non-singleton items. Note that trials in which the first fixation landed outside of the region of interest were excluded. For the non-singleton item, we further divided the proportion by eight to obtain the landing probability on an average non-singleton item. These measures were calculated based on distractor-present trials with the first fixation inside the region of interest. Such trials constituted 45% of all trials and 89% of distractor-present trials. At the level of a single participant, these landing probabilities were calculated based on an average of 262 trials, ranging from a minimum of 61 to a maximum of 391 trials.
The landing probabilities on each item in each group are shown in Figure 3A. The oculomotor capture effect was commonly defined as the difference in the landing probability between the distractor and an average non-target item on distractor-present trials (Gaspelin et al., 2017). Therefore, we conducted a mixed-effects ANOVA with item type (distractor vs. average non-singleton) as a within-participant factor, ADHD group (ADHD vs. non-ADHD) as a between-participant factor, and study version (version 1 vs. version 2) as another between-participant factor. We found a significant main effect of item type, F (1, 78) = 432.51, p < .001, ηP2=.847, BF01 < 1/100, indicating extreme evidence against the null hypothesis. The main effect of ADHD group was not significant, F (1, 78) = .17, p = .683, ηP2=.002, BF01 = 4.88, indicating moderate evidence in favor of the null hypothesis. Importantly, the interaction between ADHD group and item type was also not significant, F (1, 78) = .09, p = .768, ηP2=.001, BF01 = 4.26, indicating moderate evidence in favor of the null hypothesis.
Comparisons of the marginal means further showed that the oculomotor capture effect was similar in magnitude, with an effect of 39% in the ADHD group (t (78) = 13.70, p < .001) and an effect of 40% in the non-ADHD group (t (78) = 15.90, p < .001). Therefore, there was moderate evidence that participants with ADHD were similarly captured by the distractor compared to those without ADHD, as measured by first fixations.
We then examined the latency of saccades that resulted in the first fixation on each trial. For each participant, we computed the average saccade latency separately for those that landed on the target, on the color-singleton distractor, and on non-singleton items. This analysis was based on the same number of trials as in the analysis of first fixations.
We analyzed these data using a mixed-effects ANOVA with item type (target vs. distractor vs. non-singleton) as a within-participant factor, ADHD group (ADHD vs. non-ADHD) as a between-participant factor, and study version (version 1 vs. version 2) as another between-participant factor. There was a significant main effect of item type, F (2, 156) = 72.93, p < .001, ηP2=.483, BF01 < 1/100, indicating extreme evidence against the null hypothesis. The main effect of ADHD group was not significant, F (1, 78) = .16, p = .694, ηP2=.002, BF01 = 3.14, indicating moderate evidence in favor of the null hypothesis. The interaction between ADHD group and item type was also not significant, F (2, 156) = .19, p = .825, ηP2=.002, BF01 = 10.91, indicating strong evidence in favor of the null hypothesis.
Comparisons of the marginal means show that, for saccades that landed on the target, the average latency was 291 milliseconds for the ADHD group and 290 milliseconds for the non-ADHD group, t (78) = .04, p = .97. For saccades that landed on the distractor, the average latency was 257 milliseconds for the ADHD group and 251 milliseconds for the non-ADHD group, t (78) = .92, p = .36. For saccades that landed on non-singleton items, the average latency was 239 milliseconds for the ADHD group and 235 milliseconds for the non-ADHD group. On average, saccades that landed on the target exhibited significantly longer latency than those landed on the distractor (290 vs. 254 milliseconds, t (78) = 6.82, p < .001), which in turn had significantly longer latency than those landed on non-singleton items (254 vs. 237 milliseconds, t (78) = 5.55, p < .001).
Overall, the results indicate that saccade latencies differed depending on their destination, but there was no significant difference in saccade latencies between ADHD and non-ADHD groups.
Our subsequent analyses examined the allocation of visual attention beyond the first fixation of each trial. Here, we computed the probabilities of landing on the target, the color-singleton distractor, and an average non-singleton item for the second fixation on each trial. This analysis was based on distractor-present trials with the first two fixations inside the region of interest. Such trials constituted 42% of all trials and 84% of distractor-present trials. At the level of a single participant, these landing probabilities were calculated based on an average of 246 trials, ranging from a minimum of 23 to a maximum of 378 trials.
Figure 3B shows the landing probabilities of the second fixations for each group. We conducted the same mixed-effects ANOVA for the second landing position. There was a significant main effect of item type, F (1, 78) = 496.48, p < .001, ηP2=.864, BF01 < 1/100, indicating extreme evidence against the null hypothesis. There was also a significant main effect of ADHD group, F (1, 78) = 7.70, p = .007, ηP2=.090, BF01 = 1/3.23, indicating moderate evidence against the null hypothesis. Importantly, there was also a significant interaction between item type and ADHD group, F (1, 78) = 7.89, p = .006, ηP2=.092, BF01 = 1/14.74, indicating strong evidence against the null hypothesis.
Comparisons of the marginal means reveal that this interaction effect was driven by the significantly higher probability of distractor-landing fixations in the ADHD group compared to the non-ADHD group (30.66% vs. 24.67%), t (78) = 2.81, p = .006. The two groups did not significantly differ in their probability of landing on an average non-singleton item (3.30% vs. 3.44%), t (78) = - .52, p = .60.
Next, we examined the sequence of items visited by the first and the second fixations. The idea was that if participants with ADHD tended to linger on the distractor after initial capture, the heightened distractor-looking probability on the second fixation observed here should be mainly driven by trials with distractor re-fixations rather than trials with fixation switches (e.g., the first fixation was on the target and the second fixation switched to the distractor). As there were 3 item types, there were a total of 9 possible fixation sequences for the first two fixations. Here, we were specifically interested in three fixation distractor → distractor, target → distractor, and non-singleton → distractor. For each participant, we computed the proportion of trials with these fixation sequences. This analysis used the same set of trials as the analysis of second landing positions.
The results are visualized in Figure 4. We analyzed the data using a mixed-effects ANOVA with fixation sequence (distractor → distractor vs. target → distractor vs. non-singleton → distractor) as a within-participant factor, ADHD group as a between-participant factor, and study version as another between-participant factor. This analysis was based on the same number of trials as in the previous analysis.
We found a significant main effect of fixation sequence, F (2, 156) = 103.47, p < .001, ηP2=.570, BF01 < 1/100, indicating extreme evidence against the null hypothesis. The main effect of ADHD group was also significant, F (1, 78) = 7.91, p = .006, ηP2=.092, BF01 = 1/3.26, indicating moderate evidence against the null hypothesis. The interaction between fixation sequence and ADHD group was not significant, F (2, 156) = 2.07, p = .130, ηP2=.026, BF01 = 1.96, indicating anecdotal evidence in favor of the null hypothesis.
We then compared the marginal means to understand these results further. For “distractor → distractor” trials, participants with ADHD indeed had a higher proportion of these trials compared to the non-ADHD group (14.43% vs. 10.59%), t (78) = 2.56, p = .012. For “target → distractor” trials, participants with ADHD also had a higher proportion of these trials compared to the non-ADHD group (2.65% vs. 2.08%), although the difference was not significant, t (78) = 1.88, p = .064. Finally, for “non-singleton → distractor” trials, participants with ADHD had a slightly higher proportion of these trials compared to the non-ADHD group (13.58% vs. 12.00%), and again the difference was not significant, t (78) = 1.15, p = .255.
Overall, these results confirm that the ADHD group had a higher proportion of trials with distractor re-fixations compared to the non-ADHD group, which indicates a higher tendency to linger on the distractor after initial capture. However, because the interaction term was not significant, the results do not suggest that the group difference was more pronounced for the “distractor → distractor” fixation sequence than for the other two fixation sequences. Indeed, the group difference for the “target → distractor” fixation sequence was also marginally significant, indicating that this shift in fixation contributed, albeit to a lesser extent, to the overall increased probability of looking at the distractor on the second fixation.
Next, we examined how lingering fixations on the distractor cost search performance in terms of distractor escape time and search RT.
To quantify the total amount of time it took the eyes to disengage from the distractor after the initial capture of attention, we computed the “distractor escape time”. This measure represents the duration during which the eyes remained on the distractor for trials in which the first fixation was on the distractor until the eyes shifted away to another item. For example, in a trial with the first five fixations following the sequence “distractor, distractor, non-singleton, distractor, target,” the distractor escape time would be determined by the time from the beginning of the first fixation to the end of the second fixation. In essence, distractor escape time was the time it took the eyes to escape, or disengage from, the initial capture. If participants with ADHD were more likely to have re-fixations on the distractor after initial capture, then these additional re-fixations should likely result in a longer distractor escape time.
This analysis was conducted using capture trials only, that is, trials on which the initial landing position was the distractor. This constituted 19% of all trials and 39% of distractor-present trials. At the level of a single participant, distractor escape time was averaged over an average of 115 trials with a minimum of 18 trials and a maximum of 286 trials.
We conducted a two-way ANOVA with ADHD group and study version as between-participant factors. There was a significant main effect of ADHD group, F (1, 78) = 6.01, p = .016, ηP2=.072, BF01 = 1/3.25, indicating moderate evidence against the null hypothesis. Comparisons of the marginal means show a 31-millisecond difference, with the distractor escape time being 237 milliseconds for the ADHD group and 206 milliseconds for the non-ADHD group. Therefore, the additional fixations on the distractor led to a significant but small increase in distractor escape time for the ADHD group.
We then examined whether RTs were longer in the ADHD group compared to the non-ADHD group on trials in which the eyes were captured. This analysis used the same set of trials as in the previous analysis with the added criterion that the trials be correct. This constituted 19% of all trials and 38% of distractor-present trials. At the level of a single participant, this mean RT was computed based on an average of 110 trials with a minimum of 16 trials and a maximum of 273 trials.
The mean RT on captured trials was 1381 milliseconds for the ADHD group and 1296 milliseconds for the non-ADHD group, indicating an 85-millisecond group difference. However, a two-way ANOVA with ADHD group and study version as between-participant factors show that the main effect of ADHD group did not reach significance, F (1, 78) = 3.14, p = .080, ηP2=.039, BF01 = 1.09, which indicates anecdotal evidence in favor of the null hypothesis.
Existing research indicates that children with ADHD tend to overshoot their saccades compared to those without ADHD (Rommelse et al., 2008). This raises the question of whether the re-fixations on the distractor observed among ADHD participants were simply due to corrective saccades that corrected an inaccurate first fixation on the distractor. To address this, we compared fixation position errors between ADHD and non-ADHD participants on their first fixation. These analyses were conducted using trials in which the first fixation was on the distractor, as in the analysis of distractor escape time.
First, we computed the difference in visual angle between the distance from the fixation position to the screen center and the radius of the search array (i.e., the distance from the actual item position to the screen center). Thus, a positive value indicates an overshoot to the distractor (end position larger than the radius), while a negative value indicates an undershoot to the distractor (end position smaller than the radius). We found that both the ADHD group and the control group tended to undershoot with an error of approximately 1 degree of visual angle (ADHD: −.92, non-ADHD: −.97). A two-way ANOVA with ADHD group and study version showed that the main effect of ADHD group was not significant, F (1, 78) = .20, p = .654, ηP2=.003, BF01 = 1/4.26, indicating moderate evidence against the null hypothesis.
Second, we computed the distance in visual angle between the fixation and the actual position of the item. We found that the average distance was 1.64 degrees for the ADHD group and 1.67 degrees for the control group. A two-way ANOVA with ADHD group and study version showed that the main effect of ADHD group was not significant, F (1, 78) = .16, p = .688, ηP2=.002, BF01 = 1/4.41, indicating moderate evidence against the null hypothesis.
In summary, these results provide moderate evidence indicating that there is no difference in the accuracy of the initial fixation on the distractor between participants with and without ADHD.
Finally, we explored the correlations between the subscales of CAARS and the main dependent variables across all participants, as shown in Figure 5. There were significant but weak correlations between the CAARS Inattention score and three of the dependent measures. Specifically, those who had a higher Inattention score tended to have a higher proportion of trials with a refixation on the distractor after initial capture (r = .29, p = .009), tended to have a longer distractor escape time after capture (r = .23, p = .037), and tended to have longer RTs on trials in which an initial capture occurred (r = .23, p = .041). Although these findings should be interpreted with caution due to the relatively small sample size (N = 82), they do suggest that individuals with higher ADHD inattention symptoms demonstrate a reduced ability to disengage from initial capture.
The primary goal of this study was to examine adults with and without ADHD in their ability to resist capture by a salient distractor as well as their ability to disengage from the distractor. Although adults with ADHD were not more likely than non-ADHD adults to be captured by the salient distractor, they demonstrated a tendency to “linger” on the distractor with more refixations and longer looking times. This “linger-on” effect suggests that participants with ADHD have impaired reactive control, that is, impaired ability to reorient their attention once distracted.
For what did these additional lingering fixations try to compensate? Wright et al. (2015) proposed that a delayed disengagement from the distractor serves a functional role, encouraging additional processing of objects that match the observer’s attentional set. Based on this view, and drawing from a drift-diffusion perspective (Wolfe, 2021), the observed increase in re-fixation on the distractor and the extended distractor escape time among ADHD participants indicate two possible, non-exclusive explanations. First, the drift rate of processes responsible for matching the distractor’s features according to the observer’s attentional set is reduced, therefore prompting the programming of another saccade to the same location (i.e., a re-fixation) to make up for the reduced speed in reaching a decision threshold. Second, the attentional set adopted by individuals with ADHD during a search task might change. For example, participants with ADHD might adopt a broader attentional set that includes all items with unique features, either as a strategy to improve search efficiency by narrowing down the list of potential targets or as a result of goal neglect (Kane & Engle, 2003). Consequently, the threshold for dismissing the color-singleton distractor becomes elevated, making it harder to reject the distractor once it has attracted attention.
The current results align with previous studies suggesting a potential association between impaired reactive distractor rejection and ADHD. For example, E. Wang et al. (2016) conducted an additional-singleton task combined with electroencephalography in children with and without ADHD. They observed an early but small N2pc component (indicating capture) followed by a large PD component (indicating distractor rejection) in both groups. However, the ADHD group exhibited a significantly smaller PD component amplitude than the non-ADHD group. According to the authors, these results point to deficits in the active suppression of distractors in ADHD.
In a separate study, Gumenyuk et al. (2005) presented novel auditory distractions while children with and without ADHD performed a visual discrimination task. They found that the amplitude of the frontal LN component (indicating re-orientation of attention) was smaller in children with ADHD than in controls. This led the authors to speculate that children with ADHD might have prefrontal cortical dysfunction, causing difficulties in reorienting attention back to the task following a distraction.
More broadly, the current results align with the idea that a core symptom of ADHD is difficulty disengaging from distracting activities. For example, clinicians often observe that individuals with ADHD exhibit a state of “hyperfocus,” in which they become deeply engrossed in activities to the point of neglecting their surroundings or other responsibilities (Hupfeld et al., 2019). This hyperfocus state makes it extremely challenging for those with ADHD to cease the activity and shift their attention elsewhere, even when necessary or important to do so. Our findings are consistent with this notion in suggesting that ADHD is associated more with deficits in reactive rejection of distractors, rather than deficits in the ability to avoid initial capture.
It is crucial, however, to consider several qualifications to the present findings. First, participants with ADHD did not consistently show capture with the presence of the distractor, nor did they consistently linger on the distractor following capture. Our results indicate that the initial fixation landed on the distractor on approximately 40% of the trials with distractors (see Figure 3A). Furthermore, there were only about 14% of trials in the ADHD group in which the first two fixations landed on the distractor, in comparison to 10% of trials in the non-ADHD group. Thus, the overall probability of having two consecutive fixations on the distractor was quite low, and the difference between ADHD and non-ADHD groups was relatively small.
It is noteworthy that participants with ADHD demonstrated a marginally increased tendency to shift their focus to the distractor even after initially fixating on the target, as evidenced by the marginally significant difference in “target → distractor” trials. This numerical distinction might also imply a deficit in distractor disengagement, wherein an overt re-fixation compensates for the inability to disengage from the distractor at the level of covert attention. Alternatively, this pattern could indicate a late capture effect instead of disengagement, where the attention is drawn to the distractor even after the target has been identified. Regardless, the core of the current findings is anchored in the significantly higher occurrence of “distractor → distractor” trials, providing a clear indication of the lingering of overt attention on the distractor.
Moreover, the additional re-fixations on the distractor did not lead to a substantial increase in looking time, as the difference in distractor escape time between ADHD and non-ADHD groups was approximately 31 milliseconds. On the other hand, RTs on trials on which attention was captured showed a larger, though not statistically significant, group difference of about 85 milliseconds. The question arises as to why the 31-millisecond difference in distractor escape time was statistically significant, while the 85-millisecond difference in RT was not. The direct reason, of course, is statistical – the standard error of the mean was larger for RT (48.1) compared to distractor escape time (12.8), suggesting that the study may have lacked sufficient statistical power to detect the RT difference. However, the underlying reason could be that RT encompasses a broader range of processes, making it a less sensitive and less direct indicator of distractor engagement than distractor escape time. Specifically, RT is influenced by additional, unrelated processes, such as the motor response required to indicate the line orientation within the target, which could add noise to the measurement.
The correlational analyses revealed several significant, albeit weak, correlations between measures of distractor disengagement and self-reported inattention symptoms, as assessed by the CAARS. These findings affirm the association between ADHD—specifically, inattention symptoms—and the tendency to linger on distractors. However, it is important to note the limitation posed by the relatively small sample size of this study in detecting correlations with magnitudes of .2 to .3. In addition, while CAARS is widely used, it has faced criticism for its lack of sensitivity and specificity in diagnosing ADHD among college students (Gorlin et al., 2016; Harrison et al., 2019). The reader should bear these issues in mind when interpreting the actual magnitude of the effect for the association between the linger-on effect and ADHD symptoms on a continuous scale. We do believe that these correlational results would be useful in designing a future study with a larger sample size and more robust measurements.
Finally, we note that the additional-singleton task, as administered in this study, represents only one of many tasks that engage in reactive rejection of distractors. For example, Forster and Lavie (2016) developed a visual search task incorporating irrelevant cartoon figures as distractors. They argued that these cartoon distractors, truly irrelevant to the central task, do not compete with the task’s imperative stimuli in eliciting a response—unlike the flanking arrows in a flanker task, for example. As such, these distractors arguably represent the type of distractions people typically encounter in real life. Using this task, they discovered that adult participants with childhood ADHD symptoms demonstrated greater distractor interference as measured by reaction times (Forster and Lavie, 2016; r = .32 in Exp. 1 and r = .32 in Exp. 2). A natural extension of the current study would be to utilize eye-tracking to investigate whether this distractor interference results from greater initial capture or slower recovery from the capture.
In summary, the current study highlights a subtle yet significant pattern of visual search behavior in adults with ADHD, indicating a deficit in reactive distractor rejection. Given the complex nature of ADHD, further research employing diverse distraction tasks and more robust diagnostic tools is needed to fully understand the nature of reactive control in ADHD.