Gaze-based communication and control represent a promising approach for hands-free interaction with computers and other devices, been in development for more than three decades (Duchowski, 2018). It is particularly beneficial for paralyzed individuals (Majaranta, 2011), but also can be used by healthy people, especially as an effective input in virtual, augmented, and mixed reality (VR/AR/MR) (O’Callaghan, 2024; Plopski et al., 2022). To issue a command, a user of this technology typically keeps their gaze on a “gaze-sensitive” element of a graphical interface, such as a screen button, for a relatively short period (Duchowski, 2018; Majaranta, 2011). An eye tracker captures the gaze position, and when the gaze dwell time in the vicinity of a screen button exceeds a specified threshold, the button is activated. The action resembles pressing a computer mouse button or tapping on a touchscreen but involves a temporary suspension of natural spontaneous gaze behavior.

Despite the growing interest in gaze-based human-computer interaction, the brain mechanisms underlying eye movement control in this context have received little attention, likely because this interaction has been considered primarily an engineering challenge. However, gaze-based interaction has unique characteristics that distinguish it from other forms of voluntary behavior. First, the “stillness-based” nature of gaze-based interaction is in stark contrast to typical manual actions, which involve intentional movement. In everyday scenarios, hands remain still by default, whereas the default state for eyes is their frequent motion (primarily through saccades). Secondly, intentional gaze dwells1 used in gaze-based interaction represent a rare example of voluntary control executed with minimal effort. Thus, studying the physiological mechanisms specifically related to gaze-based control may enhance this technology. Moreover, using gaze-based actions as a model of voluntary actions could also provide insights into the brain’s mechanisms of voluntary control in general.

Voluntary control of eye movement has been extensively studied in brain research, usually involving salient visual stimuli. In these studies, participants may be required to make voluntary saccades in different directions (often opposite to the stimulus), as in the antisaccade task (Funahashi et al., 1993; Hallett, 1978; Munoz & Everling, 2004), or to maintain a gaze fixation despite an urge to make a saccade, as seen in response inhibition paradigms like the NoGo task (Brown et al., 2006, 2008; Isoda & Hikosaka, 2007) and in delayed antisaccade and prosaccade tasks [e.g., (Ettinger et al., 2008)]. These tasks often impose stronger control demands than most everyday activities, which can lead to an overemphasis on the effects of voluntary control and potential confounding with effort-related effects (Jarvstad & Gilchrist, 2019). To address this issue, Jarvstad and Gilchrist (2019) introduced a novel, less demanding paradigm for studying saccade control. Their design included prolonged fixations as a baseline (20 sec), which still demanded considerable cognitive effort from participants and, therefore, did not fully eliminate unnecessary effort. Moreover, instructions to suppress natural gaze behavior and maintain extended fixation on a target—often to an uncomfortable degree—are common even in baseline conditions of psychological and neurophysiological studies that do not specifically target voluntary eye movement control.

In contrast, gaze–interaction interfaces typically employ much shorter dwell times—often 1 sec or less—to minimize user discomfort. Leveraging gaze-based interaction paradigms could offer valuable insights into the fundamental mechanisms underlying voluntary control, without the need to account for effortful behavior. Voluntary brain control of gaze dwells used for interacting with computers is expected to differ from the control of similarly timed spontaneous dwells, which are highly automatized and basically driven by distinct goals or intentions, or even by reflective reactions. However, since cognitive effort remains minimal even for intentional dwells in this context, any differences in brain activation between intentional and spontaneous dwells during gaze-based interaction are likely to more precisely reflect the voluntary components of eye movement control than traditional paradigms. Thus, from the neuroimaging perspective, intentional dwells represent a rare example of voluntary gaze control behavior achieved without substantial effort. By offering an ecologically valid setting, a naturalistic gaze-based interaction paradigm complements traditional approaches—such as countermanding or anti-saccade tasks—that rely on external cues to inhibit gaze shifts and often impose additional demands on top-down inhibitory control to suppress prepotent responses.

Understanding the brain mechanisms underlying low-effort voluntary control of eye movements is important not only for basic neuroscience but also for advancing gaze-based interaction technologies. This knowledge could specifically address the “Midas touch” problem—the challenge faced by gaze–sensitive interfaces in distinguishing intentional eye movement patterns used to issue commands from similar patterns that occur spontaneously or without a clear intent (Jacob, 1990). In particular, when an interface is designed to respond to prolonged dwells, it inevitably risks triggering unintended actions caused by prolonged dwells that are out of conscious control.

Existing countermeasures against the Midas touch problem impose additional effort on users (Majaranta & Bulling, 2014; Velichkovsky et al., 1997). Identifying the differences between intentional and spontaneous gaze dwells could enable the development of brain signal markers capable of distinguishing intentional dwells, allowing interfaces to respond appropriately (Ihme & Zander, 2011; Protzak et al., 2013).

A straightforward approach to studying the difference between intentional and spontaneous gaze dwells is to collect them from participants engaged in tasks using a gaze-based human-computer interface (Shishkin, Nuzhdin, et al., 2016). The task and visual display should be designed to provoke a sufficient number of spontaneous gaze dwells with a duration exceeding the dwell time threshold used in the interaction. To this end, Shishkin, Nuzhdin, et al. (2016) developed a gaze-controlled game and recorded electroencephalogram (EEG) data as participants played. Intentional and spontaneous dwells were labeled based on a rule requiring the activation of gaze-based control prior to each move, using a dwell at a designated screen position (the “switch-on button”). Intentional dwells were found to exhibit the EEG phase-locked activity gradually increasing up to the dwell’s termination, likely reflecting anticipation of an interface triggering (Shishkin, Nuzhdin, et al., 2016). However, this EEG marker did not appear to be specific to the intentionality of a dwell—similar activity could also develop in spontaneous dwells perceived as potentially triggering unintended interface actions, as was likely the case in our online study (Nuzhdin et al., 2017). Unfortunately, the earlier time interval could not be effectively studied due to significant variability in the eye movement patterns prior to the dwell, which was determined using the “switch-on button.”

In the current study, we investigated whether voluntarily prolonged gaze dwells used to control a computer are associated with distinct brain activity that is not present, or at least less pronounced, in spontaneous long dwells from the onset. We employed the same gaze-controlled game as used in the study by (Shishkin, Nuzhdin, et al., 2016) but modified the gaze-based control rules to minimize the influence of preceding events on the brain activity during intentional dwells. Most of the game rules remained unchanged, as they provided a basis for comfortable interaction and stable motivation in the previous study. Instead of recording the EEG, we co-registered 306-channel magnetoencephalography (MEG) data with eye-tracking data. A special procedure for selecting intentional and spontaneous dwells was developed to ensure contrast on the “voluntariness” scale. The dataset used in the present study was previously employed to assess the feasibility of classifying spontaneous versus intentional gaze dwells using MEG data, based on a simplified labeling procedure (Ovchinnikova et al., 2021). However, that earlier work applied convolutional neural networks (CNNs) to classify the data, treating neural features primarily as abstract input vectors rather than signals reflecting interpretable neurophysiological processes. While this approach demonstrated classification potential, it did not involve a formal analysis of the underlying neural dynamics, leaving the cognitive and neurobiological mechanisms of intentional gaze control unexamined.

In the present study, we aimed to fill this gap by formally characterizing the behavioral and neural differences between intentionally and spontaneously prolonged gaze dwells. We hypothesized that the two types of dwell would differ in both behavioral characteristics and associated neural signatures. Specifically, we expected intentional dwell events—being goal-directed and embedded within the task’s control structure—to exhibit distinct gaze dynamics, including greater temporal stability and reduced deviation from central fixation during the dwell period. At the neural level, we anticipated that intentional gaze withholding would modulate activity in frontal cortical regions involved in voluntary oculomotor control, particularly the frontal eye fields (FEF), as well as in posterior parietal areas supporting spatial attention shifts. This expectation was based on the assumption that reduced oculomotor drive during intentional dwells would be accompanied by diminished covert attentional reorienting (Hafed & Clark, 2002; Rolfs et al., 2008). We further hypothesized that such differences would be reflected in regional alpha-band oscillations, given the well-established role of alpha synchronization in functional inhibition (Jensen, 2024) and its relevance to both oculomotor and attentional processes (Cruz et al., 2025; Hwang et al., 2014).

Although the evoked neural components linked to intentional gaze control remain largely underexplored, we posited that these events may engage preparatory cortical dynamics, potentially expressed as slow ERP shifts preceding dwell onset. This would be analogous to the ramping activity often observed in anticipation of voluntary saccades (Errington & Schall, 2025; Richards, 2013). Since intentional, goal-directed actions typically involve advance motor and cognitive planning (Haggard, 2008), we expected neural markers of gaze withholding to emerge early in the dwell period—or even during the pre-dwell interval—reflecting anticipatory control mechanisms. Accordingly, we employed an exploratory whole-head analysis rather than a region-of-interest (ROI) approach, aiming to identify robust group-level differences in both evoked and induced activity that could reliably differentiate intentional from spontaneous gaze behavior.