Breathing is coupled with voluntary action and the cortical readiness potential

The study addresses whether interoceptive bodily signals, particularly respiration, influence the timing of voluntary actions and the cortical readiness potential (RP) that precedes self-initiated movement. Classical interpretations view the RP as a preparatory neural process and even an unconscious precursor to intention, whereas recent accounts propose it reflects spontaneous fluctuations in background neuronal activity. Building on evidence that interoceptive signals (respiratory and cardiac) modulate sensory processing and ongoing neural activity and that breathing control engages motor areas implicated in RP generation (e.g., SMA), the authors hypothesized that respiration phases might couple with voluntary action onset and modulate RP amplitude. They further contrasted this with cardiac phase effects and tested specificity relative to externally triggered actions.

Foundational work established the RP as a slow negative cortical potential preceding voluntary action and suggested it precedes conscious intention (Kornhuber & Deecke; Libet). More recent models (e.g., Schurger’s accumulator model) argue RP reflects stochastic fluctuations of ongoing neural activity rather than specific motor preparation. Interoceptive signals have been shown to shape perception and behavior, including visual detection and emotional processing, and to contribute to ongoing brain activity at rest. Breathing is linked anatomically and functionally to orofacial actions and is coupled with locomotion and whisking in animals. The SMA and motor regions are involved in both breathing control and voluntary action, raising the possibility that respiratory cycles could structure the fluctuations that give rise to RP and the timing of self-initiated movements. Prior human studies often treated respiration as physiological noise; few examined its role in voluntary action timing.

Participants: Experiment 1 included 20 healthy adults (10 female; all right-handed; mean age 26 ± 1.3 years). Experiments 2 and 3 included 34 participants (15 female; 31 right-handed; mean age 26.5 ± 5.1 years); 2 were excluded for excessive artifacts (>50% of respiration and EEG), 1 was excluded from cardiac analysis for noisy ECG, and 2 from resting-state analyses for missing triggers. Ethics approval and informed consent were obtained.

Tasks:

• Experiment 1 (Kornhuber task): Participants made self-initiated right index finger button presses roughly every 8–12 s (3 blocks × 8 min), avoiding counting and rhythm to maximize spontaneity. Eyes closed with white noise; first trials per block excluded (<3%). Resting-state EEG (3 min) followed.
• Experiment 2 (Libet task): A red dot rotated on a clock (2560 ms per cycle). After at least one full rotation, participants pressed a button at a time of their choosing and later reported the time of first intention (W-time). 3 blocks (75 trials total; one participant 90 trials). Resting-state EEG (5 min) after Experiments 2–3.
• Experiment 3 (Externally triggered): Participants pressed as quickly as possible upon detecting a brief (200 ms) green dot at fixation while the red dot rotated. Green dot timing was individualized from Experiment 2 performance. Same block structure and ITIs as Experiment 2.

Recordings:

• Respiration: Belt transducer (Biopac MP36), 2000 Hz sampling. Bandpass 0.2–0.8 Hz; instantaneous phase estimated via Hilbert transform; inspiration peaks detected by template correlation. For phase-amplitude coupling analyses, downsampled to 512 Hz.
• EEG: 64-channel Biosemi ActiveTwo; 2048 Hz sampling; online low-pass 400 Hz. Offline downsample to 512 Hz; bandpass 0.1–40 Hz; re-referenced to common average. Epochs −4 to +1 s around movement onset; artifact rejection >3 SD. RP computed from fronto-central electrodes (Cz, FCz, Fz, AFz), selecting per-subject the electrode with largest RP; no baseline correction. Average usable epochs: 118 ± 24 (Exp 1) and 68 ± 4 (Exp 2).
• ECG: Bipolar electrodes (right shoulder, left abdomen); same preprocessing. R-peaks detected by template correlation; cardiac phase computed as ϕ(t) = 2π((t − ta)/(tb − ta)) using surrounding R-peaks; Hilbert not used due to non-oscillatory ECG morphology.

Analyses:

• Breathing–action coupling: At each button press, compute respiration phase. First, apply Hodges–Ajne (omnibus) test of circular uniformity per subject to obtain M (minimum number in any semicircle). Second, to account for longer expiration duration inflating chance expiratory counts, generate surrogate phase distributions by random phase shifts (cut and swap) of respiration per block/subject (1000 permutations) to build a null distribution of summed M across participants; two-sided permutation p-value.
• Cardiac phase–action coupling: Same circular uniformity testing using ECG-derived phases.
• RP–respiration phase-amplitude coupling: Combine EEG–respiration trials from Experiments 1 and 2. For each trial in the −4 to 0 s window, average RP amplitudes within six equal respiration phase bins (0–2π). Normalize per trial by the sum across bins. Compute Modulation Index (MI) quantifying deviation from uniform distribution across the six bins per participant; assess significance by comparing the grand-averaged MI to a null distribution from randomly phase-shifted respiration (1000 permutations; two-sided p-value).
• Resting-state control: Epoch EEG −2 to +2 s around inspiration peaks; compute respiration-phase–amplitude MI using the same procedure to test for respiration-related artefacts in resting EEG.
• Behavioral measures: Waiting times (intervals between presses) and W-times in the Libet task.

Questionnaires: After Experiments 1 and 2, participants reported awareness or use of breathing/heartbeat in timing their button presses (Q1/Q2).

• Voluntary action is coupled to respiration phase. Participants pressed more often during expiration, particularly late expiration just before inspiration. • Experiment 1 (Kornhuber task): N=20; mean interval between presses 11.20 ± 2.30 s; SD of intervals 3.26 ± 1.70 s. Respiration phase at press was concentrated in expiration (permutation p = 0.0009). Mean respiration phase fell within expiration for 19/20 participants. • Experiment 2 (Libet task): N=32; mean waiting time 6.67 ± 1.52 s; SD 2.20 ± 0.97 s; mean W-time −0.26 ± 0.17 s. Replicated expiratory coupling (permutation p = 0.0009); 30/32 participants’ mean press phase in expiration. • Awareness: 18/20 (Exp 1) and 31/32 (Exp 2) reported no awareness or use of breathing/heartbeat in their timing.
• No respiration–action coupling for externally triggered actions. Experiment 3 showed no significant association between respiration phase and button presses (permutation p = 0.32).
• Cardiac phase is not associated with voluntary action onset. Across experiments, circular uniformity tests for ECG phase at press were non-significant (e.g., Exp 1 p = 0.80 for all presses; Exp 2 p = 0.24; Exp 3 p = 0.78), indicating presses were not locked to cardiac phase.
• RP amplitude is modulated by respiration phase. The RP recorded over fronto-central electrodes began ~2 s prior to movement. RP amplitude varied with respiration phase, being smaller during expiration than inspiration. Phase–amplitude coupling quantified by MI was significant relative to phase-shifted surrogates (permutation p = 0.0009).
• Resting EEG amplitude is not modulated by respiration phase. Respiration-locked resting-state EEG over the same electrodes showed no significant MI relative to surrogates (p = 0.60), arguing against respiration artefacts driving the RP–respiration coupling.

The findings demonstrate a systematic coupling between the respiratory cycle and self-initiated actions: participants preferentially initiated voluntary movements during expiration, an effect absent for externally triggered actions and not observed for cardiac phase. Converging EEG evidence shows that the RP amplitude varies with respiration phase, being reduced during expiration, indicating that interoceptive respiratory dynamics modulate the neural processes leading up to voluntary movement. These results support models positing that the RP reflects fluctuations of ongoing neural activity rather than a dedicated preparatory motor program, and identify respiration as a structured physiological source of such fluctuations. The specificity to voluntary (vs externally triggered) actions suggests that interoceptive–motor coupling relates to endogenous action selection/timing mechanisms. Control analyses of resting EEG argue against generic respiration artefacts explaining the effect. The authors propose that coupling may minimize competition between respiratory motor commands and voluntary motor outputs at premotor (e.g., SMA) and subcortical levels, and highlight candidate networks (insula, cingulate, ventrolateral medulla) for future mechanistic studies.

Spontaneous breathing modulates when people choose to act and the cortical RP that precedes voluntary movements. By linking interoceptive respiratory cycles to voluntary action timing and RP amplitude, the study bridges accounts that attribute the RP to ongoing neuronal fluctuations with evidence that respiration is a major source of such fluctuations. The results challenge interpretations of the RP as an unconscious initiator of voluntary action, suggesting it at least partly reflects respiration-related cortical processing coupled to action onset. Future work should causally manipulate respiration (e.g., nasal vs oral breathing, loaded breathing), combine EEG with fMRI or invasive recordings to localize mechanisms, and examine individual differences (e.g., athletes, breathing awareness) to detail how respiratory control interfaces with voluntary action and cognition.

The study is correlational; respiration was not experimentally manipulated to establish causality for action timing or RP modulation. While permutation controls accounted for longer expiration duration and resting-state analyses argued against artefactual respiration effects on EEG, unmeasured confounds cannot be fully excluded. Neural sources of the coupling were not localized; proposed mechanisms (e.g., competition between motor commands at SMA or subcortical nuclei) remain speculative. Generalizability beyond the specific button-press tasks and healthy young adults requires further study, as do potential differences between nasal and oral breathing and across populations.