Eye movement patterns could predict mistakes in memory


In a recent study, scientists at Baycrest’s Rotman Research Institute (RRI) found that research participants moved their eyes to determine whether they had seen an image before, and that their eye movement patterns could predict mistakes in memory.

They obtained these results using an innovative new eye tracking technique they developed.

“Our findings indicate that eye movements play a functional role in memory retrieval,” says Dr. Jennifer Ryan, senior scientist at the RRI and Canada Research Chair in Cognitive Neuroscience of Memory. “They can tell us a lot about someone’s memory.”

This study builds on previous Baycrest research examining the link between eye movements and memory, including the role of our eye movements in memorization and the weakening connection between our eye movements and our brain activity as we age.

“When we see a picture, a face or something else that we have already seen, our eyes tend to look at the same locations as they did the first time. The brain compares important characteristics of what we are seeing to a mental picture in our memory, and it identifies the two as the same,” says Dr. Bradley Buchsbaum, senior scientist at the RRI.

“The brain is pretty good at this, even in conditions of lower visibility.”

“If we see someone in the distance, or if their face is partially hidden by branches, our brain will compare the features that are visible to a mental picture to determine whether we know that person,” says Jordana Wynn, lead researcher on this study, former PhD student at the RRI and current fellow at Harvard University.

This phenomenon is called “pattern completion.” When it goes wrong, we may end up mistakenly waving to a stranger if he or she has similar hair or a similar nose to someone we know.

In this study, published in the journal Proceedings of the National Academy of Sciences (PNAS), participants were first asked to memorize a series of 30 new images on a screen.

Next, they viewed another series, this time containing both some of the previously seen images and some new-but-similar images. They were then asked to indicate whether they had seen each one before.

Their eye movements were tracked during both stages. Each image was shown briefly, ranging from 250 milliseconds to 750 milliseconds, before the participants were instructed to visualize it while looking at a blank screen.

Participants were highly accurate in identifying previously seen images as old, scoring almost 90%. They were more likely to be correct if their eye movements were the same as when they initially saw the image.

On the other hand, they performed less well, at 70%, when faced with a new-but-similar image. In the latter case, the more participants repeated their initial viewing pattern instead of focusing on the different aspects of the image, the more likely they were to incorrectly identify the image as old.

To emulate real-world situations where we don’t have full information, the researchers also used incomplete, or “degraded,” versions of images.

This ranged from 0 to 80% degradation, in the form of grey squares covering parts of the image. Remarkably, even when the image was 80% degraded, performance was much better than pure guessing, reflecting the strength of pattern completion.

This phenomenon is called “pattern completion.” When it goes wrong, we may end up mistakenly waving to a stranger if he or she has similar hair or a similar nose to someone we know.

“Using our eye tracking technique, we were able to map the participants’ eye movements and observe that they were mentally picturing an image that they could not see,” says Wynn. “They were using pattern completion.”

Many studies have examined pattern completion over the past decades, but with one critical weakness. “These studies have all been based on the untested assumption that we can infer pattern completion is happening when participants mistakenly ‘recognize’ images that they have not seen before,” says Wynn.

“Our study is the first to use eye movement analysis, rather than behaviour, to show that people are in fact retrieving a memory of an old image when they make this mistake.”

This study’s findings have important implications in terms of assessing memory. “Some of the traditional tests used to diagnose memory impairments are quite verbal,” says Dr. Ryan. “They often require good command of the English language, which can be a problem in a multicultural city like Toronto.”

“With eye tracking, you don’t have to ask people what they remember. You can just look at their eyes. This gives us a lot more information about their memory than we thought,” says Dr. Buchsbaum.

This work was made possible with support from the Canadian Institutes of Health Research (CIHR) and the Natural Sciences and Engineering Research Council of Canada (NSERC).

With additional funding, the researchers could further examine the role of eye movements in memory retrieval. “This could lead to the development of better screening tools for dementia, which is the ultimate hope,” says Dr. Ryan.

Eye movements and memory processes

The discipline of psychology has long been interested in the role that the human visual system plays in everyday cognition. In 1972, Loftus published an article in Cognitive Psychology that reported a positive correlation between the number of eye movements that a participant made while viewing a picture and their subsequent memory for that picture [1].

Since then, researchers have linked eye movements to visuo-spatial working memory [2], to external projection of mental imagery [3], and even to direct neuroanatomical connectivity ties to the hippocampus in macaque monkeys [4]. The study of eye movements is now firmly embedded in the study of cognition, and extensions to the memory literature continue to emerge.

In 2003, Christman and colleagues were the first to investigate a new perspective on eye movements and memory—the potential benefit of bilateral eye movements for memory retrieval [5].

They undertook this work because of research in their own lab suggesting that bilateral eye movements can enhance interhemispheric interaction [6], which in turn was theorized to have a role to play in episodic memory processes [7,8].

Their prediction was that repeated horizontal saccadic eye movements, when performed immediately before a memory test, should cause increased interhemispheric interaction and lead to superior retrieval relative to no eye movements at all.

That pattern was in fact what they observed. The purpose of the current study was to first perform a close extension of this seminal work by Christman and colleagues in our own lab before ultimately moving forward to test proposed mechanisms underlying the observed memory boost following bilateral eye movements.

Seven years and several articles after Christman et al.’s (2003) work [5], Lyle and Martin published the first study that would refer to their finding as the Saccade-Induced Retrieval Enhancement (SIRE) effect [9].

Lyle and Martin argued against Christman et al.’s (2003) account of the reported memory benefit in terms of interhemispheric processes. They employed a letter-matching task both within and across visual fields to provide evidence that the SIRE effect is more likely due to intrahemispheric (i.e., within one hemisphere) rather than to interhemispheric (i.e., between hemispheres) activity [9].

Their idea was that detection of letter pairs on the matching task requires intrahemispheric interaction on within-hemisphere trials but mostly interhemispheric processing on across-hemisphere trials. Therefore, the superior performance on within-hemisphere trials that they observed following eye movements suggests increased intra- as opposed to inter- hemispheric interaction [9].

Generally, the SIRE effect has been researched using word stimuli and simple memory retrieval tasks [e.g., 5,8]. Brunyé and colleagues, however, sought to extend the literature to include picture stimuli [10].

Edlin and Lyle (2013) actually went so far as to rename the effect saccade-induced cognitive enhancement (SICE) since they have reported benefits of bilateral eye movements extending to executive functioning aspects as well [11].

Further studies have reported broader SIRE benefits in domains including autobiographical memory [12], gist-based false recognition [13], episodic future thinking [14], and childhood amnesia [15].

While there have been several publications reporting significant SIRE effects, there have also been reported failures to replicate findings. Most notably, Matzke et al. (2015) failed to replicate the SIRE effect, consistently finding strong Bayesian evidence for a null [16].

Critically, their method followed a framework of pre-registered adversarial collaboration, which involved two opposing research groups discussing and ultimately agreeing upon specific experimental and statistical methodologies before pre-registering their study on the Open Science Framework (OSF).

In addition, Samara et al. also failed to find a significant effect when using neutral valence words for stimuli (but did succeed in finding the effect for emotional words) [17]. Further, while some articles have proposed that eye movements can have the side effect of increasing false memory [e.g., 18], a recent direct replication attempt failed [19].

These failed replications are obviously of major concern and served as a main motivation for attempting to conceptually replicate the SIRE effect in our lab.

Despite previous research on the SIRE effect having varied widely with regard to its application, the fundamental underlying mechanisms are still in need of clarification.

There are currently two major proposed theories that attempt to explain the SIRE effect: Christman et al.’s theory of interhemispheric interaction [5], and Lyle et al.’s theory of top-down attentional control [9]. Because the purpose of the current study was to conceptually replicate and extend existing work before moving on to further test these leading theories, we begin by examining the background rationale and literature supporting each account.

Interhemispheric interaction

Christman et al. (2003) argued in the very first bilateral eye movement article that their observed memory benefit was due to increased interhemispheric interaction [5]. They based this hypothesis on their own previous work on handedness, which suggested that left-handed individuals have naturally greater interhemispheric interaction due to their larger corpus callosa, and therefore naturally perform better on episodic memory tests [7,20].

Christman et al. (2003) broadly based their interhemispheric interaction hypothesis of SIRE around the Hemispheric Encoding and Retrieval Asymmetry (HERA) [21,22] and Cortical Asymmetry of Reflective Activity (CARA) [23] hypotheses. While the HERA model suggests that encoding is mostly left-lateralized and retrieval is right-lateralized, the CARA model alters this slightly to propose that retrieval processes occur either primarily in the right hemisphere or in both hemispheres, depending on the task [23].

Following the logic of the HERA and CARA models, the interhemispheric interaction hypothesis argues that eye movements made before retrieval prime the hemispheres for interhemispheric activity so that tasks taking advantage of bi-hemispheric communication should receive a performance benefit [5,10].

According to this hypothesis (and the CARA hypothesis broadly), tasks with more ‘reflective activity’ (i.e., “when retrieval of additional information or more detailed evaluations of activated information are required during episodic remembering” [23, p. 3513]) will lead to the use of both hemispheres at retrieval [23,24], thus benefiting from an interhemispheric interaction priming effect.

This proposal has been supported by work from Brunyé et al. [10], which showed a memory benefit following bilateral eye movements when participants performed a YES/NO recognition test, but not when they performed a two-alternative forced choice recognition test (2AFC) [10]. Their argument was that the YES/NO test makes use of more reflective activity than does the 2AFC test, and for that reason showed a performance benefit following bilateral eye movements [10].

Propper and Christman (2008) also supported the interhemispheric interaction hypothesis, in their case by connecting it to REM sleep processes [25]. It was suggested that because the majority of eye movements made during REM sleep cycles are horizontal [26], and that there is an increase in interhemispheric coherence during REM sleep [20], the SIRE effect mimics the same processes found in REM sleep, thereby resulting in increased interhemispheric interaction [26].

Importantly, the interhemispheric interaction hypothesis predicts that only bilateral eye movements (meaning, only horizontal movements) should result in a SIRE benefit. It has been shown that unilateral saccades activate the contralateral hemisphere [27].

In the context of SIRE, this would suggest that vertical eye movements should not produce a SIRE effect because they do not produce the same bilateral activation in frontal eye fields as is produced by horizontal eye movements [22,28]. In summary, the interhemispheric interaction hypothesis predicts that only horizontal eye movements should result in a performance benefit due to the inherent alternating activation of hemispheres. Thus, to test this theory, the current experiments measured memory performance following horizontal eye movements compared to both vertical and no eye movements.

Top-down attentional control

A competing account soon emerged for the observed memory benefit following bilateral eye movements—that the boost resulted from top-down attentional control processes. Work by Lyle et al. was the first to propose that the SIRE effect could be due to the influence of eye movements on attention [29].

They suggested that since bilateral activation of frontal eye fields occurs even with vertical eye movements [30,31], this means that both types of eye movements should lead to a SIRE effect [29]. Indeed, others have reported that eye movements in any direction lead to frontal eye field and intraparietal sulcus activation, which themselves have been implicated broadly in attention processes [32,33].

Because past work has shown that frontal eye fields are so intertwined with other brain regions within the same attention network (e.g., the central sulcus, cingulate sulcus, medial frontal gyrus, and intraparietal sulcus) [32], Lyle et al. theorized that frontal eye field activity can lead to a cascading effect of activation that ultimately results in broad activation of both hemispheres [29].

As already mentioned, Lyle and Martin (2010) coined the term “SIRE effect” [9]. In the same article, they also refined the top-down attentional control hypothesis to further emphasize an attentional component as opposed to interhemispheric interaction.

Their updated account suggested that engaging in bilateral eye movements requires top-down attentional processing to successfully perform repetitive purposeful eye movements, thus preparing the participant to make fewer errors, for instance, on subsequent tasks that require higher levels of attentional control [9].

Further, they proposed that eye movements are tied to activation in frontal eye fields and the intraparietal sulcus, which are both in turn part of a frontoparietal attention-control network [32].

Since the aforementioned attention network has been implicated in top-down allocation of attention during memory retrieval, eye movements in any orientation should theoretically prime this network. Priming of this network then ultimately aids top-down attention processes—and therefore memory performance—on a retrieval test [9].

After finding evidence of a SIRE effect following vertical eye movements [see also 29], Edlin and Lyle (2013) sought to more directly test their top-down attention control theory [11].

To do this, they employed a typical eye movement paradigm as used by others in the SIRE literature but instead employed a task designed to measure executive attention. Their results indicated that participants who made bilateral eye movements, relative to controls who made no eye movements, experienced greater attentional control as measured by faster reaction times on the ANT-R [11].

In 2015, Lyle and Edlin evaluated this attention idea again, opting to employ both horizontal and vertical eye movement conditions, along with a return to more common memory retrieval tests (namely, recall and recognition) instead of the attention test used previously [28].

Critically, some items on the recall test were made more difficult than others by first having participants practise recall from certain studied semantic categories but not from others. They then administered a final recall test, with the prediction that items from non-practised categories would be more difficult to retrieve.

Lyle and Edlin performed this experimental manipulation because it has been shown in the literature that top-down attentional processes play a key role in memory tests, especially when performance is low and cognitive effort is high [34]. Results aligned with their predictions: The SIRE benefit was found when memory tasks were harder and therefore required greater top-down attentional control [28].

Crucially, they also found a memory performance benefit following vertical eye movements (for the second time), which is predicted only by their top-down attentional control account.

In summary, the top-down attentional control hypothesis suggests that purposeful eye movements (horizontal or vertical) lead to the priming of an executive function attention network that aids the allocation of top-down attentional control at retrieval, ultimately manifesting in a memory performance boost.

Our experiments, much like for the previous theory discussed, compared memory performance following each of three critical eye movement conditions (horizontal, vertical, and centered) in an attempt to provide evidence in support of or in opposition to the top-down attentional control account.

Neuroimaging evidence

Four studies to date have used neuroimaging methods (all electroencephalogram; EEG) to directly assess the interhemispheric interaction hypothesis as an account for benefits following bilateral eye movements.

The first, conducted by Propper and colleagues (2007), employed EEG following bilateral eye movements to show that, relative to controls (who made no eye movements), there was decreased gamma band coherence between the two hemispheres in the bilateral eye movement condition [35].

That is, the authors actually observed decreased interhemispheric coherence (contrary to the increased interhemispheric interaction that Christman et al., 2003 suggested). They nevertheless argued that any changes in neural activity induced by eye movements that resulted in a boost in memory performance still broadly supported Christman and colleagues’ (2003) idea that changes in interhemispheric communication are key.

In the second relevant neuroimaging study, Samara and colleagues sought to test the interhemispheric interaction hypothesis using EEG [17]. Contrary to Propper and colleagues’ article, Samara et al. found no evidence for consistent alterations in interhemispheric interaction following bilateral eye movements across six frequency bands, including gamma [17].

It should also be noted that Samara et al. did not find a significant behavioral SIRE effect for neutral valence words, although they did find one for emotional words [17]. Ultimately, though, no difference was found in bilateral EEG coherence for either stimulus valence condition [17]. Thus, their results were entirely in contrast to those of Propper et al. [35] as well as being inconsistent with the broader interhemispheric interaction hypothesis.

Finally, the most recent two EEG studies that have investigated the effects of bilateral eye movements were performed by Fleck and colleagues. In the first, the authors found reduced posterior delta coherence in the control (center-eye movement) group relative to a bilateral eye movement group, suggesting that performing eye movements led to sustained executive attention [36].

In the second article, Fleck and his collaborators failed to find any behavioral differences between groups on the revised attention network test (ANT-R) following bilateral eye movements [37].

The authors did, however, find differences in N100 and P200 event-related potentials (ERPs), which due to these ERPs’ connection with selective attention [38,39], results were taken to indicate that less selective attention is needed on the ANT-R following eye movements [37]. Taken together, these two studies from Fleck and colleagues posit that participants experience a broad benefit for multiple aspects of attention following bilateral eye movements.

Handedness considerations

In addition to eye movement orientation, the SIRE effect has also been reported to be rather sensitive to participant handedness. In fact, Christman and colleagues’ (2003) original article based their hypothesis concerning interhemispheric interaction on prior research from their own laboratory that suggested a similar neural substrate and episodic memory enhancement for participants with positive familial sinistrality (i.e., having left-handed relatives).

They argued that this was likely due to left-handers’ naturally larger corpus callosa, which allows for greater interhemispheric communication [7; also see 20].

Their idea was that right-handers’ naturally smaller corpus callosa could benefit from an increase in interhemispheric communication that comes as a result of bilateral eye movements, whereas left-handed and mixed-handed individuals (with naturally larger corpus callosa) are already at such a high level of baseline interhemispheric communication that the eye movements provide no additional benefit [10].

Because Christman et al. [5] specifically selected their participants to be strongly-right handed and made the aforementioned case for handedness considerations, most research in this domain has followed suit and similarly only tested right-handed participants [e.g., 10].

A brief report by Lyle and colleagues demonstrated—in contrast to the enhancement seen for strongly right-handed individuals—a significant reduction in memory performance following bilateral eye movements for non-strongly right-handed individuals [29]. This is odd, though, because neither theory of the SIRE effect would have predicted memory attenuation for the latter group.

It is worth noting that Lyle and Edlin’s (2015) top-down attentional control theory did not consider right- versus left-handedness; instead, they stated that consistent handedness (often considered to be a score of 0.80 or greater on the Edinburgh Handedness Index) is the truly important factor [28].

They argued that only consistent handers seem to benefit from bilateral eye movements, but unfortunately they did not provide a solid prediction as to why this handedness inconsistency occurs [28].

Similar to the Christman et al. (2003) idea, Lyle and Edlin speculated that inconsistent-handers might have larger corpus callosa than consistent-handers, seemingly following along the lines of an interhemispheric interaction account for SIRE [28].

That is, consistent-handers with smaller corpus callosa theoretically have a lower baseline level of interhemispheric interaction and can therefore benefit from bilateral eye movements. Indeed, a recent review article summarizing handedness suggested ultimately that consistency may be a better predictor of interhemispheric interaction and right hemisphere processes than traditional left- or right-handedness distinctions [40].

Within the SIRE literature more specifically, Lyle and Edlin sum up perfectly that “the effect of saccade execution on inconsistent [handed] individuals is highly variable” [28, p191]. A potential explanation for these findings tracks well with an earlier presented argument from Brunye et al. [10]: inconsistent-handers have naturally larger corpus callosa and therefore might be at a ‘ceiling’ of interhemispheric interaction making them unable to benefit from bilateral eye movements.

In summary, while Christman et al.’s (2003) interhemispheric interaction account predicts that only right-handers (or in later literature, consistent-handers) should benefit from SIRE, Lyle and Edlin’s (2015) top-down attention account makes no differential prediction.

Implications for eye movement desensitization and reprocessing (EMDR) therapy

The astute reader will have noticed that the SIRE effect literature employs a paradigm that is very similar to the technique used in Eye Movement Desensitization and Reprocessing (EMDR) therapy. When EMDR therapy is used to treat Posttraumatic Stress Disorder (PTSD)—the most common use for this type of therapy—the patient is typically instructed to make repetitive bilateral eye movements while recalling their traumatic experience [41; see 42 for a review]. This is one of the most critical factors warranting further exploration of the SIRE effect: Related therapeutic practices are basing their patient treatment on the efficacy of bilateral eye movements as a memory enhancement tool.

A common problem with implementing therapy techniques for those that suffer from PTSD is that patients’ dissociative amnesia of their traumatic memories makes it difficult to recall and work through their trauma in a therapeutic environment [see 43 for a review]. Christman et al. [5] suggested that the bilateral eye movements made during EMDR therapy play a similar role to the eye movements made in SIRE studies. That is, the role of eye movements in EMDR is to promote eased retrieval of episodic memories, and therefore to allow for therapeutic guidance to take place more effectively [see 25 for a review].

In addition to the reported episodic memory benefits following bilateral eye movements, there is some evidence that these same eye movements can also reduce an individual’s emotionality.

If this were indeed the case, it would imply that SIRE is even more applicable for PTSD-related therapeutic interventions: Eye movements could not only allow patients to remember their repressed traumatic memories better, but they could also calm the patient’s emotional state, allowing for eased therapeutic guidance to take place.

Christman and Propper (2008) reported a study in which horizontal eye movements were associated with a significant neutralization of mood among participants, such that previously happy and previously sad individuals became respectively less happy and less sad following bilateral eye movements. Bartels and colleagues (2018) reported that bilateral eye movements even caused a reduction in sexual fantasy vividness, arousability, and emotionality [44].

In relation to EMDR more specifically, Stickgold has argued that bilateral eye movements mimic rapid eye movement (REM) phase sleep processes [45]. Much like in REM sleep, it has been proposed that repetitive eye movements allow for enhanced cortical integration of memories, thus making the memories less dependent on the hippocampus (and, by extension, on the amygdala), ultimately reducing the emotionality of the memories.

Although research on the emotional neutralization effects of SIRE is rather limited, the commonality of bilateral eye movement manipulation between SIRE and EMDR is obvious. However, mixed results and interpretations within both the EMDR [e.g., 46] and SIRE [e.g., 47,48] literatures have resulted in a rather controversial discussion surrounding the efficacy of both effects. Thus, the validity of both phenomena should be interpreted with caution.

Due to both the theoretical and real-world implications of SIRE for psychological research and therapy applications, the effect has not surprisingly intrigued investigators. First and foremost, therefore, we wanted to better understand the underlying mechanisms that may be at work causing a boost in memory performance. Believing that it is always wise to start with replication of a basic phenomenon before undertaking further research, we began with a conceptual replication of the SIRE effect in our laboratory.

Baycrest Center for Geriatric Care


Please enter your comment!
Please enter your name here

Questo sito usa Akismet per ridurre lo spam. Scopri come i tuoi dati vengono elaborati.