Internal coupling: Eye behavior coupled to visual imagery ☆

Our eyes do not only respond to visual perception but also to internal cognition involving visual imagery, which can be referred to as internal coupling . This review synthesizes evidence on internal coupling across diverse domains including episodic memory and simulation, visuospatial memory, numerical cognition, object movement, body movement, and brightness imagery. In each domain, eye movements consistently reflect distinct aspects of mental imagery typically akin to those seen in corresponding visual experiences. Several findings further suggest that internal coupling may not only coincide with but also supports internal cognition as evidenced by improved cognitive performance. Available theoretical accounts suggest that internal coupling may serve at least two functional roles in visual imagery: facilitating memory reconstruction and indicating shifts in internal attention. Moreover, recent insights into the neurobiology of internal coupling highlight substantially shared neural pathways in externally and internally directed cognition. The review concludes by identifying open questions and promising avenues for future research such as exploring moderating roles of context and individual differences in internal coupling.


Internal coupling: eye behavior coupled to visual imagery
Our eyes are an important window to the environment as they enable us to find and process external visual information, such as when searching for a friend at the train station or reading a book.However, recent research is increasingly recognizing that our eyes do not just interact with the external world, but are also deeply involved in processes of our internal mental world, such as memory recall and imagination (Benedek, 2018;Theeuwes et al., 2009;van Ede et al., 2019).This review provides an overview of the accumulating evidence about the many ways in which eye behavior couples to mental processes and finally explores possible explanations of this phenomenon.
A distinction can be made between externally directed cognition, characterized by directing attention towards the external world, and internally directed cognition, characterized by focusing attention on self-generated internal (mental) representations (Chun et al., 2011;Dixon et al., 2014).Internally directed cognition encompasses cognitive processes such as memory recall, future thinking, imagination, mind wandering, mental arithmetic or navigation, and creative ideation (J.Pearson, 2019).For instance, a person might remember the beach they once visited, think of different destinations for their next vacation, or mentally map out the quickest route home.Interestingly, eye behavior was shown to distinguish between external and internal attention focus (Korda et al., 2023(Korda et al., , 2024;;Walcher et al., 2023).For example, during periods of internal focus, individuals exhibit increased blink frequency and duration, fewer but longer fixations, and greater variability in pupil diameter (Annerer-Walcher et al., 2018, 2021;Hollander and Huette, 2022;Smallwood et al., 2011;Uzzaman and Joordens, 2011).Several notions have been put forward to explain differences in eye behavior between externally and internally directed cognition.Eye behavior differences can be partly attributed to perceptual decoupling, which refers to the mind's capacity to disengage attentional processes from the sensory input (Smallwood and Schooler, 2015).As eye behavior is no longer determined by external stimuli it appears to become more undirected and spontaneous (e.g., Annerer-Walcher et al., 2018;Smallwood et al., 2011).Yet, eye behavior during internally directed cognition also shows some systematic characteristics.For example, it includes a greater tendency towards visual disengagement (Walcher et al., 2017), which is characterized by eye behavior oriented towards reduced visual input, such as increased blink rate and duration (Walcher et al., 2017), gaze aversion (Servais et al., 2023), and "staring into space" as reflected by decreased eye vergence (Benedek et al., 2017;Huang et al., 2019).Yet, these mechanisms characteristics alone cannot fully explain how internally directed cognition affects eye behavior dynamics.Eye tracking studies in diverse fields of research such as memory encoding and retrieval or numerical cognition have consistently revealed that eye behavior also systematically reflects changes in task-induced internal representations and shifts of internal attention.Hence, eye behavior is not only guided by external stimuli but also coupled to internal representations, a phenomenon that we refer to as internal coupling.
While extensive research has explored the dynamics of eye behavior in various forms of imagination, these fields have remained disparate, with each typically focusing on specific cognitive tasks or environments.This review seeks to bring together the current state of research in a unified framework of internal coupling, thereby highlighting how our eyes engage with our mental landscape.To this end we provide a comprehensive overview of the existing literature on internal coupling by synthesizing available evidence across distinct lines of research.We review the available evidence to examine how robustly internal coupling is observed across different tasks and conditions.We then consider available literature that may explain why internal coupling happens and whether it is merely an epiphenomenon or plays a functional role in cognitive performance.Finally, we explore related evidence that sheds light on the potential neurocognitive mechanisms of how internal coupling takes place.In sum, this work reviews when, why, and how internal coupling happens.

Evidence of internal coupling
In this section, we will review the evidence supporting internal coupling from seven lines of research: episodic memory, episodic simulation, visuospatial memory, numerical cognition, object movement, body movement, and brightness imagery.A common aspect across these cognitive domains is their tendency to evoke visual imagery, often accompanied by corresponding eye movements.We acknowledge that these categories, while seemingly disjoint, often exhibit considerable overlap with occasional shared underlying mechanisms, making our categorization to some extent arbitrary.Nevertheless, the type of assessments and task employed can vary significantly, thus for clarity and in line with the historical evolution of these research fields, we will explore these domains independently, reserving a detailed discussion of their interconnections for Sections 3 and 4. The review focuses primarily on works that studied internal coupling in controlled settings, where eye movements are exclusively coupled to mental representations usually in the absence of salient external stimulation, thereby isolating eye movements driven solely by internal events or sources.

Imagery associated with episodic memory
Memory research has been at the forefront of studying eye behavior in internally directed cognition in the context of retrieval processes.Various studies revealed that when a visual stimulus was present during encoding at a specific location, but is no longer available during retrieval, individuals still tend to gaze at the location where the stimulus once was (e.g., Altmann, 2004;Laeng and Teodorescu, 2002;Richardson and Spivey, 2000; see Fig. 1A, B).This behavior is sometimes referred to as the looking at nothing phenomenon (Ferreira et al., 2008;Richardson and Spivey, 2000) or as gaze reinstatement (Wynn et al., 2019).Furthermore, the looking at nothing effect has also been observed when people listen to a story (Spivey and Geng, 2001) and scanpaths, a series of fixations and saccades, during story telling (i.e., recall) resembled those enacted during story listening (Johansson et al., 2006).When participants were asked to determine the correctness of a sentence, which was accompanied by a grid containing a task-irrelevant image in one corner, they still tended to direct their gaze towards the grid region where the image was previously located (Richardson and Spivey, 2000).Furthermore, Dudschig et al. (2013) demonstrated that even the semantic content of words can affect gaze behavior.In their study, participants categorized words as either nouns or non-words by making an upward or downward saccade.Nouns associated with spatial meanings, such as 'sun' typically associated with an upward direction, decreased saccade latency when saccade direction matched the nouns' semantic direction.Together these findings illustrate common tasks used in the looking at nothing paradigm, where gaze is directed towards (previously) relevant locations during memory recall.
Studies have shown that scanpaths during memory recall are not a precise temporal replication of scanpaths during encoding (Bone et al., 2019;Damiano and Walther, 2019;Foulsham and Kingstone, 2013;Johansson et al., 2012;2022;Wang et al., 2020;Wynn et al., 2020; see Fig. 1A, B).Importantly, while some studies used the looking at nothing paradigm, i.e., with a blank screen during recall, others have instead used an old/new recognition task, where there is an image (old or new) presented on the screen during recall.While the latter paradigm does not directly test internal coupling, it is nonetheless important for the discussion on the relationship between memory and eye behavior.For example, Wynn et al. (2020) demonstrated that gaze reinstatement, in an old/new recognition task, did not consistently predict image recognition as its influence was most pronounced early on (within the first second) and diminished over time.Furthermore, temporal dynamics were also investigated in a looking at nothing paradigm.Johansson et al. (2022) showed that the timing and pattern of scanpaths was preserved between encoding and recall.On the other hand, Gurtner et al. (2019); (2021) found differences in temporal patterns as participants looked more often at the same place and repeated the same fixation sequences in recall compared to encoding.Overall, while there is a general consensus that gaze reinstatement occurs during recall, the exact nature and temporal dynamics of this process are still a matter of debate and further research.
Looking at nothing might affect memory performance, suggesting that internal coupling is more than an epiphenomenon.Research showed that restricting gaze position, by means of (central) fixation, during encoding or retrieval decreases memory performance (Damiano and Walther, 2019;Foulsham and Kingstone, 2013;Johansson et al., 2012;Johansson and Johansson, 2014;Laeng et al., 2014;Laeng and Teodorescu, 2002;Lenoble et al., 2019;Liu et al., 2020;Scholz et al., 2016) as well as vividness of internal representations (Bone et al., 2019;Lenoble et al., 2019;Liu et al., 2020).Furthermore, memory retrieval accuracy was higher when the item was shown at the same location as during encoding, while displaying the item at the interfering position decreased accuracy, suggesting that spatial similarity in gaze position during encoding and retrieval positively affects memory performance (Johansson and Johansson, 2020;Scholz et al., 2016).In line with this, scanpath similarities between encoding and retrieval were predictive of subjective memory quality (Johansson et al., 2022;Wynn, Liu, and Ryan, 2022).Overall, gaze similarity between encoding and recall has been shown to have a positive effect on memory performance.
However, not all findings consistently showed this relationship (Armson et al., 2021;Bone et al., 2019;Chiquet et al., 2021Chiquet et al., , 2022;;Gurtner et al., 2021;Kinjo et al., 2020;Richardson and Spivey, 2000).For example, restricting gaze behavior during encoding did not affect memory performance (Richardson and Spivey, 2000), looking at nothing did not differ between correct and incorrect retrievals of objects in a virtual reality setting (Chiquet et al., 2021(Chiquet et al., , 2022)), and gaze reinstatement did not correlate with memory accuracy of natural images (Bone et al., 2019).Interestingly, while gaze restriction generally did not affect memory performance, i.e., accuracy and reaction time, looking at nothing during unrestricted viewing increased memory performance (Kinjo et al., 2020).One possible explanation is that covert shifts of attention are sufficient to improve memory performance (Scholz et al., 2018).When participants were instructed to recall verbal information, while either overtly (i.e., look at) or covertly (i.e., maintaining central fixation) focusing on the quadrant where information was presented, there was no benefit of looking at nothing over covert attention.Hence, internal coupling is not always related to improved performance in episodic memory tasks.
Several explanations have been proposed to account for these differences in outcomes.For example, the effect of looking at nothing appears to be subject to individual differences in cognitive capacities.Gurtner et al. (2021) showed that the relationship between fixation locations during encoding and recall was positively moderated by differences in visuospatial working memory, where higher visuospatial working memory capacity led to a higher fixation reinstatement.Furthermore, the number of autobiographical details remembered was predicted by fixation rate only in individuals with high autobiographical memory specificity (Armson et al., 2021).On the other hand, a negative relationship was observed between spatial imagery score and spatial dispersion of eye movements, indicating that participants with lower spatial imagery abilities exhibited a larger looking at nothing effect (Johansson et al., 2011).Lastly, Wynn et al. (2019) proposed that task demands play a significant role in gaze reinstatement, as memory retrieval facilitation is only necessary when demands exceed memory capabilities.In line with this assumption, Wynn et al. (2020) have shown that for more difficult images, i.e., more degraded, gaze reinstatement was a better predictor of accurate recognition compared to the easy ones in an old/new recognition task.Similarly, Johansson et al. (2022) have shown that elaborate spatial and sequential gaze reinstatement was seen for the object arrangements, which impose higher demands on relational memory, but not for scenes.Dewhurst et al. (2018) further investigated how different levels of target visibility, set size, and noise affect gaze position in a visual search task.Across all task conditions, scanpaths showed increasing variability in saccade length and position with increasing task demands.Hence, the impact of eye behavior on recall seems to be context-dependent and vary based on specific conditions and individual differences such as memory capacity.
Lastly, neuroscientific findings have shown that regions like the medial temporal lobe, hippocampus, parahippocampal place area, and fusiform gyrus are tied to both episodic memory and to reinstating eye movement patterns (Bone et al., 2019;Liu et al., 2020;Wynn, Liu, and Ryan, 2022).In a looking at nothing paradigm, neural reactivation in posterior parts of the brain, as well as within dorsal, occipital, and ventral cortical regions of interest, were positively correlated with gaze reinstatement, memory performance and image vividness during recall (Bone et al., 2019).Thus, brain regions involved in memory, which is known to be a (re)constructive process (Schacter and Addis, 2007), are also actively involved in ocular behavior associated with memory encoding and recall.
This section reviewed evidence for internal coupling in the context of episodic memory retrieval.From pioneering work by Noton and Stark (1971) to the recent study on neuro-ocular coupling (Bone et al., 2019), extensive evidence supports that recall of episodic memories is tightly associated and often supported by internal coupling of eye behavior to remembered locations or features of encoded representations.Subsequent sections will continue to explore internal coupling for different forms of visual imagery.

Imagery associated with episodic simulation
Episodic simulation refers to the process of mentally imagining or simulating a specific action or scenario, which allows individuals to explore, anticipate, or even rehearse possible outcomes (Schacter et al., 2007, see Fig. 1C, D).Akin to episodic recall, episodic simulation involves the dynamic reconstruction of past experiences to construct hypothetical future scenarios (Abraham, 2016;Hassabis and Maguire, 2007;Schacter and Addis, 2020;Schacter et al., 2007).While primarily an internal cognitive activity, episodic simulation can also manifest in observable physical responses, particularly in eye movements (Conti and Irish, 2021;El Haj and Lenoble, 2018;Gautier et al., 2022;Wynn, van Genugten, et al., 2022).
Research into gaze behavior during future thinking has yielded insightful findings.Vito et al. (2015) demonstrated that irrelevant eye movements negatively impacted future thinking, but not episodic memory, highlighting the greater importance of eye movements for constructing hypothetical episodic simulations.This is further supported by findings that episodic simulation demands more cognitive resources than recall, primarily due to heightened spatial and constructive requirements rather than an increase in personal detail (Wiebels et al., 2020).Moreover, studies comparing eye movements across future thinking and episodic memory have identified that recall prompts a greater frequency of fixations and saccades, which correlates with enhanced imagery vividness (El Haj and Lenoble, 2018).This suggests that the complexity of episodic simulation is mirrored in eye movement patterns as decreased fixation rate and gaze dispersion have been previously related to decreased autobiographical recall and imagery vividness as well as increased task difficulty (Armson et al., 2021;Cui and Herrmann, 2023;Sheldon et al., 2019).Consistent with how restraining eye movements during recall can impair memory performance, it similarly diminishes the vividness of an episodic simulation (Ladyka-Wojcik et al., 2022), leading to a reduction in spatiotemporal detail, extended retrieval times, and shorter imagined content descriptions (Gautier et al., 2022).Expanding the theory on gaze reinstatement, Wynn, van Genugten, et al. ( 2022) compared individual's gaze patterns to gaze templates, that is, across-participant gaze patterns for a specific episodic simulation (see Fig. 1C, D for a depiction of individual's gaze pattern).Gaze patterns across participants showed high similarity for a specific simulation and dissimilarity between simulations.Furthermore, a higher number of details reported correlated with a higher degree of consistency with the gaze template for the corresponding simulation.Finally, task demands also played a role, as eye movements were indicative of successful simulation only under more challenging conditions (i.e., with visual noise), in contrast to scenarios with a blank screen.Together, these findings suggest that eye movements during simulation likely mirror their role in episodic memory, both reflecting and potentially aiding the simulation based on the recall of stored spatiotemporal contextual details.

Imagery associated with visuospatial memory
Visuospatial memory involves the ability to remember the location  (A, B), Episodic Simulation (C, D), and Visuospatial Memory (E, F, G).Note.A, B: A prototypical paradigm for the looking at nothing paradigm.During memory encoding, an image is present on the screen (A), while during memory recall, the screen is blank (B).Eye behavior is represented as scanpaths consisting of fixations (red dots) and saccades (blue lines).Internal coupling is evidenced by the similar eye behavior observed during encoding (A) and recall (B).Example study: Johansson et al. (2012).C, D: A prototypical paradigm for the episodic simulation paradigm.Imagination of a hypothetical (future) event involves gradual build-up to an increasingly detailed scene.Eye behavior is represented as scanpaths consisting of fixations (red dots) and saccades (blue lines).Internal coupling would be evidenced by predictable eye behavior with more fixations in more detailed scenes.Example study: Wynn et al. (2022).E, F, G: A prototypical paradigm for the retro-cue paradigm.During memory encoding, two differently colored and oriented bars are presented on the screen (E).A retro-cue is then presented as a change in the color of the fixation cross, which now matches one of the previously shown bars, i.e., the right, yellow bar (F).Internal coupling is evidenced by eyes moving in the direction where the corresponding bar was presented (here, to the right).The orientation of the corresponding bar is then reported on the response screen (G).Example study: van Ede et al. (2019).and spatial relationship of objects.In the literature, the terms working memory and short-term memory are not used consistently, often referring to similar or overlapping constructs (e.g., Cowan, 2008); however, for the purposes of this review, we will not delve into the distinctions between these concepts.Given that similarity between the tasks used, we will review them together in this section.
Studies investigating visuospatial working memory mostly use simple stimuli like Gabor patches and bars to report color and/or location.Similarly to episodic memory, and perhaps to even a larger extent, eye behavior plays a role in the visuospatial working memory, with some authors suggesting a functional integration between the two systems (e. g., Jonikaitis and Moore, 2019;van der Stigchel and Hollingworth, 2018).Furthermore, visuospatial working memory consists of visual and spatial memory, retaining the information about object features (e.g., color, orientation) and the spatial features (e.g., location, relationships between stimuli), respectively (Klauer and Zhao, 2004;Tresch et al., 1993).Interestingly, both prohibiting eye movements as well as concurrent task-irrelevant eye movements have been shown to negatively affect spatial but not object working memory (Ball et al., 2013;McAteer et al., 2023;D. G. Pearson et al., 2014).For example, in an abducted eye paradigm, where participants were not able to plan or execute a saccade, performance was impaired on a spatial but not visual and verbal working memory task (Ball et al., 2013).Further investigation of encoding, maintenance, and retrieval phases of working memory with the abducted eye paradigm showed that eye behavior is functionally involved in encoding and maintenance but not retrieval of spatial information (D.G. Pearson et al., 2014).Similarly, task-irrelevant eye movements were shown to disrupt spatial but not object visual imagery (Vito et al., 2014).This supports the notion that eye behavior is not just an epiphenomenon but may play a functional role in retrieving the spatial features of internal representations.
Studies by Ohl and Rolfs (2017), ( 2018), (2020) used a visuospatial working memory task, Gabor patches at multiple locations, with a spatial cue prompting the observers to execute a saccade towards one of the locations.Results showed that memory retrieval was more accurate for items located at the saccade target location compared to those at incongruent positions.Memory performance was better for shorter delay periods, suggesting that the more recent the internal representation is in memory, the more beneficial the eye movements are in refreshing and maintaining this representation (Ohl andRolfs, 2017, 2018).Similar results were reported by Loaiza and Souza (2022), showing that spontaneous saccades towards targets' locations during the delay period did not affect memory performance, however, instructing participants to look at the relevant location resulted in improved memory performance.Hence, saccades following memory encoding may enhance the recall of connected features by activating associated memory components.
A looking at nothing effect was also observed in studies investigating how internal attention selects information from the visuospatial working memory (Linde-Domingo and Spitzer, 2023; B. Liu et al., 2022;Sahan et al., 2023, van Ede et al., 2019, 2020, 2021;Vries and van Ede, 2024).Selective attention operates in a similar manner over external and internal information (e.g., Awh and Jonides, 2001;Griffin and Nobre, 2003;van Ede and Nobre, 2023;Verschooren and Egner, 2023;Zhou et al., 2022), which can also be observed in eye movements (Theeuwes et al., 2009).A common way to study attentional mechanisms in this context is through the use of a retro-cue paradigm (Griffin and Nobre, 2003; see Fig. 1E, F, G).This approach typically entails the concurrent presentation of multiple stimuli, such as colored bars, during the encoding phase.Subsequently, participants receive a spatially informative cue, such as color, which directs them to recall and report the orientation of the corresponding bar.Internal coupling, i.e., looking towards the remembered location, is expected during the retro-cue presentation and not during the recall period as it is common in the episodic memory studies.For example, van Ede et al. (2019) found that selecting an item from visuospatial working memory biases gaze direction towards the item's memorized location.Importantly, these gaze biases were found in every participant.In their studies, participants were required to maintain fixation, hence, under this restricted gaze participants exhibited microsaccades rather than full gaze shifts to the original locations of the remembered items.Small gaze shifts were also shown to reflect the format of the visuospatial memory content (Lin- de-Domingo and Spitzer, 2023).In this study, images of objects in various orientations were sequentially presented in pairs, followed by a retro-cue indicating to report the orientation of one of the objects.Interestingly, orientations of both objects were reflected in gaze behavior during encoding and maintenance.This illustrates the link between ocular behavior and the internal process of attention shifts in visuospatial working memory retrieval.
Furthermore, a study by Walcher et al. (2024) showed that eye movements correspond to internal attention navigating through the mental representations during a continuous visuospatial working memory task.Participants were tasked with mentally moving a patch in a matrix, guided by auditory cues.The findings revealed that saccade directions aligned with these imagined movements triggered by the audio cues, highlighting the phenomenon of internal coupling in the navigation through the mental space.Yet, this relationship was not present in all trials or all participants, suggesting that individual differences or different cognitive strategies might influence the degree to which this coupling occurs.A similar finding was reported in an eye tracking and EEG co-registration study, where shifts of covert attention, as seen by alpha activity and microsaccades, towards a memorized location had a positive relationship, although trials without microsaccades still showed expected spatial modulation of alpha activity (B.Liu et al., 2022).Furthermore, neither of the studies found a direct relationship between task performance and internal coupling, suggesting that eye movements might be epiphenomenal (B.Liu et al., 2022;Walcher et al., 2024).The authors speculated that shifts of attention facilitate or coactivate eye movements, but the threshold for outputting an eye movement is not always reached, leading to the discrepancy between attention shifts and eye movements (B.Liu et al., 2022).Together, these findings underscore the complexity of the cognitive mechanisms at play and highlight the variability of these processes, not only across different tasks and contexts but also among individuals.
Finally, Masson et al. (2021) explored how eye movements contribute to attention shifts in a retro-cue paradigm through a study involving a patient with congenital gaze paralysis, who is unable to execute eye movements along the paralyzed axis.Surprisingly, the patient showed typical retro-cue effects with faster and more accurate responses to informative versus misleading or no cue, suggesting that eye movements do not play a functional role in directing attention towards internal representations.Hence, it appears that mere shifts in attention are sufficient to generate the benefits associated with retro-cues, with eye movements emerging as epiphenomenon resulting from the underlying change in attentional focus.
In visuospatial working memory tasks, internal coupling can enhance performance, particularly when eye movements are experimentally directed toward relevant spatial locations.Evidence also points to the presence of spontaneous internal coupling, though its impact on memory performance is less consistent.It appears that especially spatial memory might benefit from internal coupling, as task-irrelevant eye movements tend to interfere specifically with spatial, rather than object, working memory and visual imagery.Moreover, individual differences in the presence of internal coupling have been observed, though the specific contributing factors remain unidentified.

Imagery associated with numerical cognition
Numeric cognition is intricately intertwined with spatial cognition and visual imagery (Hawes et al., 2019;Hevia et al., 2008;Hubbard et al., 2005).Multiple spatial orientations have been suggested for the mental number space, including horizontal and vertical mental number lines as well as near/far space (Winter et al., 2015).The interplay between numbers and space has been demonstrated by respective eye movements, revealing a coupling of eye behavior to numerical cognition (for a comprehensive review on eye tracking studies in mathematical education see Strohmaier et al., 2020).Even basic numerical activities, such as simple counting, elicit notable shifts in gaze behavior, with ascending counts shifting the gaze rightward and upward, and descending counts shifting it downward (Hartmann et al., 2016).Similar eye behavior was observed during number comparison (Loetscher et al., 2008;Salvaggio et al., 2019).Furthermore, Salvaggio et al. (2019) showed that visual imagery is involved in magnitude comparison and that gaze behavior couples to attentional shifts between mental number representations.When colorful landscape images were used as backgrounds, internal coupling was evident early on, implying active cognitive engagement during the comparison task.Contrarily, with a blank screen, this coupling manifested only after completing the task, suggesting that active comparison did not necessitate immediate eye movements but rather occurred afterwards.Using a non-blank screen during episodic memory and simulation was shown to disrupt visual imagery processes and hence increase imagery task demands (Sheldon et al., 2019;Wynn, van Genugten, et al., 2022), which aligns with the gaze reinstatement theory postulating that internal coupling becomes more prominent as task demands increase (Wynn et al., 2019).The observation that gaze shifts during the active number comparison stage occurred exclusively with landscape picture backgrounds (Salvaggio et al., 2019), underscores the influence of imagery task demands on internal coupling in numerical cognition.
Apart from simple counting and number comparison tasks, gaze behavior was also linked to mental arithmetic (Blini et al., 2019;Hartmann, 2022;Hartmann et al., 2015;Masson and Pesenti, 2023;Salvaggio et al., 2022; see Fig. 2A, B).For example, Blini et al. elicited optokinetic nystagmus (OKN) to investigate the potential functional role of eye movements in addition and subtraction.While horizontal OKN did not influence mental subtraction or addition, vertical OKN did impact mental subtraction so that fewer errors were made during the downward as compared to the upward or static OKN.In contrast, vertical OKN was implicated in solving addition problems, as problems were solved faster when upward eye movements were elicited compared to downward eye movements, fixed gaze position, and unrestricted viewing (Hartmann, 2022).Another study found, however, that restriction of horizontal eye movements impairs performance in addition and subtraction problems, suggesting a functional relevance of internal coupling (Masson and Pesenti, 2023): When rightward gaze shifts were restricted, the performance of addition with carrying decreased, while the prevention of leftward gaze shifts impaired subtraction with borrowing.Calculations without borrowing or carrying were not affected by the gaze obstruction, further implicating the importance of task demands on internal coupling.
In conclusion, evidence suggests that eye behavior during numerical cognition couples to movements along a mental number line, indicating a spatial positioning system at work.While the direction of these movements can vary based on factors like experimental design and task type, the consistent finding across studies is the presence of internal coupling.It is plausible that eye movements during mental arithmetic mirror the spatial locations we might use when visualizing computations, akin to performing arithmetic on paper.Nevertheless, there may be individual and cultural variations in what is visualized during mental arithmetic, adding complexity to the interpretation of the precise way in which eye behavior couples with these cognitive processes.

Imagery of object movement
Several studies have explored the role of oculomotor behavior in dynamic motion imagery.Dynamic imagery, as opposed to static, involves imagining an object's motion, with studies mainly investigating smooth pursuit eye movements (SPEM) or target's location prediction in the absence of a target.During normal perception, SPEM are produced instead of saccades when tracking a moving object (Lisberger et al., 1987).Some studies have attempted to elicit SPEM during imagery, such as Deckert (1964) using of a pendulum to elicit SPEM and then instructing participants to imagine its movement.Eye movements generally coupled to the internal representation of motion, with more SPEM like behavior in closed-eye imagery than open-eye imagery.Compared to the SPEM elicited by the visually-present pendulum, SPEM during closed-eye imagery showed distorted amplitude and large overshot (Lenox et al., 1970).
A more recent study reexamined smooth pursuit imagery by comparing closed-eye imagery, lucid dreaming, and perception conditions (LaBerge et al., 2018).Smooth pursuit was elicited by asking the participants to rotate (or imagine rotating) their thumb and follow the movement with their eyes.The study found similar saccades and pursuit ratio in lucid dreaming and perception, whereas the closed-eye imagery condition yielded more saccades and a decreased pursuit ratio.LaBerge et al. speculated that sufficient activation of the middle temporal visual area (the primary pursuit pathway) is needed to elicit real SPEM in the absence of a stimulus, which might only be achieved during lucid dreaming.Overall, the studies so far support that eye movements are indicative of imagined object movement, albeit real pursuit movements may still require a visual percept or perception-like imagination (lucid dreaming).
Although our eyes are typically not capable of producing real pursuit movements during visual imagery, they are still involved in, and perhaps even functional for, generating a clear mental motion representation.For example, an early study without eye tracking reported that when participants were first instructed to imagine a horse running and then to stop moving their eyes, they reported a disappearance or modification of the imagined movement, indicating that the internal representation of the horse slowed down or completely stopped (Ruggieri, 1999).Similarly, when participants had to mentally follow the horizontal direction of a disappearing moving target while maintaining central fixation, they reported this condition as more challenging compared to unrestricted viewing (Jonikaitis et al., 2009).Furthermore, the main findings of Jonikaitis et al. demonstrated a clear relationship between eye position and imagined movement based on reported imagined target position (see Fig. 2C).As anticipated, the imagined movement was characterized by a sequence of saccades along the target's trajectory instead of SPEM.However, the sequence of saccades following the imagined trajectory exhibited a speed that matched that of the disappeared target.Similar findings were obtained by de'Sperati (2003) who reported a consistent gaze velocity (i.e., speed of a sequence of saccades) during circular motion imagery comparable to that observed during SPEM.Another study investigated the role of covert attention during motion imagery (de' Sperati and Deubel, 2006).Participants were instructed to covertly follow the target's trajectory and make a saccade towards a flash, which appeared at various locations around the predicted target position.Saccades had shortest latencies when the flash was near the predicted target's position, indicating that participants covertly attended to the imagined trajectory of the target.Finally, participants demonstrated greater accuracy in estimating movement duration when they were instructed to imagine the movement rather than merely reason about it (Huber and Krist, 2004).Although those who displayed eye movements showed enhanced performance, the fact that restricting gaze did not diminish performance suggests that while eye movements may facilitate motion imagery, they are not obligatory for its effectiveness.
Taken together, research on motion imagery indicates that continuous smooth eye motion is not possible in the absence of visual input.Instead, sequences of saccades dominate, with their gaze velocity often resembling the imagined motion trajectory.This suggests that even in the absence of an overt visual stimulus, our oculomotor system is tightly linked with our internal representations of motion.Eye movements seem to play a facilitative role in motion imagery, which is enhanced when eyes are allowed to move in sync with the imagined motion (Jonikaitis et al., 2009;Ruggieri, 1999).However, the research also underscores the capacity for covert motion imagery (de' Sperati and Deubel, 2006;Huber and Krist, 2004) as even without eye movements, participants were able to accurately predict the trajectory of an imagined target.In sum, eye movements show robust coupling to imagined object movement, however, following the object in a saccadic rather than smooth pursuit fashion.

Imagery of body movements
Eye movements are known to exhibit a temporal and spatial coupling to motor execution, particularly during goal-directed movements (Helsen et al., 2000;Hollands et al., 2004;Sailer et al., 2000;Wilson et al., 2007).Despite being controlled by separate neural pathways (Hwang et al., 2014) and attention networks (Jonikaitis and Deubel, 2011), studies have demonstrated that motor imagery is also associated with changes in eye movements, indicating a potential link between these two systems (Gueugneau et al., 2008;Heremans et al., 2008;Lanata et al., 2020;Poiroux et al., 2015).Motor imagery can be categorized as visual or kinesthetic, with the former involving the visualization of observing the movement and the latter involving the mental sensation of performing the movement yourself.In one of the initial investigations into eye behavior during motor imagery, participants were instructed to mentally rotate their entire body to the left or right without relying on visual cues (Rodionov et al., 2004).The imagery task resulted in pronounced nystagmus activity on a horizontal level, similar to what occurs during actual body rotation, indicating a coupling of eye movements with the imagined environment.Heremans et al. (2008) explored visual motor imagery by instructing participants to either execute or mentally visualize themselves moving their hand between two targets (see Fig. 2D).The results revealed a clear coupling between eye behavior and hand movement during the imagery conditions, as evidenced by the number and amplitude of saccades, whereas this coupling was absent during rest.However, not all participants showed these task-related eye movements, highlighting individual differences and suggesting that eye movements may not be as functional for motor imagery as for motor execution studies (Jonikaitis and Deubel, 2011;Sailer et al., 2000).Nonetheless, building on their earlier work, Heremans et al. (2011) investigated the effect of eye movements during motor imagery training on later physical execution by dividing the participants into unrestricted viewing, fixed viewing, and no training groups.They showed that while eye movements did not alter the timing aspects of the imagined movements, they were instrumental for the spatial coordination as seen by the enhanced movement precision.Particularly, the unrestricted viewing condition resulted in higher accuracy and efficiency in target hits compared to both the control and fixed view conditions.Together, this shows that eye behavior is similar between visual motor imagery and execution and suggests that eye movements can contribute to the refinement of motor skills.
Regarding kinesthetic motor imagery, Gueugneau et al. ( 2008) investigated the role of eye movements in imagining arm movements compared to movement execution in either unrestricted or fixed gaze conditions.Movement durations were prolonged in the fixed gaze condition for both imagined and executed movements.Additionally, movement durations were highly correlated between the imagined and executed conditions.Hence, just like visual motor imagery, eye behavior in kinesthetic motor imagery is similar to movement execution and restraining eye movements seem to negatively affect motor planning.
McCormick et al. ( 2012) investigated number of fixations and fixation duration between movement observation and motor imagery, without specifying whether the imagery should be visual or kinesthetic.Their results showed similar fixation characteristics in both activities, though longer fixation durations were observed during observation, which might reflect higher task demands as the observed movement had to be memorized in order to be mentally repeated later.These findings were further extended by including movement execution as well as guided and unguided imagery (McCormick et al., 2013).The results indicated that the number of fixations increased only in the movement execution condition, while fixation duration did not differ significantly between conditions, lending further support to internal coupling during motor imagery.Poiroux et al. (2015) directly compared movement observation to visual and kinesthetic motor imagery in a task that required transferring blocks from one box to another.The study found that kinesthetic motor imagery was more time-consuming than visual motor imagery, suggesting higher cognitive demand for imagining the sensation of movement compared to visualizing it.A key finding was the difference in eye movement patterns between visual and kinesthetic motor imagery with the former resulted in more active eye movements possibly reflecting more dynamic cognitive processing.Similarly, Lanata et al. (2020) reported more irregular and complex patterns of eye movements in the kinesthetic compared to visual motor imagery.Furthermore, individuals with low imagery abilities exhibited higher fixation sequence re-enactments, a greater number of fixations in the same area, and a more irregular pattern of eye movements during high but not low task demands.Thus, it is suggested that kinesthetic imagery and low imagery abilities may contribute to more complex eye behavior, both potentially reflecting effects of higher cognitive demand.Lastly, a study exploring a brain-machine interface using EEG signals demonstrated that brain activity during kinesthetic motor imagery overlapped with that observed during actual movement, but they differed from the brain activity in visual motor imagery and movement observation (Yang et al., 2021).Overall, these findings underscore the distinct relationship between the type of motor imagery-kinesthetic versus visual-and eye behavior, which are further corroborated by distinct patterns of brain activity.
The presented evidence, suggests a tight relationship between eye movements and imagining body movements, mirroring the coordination observed during actual physical execution.Findings revealed internal coupling during both visual and kinesthetic motor imagery.Nonetheless, there are also differences in both eye behavior and brain activity between the two forms of motor imagery, which warrant further research.E, F).Note.A, B: A prototypical paradigm for the mental arithmetic paradigm.Subtraction (A) and addition (B) of given numbers are preformed mentally.Internal coupling is evidenced by eyes moving to the left during subtraction and right during addition.Example study: Salvaggio et al. (2022).C: A prototypical paradigm for the object movement paradigm.The bird on the top represents a moving target that disappears for part of its trajectory (dashed line).Internal coupling is evidenced by eyes moving along the imaginary trajectory while the target is absent, and closely matching the target's position upon its reappearance.Importantly, smooth pursuit appears only possible when the target is visible, while during imagery it is replaced by saccades, which continue to match the target's velocity.Eye behavior during imagery is represented as fixations (red dots) and saccades (blue lines).Example study: Jonikaitis et al. (2009).D: A prototypical paradigm for the body movement paradigm.During motor execution (left) the hand alternates between the right and left target (red circles), with the eyes spontaneously matching the direction of these movements.During motor imagery (right), one imagines moving the hand between the targets.Internal coupling is evidenced by the eye movements alternating between imagined targets with similar saccade duration and amplitude between execution and imagery conditions.Example study: Heremans et al. (2008).E, F: A prototypical paradigm for the brightness imagery paradigm.Internal coupling is evidenced by pupil constriction during the imagery of a bright stimulus like a sunny day (E) and pupil dilation during imagery of a dark stimulus like a dark forest (F).Example study: Laeng and Sulutvedt (2014).

Imagery of brightness
The pupil regulates the amount of light entering the eye through the pupillary light response (PLR), which constricts or dilates the pupil in response to increases in brightness or darkness, respectively.However, pupil size is not a mere reflexive adaptation to the lighting conditions, but increasingly recognized as a read-out of different mental activities.For example, pupil diameter (PD) is affected by cognitive workload, with increased workload resulting in increased PD, which is an established objective measure of workload demand (for reviews see Beatty and Lucero-Wagoner, 2000;van der Wel and van Steenbergen, 2018).Furthermore, pupil adapts to the brightness of the covertly attended stimulus even in the absence of (overt) eye movements (for reviews see Binda and Murray, 2015;Mathôt and van der Stigchel, 2015) and internal attention directed towards continuous task performance reduced the PLR to luminance changes of medium but not strong intensity (Korda et al., 2024).
Subjective luminance perception can also influence PD (Binda et al., 2013;Laeng and Endestad, 2012;Naber and Nakayama, 2013).For example, in an experiment with visual illusions of brightness PD adapted to the perceived brightness, that is, some of the stimuli were rated as brighter than the others even though their luminance remained constant, and these perceived differences were reflected in PD (Laeng and Endestad, 2012).
Furthermore, the influence of conscious awareness on pupil modulation was demonstrated using the continuous flash suppression (CFS) approach, a technique that renders stimuli requiring top-down visual processing invisible to awareness (Sperandio et al., 2018).In the absence of CFS, the pupil constricted during the presentation of sun pictures compared to phase-scrambled controls, as expected.However, in the presence of CFS, there were no differences in PD between the types of pictures.Moreover, Sperandio and colleagues compared trials without CFS to those with failed and successful awareness suppression, showing that the distribution of PD change in trials with failed awareness suppression corresponded to those without CFS, but differed from the successful trials.Thus, the difference in PD between sun and control pictures is only present when conscious awareness of the picture is intact.Overall, these studies provide evidence for the influence of top-down processes on pupil modulation, highlighting the complex interplay between cognitive factors, visual perception, and the pupillary response.
In addition to top-down influence of visual processing, the pupil also responds to purely internal representations in the absence of perceptual input.Laeng and Sulutvedt (2014) conducted a series of experiments using simple and complex shapes with distinct luminances (e.g., lowest: 0.9 cm/m 2 , highest: 78.5 cd/m 2 ) to investigate the pupillary response during visual imagery.Participants were instructed to imagine the previously seen stimulus, and it was found that the pupil reacted similarly during visual imagery as it did during perception.Furthermore, internal coupling was observed during the visual imagery of a complex scene (see Fig. 2E, F).Imagining a sunny sky led to a decrease in PD from baseline, while imagining a dark room increased it.In addition to pictures, the pupil has even been shown to react to the meaning of words (Mathôt et al., 2017).Hearing or reading words associated with brightness (e.g., "illuminated") resulted in a smaller PD compared to words associated with darkness (e.g., "dark"), indicating that the pupil also couples to internal representations during semantic processing.These findings suggest that the effect of internal coupling is not exclusive to working memory and short-term memory representations.
Furthermore, PLR during visual imagery depends on the imagery vividness and strength, as measured by the trial-by-trial vividness ratings and binocular rivalry, respectively (Kay et al., 2022).Specifically, PLR was larger in trials with higher vividness ratings, and greater differences in PD between dark and bright stimuli were observed with higher imagery strength.Interestingly, when the same experiment was performed by individuals reporting aphantasia, which refers to a condition where the individual is incapable of visual imagery, neither imagery of single nor multiple (four) triangles elicited a PLR.However, visual imagery of multiple triangles increased PD compared to a single triangle, which aligns with the higher workload demands and reflects active task performance (Kay et al., 2022).
Lastly, PD does not only respond to visual representations when actively trying to imagine a shape but also when accessing information from the visuospatial working memory (Blom et al., 2016;Hustá et al., 2019;Koevoet et al., 2024;Zokaei et al., 2019).Although initially, Blom and colleagues reported that only the encoding, but not the maintenance of the visuospatial stimulus affects PD, subsequent studies challenged these findings (Hustá et al., 2019;Zokaei et al., 2019).For instance, Husta and colleagues used a retro-cue paradigm in their VSWM task, eliminating the possibility of covert attention influencing PD during item selection, as the cued item had to be retrieved from memory.This demonstrated that PD adapts to the brightness of the object that is being maintained in VSWM.Zokaei and colleagues extended these findings by showing that the pupil reacts to the brightness of the maintained stimulus even when the brightness itself is unrelated to the task.
Internal coupling in pupillary responses is evident under varied brightness scenarios, from the adaptation to perceived brightness in visual illusions to the changes in pupil diameter during visual imagery of different luminance levels.Furthermore, cognitive processes such as attention and semantic interpretation of words related to brightness also influence pupillary responses, indicating that the pupil's reaction is not merely a reflex but a complex interaction of sensory perception and higher cognitive functions.These findings underscore the importance of top-down processing, such as the modulation of the pupillary light response by visual attention and the impact of conscious awareness on pupil size, in determining the extent and nature of internal coupling.

Summary of internal coupling evidence
The previous sections demonstrated that internal coupling-eye behavior coupled to features of internal cognition-has been consistently observed during such diverse cognitive activities as episodic memory recall, episodic simulation, visuospatial memory processes, numerical cognition, imagined object and body movements, and object brightness (see Table 1, Figs. 1 and 2).In these diverse tasks and paradigms, the imagery of internal representations was shown to elicit eye behavior akin to that during corresponding visual experiences.Hence, we can conclude that internal coupling appears a robust phenomenon across different forms of visual imagery.Importantly, several studies suggested that internal coupling may not just accompany but also facilitate internal cognition.Moreover, some findings also point to potential moderating factors such as individual differences in cognitive capability and specific task requirements.Together, these observations may shed light on the underlying mechanisms and potential functional role of internal coupling.The following sections thus explore the questions of why and how internal coupling occurs.

Why does internal coupling happen?
We see that eye behavior during imagination often corresponds to that during actual visual perception.But why would this happen?Is it not a waste of resources when eyes are directed to a location or adapted to luminance of stimuli that are imagined and thus do not require sensory processing?Importantly, internal coupling does not just happen but preventing it can impair cognitive performance.This raises the question whether internal coupling has a functional role in supporting visual imagery.To address this question, this section first revisits evidence on the relationship between internal coupling and task performance to evaluate how consistently the facilitative role of internal coupling has been observed so far and to explore factors that may moderate this effect.In the next step, we consider theoretical accounts and hypotheses that address why internal coupling takes place and examine how well they fare in explaining the accumulated evidence.

Does internal coupling have a functional role?
Table 2 revisits research from Section 2, which investigated the relationship between internal coupling and task performance.This analysis included 44 articles with a total of 66 distinct results (usually from separate experiments) that speak to a potential facilitative role of internal coupling for task performance.The studies employed three different approaches: comparing task performance between unrestricted gaze and restricted gaze conditions preventing internal coupling (i.e., coupling interference; 41 results), examining whether internal coupling relates to higher task performance as opposed to no or less accurate coupling behavior (i.e., spontaneous coupling; 17 results), or experimentally guiding eye behavior in congruent versus incongruent ways with internal coupling (i.e., guided coupling; 11 results).Hence, when a study demonstrated that preventing internal coupling impaired task performance or that coupled eye behavior increased performance, this was interpreted as indicative of a functional role for eye behavior in facilitating cognitive performance.Conversely, the absence of such relationships indicates that coupled eye behavior may be just an epiphenomenon of internal cognition.A count of the reviewed evidence indicates a predominance of supportive findings, with 39 results showing a beneficial effect of internal coupling on performance compared to 27 that do not (see Table 2).We should note that the specific ratio needs to be treated with caution as it resulted from a broad, yet nonsystematic review of the literature.As such, we also refrained from any further formal analyses.The general observation suggests there is substantial empirical support for the functional role of internal coupling in enhancing cognitive performance, but evidence still appears mixed.We thus can proceed to have a closer look at potential patterns within the reviewed data.
Research within the episodic simulation domain (n = 4) consistently demonstrates that internal coupling during the visual imagery of complex scenes enhances the vividness of imagination and increases the number of details retrieved.Similarly, in numerical cognition (n = 4), internal coupling along the mental number line was consistently related to faster or more accurate responses; specifically, directing the gaze downwards tends to facilitate subtraction tasks, but not addition tasks.
In other domains, however, studies found not such a reliable association between eye behavior and task performance, suggesting that the role of internal coupling is context-dependent and not uniformly straightforward.For example, research on object movement imagery (n = 3) generally indicate that internal coupling does not significantly impact cognitive performance.While one study initially found enhanced accuracy of movement prediction associated with internal coupling, a subsequent experiment revealed that similar improvements could potentially be attained through covert attention alone, as demonstrated by the fact that enforced gaze fixation did not decrease prediction accuracy (Jonikaitis et al., 2009).The findings from episodic (n = 21) and visuospatial memory (n = 18) studies are particularly mixed, a trend that might be attributed to the substantial volume of research in these domains.A closer examination of whether experimental designs or stimulus types moderate the effect, did not reveal a discernible pattern.Mixed findings were even observed within studies that interfered with gaze position or saccade planning (episodic memory: n = 12; visuospatial memory: n = 5) in terms of gaze fixation, abducted eye paradigm or selectively disrupting spatial/object working memory, respectively (see Table 2).It is further important to consider that observed performance in restricted gaze designs could, in part, be attributed to the increased total demands.People might need to engage additional cognitive resources to ensure or cope with gaze restriction that may eventually affect task performance.Conversely, the absence of observed impairments in some cases might be explained by the individuals' ability to employ effective compensation strategies, thereby maintaining performance levels despite the limitations imposed on internal coupling.
Nonetheless, one potential moderator could be task modality since the evidence is considerably stronger for spatial versus visual memory processes or for relations between the objects versus the characteristics of an object (n = 6 vs. n = 1).This suggests that internal coupling might reflect a spatial pointer, with saccades being directed to relevant locations within our internal representations (see Section 2).Another fairly consistent finding was that guiding internal coupling (episodic memory: n = 4; visuospatial memory: n = 4; numerical cognition: 2), i.e., guiding the gaze position towards the location where information was presented, during memory recall or a delay period in working memory tasks benefits memory performance.However, Scholz et al. (2018) investigated both gaze restriction and gaze to congruent or incongruent locations and found that covert attention shifts of attention could explain the positive effect on episodic memory performance in gaze congruent conditions.
Internal coupling was further found to be affected by individual differences in traits including autobiographical memory, visuospatial working memory capacity, and preferential imagery style.Armson et al. (2021) showed that fixation rate was predictive of memory performance only for people with high autobiographical memory ability.Another study found that internal coupling (more similar eye behavior between encoding and recall) was higher in people with higher visuospatial working memory capacity (Gurtner et al., 2021).People who preferentially used object imagery showed a smaller looking at nothing effect (Chiquet et al., 2022), suggesting that they might rely less on the spatial location and more on vividness and detail of internal visual representations; however, neither imagery style nor looking at nothing affected memory performance.Furthermore, two studies reported heterogeneity of internal coupling across trials and participants (B.Liu et al., 2022;Walcher et al., 2024), and assumed that individual differences in working memory capacity or imagery vividness might be at play.Such individual differences could serve as moderating factors in the complex relationship between eye behavior and behavioral performance.For example, a person's memory capacity might influence the level of details in internal representations, thereby affecting the utility of eye movements in memory tasks, and potentially making internal coupling more effective for people with higher memory capacities and thus more detailed representations (Armson et al., 2021;Gurtner et al., 2021).This would suggest that internal coupling is driven by high vividness of internal representations, thereby allowing the eyes to navigate through them as if they were physical entities, enhancing the recall process in memory tasks.On the other hand, one could also assume that internal coupling may serve as compensatory mechanism in individuals with less vivid imagery, by facilitating the construction or reinstatement of internal representations through eye movements (Chiquet et al., 2022).In sum, individual differences seem to moderate internal coupling, but the extent to which these differences affect the relationship between internal coupling and performance is less evident and represents an important avenue for future research.Studies on people with aphantasia could be especially relevant in this context, as a single case study with an aphantasia patient indeed suggested a functional role of visual imagery in working memory (Jacobs et al., 2018).Specifically, lower patient performance was observed only in working memory tasks that required high levels of visual precision.However, more recent studies with larger sample sizes have shown that participants with aphantasia do not show (large) deficits in visual working memory tasks (Keogh et al., 2021;Pounder et al., 2022).In terms of eye behavior, people with aphantasia do not show the expected PLR during the visual imagery tasks (Kay et al., 2022).To the best of our knowledge, no studies investigated gaze behavior during working memory or visual imagery in this population.Such findings could shed further light on the role of eye movements during internally directed cognition.

Internal coupling as a memory reconstruction mechanism
Memory of visual scenes reflects an internal representation of previously encoded features which have typically been perceived through a series of fixations and saccades.The scanpath theory assumes that eye behavior does not only play a crucial role during encoding but more generally for the (re)construction of these internal representations (Noton and Stark, 1971).In their seminal work, Noton and Stark demonstrated that eye movement patterns during the memorization of line drawings were repeated during subsequent recognition tests.Subsequent research, as reviewed in Section 2.1, has shown that such repetition of eye movement patterns occurs also in conditions where the original visual stimulus is absent from the screen during retrieval, leading individuals to gaze towards the locations where the stimuli were previously encountered.This observation is commonly referred to as looking at nothing (Ferreira et al., 2008).Ferrerira and colleagues posit that humans create detailed internal memory representations of the external world, which include visual, spatial, linguistic, and conceptual information that is bound together into a coherent memory trace.When part of this integrated representation is reactivated, such as through language, it facilitates the retrieval of the other components of the memory.Hence, when eye movements return to the locations associated with those memories this should improve memory performance due to matching spatial information.Eye behavior during looking at nothing thus can be viewed as "a cause and a consequence of memory retrieval" (Ferreira et al., 2008, p. 407), and consequently internal coupling could be attributed to highly integrated memory representations.
A related notion, known as gaze reinstatement theory, was put forward by Wynn et al. (2019).It proposes that eye movements support memory recall by actively reinstating the spatiotemporal encoding context, thereby aiding the (re)construction of internal representations.During such constructive processes, gaze reinstatement is thought to support relational memory, and thereby serves as a scaffold for the integration of spatial and temporal details (Ryan et al., 2020;Wynn et al., 2019).Hence, while scanpath theory and the looking at nothing research focus on the retrieval of specific visual stimuli, gaze reinstatement would also manifest in the absence of specific visual stimuli, such as in the context of imagining more complex, dynamic scenes.Moreover, it can also function as a recognition mechanism when external stimuli are present, guiding the eyes in response to visible cues rather than solely internal ones.Furthermore, contrary to the notion that any feature of an event can trigger the retrieval of internal representations (Ferreira et al., 2008), the gaze reinstatement theory posits more explicitly that eye movements aid recalling spatiotemporal features, which facilitates the recall of other event features like linguistic and conceptual elements (Wynn et al., 2019).Nonetheless, both theories posit a functional role of eye behavior in memory recall.
The role of gaze reinstatement in reactivating and navigating through spatiotemporal dimensions of past experiences finds a compelling parallel in the domain of episodic simulation.Episodic simulation plays a role in several cognitive processes including future thinking, counterfactual thinking, decision making, and creativity (Abraham, 2016;Benedek et al., 2023;Schacter and Addis, 2020).Both episodic memory and simulation depend on reconstructive processes, where memory elements are recombined to construct previously experienced events or simulate hypothetical scenarios (Addis, 2018;Hassabis et al., 2007;Schacter and Addis, 2007;2020).Gaze reinstatement has been suggested to also play a role in episodic simulation, aiding the construction of hypothetical events and scenes (Conti and Irish, 2021;Ryan et al., 2020).Constructive demands of imagining future events are associated with spatial processing, where spatial context acts as a scaffold, providing the necessary framework upon which the elements of an event are organized (Wiebels et al., 2020).In this context, internal coupling might facilitate episodic simulation by supporting the positioning of each part within the imagined scene.Conversely, restricting internal coupling can interfere with the effective use of spatial imagery (D.G. Pearson et al., 2014;Vito et al., 2014Vito et al., , 2015)).This interference suggests an integral role of internal coupling in scaffolding the mental construction of future events, highlighting its importance in the effective utilization of spatial context and the overall process of episodic simulation.Importantly, as episodic simulation supports imagination of novel and even counterfactual events, this opens the possibility of testing the role of memory in internal coupling.For instance, when imagining raindrops falling upwards, a close integration of eye movements during reconstruction of imagined scenes (as assumed by gaze reinstatement) would predict that eye movements align with this counterfactual direction, thereby coupling to the simulated scenario although this opposes any previously encoded memory.
Wynn et al. ( 2019) further specified the role of gaze reinstatement for task performance by assuming that the oculomotor system may get recruited to support memory only when the task demands exceed memory capacity.So far, only a few studies have explored the effect of varying task demands on internal coupling.Johansson et al. (2022) investigated scanpaths similarity between encoding and recall of scenes and object arrangements, where the latter impose higher demands on relational memory, and found that greater scanpath similarity predicted better memory quality regardless of the image type.These results did not support the specific assumption by Wynn et al., but it also seems possible that differences in task demands between scenes and object arrangements were not substantial enough, or that the recruitment threshold was already met even by the less demanding task.Alternatively, these observations could imply a difference between gaze reinstatement and internal coupling, as effects of task demands on gaze reinstatement were found in studies with visual input available throughout the experiment (Dewhurst et al., 2018;Wynn et al., 2018;Wynn et al., 2020).This is particularly relevant when considering tasks like visual search or old/new recognition tasks, where the direct visual presence of stimuli during both encoding and recall phases does not require internal coupling as opposed to tasks dependent only on memory recall.On the other hand, a simultaneous investigation of episodic memory and simulation showed that only simulation performance was affected by concurrent SPEM (Vito et al., 2014), suggesting that episodic simulation imposes higher demands on visual imagery and relation memory (Wiebels et al., 2020).This would suggest that episodic simulation is more demanding than episodic memory, which could explain the more coherent trend in functional relevance of internal coupling in this domain (i.e., all four studies demonstrated this trend, see Table 2).Nonetheless, future studies are needed to provide a clear support for the role of task demands in internal coupling.
Overall, theoretical accounts like the gaze reinstatement hypothesis provide a compelling framework for understanding the role of eye behavior in memory processes.Much evidence on internal coupling could be understood in terms of the crucial role of eye behavior in facilitating memory (re)construction processes in the domains of episodic recall and simulation.Yet, internal coupling has also been observed during processing of more abstract information, which points towards its relationship to internal shifts of attention.

Internal coupling as a marker of internal attention shifts
Emerging evidence suggests a fundamental similarity in how information is chosen from external sensory input and internal representations (for recent reviews see van Ede and Nobre, 2023;Verschooren and Egner, 2023).Just as focusing attention on a stimulus in the visual environment enhances its processing, the attentional focus on a specific location in working memory also enhances the processing of that specific working memory content (Theeuwes et al., 2009;van Ede and Nobre, 2023).Moreover, visual attention, working memory, and the control of eye behavior are known to be highly integrated (Jonikaitis and Moore, 2019;Theeuwes et al., 2009): Working memory interacts with visual search, and external and internal attention share cognitive resources and a neurocognitive basis (Dixon et al., 2014).This suggests that internal coupling might be a mechanism through which attention facilitates the retrieval of information from working memory by shifting the focus of attention between internal representations.Hence, internal coupling could be viewed as marker of internal attention shifts.
The concept of shared attentional resources describes how people distribute the limited attention capacity across both external and internal information (Awh and Jonides, 2001;Gazzaley and Nobre, 2012;Gresch et al., 2024;Kiyonaga and Egner, 2013;Verschooren et al., 2019).Given the interplay between externally and internally directed attention, it seems possible that the external world could even be used for off-loading internal load as proposed by the embodied account of visual working memory (van der Stigchel, 2020).This perspective implies that eye movements may be biased to locations in the external world that correspond to the spatial locations of internal representations, effectively bridging internal cognitive processes with external sensory resources, hence preserving shared attentional resources.
Various studies exploring internal attention shifts in the domain of visuospatial memory have consistently observed a gaze bias towards the location of an internal representation; however, they failed to demonstrate a clear functional impact on memory performance (B.Liu et al., 2022;van Ede et al., 2019;Vries and van Ede, 2024;Walcher et al., 2024).This suggests that while internal coupling reflects the underlying cognitive processes, it may not directly enhance the effectiveness of internal selection.Given the close integration among attention, working memory, and oculomotor behavior, it is plausible that shifts in internal attention occasionally surpass a saccade threshold, inadvertently triggering a saccade (e.g., B. Liu et al., 2022).On the other hand, studies interfering with internal coupling in working memory showed a detrimental effect on spatial working memory when eye movements were either constrained or diverted towards a secondary, unrelated task (Ball et al., 2013;McAteer et al., 2023;D. G. Pearson et al., 2014;Vito et al., 2014).Interestingly, unlike spatial working memory, visual working memory was not affected by gaze interference, suggesting that internal coupling may be generally more relevant to spatial working memory.This would also explain why retro cue paradigms, which primarily require to report a target's orientation, often yielded no internal coupling effects on task performance, whereas in the context of numerical cognition, where numbers are often conceptualized and manipulated within spatial frameworks like the mental number line, internal coupling seems to play a supportive role in task execution (Hartmann et al., 2016;Salvaggio et al., 2022).
Furthermore, the dynamics between saccades and fixations are crucial for transforming spatial information into spatiotemporal flow and hence visual processing (Rucci et al., 2018).Specifically, saccades enhance low spatial frequencies in visual input immediately following their occurrence, which helps in encoding the general structure of the scene, while during fixations, ocular drift enhances high spatial frequencies, which is essential for detailed spatial analysis.In other words, saccades bring new information into focus, while fixations allow for detailed processing of that information.Extending this concept to internal coupling, we can hypothesize that similar functional connections exist between "internal saccades" (shifts of attention within internal representations) and "internal fixations" (focused attention on specific details of internal representations), and that internal coupling involves temporal sequences of attention shifts and focused processing that support the reconstruction and mental exploration of complex, dynamic scenes.
Interestingly, research on the pupillary light response found it to be modulated by covert attention and visuospatial working memory (Binda and Murray, 2015;Hustá et al., 2019;Mathôt and van der Stigchel, 2015;Zokaei et al., 2019), again highlighting the profound connection between attention and eye behavior.The pupil diameter was shown to reflect phases of encoding, maintenance, and prioritization within the visuospatial working memory (Koevoet et al., 2023).This suggests that even the pupillary response is indicative of attentional shifts, whether directed towards external stimuli or internal representations.In sum, we can conclude that internal coupling appears functionally engaged in the processes of spatial working memory, yet it may be less relevant or even reflect an epiphenomenon in the realm of visual working memory.
It is worth considering the possibility that the two proposed roles of internal coupling-facilitating memory reconstruction and indicating shifts in internal attention-are interrelated.Specifically, internal coupling may facilitate memory reconstruction through its role in spatial attention control.When an individual recalls a memory, the act of shifting internal attention to the spatial locations of internal representations in working memory or activated long-term memory could aid in reconstructing the memory itself.This suggests that, regardless of the task, internal coupling might reflect internal attention shifts that help to navigate and retrieve the spatial components of a memory trace.Therefore, the effectiveness of memory reconstruction could be partially due to the underlying mechanism of internal attention shifts to these spatial locations.This integrated perspective highlights the shared functionality of internal coupling in both memory reconstruction and attention control, suggesting a more unified mechanism underlying these cognitive processes.
This section examined why internal coupling is happening, considering empirical evidence and theoretical accounts that may explain the conditionality of its occurrence and its potential functionality.The accumulation of evidence points towards a complex picture, where internal coupling is consistently observed across a range of cognitive tasks but does not always correlate with enhanced task performance.Its functional relevance appears to be context-dependent, and more evident in tasks involving spatial working memory and relational episodic memory, whereas its functional role in visual working memory has been less clearly established and may even reflect an epiphenomenon.Theoretical accounts suggest that internal coupling may serve at least two different potentially complemental purposes: Aiding memory reconstruction and acting as a marker of internal attention shifts.In sum, it seems that internal coupling is functionally relevant in certain forms of visual imagery, whereas in other contexts it could also be viewed as an epiphenomenon attributed to the shared neural underpinnings of external and internal cognition, a topic we will explore further in the following section.

How does internal coupling happen?
Having discussed the 'why' behind internal coupling, we now turn to the 'how'-the underlying neurobiological mechanisms that facilitate this phenomenon.Research suggests that when we 'see' with our mind's eye, we activate many of the same areas involved in sensory visual processing, which may have important implications for concurrent eye behavior.So far, only a handful of investigations have employed neuroimaging methods to directly explore internal coupling, most of which were studies on episodic and visuospatial memory (Sections 4.3 and 4.4,respectively).However, neuroimaging outside of internal coupling domain has revealed substantial overlaps between internally and externally directed cognition (Sections 4.1 and 4.2), enabling us to draw inferences or hypothesize about internal coupling based on this evidence.Hence, this section delves into the neuroscientific findings on the overlap between visual perception and visual imagery, and explores how the brain's visual, memory, and attentional systems are thought to interact during internal coupling.

Neurobiology of visual imagery
Visual imagery activates a wide network in the brain, including frontal, (medial) temporal, parietal, and occipital regions, highlighting the overlap between the areas involved in creating mental images and those used for processing sensory information from our environment (for reviews see Andrews-Hanna and Grilli, 2021;Dijkstra et al., 2019;J. Pearson and Keogh, 2019).This suggests that the process of imagining something and actually seeing it might involve similar brain mechanisms, but with a key difference: the direction in which information flows through the brain networks.The idea of a reverse visual hierarchy proposes that visual imagery might work by pulling information from memory storage in a top-down manner, which is (almost) the opposite of how we process new visual information from our surroundings in a bottom-up fashion (Dentico et al., 2014;Dijkstra et al., 2019;J. Pearson, 2019).Research has shown that during perception there is an increased bottom-up and top-down coupling between fronto-parietal and visual areas, while during visual imagery only top-down coupling was observed (Dijkstra et al., 2017).The most prominent top-down coupling was between inferior frontal gyrus and early visual areas, a connection that is known to play a role in attention and visual working memory.This coupling was stronger during visual imagery compared to perception and positively associated with imagery vividness, leading to the conclusion that it compensates for the lack of perceptual input by reinforcing visual representations.Overall, these findings shed light on the neurobiological mechanisms of visual imagery, which enable the brain to vividly reconstruct and manipulate internal representations in the absence of external sensory input.
Representations of perceived as well as imagined objects and scenes were found in alpha frequency band, suggesting that alpha oscillations are involved in top-down activation of visual representations (Stecher and Kaiser, 2023;Xie et al., 2020).Similarly, higher alpha power has been consistently observed during creative thinking which was interpreted as top-down controlled, internally directed attention to support complex, vivid imagination (for reviews, see Benedek, 2018;Fink and Benedek, 2014).Notably, creative thinking also involves eye behavior representing an internal focus of attention (Salvi and Bowden, 2016;Salvi et al., 2015;Walcher et al., 2017).A systematic relationship between alpha oscillations and gaze behavior also emerged as an important mechanism in memory encoding, which was predictive of later recall (Nikolaev et al., 2023;Popov and Staudigl, 2023).Jointly, these findings implicate alpha oscillations in the orchestration of internal coupling, particularly through their influence on the networks governing visual perception, attention regulation, and memory processing (Jensen et al., 2012;Klimesch, 2012).That is, alpha oscillations could play a key role in synchronizing the activity of these neural circuits during the engagement with internal representations, potentially aligning eye movements with the dynamics of visual imagery and memory recall.Such synchronization might enhance the coherence and quality of the visual imagery or memory being reconstructed.Unfortunately, studies investigating internal coupling through EEG are scarce, underscoring a critical gap in our understanding that future research should address.Internal coupling fits well with the idea of top-down processing during visual imagery.Just as top-down signals guide the interpretation and focus within visual scenes, eye movements during visual imagery might reflect a top-down influence on how internal representations are navigated and explored.This suggests a more dynamic and interactive model of visual imagery, where internal visualizations are not just passively 'viewed' but actively examined and manipulated through eye movements.
The idea of full reverse visual hierarchy in visual imagery has been challenged by meta-analyses showing no consistent activation of early visual regions (Spagna et al., 2021;Spagna et al., 2023;Winlove et al., 2018).Spagna et al. (2023) suggested that visual imagery may rely on specific network connections rather than a simple reverse of vision, which would allow for both top-down and bottom-up processes to work together.They propose a fusiform imagery node located in the left fusiform gyrus as a central hub for visual imagery as it connects high-level visual areas with memory related regions in the temporal lobe.However, Dijkstra (2024a), b) suggests that the absence of the early visual regions' activity could be due to the use of specific fMRI analysis techniques, or task conditions where low-level visual details are not imagined, and thus the early visual regions are not recruited.For example, when using more advanced decoding techniques based on machine learning algorithms, the brain activity patterns during imagery were shown to be similar to those during perception (Dijkstra, 2024b).Furthermore, when the task instructions do not specify the size or location of the to-be-imagined stimulus, differences in these aspects can lead to different patterns of activity in the early visual regions.Hence, future studies should focus on the specific conditions of visual imagery that lead to activation of early visual regions (Dijkstra, 2024a, b).
A meta-analysis of 40 visual imagery studies revealed that regions controlling eye movements, i.e., the frontal, supplementary, and cingulate eye fields, were consistently activated during visual imagery (Winlove et al., 2018).These findings highlight the neural overlap between visual imagery, attention, and eye movements, and point towards a potential neural representation of internal coupling.It should be noted, however, that the majority of these studies suppressed internal coupling by limiting the gaze position to a central fixation.To address this gap in the literature, future research should explore visual imagery and perception without such constraints, potentially revealing more about the natural dynamics of internal coupling.Insights from retinotopic connectivity studies further illustrate how allowing free eye movements could deepen our understanding of these processes, a topic explored in greater detail in the following subchapter.

Retinotopic connectivity and internal representations
Modelling retinotopic connectivity has shed further light on how mental images are represented in the human brain.Retinotopic coding refers to the way in which visual information is represented in the brain in a manner that mirrors the spatial layout of the retina.This organizational principle means that the brain's representation of the visual field maintains the spatial relationships found in the real world, allowing for a coherent perception of visual scenes.Retinotopic coding was classically thought to be reserved for visual areas, while higher cortical regions were presumed to use amodal coding.However, recent findings suggest that even the default mode network and hippocampus exhibit retinotopic coding, suggesting that the brain uses a consistent visuospatial representation of external and internal world (Knapen, 2021;Steel et al., 2024).This suggests that brain areas related to memory, imagination, and internal thought processes possess an explicit representation of the visual world, i.e., a reversed retinotopic coding system to structure their functional connections with visual regions (Steel et al., 2024).Furthermore, it shows that the brain uses shared mechanisms for processing spatial information, whether it is perceived directly from the environment or recalled from memory.The presence of retinotopic coding in these higher-order regions supports the idea that our internal cognitive processes are grounded in the same spatial coordinates used during actual perception.Steel et al. (2024) further observed that the strength and precision of these retinotopic maps in the hippocampus were predictive of memory performance, indicating a functional role for this spatial coding in memory recall and visualization tasks.A common organization of how internal and external worlds are represented in the brain could explain the neural architecture behind internal coupling.This organization principle would facilitate the navigation among internal representations, where eye movements could act as spatial pointers and help to navigate through mental images or scenes in a manner akin to how they guide us through physical environments.Speculatively, the brain's use of a consistent visuospatial representation could also suggest that internal coupling is not just reactive or epiphenomenal but predictive.For example, if someone is planning a route through a familiar environment, they might predict the visual and spatial aspects of the route, adjusting the mental scene accordingly.Memory-related areas could then interact with the visual system to inform it of expected changes in the visual field, thereby guiding eye movements in a way that aligns with these expectations.Eye movements, under this assumption, would then not just be random or reactive but would be directed in a way that anticipates changes in the imagined or remembered visual scene.This predictive processing might help in smoother navigation through mental images, enhancing cognitive functions such as planning, memory recall, and even creative visualization.
One limitation of studies using retinotopic modelling is that they usually rely on gaze fixation, which does not fully reflect natural behavior.There is extensive evidence from studies investigating memory encoding and recall that gaze fixation decreases memory performance, imagery vividness, as well as medial temporal lobe (MTL) connectivity (Bone et al., 2019;Damiano and Walther, 2019;Gautier et al., 2022;Johansson et al., 2012;Johansson and Johansson, 2014;Ladyka-Wojcik et al., 2022;Laeng and Sulutvedt, 2014;Laeng and Teodorescu, 2002;Lenoble et al., 2019;Liu et al., 2020;Scholz et al., 2016;Vito et al., 2014).Hence, experiments allowing free eye movements during perception and visual imagery could offer a more comprehensive understanding of neurocognitive mechanisms involved in internal coupling.In this context, the work by Fabius et al. (2022) is particularly relevant as they investigated how the strength of a neuron's response to the same visual input can vary with different eye positions.The findings suggest that the visual system integrates eye position information early on, aiding in the accurate localization of objects regardless of eye movements.Hence, internal coupling could be speculated to involve activation of the retinotopic maps, allowing the visual system to simulate the experience of viewing actual scenes, thereby enhancing the vividness and spatial coherence of internal representations.Although retinotopic connectivity and mapping studies generally require fixation maintenance to provide a steady image on the retina, recent studies described above have introduced methodological approaches like retinotopic connectivity analysis that are robust with regard to eye movements (e.g., Fabius et al., 2022;Knapen, 2021).Future investigations of retinotopic mapping and connectivity during visual imagery could elucidate whether internal coupling acts as an internal attention mechanism to adjust the 'viewpoint' or focus within the mental image.

Neurobiology of eye movements and episodic memory
The hippocampus and MTL are implicated not only in visual processing but also in eye behavior, a linkage that has been proposed to serve a functional role in facilitating memory encoding (Meister and Buffalo, 2016).Interestingly, signals of eye movements made in the dark have been found in the MTL, suggesting that the role of eye movements in memory processes goes beyond bare visual input (Sobotka et al., 2002).Saccade related activity was also observed in the hippocampal cells, further establishing the importance of eye movements for hippocampal processing (Vericel et al., 2023).Furthermore, saccades were shown to facilitate a general increase in connectivity between visuo-oculomotor regions and the hippocampus as seen by the increased excitatory activity from the visual cortex and frontal eye fields (FEF) towards hippocampus, independent of memory outcomes (Wagner et al., 2022).Hence, the authors suggest that saccade-related activity seems to drive the network dynamics necessary for encoding visual information into memory.It is thus possible that eye movements are affecting connectivity patterns during internal coupling in a similar way.
Furthermore fixations, the end points of eye movements, are crucial for memory processes.Memory performance depends on the number of fixations during encoding and subsequently guides viewing behavior (Meister and Buffalo, 2016).Among other factors, viewing behavior can be guided by a non-retinal neural spatial map, also known as an allocentric map, which is referenced to environmental features rather than the body.Unlike the egocentric frame of reference used by visual cortex neurons that reflect stimuli relative to the eye position, the allocentric map allows for a more stable spatial representation, unaffected by eye movements.It is therefore likely that these memory-based spatial maps are relevant for guiding internal coupling as they are for guiding viewing behavior.
These processes are mediated by regions in the medial temporal lobe, including the entorhinal cortex and hippocampus.Entorhinal neurons of macaque monkeys show selectivity for both gaze and head position, i.e., some of them align to the allocentric and others to the egocentric frame of reference, respectively (Meister and Buffalo, 2018).Hence, entorhinal cortex integrates information from both allocentric and egocentric reference frames to provide a coherent representation of spatial information.Furthermore, the MTL neurons are engaged in the representation of a view-centered background image, while hippocampal neurons respond to observed as well as anticipated landmarks (Chen and Naya, 2020;Vericel et al., 2023;Wirth et al., 2017).Wirth et al. (2017) suggested that the observed hippocampal activity to the anticipated landmark hints at the projection of memory-based spatial maps onto physical space.This projection mechanism could be key in internal coupling, where eye movements are directed towards anticipated locations within a mental representation, aiding in the retrieval and reconstruction of memories.Additionally, it reinforces the idea that internal coupling is predictive, preparing our visual system for expected changes.Together these findings support the looking at nothing effect (Ferreira et al., 2008), suggesting that eye movements are actively encoded in a cohesive memory trace, which is reactivated during memory retrieval.Specifically, the stable spatial representations provided by allocentric maps could serve as spatial reference points for internal coupling.When individuals recall visual scenes or navigate through mental representations, eye movements are triggered due to the memory reactivation and likely guided by memory-based spatial maps.
The literature on gaze reinstatement has provided a body of research underscoring the pivotal role of brain regions traditionally associated with memory and spatial navigation, such as the MTL, hippocampus, parahippocampal place area, and fusiform gyrus, in modulating eye movement patterns during episodic memory and simulation (Bone et al., 2019;Ladyka-Wojcik et al., 2022;Liu et al., 2020;Wynn, Liu, and Ryan, 2022).Importantly, allowing eye movements during episodic simulation was associated with enhanced top-down connectivity from the MTL to oculomotor control regions, and simultaneous suppression of bottom-up visual inputs compared to gaze fixation (Ladyka-Wojcik et al., 2022).This reciprocal modulation of neural pathways suggests a mechanism by which the brain prioritizes internal representations over external visual inputs during the construction of visual imagery (cf.Verschooren and Egner, 2023), potentially mediated by the hippocampus's regulatory influence on visual cortical activity.Based on the extant literature on the involvement of hippocampus in visual and ocular dynamics in episodic simulation, Conti and Irish (2021) suggested that the hippocampus forms transient process specific alliances (PSAs) with memory, ocular, and visual systems.PSA refers to a task dependent temporary organization of cortical activity, when groups of brain regions come together temporarily to support a particular cognitive function and disband once the task is completed (Cabeza et al., 2018).According to this framework, the hippocampus may form different PSAs with either the core memory network or the visuo-oculomotor network to assist in different aspects of episodic simulation, influenced by the novelty of the event and the demands of the task (Conti and Irish, 2021).For example, coupling of the hippocampus with the oculomotor system could facilitate the retrieval of sensory-perceptual details from episodic memory for future thinking, whereas a PSA involving the hippocampus and the ventromedial prefrontal cortex (vmPFC) could be engaged for constructing future events based on schemas.Future research should continue examining the neuro-ocular coupling in internal coupling to draw further conclusions on the role of eye behavior during internal cognition.

Neurobiology of eye movements and internal attention shifts
Early studies focusing on retro-cue methodologies laid the foundational understanding of how humans orient attention toward content within working memory.Contrasting the neural activations instigated by pro-cues and retro-cues revealed a significant neural overlap in the mechanisms that govern both external and internal attention orientation (Griffin and Nobre, 2003;Nobre et al., 2004).It was discovered that directing attention to specific locations in working memory representations activates the same dorsal frontal-parietal network responsible for external spatial attention (Nobre et al., 2004).However, there were also direction-specific effects, with external attention additionally activating the right inferior parietal cortex, while internal attention additionally activated regions in the PFC.Importantly, both internal and external attention activated the FEF suggesting that the oculomotor system is also involved in coordination of the internal focus.Further research on rhesus monkeys demonstrated that the neural representations of internal and external attention are overlapping within the lateral PFC (Panichello and Buschman, 2021).In this study, a classifier was trained on neural activity from the task eliciting internal attention, i.e., the retro-cue task, and tested for its ability to decode external attention from the external attention task, i.e., the pro-cue task, and vice versa.This demonstrated that within the lateral PFC, neural representations for internal and external attention significantly overlap, allowing the classifier trained on one type of task to effectively predict the other.However, this was not the case in the FEF, parietal cortex or intermediate visual area V4.Interestingly, in the FEF, the ability of the classifier to generalize between internal and external attention was delayed-decoding became reliable only about 400 ms after cue presentation, in contrast to around 200 ms in the lateral PFC.This suggests a temporal differentiation in how the FEF integrates and processes attentional information.The delayed generalization in the FEF may indicate that while it potentially plays a role in mediating both external and internal attentional shifts, it might do so to a different extent.Consequently, the FEF's participation in internal and external attention might be more closely tied to the implementation phase of attentional shifts, preparing the system for potential eye movements, even when such movements are not overtly executed.Furthermore, both eye movements (overt) and covert shifts of attention activate a network of brain regions spanning the parietal, frontal, and temporal cortex, notably including the FEF and SEF (Corbetta et al., 1998).This underscores the FEF's function in facilitating shifts in attention, even in the absence of eye movements.This convergence of findings was further validated by a finding that a classifier trained on fMRI data predicted the selection of both external and internal location, based on the pro-and retro-cue tasks, respectively, confirming the overlapping neural representations of external and internal attention (Zhou et al., 2022).However, the control analysis on the eye tracking data did not reach significance, suggesting that ocular behavior patterns between external and internal attention might be more distinct.Furthermore, in a study combining EEG and eye tracking it was found that internal attention elicits alpha activity modulation towards the direction of attended location, however, only some of the trials involved a corresponding microsaccade (B.Liu et al., 2022).The authors assumed that, while the microsaccades sometimes align with the direction and timing of EEG alpha shifts, this might only occur when a certain threshold for the eye movement is crossed, reflecting a tangible output of the oculomotor system's engagement in attentional processes.Microsaccades and EEG alpha modulation might thus represent distinct facets of covert attention mechanisms potentially driven by similar oculomotor circuits.Future research should investigate under which conditions the threshold for outputting a microsaccade is passed.Several studies reviewed in Section 2.3 reported coupled saccades to working memory representation, however, studies combining such approaches with fMRI or EEG are largely missing.
Overall, internal coupling seems to depend on regions such as the frontal, occipital, and particularly the temporal lobes, which are integral to creating mental images.Moreover, the involvement of the FEF in directing both external and internal attention highlights a crucial intersection where eye movements are not only reactive but actively support the cognitive process of navigating internal representations.The FEF's dual role in managing attention shifts further complements the findings on eye movements in episodic memory, where saccades are linked with enhanced hippocampal connectivity, reinforcing the memory's visual and spatial details.Collectively, these findings could point to a coherent neural and cognitive framework where internal representations are dynamically constructed, navigated, and manipulated, leveraging common brain regions and mechanisms that bridge oculomotor behavior, visual processing, memory recall, and attentional control.Future studies should employ neuroimaging techniques coupled with eye-tracking, allowing unrestricted eye movements to elucidate the neural mechanisms of internal coupling.

Concluding remarks and future directions
This review documents that internal coupling has been reliably observed during a wide range of visual imagery-related activities, from episodic memory reconstruction, over numerical cognition, to spatial navigation and beyond.It becomes clear that eye behavior during these forms of internally directed cognition does not just exhibit idle activity due to perceptual decoupling from external stimulation but instead represents a systematic coupling to the features of ongoing internal events.A considerable amount of findings further suggests that internal coupling may not only cooccur with internal activities but actually facilitate cognitive performances, particularly in tasks relying on spatial working memory and numerical cognition.Yet, the functional significance of internal coupling remains elusive in cognitive domains such as visual working memory tasks and appears further moderated by individual differences in memory capacity and imagery vividness.Available theoretical accounts emphasize the role of internal coupling for the reconstruction of episodic memories such as by reinstating gaze behavior associated with memory encoding and as a spatial marker reflecting internal attention shifts.These models assume that direction of attention in external and internal worlds are supported by similar neurocognitive mechanisms, which is substantiated by the findings highlighting shared neural substrates for external and internal attention and their overlaps with eye behavior control.
Looking ahead, several key areas warrant further investigation to deepen our understanding of internal coupling and its implications for cognitive science.Future studies should target unique populations, such as individuals with aphantasia, as this research could provide invaluable insights into the neural and cognitive underpinnings of internal coupling by contrasting typical and atypical cognitive profiles.Furthermore, there is a need for more neuroscientific research to delineate the specific brain circuits involved in internal coupling.This includes identifying the neural correlates of top-down and bottom-up processes that facilitate visual imagery and understanding how these processes interact during different cognitive tasks.Specifically, we need to better understand the conditionality of when internal coupling facilitates cognitive performance.This requires larger-scale studies to enable powerful tests of how individual differences in specific cognitive abilities and cognitive strategies moderate the benefits of internal coupling across different contexts.Such work could have important implications for facilitating learning and imagination, especially in a world that offers constant highly salient, sensory stimulation that may easily interfere with ongoing internal activities.Lastly, further research should explore how the brain integrates information across different sensory modalities as internal coupling might not be limited to purely visual tasks but could extend to how visual spatial information is integrated with other senses and types of information (e.g., auditory scenes, tactile environments) in the construction of rich, multimodal internal representations.In conclusion, internal coupling represents a fascinating aspect of human cognition, representing an oculomotor mechanism through which the brain orchestrates its internal narrative and navigates its mnemonic landscape.

Declaration of Generative AI and AI-assisted technologies in the writing process
During the preparation of this work the author(s) used ChatGPT 4o in order to improve language and readability.After using this tool, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.

Declaration of Competing Interest
None

Fig. 1 .
Fig. 1.Prototypical Research Paradigms for Imagery Associated with Episodic Memory(A, B), Episodic Simulation (C, D), and Visuospatial Memory (E, F, G).Note.A, B: A prototypical paradigm for the looking at nothing paradigm.During memory encoding, an image is present on the screen (A), while during memory recall, the screen is blank (B).Eye behavior is represented as scanpaths consisting of fixations (red dots) and saccades (blue lines).Internal coupling is evidenced by the similar eye behavior observed during encoding (A) and recall (B).Example study:Johansson et al. (2012).C, D: A prototypical paradigm for the episodic simulation paradigm.Imagination of a hypothetical (future) event involves gradual build-up to an increasingly detailed scene.Eye behavior is represented as scanpaths consisting of fixations (red dots) and saccades (blue lines).Internal coupling would be evidenced by predictable eye behavior with more fixations in more detailed scenes.Example study:Wynn et al. (2022).E, F, G: A prototypical paradigm for the retro-cue paradigm.During memory encoding, two differently colored and oriented bars are presented on the screen (E).A retro-cue is then presented as a change in the color of the fixation cross, which now matches one of the previously shown bars, i.e., the right, yellow bar (F).Internal coupling is evidenced by eyes moving in the direction where the corresponding bar was presented (here, to the right).The orientation of the corresponding bar is then reported on the response screen (G).Example study:van Ede et al. (2019).

Fig. 2 .
Fig. 2. Prototypical Research Paradigms for Imagery Associated with Numerical Cognition(A, B), Object Movement (C), Body Movement (D), and Brightness (E, F).Note.A, B: A prototypical paradigm for the mental arithmetic paradigm.Subtraction (A) and addition (B) of given numbers are preformed mentally.Internal coupling is evidenced by eyes moving to the left during subtraction and right during addition.Example study:Salvaggio et al. (2022).C: A prototypical paradigm for the object movement paradigm.The bird on the top represents a moving target that disappears for part of its trajectory (dashed line).Internal coupling is evidenced by eyes moving along the imaginary trajectory while the target is absent, and closely matching the target's position upon its reappearance.Importantly, smooth pursuit appears only possible when the target is visible, while during imagery it is replaced by saccades, which continue to match the target's velocity.Eye behavior during imagery is represented as fixations (red dots) and saccades (blue lines).Example study:Jonikaitis et al. (2009).D: A prototypical paradigm for the body movement paradigm.During motor execution (left) the hand alternates between the right and left target (red circles), with the eyes spontaneously matching the direction of these movements.During motor imagery (right), one imagines moving the hand between the targets.Internal coupling is evidenced by the eye movements alternating between imagined targets with similar saccade duration and amplitude between execution and imagery conditions.Example study:Heremans et al. (2008).E, F: A prototypical paradigm for the brightness imagery paradigm.Internal coupling is evidenced by pupil constriction during the imagery of a bright stimulus like a sunny day (E) and pupil dilation during imagery of a dark stimulus like a dark forest (F).Example study:Laeng and Sulutvedt (2014).

Table 1
Overview of visual imagery domains including typical tasks, mental representation guiding internal coupling with a common eye parameter, and a representative study.

Table 2
Overview of studies investigating functional relevance of internal coupling across visual imagery domains.
(continued on next page)

Table 2
Johansson and Johansson (2014)onflicting outcomes across different measures, results supporting and refuting the functional role are reported together, clearly distinguished within the same row.To facilitate interpretation, findings fromJohansson and Johansson (2014)are presented in two separate rows.Exp.= experiment, IC = internal coupling, LAN = looking at nothing, ROI = region of interest, RT = reaction time, SPEM = smooth pursuit eye movements, VR = virtual reality, WM = working memory.