Introduction

The interoceptive rhythms of the body are essential for maintaining homeostasis [1,2]. While our breathing, heartbeat, and other visceral oscillations are critical for keeping us alive, less is known about how these rhythms influence our perception of the world. Indeed, until recently, breathing in the brain was thought to be largely constrained to basic processes concerning only the maintenance of respiratory drive and gas exchange [3,4]. This idea has recently been challenged by novel neurophysiological and behavioural discoveries [for reviews see, 5,6]. These studies have shown that breathing drives alterations in both neural oscillations and behaviour across numerous tasks and animal models.

The finding that perceptual decision-making depends on the rhythms of breathing span multiple cognitive and sensory domains. One recent study found that across visual, auditory, and memory tasks, reaction times differed substantially if participants responded during inspiration as compared to expiration [7]. Other studies have found that inspiration increases the speed of emotional processing [8], and that participants make more volitional movements during expiration [9]. These and many more recent reports suggest that breathing plays a fundamental role in shaping exteroceptive, affective [8,10,11], and cognitive behaviour [for reviews, see 5,6,12,13].

Theoretical proposals suggest that this modulation operates by adaptive gain control through the respiratory tuning of neural excitability, a mechanism that is critical for evidence accumulation during perceptual decision-making [5,14–17]. While some studies have demonstrated psychological and neural effects of the respiratory cycle that are broadly consistent with shifts in neural gain [17,18], to our knowledge, no study has directly examined the underlying cognitive mechanisms using computational modelling. As studies have highlighted consistent modulation of reaction time by the breathing cycle [7], we hypothesised that breath-brain coupling may alter the underlying latent variables governing perceptual and affective decision-making.

To test this hypothesis, we conducted a psychophysical experiment in which participants performed stimulus discrimination tasks in the domains of visual random dot motion and affective face perception. By modelling the influence of the inspiratory-expiratory cycle on perceptual choices with a hierarchical drift diffusion approach [19], we evaluated which specific decision variables in each modality were modulated by the breathing state. This enabled us to test whether respiratory state effects are primarily elicited by changes in stimulus processing or response execution, and to evaluate the cognitive mechanisms underlying these putative effects. We confirmed that participants respond more quickly in both domains when responding during inspiration, and also found that this acceleration arises from the modulation of different decision variables across these domains. Our findings highlight distinct mechanisms by which the respiratory cycle modulates visual motion and affective face perception.

Results

We tested the effects of respiratory state on visual and affective processing using a perceptual decision-making task which alternated between each stimulus domain across blocks. These two domains were selected because they represent fundamental dimensions of visual perception - basic sensory processing and emotional interpretation - allowing us to examine whether respiratory state effects generalise across different cognitive domains. Additionally, the careful control of perceptual and affective content in our stimuli enabled us to minimise potential biases and isolate the influence of respiration on decision-making processes in both modalities.

To maximise the ability to detect respiratory state effects on perceptual behaviour, all stimuli were individually thresholded using a Bayesian psychophysical approach [20]. Random dot motion (RDM) stimuli were calibrated to the individual motion coherence threshold for discriminating upwards vs downwards motion (mean coherence threshold = 0.21, SD = 0.14). For the affective domain, we adapted a face affect discrimination (FAD) procedure [21], which utilised face morphing to present stimuli conveying a range of emotional expressions between 100% happy and 0% angry and vice versa. This enabled us to titrate stimuli to the individual threshold for perceiving a facial expression as happy vs angry (mean face morph threshold = 53% happy, SD = 9). Following this staircasing procedure, participants completed a one-alternative forced choice paradigm during presentation of 320 near-threshold stimuli in each modality (see Fig 1 for overview).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. A) Trial structure of the random dot motion (RDM) and face affect discrimination (FAD) task.

The task is to identify the coherent motion direction (blue arrows denote coherent motion, red arrows denote random motion. Arrows are shown here for illustrative purposes) or decided whether face expressions are happier or angrier. B) Estimated psychometric functions for both RDM and FAD stimuli. Grey lines show individual participants’ traces, dashed red shows the group mean. Green and purple symbols indicate examples of the selected stimulus intensities used in the test trials. C) Each test trial was labelled both by the respiratory state at stimulus onset and at the time of response as indicated by face and hand icons, respectively. Face icons show examples of happy and angry stimuli. Red and blue lines represent periods of inspiratory and expiratory states respectively. ITI: Inter-trial interval.

https://doi.org/10.1371/journal.pcbi.1013086.g001

We first assessed respiratory effects on choice behaviour.

As the perceptual decision process unfolds over time,different parts of this process may pertain to different respiratory states. To establish whether respiratory-behavioural coupling is primarily associated with perceptual or motor systems, we analysed trials grouped independently by respiratory state at stimulus presentation (stimulus-locked) and at the time of response (response-locked). It should be noted that these were analysed separately, in order to assess the independent contributions of respiration to stimulus processing and response modulation, as the study was not designed to directly compare these conditions.

In the visual motion domain, we found a significantly higher hit rate (HR) for responses made during expiration compared to inspiration (mean difference = 2.0%, 95% Confidence Interval (CI) = [0-4%], p = 0.044). This effect was not observed when grouping trials based on respiratory state at stimulus presentation (Bayes factor in favour of the null, BF01: 4.67). For FAD, the proportion of happy vs angry responses did not differ significantly for either stimulus or response grouping (BF01: 3.44 and 5.46, respectively). See S1 Fig and S1 Table for all statistics.

We next determined whether response speed for the two domains differed as a function of respiratory state, again using both stimulus and response locking. Here we found a predominant effect for response-locked median reaction times, such that responses were faster when they occurred during inspiration vs expiration for both RDM, mean difference = 18.9 milliseconds (ms), 95% CI = [10-30 ms], p < .001, and for FAD, mean difference 21.2 = ms, 95% CI = [10-30 ms], p = .001 (see Fig 2). In contrast, no effect for either domain was observed when analysing reaction times based on stimulus locking (BF01, RDM: 5.21, FAD: 3.4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Boxplots depicting the difference (delta) in median reaction time between inspiration and expiration, based on response grouping for each domain.

Positive values indicate longer reaction times during inspiration compared to expiration. Dots indicate individual participant delta medians. Notches indicate 95% CI of the median. RDM: Random dot motion, FAD: Face affect discrimination. **p < 0.01, ***p < 0.001.

https://doi.org/10.1371/journal.pcbi.1013086.g002

These results are in line with the descriptive statistics across subjects where RDM responses made during inspiration and expiration respectively had means of 675 ms ± 275, and 694 ms ± 274, just like response-locked FAD reaction times for inspiration and expiration were 737 ms ± 295, and 763 ms ± 309 respectively (±: SD). Stimulus-locked reaction times for inspiration and expiration respectively were 689 ms ± 273, and 688 ms ± 275, for RDM and 758 ms ± 306, and 753 ms ± 303 for FAD.

In summary, responses made during inspiration were faster for both stimulus domains and less accurate only for the visual motion domain, pointing to a complex and heterogeneous coupling between respiratory state and decision-making. To establish the underlying mechanisms driving the observed coupling effects we employed drift diffusion modelling (DDM) [22]. The drift diffusion model jointly analyses choice and reaction time data. The model postulates that speeded decisions can be characterised as a noisy accumulation of evidence originating from a starting point positioned between two decision boundaries (e.g., reporting a happy or an angry face), and additional latencies not related to the decision process itself. Fitting this model to choices and reaction timesacross respiratory states allowed us to investigate whether the observed coupling effects were most effectively accounted for by changes in the rate of evidence accumulation (drift rate, v), participants’ biases towards particular responses (starting point bias, z), the balance between speed and accuracy (boundary separation, a), or by extra-decisional variables such as attention or motor preparation (non-decision time, t). For a visual representation of these parameters, see Fig 3A.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Drift Diffusion Modelling of Respiratory State Effects on Decision-Making.

A) Schematic representation of the drift diffusion model (DDM) visualising the drift rate (v), non-decision time (t), bias (z), and decision threshold (a). B) Fitted group level DDM parameters and posterior predictive checks for the random dot motion (RDM) model and C) face affect discrimination (FAD) model. Drift rate was fitted for each stimulus class as shown by different coloured lines. Bold lined histograms represent the empirical reaction time histogram for each stimulus type and for each response option accumulated over time across all participants. Shaded lines represent the histograms of 100 data sets simulated based on the estimated model parameters. Note that the empirical and simulated data contain reaction times faster than the depicted group level non-decision time since some participants have shorter non-decision time than the group. D) Posterior predictive distributions for differences in DDM parameters by respiratory state. We found a significant reduction in decision threshold (P = 0.008) and drift rate of upwards stimuli (P = 0.013) for RDM, and a positive affective bias towards ‘happy’ responses during inspiration for FAD (P = 0.048). No significant shifts were found in drift rate of other stimuli, decision threshold in FAD or for non-decision time (all P > 0.05). Each parameter’s 95% highest posterior density interval (HPDI) is shown, reflecting the effect of respiration on decision dynamics. AL: angry, low ambiguity, AH: angry, high ambiguity, HH: happy, high ambiguity, HL: happy, low ambiguity.

https://doi.org/10.1371/journal.pcbi.1013086.g003

We fit hierarchical DDMs where these parameters were modulated by the inspiratory vs expiratory state of respiration. Suitable model convergence and fit was assessed by calculating the Gelman-Rubin statistic (see Methods), and by posterior predictive checks (Fig 3B and 3C). Inspection of the group-level posterior probability distribution of each decision variable’s modulation by respiratory state (as illustrated in Fig 3D) revealed a significant decrease in boundary separation during inspiration for visual motion (P = 0.008), consistent with observed changes in speed and accuracy described above. For the affective domain, we observed a significant shift in starting point bias towards ‘happy’ responses during inspiration (P = 0.048). For both domains, the state-related changes in evidence accumulation showed a complex, stimulus dependent pattern of respiratory modulation. That is, we found a significant reduction of drift rate for upwards motion stimuli (P = 0.013) during inspiration with no evidence for changes in drift rate for downwards motion stimuli (P = 0.889). For FAD we observed weak evidence that inspiration increased evidence accumulation for both low ambiguity stimuli (happy and angry) and for angry high ambiguity stimuli, as well as a weak trend towards a decrease in evidence accumulation for happy high ambiguity stimuli, however, none of these effects passed standard evidentiary boundaries and should thus be interpreted with caution.

These effects are illustrated in Fig 3D. Additionally, we report 95% highest posterior density intervals (HPDIs) and P-values in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Respiratory Delta Posteriors.

https://doi.org/10.1371/journal.pcbi.1013086.t001

Discussion

Numerous studies have demonstrated a close linkage between the rhythms which govern brain, body, and behaviour [5,17,18,23]. While the influence of respiratory states on cognition is by now well established, so far little direct evidence has emerged to explain the computational mechanisms underlying these effects. Here, we confirmed previous findings that responses made during inspiration are associated with faster reaction times across visual and affective stimulus processing domains [7]. However, through computational modelling, we found that these surface level phenomena are explained by distinct respiratory modulation of latent decision variables. Our study sheds new light on how respiratory-brain interactions control behaviour and highlights the complexity of embodied modulation of exteroceptive perception.

In the context of visual motion discrimination, we found an inspiratory reduction of decision threshold. This indicates that responding during inspiration shifts decision-making to prioritise speed over accuracy. Previous studies have indicated that respiratory state modulates cortical motor readiness [9], cortico-spinal excitability [24] and cortico-muscular communication [25], all of which point towards a possible effect of motor system activation. As such, one interpretation of our result is that the respiratory state can shift the threshold of executing actions, leading to shifts in both the frequency of spontaneous actions, as shown by Park and colleagues [9], and in the amount of evidence needed to trigger a decision response, as our findings show. This suggestion is in line with recent developments in the DDM literature that increasingly posits motor areas not merely as effectors, subject to the output of higher order decision variables, but rather as important integrators of evidence, gating the decision and thereby contributing to the setting of evidence threshold [26,27]. On this basis, we suggest that respiration may alter motor readiness during visual processing.

In the context of discriminating between angry and happy faces, we found a unimodal increase in bias towards positive emotional judgments during inspiration. This selective modulation of affective decision bias by inspiration adds further nuance to the findings of Zelano et al. [8] and Mizuhara and Nittono [11], who reported modulation of affective processing by respiratory phase. In rodents, the respiratory rhythm has also been linked to the maintenance of fear-conditioned responses, suggesting a significant role for breathing in emotional regulation [10]. Our findings highlight a distinct mechanism through which respiration may bias evidence accumulation to bring about these and other changes in affective decision-making.

Numerous authors have hypothesised that breath-brain coupling effects may depend in part on the modulation of neural excitation and perceptual tuning [17,18,23,28]. On these grounds, one might have expected to observe robust modulation of evidence accumulation across stimulus domains. In contrast, here we report a complex pattern of effects with respect to drift rate, such that evidence accumulation is selectively enhanced or suppressed in a stimulus-specific manner. For example, for visual motion discrimination, we found that inspiration is associated with a decrease in the rate of evidence accumulation only for upwards motion stimuli, whereas drift was unaffected by respiratory state for downwards stimuli. For the affective modality, no significant changes were seen for any of the stimulus types. However, we observed trends towards increased evidence accumulation during inspiration for happy low ambiguity (HL) stimuli, as well as weak decreases in evidence accumulation for all other stimulus types. While these results should be interpreted with caution, they may indicate the possibility of stimulus specific neural tuning by the respiratory cycle across perceptual domains.

What neural mechanism could explain our findings? Breathing influences the brain through various pathways see [5] for review. Relevant examples include the brainstem mediated modulation of tonic noradrenaline [29,30], olfactory bulb propagated phase amplitude coupling [8,31–33], as well as respiratory modulation of affective and prefrontal circuits [34–37]. Other possible complementary mechanisms include respiratory coupled fluctuations in the physicochemical properties of the central nervous system, such as cardiorespiratory modulation of baroreceptors [38] and blood gas concentrations [39,40]. Given the complexity and heterogeneity of these pathways, future studies will need to directly investigate the neural mechanisms underlying respiratory modulation of decision dynamics. For example, one recent study found that the flexible modulation of decision bias depends upon alpha-band neuromodulation [41]. An intriguing avenue for future research could be to combine these approaches, for example by examining the neural correlates of respiratory shifts in affective bias, which could depend on a similar mechanism.

Our study has a few key limitations that could be addressed in future investigations. Participants viewed stimuli which were presented at peri-threshold levels. While this manipulation ensured that respiratory effects could be observed while controlling for stimulus-specific biases, it potentially introduces confounds related to drifting participant performance over time due to cognitive fatigue. While inspection of performance across blocks of trials suggested that performance was relatively stable, future studies could utilise adaptive thresholding, which would offer the further benefit of enabling estimation of respiratory-modulated psychophysics [17].

Another important limitation is that our focus on visual perception inherently restricts the generalizability of our findings. Future research could overcome this by employing varied experimental protocols and including a broader range of sensory modalities to enhance applicability across cognitive domains. While we here focus on the commonly reported effects concerning the binary treatment of the respiratory cycle, future studies could make use of more complex methods examining the link between continuous respiratory phase and behaviour.

Additionally, we did not collect eye tracking data. Prior work has shown that oculomotor behaviour, including microsaccades and saccades, is coupled to the cardiac cycle [42,43], which itself is modulated by respiration. It is therefore possible that our observed effects are partially mediated by breathing-induced changes in oculomotor rhythms. Future studies incorporating eye tracking will be needed to disentangle these contributions.

Finally, although our results suggest that breathing exerts distinct rhythms on visual and affective perception, comparable non-significant effects (e.g., bias in RDM and threshold in FAD; see Fig 3) were also observed. While we utilised a relatively large number of trials and participants for a psychophysical design, these effects might have reached significance with increased power. It may be that the overall magnitude—but not the qualitative nature—of these effects is differentially modulated. Future studies using a confirmatory design (e.g., registered reports) with enhanced within- and between-subject power could help resolve this issue.

More generally, understanding the respiratory modulation of the brain and behaviour ultimately requires a multi-modal and multi-organ perspective [See for example, 44–46]. In this study, we utilized computational modelling to explain how breathing influences choice, accuracy, and decision speed, extending and replicating previous findings. However, brain-body interactions encompass a complex array of physiological processes, including various neurotransmitters such as noradrenaline and autonomic signals from the heart, gut, and other visceral systems [43,47–49]. This complexity highlights a limitation of our current work and suggests directions for future research. Future studies could employ multimodal recordings — such as pupillometry, electrocardiograms, and electrogastrography — alongside respiratory measurements, thereby expanding computational models to account for both uni- and multimodal brain-body interactions. Indeed, recent findings on eye movements and pupillary dynamics [50], as well as animal studies demonstrating that interoceptive vagal pathways involve both multimodal and unimodal afferents [51], indicate that a comprehensive understanding of respiratory-brain coupling necessitates a multimodal approach. Note, however, that both unidirectional anatomical connections from the preBötzinger complex to LC [30] as well as recent findings from directed functional connectivity analyses [50] suggest respiration as the primary driver of both behavioural changes and arousal markers.

Finally, while our study builds on numerous studies suggesting a causal role for breathing in modulating the brain, future work could benefit from the application of respiratory manipulations to further elucidate the directionality of breathing on specific decision variables.

Methods

Analysis

Supporting information