Mild cognitive impairment (MCI) is a syndrome defined as cognitive decline more significant than expected for an individual’s age and education level, but that does not interfere notably with daily life activities (Gauthier et al., 2006). Cognitive or behavioral impairment involves memory, executive function (EF), visuospatial abilities, attention, language, and changes in personality, behavior, or comportment (Petersen, 2004). MCI may be a prodrome to several degenerative, vascular, psychiatric, and medical conditions. It could progress to degenerative diseases, such as Alzheimer’s dementia (AD), frontotemporal dementia (FTD), or dementia with Lewy bodies (DLB); it is also possibly a symptom of non-degenerative conditions, such as vascular cognitive impairment (VCI), major depression, generalized anxiety disorders, uncompensated heart failure, and poorly controlled diabetes mellitus (Petersen, 2016). MCI is characterized as either amnestic MCI (aMCI) or non-amnestic MCI (naMCI), depending on the cognitive domains affected. Both MCI phenotypes could be further divided into single- or multiple-domain MCI depending on the number of cognitive domains affected. The number of domains impacted has crucial implications for determining the degree of the underlying brain pathology, burden of disease, and likelihood of progression to dementia. The annual rate at which MCI progresses to dementia varies between 8 and 15% per year, implying that it is essential to identify and treat patients with MCI (Petersen, 2016). The prevalence of MCI is estimated to be between 15 and 20% in persons 60 years and older, making it a common condition encountered by clinicians. Generally, patients with aMCI eventually develop into AD, and those with naMCI progress into non-AD (Gauthier et al., 2006). Furthermore, since MCI is a clinical diagnosis informed by neuropsychological data, some individuals are never diagnosed; in contrast, other individuals, especially older persons, are often diagnosed late because neuropathologic changes are often present without signs or symptoms (Jack et al., 2018). Sharp distinctions between adults with normal cognition and patients with MCI are complex, and clinical judgment must be used to make these differences (Albert et al., 2011). Therefore, researchers may use biomarker abnormalities that can be detected several years before clinical symptoms appear to aid clinical judgment (Jack et al., 2018). Although these biomarkers refer to those individuals with MCI on the AD spectrum, it is likely that as biomarkers for other disorders evolve, such biomarkers will be matched with a particular phenotype of MCI and increase the ability to predict clinical outcomes.

Several studies suggest that neurodegenerative pathophysiology may be present earlier in vision-related brain structures than in other structures (Albers et al., 2015; Kusne et al., 2017). A previous study found neurofibrillary tangles and neuritic plaques in the occipital cortex of healthy adult subjects and patients with MCI (McKee et al., 2006). These findings provide histopathological evidence that MCI caused by degenerative disease processes disrupts the visual system much earlier than previously thought and thus substantiates the use of visual biomarkers in early disease detection. Many studies have also reported significant differences in structural and functional visual measures between patients with MCI and those with normal cognition (Javaid et al., 2016; Chan et al., 2019), with some tests being capable of detecting MCI (Galetta et al., 2017). Eye tracking, especially saccadic eye movement (SEM) evaluation, is beneficial in determining the disease stage in patients with minor motor dysfunction and cognitive deficits (Anderson and MacAskill, 2013). There are two main types of SEMs: visually guided saccades and voluntary (or volitional) saccades. A visually guided saccade is an involuntary positioning reaction to a new event in the field of vision. In contrast, voluntary saccades result from purposeful activity in various paradigms, such as antisaccade (AS), memory-guided saccade, and predictive saccade paradigms. More extensive reliance on higher-level executive control during volitional saccades results in increasingly complex brain stimulation patterns. The network involved in saccade generation includes subcortical and cortical regions (Leigh and Kennard, 2004; Pierrot-Deseilligny et al., 2004). Central to these executive control functions is the ability to appropriately select behaviorally advantageous actions and withhold or suppress inappropriate actions in a given behavioral context (response inhibition) because they interfere with completing motor and cognitive goals. Response inhibition is an essential EF implemented by the prefrontal cortex (PFC). Various cortical subregions, such as the ventrolateral prefrontal cortex (VLPFC), presupplementary motor area (pre-SMA), dorsolateral prefrontal cortex (DLPFC), lateral orbitofrontal and anterior cingulate (ACC) areas, play vital roles in inhibitory behavior control, and both brain hemispheres are involved in inhibitory cognitive function (Bokura et al., 2001; Simmonds et al., 2008; Chikazoe, 2010). Pertinently, inhibitory control is a heterogeneous construct (Barkley, 1997; Nigg, 2000; Aron, 2011), and therefore, researchers use a range of paradigms, such as AS and Go/No-go, to assess specific aspects of this function.

In AS trials, at least two steps are necessary: inhibition of the reflexive response for a visually guided saccade to the target and the reversal of the stimulus location into a voluntary motor command to look in the direction opposite from the stimulus. AS trials activate the oculomotor network more than visually guided trials (e.g., PS and Go trials) and may also recruit additional brain areas, such as the DLPFC and ACC, which are unnecessary in visually guided trials. The AS task encompasses various cognitive processes, such as decision making, working memory, goal-oriented behavior, learning, and attention (Jamadar et al., 2013). The traditional Go/No-go task design involves only two stimuli: a “Go” stimulus and a “No-go” stimulus. In Go/No-go tasks, a “Go” stimulus (or stimuli) requires executing a response, and a “No-go” stimulus (or stimuli) requires withholding a response (response inhibition). Typically, the task is weighted toward Go stimuli to build up a prepotent tendency to respond, thereby increasing the inhibitory effort necessary to withhold responding to No-go stimuli successfully (Simmonds et al., 2008). Performance of a Go/No-go task requires the recruitment of various cognitive components, including working memory, goal-oriented behavior, and attention (Chikazoe, 2010). Studies employing Go/No-go and AS tasks have shown very similar activation patterns, suggesting that these tasks inherently test the same cognitive processes (i.e., response inhibition) (Chikazoe, 2010). Additionally, response inhibition is vital to isolate and identify as it relates to self-regulatory processes in behavior and behavior changes (Simmonds et al., 2008). Crawford et al. (2005) showed that patients with AD have impaired inhibitory control and error correction that exceed the effects of normal aging; these impairments are related to the severity of dementia. To the best of our knowledge, no study has shown the extent of inhibitory control using both PS/AS and Go/No-go paradigms in patients with MCI. To examine the specific self-monitoring process, we distinguished between the inhibition errors that were corrected and uncorrected. We defined inhibition errors as the combined sum of corrected and uncorrected errors. Uncorrected-inhibition errors consisted of the trials in which the eye was captured inappropriately by the target, but participants did not make an error-corrected saccade.

Previous work has shown that PS/AS paradigms could distinguish controls from patients with MCI using latency and error saccade parameters (Opwonya et al., 2021). However, most studies have had few participants and used a blocked administration approach while conducting eye-tracking tasks. It is not well established whether the outcomes between these two methods are similar since the administration approach (block vs. interleaved) affects saccade latencies and errors; the interleaved approach produces larger PS/AS differences in error rates, while block administration produces larger latency differences (Zeligman and Zivotofsky, 2017). To determine the saccade parameter differences between subjects with MCI and controls, we tested patients in a series of interleaved saccadic paradigms: PS/AS and Go/No-go. To date, most eye-tracking studies have examined Western and educated populations. A drawback of this constraint on the studied population is that it limits the understanding; hence, they cannot be generalized to other populations since eye movements vary across different cultural and ethnic groups (Rayner et al., 2007; Knox et al., 2012). Here, we studied saccadic behavior patterns in East Asia to bridge this gap.

In summary, this study aimed to broaden the current knowledge of saccade behavior and executive control deficits related to eye-movement control in patients with MCI using interleaved saccade paradigms and neuropsychological assessment. Additionally, this paper discusses specific features of eye-movement control–saccadic latency and inhibitory control in patients with MCI and how these features relate to the typical effects of healthy aging in native Koreans. Considerable knowledge of the neural basis of eye movements and the overlap of structures implicated in inhibition with those suggested to be dysfunctional in patients with MCI make inhibitory oculomotor outcomes potential biomarkers for screening and identifying MCI. The analysis of eye-tracking measures provides information on the cognitive and neural mechanisms involved in the volitional control of behavior. We speculate that brain changes occurring in MCI patients may influence their saccade behavior, providing the potential for cost-effective and straightforward objective measures to assess executive control deficits in patients with MCI.

In this study, our main objectives were to increase the current knowledge of saccade behavior and executive control deficits related to eye-movement control and to examine how the saccade parameters in patients with MCI relate to the typical features of cognitively healthy aging. We examined visually guided saccades and voluntary/volitional saccadic eye responses between patients with MCI and HC using interleaved PS/AS and Go/No-go paradigms. We also investigated the link between saccadic eye movement parameters and several neuropsychological tests frequently used during MCI screening.

Previous eye-tracking studies have shown a close relationship between cognitive processes and fixation duration, with longer fixations associated with greater attentional focus (Rayner, 2009; Henderson et al., 2015). Fixation duration represents the relative engagement with the object, i.e., the greater the average fixation duration, the greater the level of engagement. Our study found that the MCI group had a shorter fixation duration than controls, which may indicate poorer attentional focus in patients with MCI than healthy controls.

In our study’s visually guided tasks (PS and Go), the MCI group had a lower proportion of correct saccades than healthy controls had in all valid trials. There are several possible explanations for this result. First, this outcome could indicate that subjects with MCI have deficits in the involuntary control of saccades resulting from dysfunction of the neural circuitry. The visually guided saccade generation network includes subcortical and cortical regions affected by MCI due to degeneration (Braak and Braak, 1995). Another possible explanation for the poorer performance is that patients have cognitive control deficits, and some cognitive processes such as attention and learning influence performance on visually guided saccades (Hutton, 2008). The AS paradigm requires a volitional saccade away from the target plus inhibition of a visually guided saccade. In contrast, the No-go task requires only the inhibition of a visually guided saccade. The proportion of correct responses in AS and No-go tasks revealed a significant difference between the MCI group and controls. This significantly lower number of correct responses in the MCI group may suggest a deficit of voluntary saccadic control in patients with MCI as volitional control tasks have increasingly complex brain stimulation patterns that recruit additional neural regions such as the prefrontal cortex not activated in visually guided tasks (McDowell et al., 2008). Coupled with the executive control deficits seen in patients with MCI, they may be unable to cope with these additional demands and hence the poorer performance when compared to healthy controls.

In the current study, when we compared the latency of correct saccades in the MCI group to that of HC in the visually guided paradigms, we found that patients had longer latencies when initiating saccades in the PS task. In contrast, the MCI group had similar latencies with controls in the Go task. These differences can partly be explained by the weaker attention-demanding nature of the Go/No-go paradigm, enabling MCI participants to automatically orient their gaze to the target with similar reaction times as controls. It is also fundamental to note that during the interleaved PS/AS trials, the subjects do not know the trial type and, aware of the possibility of an AS, engage proactive inhibitory control to tamper with the involuntary PS. The longer latencies in patients with MCI during PS suggest that during the gap period, attentional disengagement which is regulated by the superior colliculus (SC), occurs at a slower rate in patients. In other words, the inhibition of the fixation neurons is more prolonged, and the activation of the saccade neurons to aid the beginning of a successive saccade is slower in patients with MCI than in HC. The results point to the likelihood that the subcortical oculomotor network is affected in the early stages of the disease (Braak and Braak, 1995). Notably, when we compared the AS latency of correct saccades in patients with MCI to HC, there was no significant difference between the groups. We previously showed that in PS/AS block administration, PS latency did not reliably distinguish between patients with MCI and age-matched controls, but AS latency was significantly different between the groups (Opwonya et al., 2021). The differences in saccade latency suggest that latency differences between patients with MCI and HC are more pronounced in the PS interleaved and AS block modes, indicating a possible effect of the mode of administration (blocks vs. randomly interleaved).

In this study, in both visually guided paradigms, the MCI group committed a higher number of errors than HC. When we assessed all errors by subtype, the proportion of self-corrected errors, self-corrected latency, and time-to-correct erroneous saccade showed little effect of disease, as there was no significant difference between the groups. The proportion of inhibition errors in AS trials is a robust measure of inhibitory control, with fewer errors signifying better response inhibition (Munoz and Everling, 2004). The primary dependent measure in Go/No-go tasks is the error commission rate, i.e., making a “Go” response on “No-go” trials. This study indicates that patients with MCI had higher inhibition errors in both No-go and AS trials. In the Go/No-go paradigm, patients with MCI produced approximately 10% more inhibition errors than controls. In the PS/AS paradigm, patients with MCI produced approximately 5% more inhibition errors than controls. Pertinently, Go/No-go task performance is also subject to a high cognitive load. It depends on the ability to switch between inhibitory and movement modes on a trial-by-trial basis, contributing to the high error rates in this task (Crawford et al., 2005). Overall, there was a higher proportion of inhibition errors in the MCI group, probably resulting from alterations in the PFC circuitry leading to changes in the input to other oculomotor regions involved in the initiation and execution of voluntary saccades (Leigh and Kennard, 2004). The findings observed in this study mirror those of previous studies that have examined the effect of inhibitory control in patients with MCI and early AD (Crawford et al., 2005; Alichniewicz et al., 2013; Heuer et al., 2013; Holden et al., 2018; Wilcockson et al., 2019).

When people commit errors, the deviation from intended behavior may require a minimal behavioral adjustment, or it may entail reassessing an entire behavioral plan. An efficient error-monitoring system should be able to evaluate the significance of an error, in addition to detecting that an error has occurred. In both the AS and No-go tasks, the MCI group had more uncorrected-inhibition errors than HC. Additionally, in the AS task, the MCI group corrected significantly fewer errors than HC. The weaker attention-demanding nature of the No-go paradigm revealed some preservation of inhibitory capacity in both groups of participants as there was no significant difference in corrected errors. A possible explanation for these findings might be that some errors, such as rapidly corrected saccades, may not even reach the level of consciousness (Nieuwenhuis et al., 2001). We believe that patients with MCI are more susceptible to not correcting errors committed due to alterations in the self-monitoring and error correction network, which recruits the PFC and ACC regions. This finding supports existing research highlighting the relationship between uncorrected-inhibition errors and dysregulation of the self-monitoring and error correction network (Crawford et al., 2013; Peltsch et al., 2014; Wilcockson et al., 2019). Overall, these findings are consistent with the dysfunction of the error-monitoring processes in patients with MCI.

The present study results determined the correlation between specific saccade parameters and neuropsychological measures. We found weak to moderate correlations across various measures in line with previous studies (Heuer et al., 2013; Yang et al., 2013; Peltsch et al., 2014; Chehrehnegar et al., 2019; Wilcockson et al., 2019; Polden et al., 2020). Some of the correlations were likely related to demographic factors (e.g., age, sex, and years of education) because controlling for them diminished the effects, especially for the MCI subjects. In general, saccade parameters correlated most strongly with executive function, language, and attention test scores, which index the functioning of these cognitive domains. The correlation between executive function, language, and attention scores and visually guided saccades is noteworthy because these saccades might be considered a reflexive movement toward a visual target onset that would not place demands on higher-order cognitive processes. However, the findings suggest that in interleaved trials, aware of the likelihood of an AS or No-go, the participants engage proactive inhibitory control processes that permit saccade planning to be adjusted based on the cue’s cognitive processing, indicating the appropriate response for each trial. Consequently, the visually guided saccades are not purely involuntary, and cognitive control processes are involved in executing these saccades.

Furthermore, our study found a significant positive correlation between neuropsychological scores and correct responses and a negative correlation with error responses. Specific brain regions are implicated in the control of these eye responses. The parietal eye field (PEF), which projects to the SC, plays a vital role in triggering reflexive saccades and is essential in preparing for correct saccades (Leigh and Kennard, 2004). Other brain areas vital for correct and error responses include the supplementary eye field (SEF) involved in sequencing and planning saccades and the DLPFC crucial for inhibiting unwanted saccades (Leigh and Kennard, 2004). Negative correlations between cognitive measures and error responses in the MCI group may suggest that neurodegenerative changes in areas responsible for high-level cognitive functions and saccade execution are progressing in parallel. Also, the negative correlations may indicate an association between early cognitive decline and inhibitory impairment. These results build on existing evidence of a relationship between executive functions and error responses (Heuer et al., 2013; Peltsch et al., 2014; Koçoğlu et al., 2021). Overall, our data offers evidence for a direct relationship between cognitive measures of executive function, language, attention, and correct and error responses modulated by higher cortical regions.

Limitations of our study are as following. Eye-movement control may be influenced by culture and ethnicity factors (Knox et al., 2012; Polden et al., 2020), and we examined saccades in only native Korean participants, which may limit the generalizability of our findings. Secondly, our results were derived from a single eye-tracking recording, and MCI patients may revert to normal or progress to other conditions. Therefore, studying how the saccade parameters change over time is worth investigating. Future longitudinal studies will help to determine the clinical implications of these findings.

Our data provided a detailed description of saccade performance changes in MCI subjects who underwent rigorous neuropsychological assessment. This study provided further evidence that in the interleaved design, visually guided saccades can be modified based on cognitive processing of the cue indicating the appropriate response for each trial and shows a relationship between EF and saccade error responses. Additionally, we highlighted the importance of the mode of administration (blocks vs. randomly interleaved) on saccade latency between patients with MCI and HC, substantially adding to the understanding of the effects of administration mode on saccade behavior. Furthermore, the results of this study indicate that PS/AS and Go/No-go paradigm data may reveal distinctive deficits in latency, inhibitory control, and error monitoring in MCI patients compared to these aspects in cognitively healthy age-matched controls.

Taken together, the evidence from this study suggested that performance in the PS/AS and Go/No-go paradigms is sensitive and objective for detecting subtle cognitive deficits. Combining these results with future longitudinal studies that track which MCI patients progress to other conditions has strong potential for clinical application.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

The studies involving human participants were reviewed and approved by the Institutional Review Board of Chonnam National University Hospital (IRB No. CNUH-2019-279). The patients/participants provided their written informed consent to participate in this study.

JUK, JIK, CW, K-MJ, KL, and JO made conceptualization, methodology, software, validation, investigation, writing—review and editing, and project administration contributions. JO, CW, and JIK made formal analysis, data curation, and writing—original draft preparation contributions. JUK conducted supervision and funding acquisition. All authors read and agreed to the published version of the manuscript.

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.