Pupillary unrest is attenuated in Attention Deficit Hyperactivity Disorder (ADHD)

Abstract

We investigated the phenomenon of pupillary unrest in individuals with Attention Deficit Hyperactivity Disorder (ADHD) compared to neurotypical controls. We measured the power of low-frequency pupil oscillations under two experimental conditions: a passive condition with minimal distraction and a resting condition with no distraction. The study included 76 adult participants (42 controls and 34 with ADHD) aged 18–40. The results show that individuals with ADHD exhibit reduced power in pupillary oscillations, suggesting a suppression of general catecholaminergic activity. The nature of the experiment indicates that this suppression is endemic in the background and independent of the visual task or the ongoing cognitive effort. This finding is consistent with our previous observations of reduced pupil dilations in ADHD during active tasks (Privitera et al. 2024) and provide basic insights for future research aimed at developing and refining a psychophysical paradigm that could serve as a biomarker to enhance ADHD evaluation and classification.

Introduction

In his pioneering work on communication engineering and stochastic biological servomechanisms, Lawrence Stark used the small and low frequency fluctuations (unrest) of pupil diameter under steady illumination as a prime example of noise in biological systems, and his contributions to the experimental, theoretical, and modeling studies of such phenomena are still foundational to this area of research (Stark et al. 1958). Since those early times, the phenomenon of pupillary unrest has been investigated in many applications and areas of medicine and neuroscience (Loewenfeld 1993). For example, pupillary unrest decreases in older age groups and in various forms of autonomic nervous system impairment (Hreidarsson et al. 1988; McKay et al. 2024), but it increases during mindful meditation (Pomè et al. 2020) and is associated with cortical plasticity and excitability (Binda & Lunghi, 2017).

Attention Deficit Hyperactivity Disorder (ADHD) is characterized by difficulties with attention, hyperactivity, and impulsivity, often persisting into adulthood and leading to significant personal and societal consequences. Despite its prevalence, accessible neurophysiological assessment methods and tools for monitoring treatment outcomes remain largely unavailable outside controlled research environments (Hinshaw & Scheffler, 2014).

In this study, we measured pupil unrest in two groups of participants—neurotypical controls and individuals with ADHD—under two different experimental conditions: one with minimal distraction (passive) and one with no distraction (resting), and we identified a range where the power of these oscillations differentiates between the two groups.

Methods

The study included 76 adults (42 Controls, 34 ADHD) aged 18–40 at the University of California, Berkeley. ADHD participants were recruited from university clinics and local ADHD support groups, while neurotypical control participants were volunteers from UCB students. The inclusion criteria for ADHD required participants to meet DSM-5 criteria, have a CGI-S score of 4 or higher, and an IQ of at least 80 (American Psychiatric Association, 2022). Exclusion criteria included a history of specific psychiatric disorders, substance abuse (except nicotine), and severe untreated psychopathology. Neurotypical controls were matched with the ADHD group based on age, gender, IQ, and educational level (see Privitera et al. 2024, for further details on the inclusion and exclusion criteria).

To maintain the clinical relevance of the sample, the study permitted stable antidepressant use but excluded medications known to affect cognition (antipsychotics, mood stabilizers, benzodiazepines, anticonvulsants). Participants were asked to abstain from ADHD medication for approximately 24 hours before testing to minimize any acute effects. Following an initial diagnostic evaluation, eligible individuals were invited to participate in the experimental protocol within a month of the assessment.

Subjects were measured under two levels of attentional demand. In one, subjects were asked to maintain fixation on a black cross in the center of the screen while two distracting checkerboard patterns flickered on circular light gray patches of 1.5 degrees in diameter, located 12 degrees apart horizontally. The frequency of the two distractors was well beyond the break frequency of the pupil servomechanism (Stark & Sherman,1957). This protocol was measured for a total of 2 minutes, and it was called “passive.” A subgroup of subjects (13 ADHD and 24 Control) participated in a second experiment where they were simply asked to maintain fixation on the cross in the center with no peripheral distraction. We called this protocol “resting,” and it was measured for 5 minutes. Experiments were conducted under mesopic conditions in a booth measuring approximately 2×2 square meters, designed to isolate participants from the surrounding laboratory environment. Inside the booth, there was a chair and a small desk, with the stimulus monitor serving as the only light source, along with the EyeLink 1000 head-supported eye-tracker system (http://www.sr-research.com/EL_1000.html) for pupil tracking. Pupil diameter was continuously recorded at a sampling rate of 1000 Hz during the entire measurement period. Blink artifacts were later detected offline during post-processing, using data from the EyeLink system, and were corrected using a linear interpolation method.

Results

An example of pupil unrest for one Control subject is shown in the top panel (A). The power spectrum of the diameter was measured for each subject and then averaged for the two cohorts. The two insets (B) and (C) on the left show the averaged power spectrum for the Control group (solid blue line) and the averaged power spectrum for the ADHD group (solid gray line). The narrow shaded area around the solid lines represents the standard error. The shaded orange area indicates where the two distributions are statistically significantly different, showing larger power in the Control group. The frequency bandwidth where this occurs is approximately 0.5 to 4.2 Hz for the passive protocol and 0.5 to 5.5 Hz for the resting protocol. This lowpass bandwidth is consistent with the findings from the original experiments conducted by Stark & Sherman (1957) in their well-known Maxwellian open-loop condition. Those experiments demonstrated the third-order lag nature of the pupil transfer function and showed that the Bode plot frequency response of the pupil servomechanism remains significantly responsive up to the 5–6 Hz range. The area of significance of the difference between the two distributions of power spectra (shaded orange in B and C) was determined using a cluster-based permutation test which avoids the problem of multiple t-tests and the severe penalty of the Bonferroni correction as used and discussed in previous works (see, for example, Privitera et al., 2014). We further validated these results using a bootstrapping approach and wavelet analysis instead of the Fourier Transform for the power spectrum.

We then divided each recording into four consecutive time intervals of approximately 30 seconds for the passive protocol and 75 seconds for the resting protocol. We calculated the power spectrum for each time interval, but this time summed the total power within the bandwidth of interest (0.5–4.2 Hz for passive and 0.5–5.5 Hz for resting) and reported this sum in the insets on the right (D and E). The sum of the power spectrum has been discussed and legitimized in other applications and referred to as pupil unrest under ambient luminance, PUAL-score (McKay & Larson, 2021). As expected, based on the observations of the power spectra B and C, the Control group exhibited higher power throughout the entire duration of the recording (asterisks indicate p<0.05 differences), which appeared to increase over time in both cohorts. The mean pupil size for the Control and ADHD groups was 4.3 mm and 4.1 mm, respectively, in the resting protocol, and 3.7 mm and 3.6 mm in the passive protocol; in both cases, the differences between the two cohorts were not statistically significant.

Discussion

An important characteristic of pupil unrest is that it is highly correlated between the two eyes, which suggests that the point of origin must be in the portion of the pupil pathway common to both irises at the level of the Edinger-Westphal (EW) complex in the midbrain (Stark et al. 1958). This complex is the main parasympathetic preganglionic motor hub for the pupil sphincter musle and is connected to the distributed catecholaminergic network through various pathways; of particular interest for this study is its association with the Locus Coeruleus (LC) (Aston-Jones & Cohen, 2005; Joshi et al., 2016).

LC serves as a major source of norepinephrine in the brain, greatly affecting neural dopamine function and dopaminergic receptors, and has been linked to the development of ADHD (Arnsten, 2011). Catecholamines, including dopamine and norepinephrine, regulate perception and attention, and deficiencies in these neurotransmitters are the primary focus of pharmacological treatments for ADHD, which often involve stimulant drugs (Aboitiz et al. 2014, Hinshaw & Scheffler, 2014).

Although no direct neurological projections between the LC and the EW nuclei are known, there is abundant evidence that adrenergic activity in the LC significantly affects pupil diameter, likely through parallel pathways involving intermediary circuits (Aston-Jones & Cohen, 2005; Joshi et al., 2016). Other sources, such as the neocortex and the hypothalamus, also play a role in the modulation of the EW nuclei (Larson et al. 1996). Finally, it is well known that attention, in general, can elevate the baseline tension of the pupil dilator muscle via the pupillary sympathetic pathway. This, in turn, may amplify the push-pull biomechanical mechanism of the sphincter-dilator pupillary muscle system, increasing sensitivity and magnitude of pupillary oscillations (Loewenfeld 1993).

In our experiment, the steady-state activity of pupil unrest is associated to a moderate suppression in the ADHD group. The nature of the experiment, where subjects were engaged with either a minimal level of distraction (passive) or no distraction at all (resting), indicates that this suppression is endemic in the background and independent of the visual task or the ongoing cognitive effort. This result complements earlier findings (Privitera et al. 2024), where larger pupil dilations were observed in the Control group compared to the ADHD group. However, in that experiment, subjects were engaged in an active target vs. distractor visual discrimination task. To summarize, these results suggest a general and distributed depression of catecholaminergic activity in ADHD, expressed by reduced task-related pupillary dilations (Privitera et al. 2024) and pupil unrest power (this report). This depression may be linked to the decrease in attentional performance in individuals with ADHD.

The small group of subjects in this pilot study and the highly heterogeneous characteristics of ADHD symptoms warrant further research and experimentation. The findings to date suggest promising applications, not only for exploring the neurodynamic mechanisms underlying ADHD but also for creating diagnostic tools that allow for more objective clinical evaluation and treatment of individuals with the disorder.

Figure.

An example of pupil unrest is shown for one Control subject in the passive protocol (A). The two insets on the left (B and C) show the averaged power spectrum for the Control group (solid blue line) and the averaged power spectrum for the ADHD group (solid gray line). The narrow shaded area around the solid lines represents standard error. The large shaded orange area indicates where the difference between the two distributions is statistically significantly indicating more power in the Control group. The frequency bandwidth where this difference occurs is approximately 0.5 to 4.2 Hz for the passive protocol and 0.5 to 5.5 Hz for the resting protocol. For each subject and measurement, all recordings were divided into four consecutive time intervals of approximately 30 seconds for the passive protocol and 75 seconds for the resting protocol. We then calculated the power spectrum of each time interval and summed the total power within the bandwidth of interest (0.5–4.2 Hz for passive and 0.5–5.5 Hz for resting). The average sums for the two cohorts are displayed in the two right panels for each time interval (D and E), with asterisks indicating p-values < 0.05 and error bars indicating SE.