The healthy human blinks around 15–20 times per minute (Tsubota et al., 1996), however the precorneal tear film, which lubricates the eye, only begins drying up approximately 25 s after a blink ends (Norn, 1969). This suggests that we blink more often than needed to maintain a lubricated precorneal tear film. Previous research has found task-related modulation of spontaneous eye blink rate (sEBR), which indicates that blinking could be reflective of cognitive factors (Siegle et al., 2008; Oh et al., 2012). For example, reading is accompanied by low levels of sEBR, while high levels of sEBR have been reported during conversation (Bentivoglio et al., 1997). More recent studies have found that sEBR correlates with attentional load and fatigue (Maffei and Angrilli, 2018), attentional control (Colzato et al., 2009; Unsworth et al., 2019a), can track working memory updating and gating (Rac-Lubashevsky et al., 2017), and can predict differences in exploration during reinforcement learning (Van Slooten et al., 2019). In addition, a growing body of literature continues to provide evidence supporting sEBR as an effective measure of striatal dopamine (DA) activity (Jongkees and Colzato, 2016). However, whether sEBR does indeed reflect DA activity is still debated today (Dang et al., 2017; Sescousse et al., 2018).

The connection between sEBR and DA first came from observations of neurological and psychiatric disorders that found decreased sEBR in patients with Parkinson’s (Hall, 1945; Reddy et al., 2013), a neurodegenerative disorder that affects the dopaminergic system in the brain, causing symptoms like rigidity (Dauer and Przedborski, 2003). Schizophrenia has also been suggested to provide evidence for a connection between sEBR and DA due to excessive DA activity in the striatum (Howes et al., 2015) and increased sEBR in schizophrenia patients (Adamson, 1995; Swarztrauber and Fujikawa, 1998). Additionally, sEBR and DA has previously been investigated in pharmacological studies, which have observed an increase in sEBR after administration of DA agonists, while DA antagonists decreased sEBR in primates (Elsworth et al., 1991; Jutkiewicz and Bergman, 2004). In one study, researchers found sEBR was correlated with dopamine levels specifically in the caudate nucleus in monkeys, suggesting that DA could regulate blink rate (Taylor et al., 1999). This is further supported by another study that found sEBR to be closely related to in vivo and positron emission tomography measures of striatal D2 receptor density in the ventral striatum and caudate nucleus of adult male vervet monkeys (Groman et al., 2014). These findings provide valuable evidence for sEBR being a viable measure of striatal DA activity and have led to many researchers to adopt sEBR as a measure of DA activity. Moreover, sEBR is an easy-to-record physiological measure that is non-invasive and affordable.

One cognitive process of interest, that is also closely related to DA activity, is working memory (WM) which is the process of actively holding information online and manipulating it to meet task demands (Baddeley, 1992). Prior research has found substantial evidence that demonstrates the importance of dopaminergic neurotransmission and the role of the prefrontal cortex during WM function (Fuster and Alexander, 1971; Funahashi et al., 1989; Courtney et al., 1998; Wager and Smith, 2003; Cools and Robbins, 2004), especially during WM maintenance (Fuster and Alexander, 1971; Funahashi et al., 1989; Constantinidis et al., 2018). Specifically, human studies investigating DA in WM tasks have found both caudate dopamine activity during WM maintenance and DA synthesis capacity to be positively correlated with WM capacity, a measure of the amount of information that can be held in WM (Cools et al., 2008; Landau et al., 2009). Though it is widely accepted that the PFC plays an important role in WM function (Roberts et al., 1998), many researchers still debate the PFC’s role in WM (Seamans and Yang, 2004). One model that attempts to elucidate the PFC’s role in WM function is the prefrontal cortex basal ganglia WM model (PBWM; Frank et al., 2001; Hazy et al., 2006). PBWN is a computational neural network model that suggests that WM requires robust maintenance and rapid selective updating. This model states that the frontal cortex facilitates robust, active maintenance through recurrent excitation in frontal neurons, while the basal ganglia orchestrates a gating mechanism that controls the flow of information into WM (Frank et al., 2001). Previous research has pointed toward DA being important for this sustained firing activity in the PFC during WM maintenance (Sawaguchi, 2001; Durstewitz and Seamans, 2008; De Frias et al., 2010). The relationship between DA and WM performance is believed to follow an inverted U-shape, in which too little or too much dopamine impairs performance, as seen in psychopharmacological studies (Stewart and Plenz, 2006). In one study, the effects of administered dopaminergic drugs on PFC function depended on baseline levels of performance, whereas administration of bromocriptine, a dopamine agonist, impaired performance for individuals with higher working memory abilities while improving performance for individuals with lower working memory abilities (Kimberg et al., 1997).

Although sEBR has been used in prior research to investigate cognitive functions like WM, many of these studies relied on baseline levels of sEBR to investigate these relationships (Tharp and Pickering, 2011; Zhang et al., 2015; Unsworth et al., 2019b). Few studies have investigated the relationship between phasic sEBR during a WM task. Phasic sEBR refers to the measuring of sEBR in response to stimulus conditions while tonic sEBR refers to baseline levels of blinking (Bacher and Allen, 2009). To the best of our knowledge, only one other study has examined sEBR as a function of the different task phases (e.g., stimulus encoding, maintenance during the delay period, and stimulus probe periods) of a WM task (Bacher et al., 2017). Bacher et al. (2017) found modulation of sEBR across these different phases are developed in infants as young as 10 months, indicating that sEBR can reflect dopamine function in early human development. They also observed higher sEBR during the Hide (delay) phase of the task in relation to the Reveal phase, which is when the experimenter revealed the toy’s location to the child. This modulation of sEBR was suggested to reflect the engagement of cognitive resources that have become available during the Hide phase and transiently elevated DA activity that is needed to update and maintain mental representations (Bacher et al., 2017).

The goal of the current study was to investigate how fluctuations in sEBR during different phases of a Sternberg visual WM task (Figure 1) relate to performance, and how sEBR fluctuations change across task demands. First, we hypothesized that sEBR during the WM Delay period, when stimuli are being maintained, would be positively correlated with task performance and that there would be a non-linear relationship such that low and high sEBR would correlate with worse performance. Second, we hypothesized differences in sEBR across phases of the WM task with differences between phasic sEBR during the WM delay and tonic sEBR during non-task rest periods.

In the present study, we investigated the temporal fluctuations in sEBR across a WM paradigm and its relation to WM task accuracy in two experiments, inside and outside an MRI scanner, and using two methods of collecting sEBR. Using the same Sternberg working memory paradigm, we observed a strong positive relationship between sEBR and task performance only during the WM task delay in both experiments. We also found a significant difference in sEBR between task phases and a difference between Delay period sEBR and baseline sEBR.

Our first hypothesis was that phasic sEBR during the Delay period of the WM task would be positively correlated with task accuracy and that we would also observe a non-linear relationship where high and low sEBR would be predictive of low performance. We observed a strong positive correlation between sEBR during the Delay period in both Experiment 1 and Experiment 2 with task accuracy. However, only in Experiment 2 was this relationship significant. We believe that the lack of significance in Experiment 1 is due to the smaller sample size and thus lack of power, which our formal post-hoc power analyses confirmed. While the sample size in Experiment 2 was also small, we observed a similar correlation and reliability statistic as well as higher power while recording sEBR using a different method (electrooculogram instead of camera-based eye-tracking hardware). Nevertheless, a correlation between sEBR and task performance of approximately 0.50 is a very high correlation in psychology, especially between a behavioral measure and a physiological measure (Gignac and Szodorai, 2016). The replication of a similar correlation between Delay period sEBR and WM performance across two separate experiments strengthens our findings but with the confidence interval being so wide with the relatively small sample size one cannot be certain the true correlation is so large. Previous research has found that correlations begin to stabilize at even larger sample sizes (Schönbrodt and Perugini, 2013). Thus, future studies should include an additional experiment with high statistical power to replicate these findings and to determine whether the observed effect stabilizes with even larger sample sizes.

If we interpret sEBR as an indirect measure of striatal DA activity, as other studies have postulated, we can speculate that higher sEBR during the WM delay was correlated with task accuracy due to DA regulating the maintenance and updating of representations in WM (Westbrook and Braver, 2016). The other results support this idea since no other task period was significantly correlated with task accuracy. Many studies that have investigated the relationship between sEBR and cognitive functions have used baseline levels of sEBR taken before or after tasks in their analysis (Tharp and Pickering, 2011; Zhang et al., 2015; Unsworth et al., 2019b). However, we show that while the WM task Delay period sEBR was correlated positively with task accuracy, baseline levels of sEBR were not. Our results highlight the importance of examining phasic and tonic sEBR when investigating the relationships between sEBR and other cognitive functions. The results also highlight that blinking may be an important component of working memory function; however, future studies, including within-subject analyses using larger number of trials, are needed to understand the role of blinking during WM maintenance. Additionally, future studies should investigate whether higher blink rates during the WM delay lead to a correct response. Since task difficulty was not controlled for in this study, participants’ task performance in both experiments was relatively high (see Tables 1 and 2). These limitations make our current dataset incapable of investigating these questions.

We also investigated the proposed “Inverted U-shape” relationship between DA and WM performance by computing a polynomial regression model on sEBR during the delay and task accuracy (Cools and D’Esposito, 2011). Though the model showed a non-linear trend in Experiment 2, the model was not significant. We believe that failure to achieve non-linear model significance was due to lack of extreme (sub- and supraoptimal) sEBRs observed in the pool of participants, which are typically found in clinical populations (e.g., with Schizophrenia; Adamson, 1995; Swarztrauber and Fujikawa, 1998). Future studies should investigate sEBR with healthy subjects and with subjects that have been observed to have extreme sEBR in order to have a wider variety of sEBRs and to better understand its connection with DA. Additionally, other methods of DA measures could be used to investigate DA during the delay period, such as correlations with neuromelanin-sensitive MRI, which can detect neuromelanin, a product of dopamine metabolism (Cassidy et al., 2019).

Our second hypothesis was that we would see significant differences in sEBR across the WM task as well as between sEBR during Rest and during the Delay period. We found sEBR to be the lowest during periods like Encoding and Probe in both Experiments, while sEBR during the Delay was the highest. Our results support previous findings which found task-related modulation of sEBR (Siegle et al., 2008; Oh et al., 2012). Prior work has found sEBR to be lower during tasks that require visual attention (Fukuda et al., 2005; Oh et al., 2012). This would explain the lower sEBR’s observed during the Encoding period when participants are encoding information into WM and during the Probe period where participants are retrieving information from WM. We also found that sEBR was the highest during the Delay period when participants were maintaining information in WM. This was also demonstrated in a different study which investigated sEBR during an A-not-B WM task where infants had to search for a hidden toy by making an eye movement to one of two locations (Bacher et al., 2017). Higher sEBR during the WM delay could be due to DA regulating the maintenance and updating of representations in WM (Westbrook and Braver, 2016), but this remains speculation until further studies directly measure dopaminergic activity during task performance. Our results further support this interpretation since Delay period sEBR was significantly higher than baseline sEBR during the Rest period. Lower sEBR during the Rest period could be explained since there is no need to update or maintain WM during this period.

To conclude, we investigated temporal changes of sEBR during different phases of a WM task and its relation to WM performance. We observed a significant positive correlation between sEBR and WM task performance during the Delay period, but not during the other phases of the task. Additionally, we found evidence for an association of sEBR during both stimulus encoding and WM probe retrieval, potential reflecting visual attention. To the best of our knowledge, this is the first study to investigate phasic and tonic sEBR during different phases of a WM task using complex visual scenes. Future studies should continue to investigate sEBRs in relation to direct measures of cortical (especially PFC) and subcortical dopamine and assess linear and non-linear relationships to task performance in healthy and clinical populations (e.g., Schizophrenia and Parkinson’s disease).

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Institutional Review Board of the City University of New York Human Research Protection Program (CUNY HRPP IRB). The participants provided their written informed consent to participate in this study.

TE designed the study. JO, CR, and BG performed the research. JO analyzed and interpreted the data, prepared the figures, and wrote the final manuscript. CR, BG, and TE edited and reviewed the manuscript. All authors contributed to the article and approved the submitted version.

Research reported in this publication was supported by the National Institute of Mental Health of the National Institutes of Health under Award Number R56MH116007 (TE).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.