We set out to review and synthesize the literature regarding the fundamental question of how breathing influences pupil dynamics in humans. Our main aim was to capture the effects of breathing phase, with the additional aims of describing results for breathing depth, rate and route. Based on state-of-the-art reviews [3, 45] targeting the relationship between breathing and higher functions, it is easy to get the impression that the theoretical and factual link between breathing and pupil dynamics is established and well-studied. However, our systematic review, which summarizes research from more than 30 articles, spanning over 80 years of inquiry, tells a different story. We find that since Golenhofen and Petrányi’s (1967) study, the first to explicitly test this question in humans, only 11 studies have been conducted that asses this relationship directly. Below we will address the current evidence for the theory that breathing can influence pupil dynamics for each breathing parameter separately, and the limitations and outstanding questions that need to be solved before a conclusive answer can be given of its validity.
Phase
Six of the included studies looked at breathing phase, and among these studies there was a diversity of study methods and aims; however, all of them aligned with the hypothesis that breathing phase influences pupil dynamics. When setting out to conduct this systematic literature review, we expected to find how inhalation and exhalation specifically affect pupil function. However, focus in the reviewed literature was instead on changes in pupil dynamics over the course of the breathing cycle in general. Therefore, we can only say that from the studies included in this systematic review, breathing phase appears to correlate with pupil size and pupil size fluctuations.
Generally, these studies have small sample sizes which reduces the strength of their findings. Furthermore, the wide variety of study methods and outcome measures made it difficult to summarize their findings. The final GRADE score for this category rated the strength of evidence as “Low”, and with that the evidence for an effect of breathing phase on pupil dynamics is not negligible but far from conclusive.
Depth
The included studies assessing breathing depth were not as homogenous in their results as those assessing phase, with only four out of six studies finding evidence for the theory that breathing depth affects pupil dynamics. Of the four studies that found an effect of breathing depth on pupil dynamics, three studies reported that an increase in breathing depth also led to an increase in pupil size, and one study even reported a linear correlation between breathing depth and pupil size [47]. However, since He and colleagues (2019) reported the opposite effect, and two studies did not find an effect at all, it is not clear if, and how, breathing depth influences pupil dynamics [13, 19, 56]. In addition, the majority of these studies did not set out to investigate the effect of breathing depth on pupil dynamics and further suffered from statistical shortcomings. This led to the quality of the evidence surrounding effects of breathing depth on pupil dynamics to be classified as “Very Low” according to the GRADE criteria. Therefore, even though the majority of studies that measure breathing depth and pupil dynamics are in favor of a correlation between these two measures, until further research is published, we are hesitant to place much confidence in the existence of such an effect.
Rate
Breathing rate was the most common breathing measure reported in the studies included in this systematic review (20 out of 31). Because we had such a large number of relevant studies, we decided to split them into two categories; the studies that reported on effects of breathing rate on pupil dynamics directly, and those that did so only indirectly. We made these classifications based on the main study outcome and the reported results of the included studies.
Of the eight studies classified as providing direct evidence on the question of whether breathing rate influences pupil dynamics, four studies reported results in favor of an effect, whereas the other four studies did not (see Fig. 4).
Importantly, while the absence of evidence in these studies should not be taken as evidence of absence, the studies that did not find an effect were generally rated as being of higher quality, and had larger sample sizes than those that found an effect. Consequently, the overall evidence for the hypothesis that breathing rate affects pupil dynamics, received a GRADE score of “Very Low”.
The studies that indirectly reported breathing rate and pupil dynamics measures are larger in quantity (12), and on average have a larger sample size and better rated study quality for their intended study aim, than the studies that directly investigate breathing rate and pupil dynamics (see Fig. 5). The majority of these indirect studies (8 out of 12) lend support to the hypothesis that breathing rate influences pupil dynamics. However, due to the indirect nature of these studies and the inconsistency of their findings, these studies still received an overall GRADE score of “Very Low”. In addition, the majority of the studies giving indirect support to the theory that breathing rate influences pupil dynamics, did so by reporting a simultaneous increase in breathing rate and pupil size. In contrast, the studies that found an effect when directly investigating breathing rate and pupil size, mainly reported the presence of a peak in the power spectrum of the pupil size at the breathing frequency. This difference in the kind of measured outcomes makes it hard to integrate both kinds of findings. However, despite the weakness of the evidence in favor of an effect of breathing rate on pupil dynamics, the relatively large number of studies finding some sort of covariance between these two measures cautions us against discarding this hypothesis altogether.
Pupil parameters
For this systematic review, we grouped studies according to the breathing parameters, given that our main aim was the effects of breathing on pupil dynamics. However, it is also possible to do the reverse, and group the study results by type of pupil measurements, which could contribute with more pupil-specific information. Specifically, the pupillary outcomes included in this review are: pupil size, pupillary unrest (hippus), pupil size fluctuations, and pupillary light reflex dynamics.
Shifting to a pupil-centered perspective, out of the five studies that discussed pupillary unrest indices or hippus (see Tables 1 and 2), only one study reported an effect of breathing (breathing rate) [12]. However, it should be noted that they did discuss pupillary unrest more or less interchangeably with pupil size and pupil size fluctuations. The other four studies measuring breathing rate did not find an effect, and one study also looked at breathing depth on pupillary unrest and found no effect [61]. Among studies measuring pupillary unrest, there was variation in how they measured this, as well as their assessed quality; however, since this was not one of our main study outcomes we did not formally assess the quality of this evidence.
All four studies reporting pupil size fluctuations found that these were influenced by either breathing rate or breathing depth [9, 12, 47, 71]. However, again, there was an overlap with both pupillary unrest and pupil size in their measurements, but it is unclear how much pupillary unrest and pupil size fluctuations are distinct from pupil size in these cases.
Finally, three of the articles measured the effect of various aspects of breathing on the response of the pupil to sudden light stimulation, which we categorized as pupillary light reflex dynamics [15, 16, 26]. Together they looked at breathing phase, rate, and the effect of hyperventilation, and all reported some effect on pupillary light reflex dynamics. However, these studies were inconsistent in their findings, and all suffer from various methodological shortcomings (e.g., small sample sizes), thereby weakening the strength of their findings. Note, though, that, as above, we did not formally assess the strength of evidence for this outcome.
Influence of experimental paradigms
A wide variety of studies have been included in this review, some that support the theory that breathing influences pupil dynamics, and some that do not. Importantly, whether or not these studies measured breathing and pupil at rest, or employed a certain set of tasks or stimuli, does not seem to be a good predictor of whether indications of breathing influences on pupil dynamics will be found. However, a few studies reported a change in both breathing and pupil measures, and sometimes a change in their interaction, as a result of increases in cognitive load or task difficulty [19, 43, 44, 56]. Two studies report significant increases in both pupil size and breathing rate in response to increasing cognitive demands during driving simulation tests [19, 56]. On the other hand, two other studies found that the interaction between breathing rate and pupil size disappeared when the participants task demands increased [43, 44]. However, there were several other studies that included cognitive tasks of various difficulty into their study design and did not report significant simultaneous effects on breathing and pupil measures. A firm conclusion cannot be made based on these results, but it is salient to highlight that potential interactions with study paradigms warrants further investigation.
Limitations of the review processes
This review follows the PRISMA guidelines and the assessment of the certainty of the evidence was based on the GRADE approach. As both these methods have been created with clinical trials in mind, we modified some of the steps of these methods and excluded others. The included steps on the PRISMA guidelines can be seen in Table 1 in the Supplementary materials while the utilization of GRADE is discussed in the Methodology section with a more detailed discussion in the "Detailed GRADE assessment” section of the Supplementary materials.
Our modifications to the GRADE approach were influenced by the large variability in study design due to the broad inclusion criteria, and the lack of statistical assessment of the relationship between breathing and pupil dynamics. This made the GRADE process somewhat more subjective and imprecise, such as the initial GRADE scoring of the study design, and the assessment of the imprecision of the reported results. We did not assess publication bias. However, publication bias is another potentially very large source of bias. Importantly, because inclusion of the “publication bias” criterion would only have reduced the GRADE score of any outcome measure, we do not see it as likely to change the results of the GRADE, which were all “Low” or “Very low”.
Moreover, as we did not do a metanalysis and perform any statistical analysis of the body of evidence, this could potentially bias the findings by making it harder to give appropriate weighting to low quality studies with low sample sizes. However, the quality assessment of the individual articles, and the GRADE approach to assess the overall evidence for each parameter, were designed to take this bias into account.
While the experimental setup and processing of the data was evaluated through the quality assessment, it was beyond the scope of this review to assess this thoroughly. For example, we did not assess how the different software used for pupil measurements, estimated pupil size. However, because some methods to assess pupil diameter are more reliable than others, looking at the measurement setup and analysis in more detail could provide a more accurate account of the reliability of the study outcomes [11, 33].
Limitations of included studies and implications for future research
During the last few years, the potential impact of breathing on behavior and brain function has received a flurry of attention. It is therefore critical to understand the extent of the interactions between breathing and pupil dynamics which has often been taken as a given. On the face of it, the question of whether breathing influences pupil size seems relatively simple and straightforward to answer. What we show in this systematic review is that this does not seem to be the case. There is no conclusive evidence that breathing phase, depth, or rate influences pupil dynamics. Surprisingly, despite research spanning many decades the fundamental question of whether breathing route impacts pupil function has, to the best of our knowledge not yet been investigated.
To clearly answer if there is a connection between breathing and pupil function, it is necessary to conduct research with a higher quality than before and a direct focus on this relationship. We believe that pre-registered studies using modern equipment, adequate sample sizes, proper statistics, and a lack of confounding stimuli should be able to answer the fundamentals of this question with relative ease. In Box 1, we outline a series of outstanding questions and potential confounds that future studies should address. Until such research has been carried out, we caution against strong trust in an effect of breathing on pupil dynamics. For example, the majority of the studies in this review measure breathing with a breathing belt. This is unfortunate as breathing belts have both poor reliability and validity [23]. Moreover, they do not separate nose from mouth breathing, which is of critical importance as the underlying central mechanisms and behavioral outcome between these two formats of breathing are very different [2, 20, 25, 65]. We also noticed that many studies do not measure or take account for eye blinks. This is critical because breathing affects the probability of eye blinks, which in turn affects pupil size. For example, blinking is more likely during exhalation [59] which means that the probability that inhalation and pupil dilation (when you open your eyes) will coincide is larger compared to exhalation. Likewise, the potential of cardio-respiratory coupling [52] also stresses the importance of controlling for heart rate, which is known to influence pupil size [67].