After providing written informed consent, subjects were enrolled in the study. All subjects were healthy and professional speech and drama therapy practitioners. The practitioners had a personal interest in understanding the physiological mode of action underlying ATS and keen to aid in this research. The study protocol was approved by the Ethics Committee of the Canton Zurich (KEK-ZH-Nr. E-50/2002).

From originally ten subjects, one had to be excluded due to frequent ectopic heartbeats. The remaining nine subjects were included and had no history of cardiovascular diseases, especially no hypo- or hypertension. The nine subjects consisted of five female and four male volunteers, age 41 ± 7.6 years, BMI 22.8 ± 3.2 kg/m². All subjects were familiar with OM chanting.

Each subject took part in one experimental session of 30 min. The session consisted of a pre-resting period of 10 min (baseline), three sequences of five subsequent OM chantings with a 2 min pause between the sequences and a post-resting period of 10 min (recovery). All phases of the experiment were performed in standing position. During pauses between the OM chanting and during the resting periods the subjects breathed spontaneously. OM chanting was instructed to be full-toned at normal pitch and on one breath and the duration of each OM to be approximately of the same length within an individual subject. No pacing was introduced to allow for a natural interindividual variance in chanting pace.

During each session, the ECG (lead II) with automatic R-peak detection (sampling rate 128 Hz for ECG and 4,096 Hz for R-peaks) and the nasal/oral airflow (sampling rate 64 Hz) were recorded by a Holter device (MK-3 Medikorder, Tom Medical, Graz, Austria). The uncalibrated continuous blood pressure using a finger-cuff measurement device (Portapres; Finapres Medical Systems, Amsterdam, Netherlands, sampling rate 100 Hz) was simultaneously continuously recorded during the entire session.

The automatically detected R peaks were visually examined using both the ECG and the RR tachogram and manually corrected for ectopic beats. The beat-to-beat systolic blood pressure and a blood pressure based interbeat interval series (corresponding to the RR tachogram) were derived from the continuous blood pressure signal using dedicated software. The blood pressure derived interbeat interval series was visually inspected and manually corrected for outliers as well. Subsequently, ECG and respiration were time matched with the continuous blood pressure signal by matching their RR tachograms, after both had been resampled to 10 Hz. Respiratory cycles were derived from the nasal/oral airflow using successive local minima.

BRS was computed according to the sequence method (Bertinieri et al., 1988; Di Rienzo et al., 2001), hence x_SBP (n-1) was paired with x_RR (n), with x_SBP (n) and x_RR (n) being the nth sample of the systolic blood pressure beat-to-beat series and RR tachogram, respectively. Furthermore, the minimal number of elements in each BRS sequence amounted to 3, the minimal absolute difference in SBP between two consecutive samples was 1 mmHg and the minimal absolute difference in RR equalled 5 ms between two consecutive samples. Moreover, the minimal coherence in each sequence had to be at least 0.8. It was verified that a sufficient number of data elements were available for computing the BRS (between 15–121 for OM and the resting periods), so that the abovementioned method led to reliable statistical results (Gouveia et al., 2009). One median BRS value was computed for each subject during OM chanting (BRS_OM), pre-resting (BRS_PRE) and post-resting (BRS_POST).

Synchronisation between respiration, RR tachogram and beat-to-beat systolic blood pressure was analysed using the oscillations in heart period as well as in systolic blood pressure induced by respiration (respiratory sinus arrhythmia, RSA). To determine synchronisation between two signals, the cross-correlation function may be used. However, a parameter of synchronization γ has been proposed to better discriminate synchronisation during varying activity levels (Cysarz et al., 2004). This parameter denotes the amount of synchronisation between two signals, 0 ≤ γ ≤ 1, where signals that are fully synchronized in a statistical sense have a value of γ = 1 and signals that are completely desynchronized have a value of γ = 0. Without going into detail, this parameter was calculated by first performing the Hilbert transform of each of the two band-pass filtered signals, resulting in their phase signals as a function of time. Thereupon, the difference of these two-phase signals was analysed for variation over time (no variation implies full synchronisation), where phase jumps of 2π had to be corrected for. The measure of this variation over time is γ. A detailed description of this method is found in (Cysarz et al., 2004).

A flow diagram for computing synchronisation between RR and RESP during OM is depicted in Figure 1. It starts with finding the first harmonic of RR, f_0,RR, and applying a band-pass filter with centre frequency f_c = f_{0, RR} and bandwidth bw = (f_{0, RR} + f_{0, SBP} + f_{0, RESP})/3 to both RR and RESP. Next, for each of the 3 OM sequences synchronisation between the filtered RR and RESP signals is determined and its average value equals γ₁. For the second harmonic of RR, the same procedure is repeated, except that f_c = 2f_{0, RR}. This results in synchronisation γ₂. The average of γ₁ and γ₂ is γ_RR. Subsequently, the same procedure as for RR is applied to RESP as a reference, resulting in synchronisation γ_RESP. Synchronisation γ_OM between RR and RESP is taken as the maximum value of γ_RR and γ_RESP.

For computing synchronisation between RR and RESP during pre- and post-resting periods, we applied a sliding window of length T_OM, with 50% overlap to RR and RESP, where T_OM is the sum of the time lengths of the 3 OM sequences. To each pair of corresponding RR and RESP data window, the same procedure as during OM was applied for computing synchronisation, with two exceptions. First, we created 3 separate sequences by splitting the data in the sliding window in 3 non-overlapping windows of equal length, to proceed identically as with the 3 OM sequences. Secondly, because RR and RESP hardly showed harmonics during the resting periods, we took for the first spectral peak the largest in the frequency range between 0.0375 and 0.2 Hz (the first spectral peak during OM was located between 0.046 and 0.095 Hz). This resulted in synchronisations γ_PRE and γ_POST.

Using the method described above, one value per subject for γ_OM was obtained and a set of values for γ_PRE and γ_POST (depending on the length of the sliding data window 3–6 values each).

The combination RR and SBP and the combination RESP and SBP were analysed analogously. We therefore obtained for each subject: 3 values for γ_OM (1 value for RR-RESP, RR-SBP and RESP-SBP each), between 9 and 18 values for γ_PRE (3–6 values for RR-RESP, RR-SBP and RESP-SBP each) and between 9 and 18 values for γ_POST.

All statistical analysis was performed using non-parametric statistical procedures. Hence, median values and the interquartile range (IQR) were used to quantify the distributions. Differences in the RR interval, the respiratory cycle (RESP), BRS and synchronisation γ during pre-resting, OM and post-resting periods were analysed by the Friedman test. In case of significant difference, the Wilcoxon signed rank test was used to examine which of the three pairs (OM—pre-resting, OM—post-resting and pre-resting—post-resting) were significantly different, using the Bonferroni correction for multiple testing.

Since uncalibrated SBP values were recorded, i.e., each blood pressure measurement (subject) had a different offset, we performed a Wilcoxon signed rank test to SBP differences: we calculated for each subject ΔSBP_OM-PRE = SBP_OM—SBP_PRE, ΔSBP_OM-POST = SBP_OM—SBP_POST and ΔSBP_PRE-POST = SBP_PRE—SBP_POST, where SBP_PRE, SBP_OM and SBP_POST are the median systolic blood pressure during pre-resting OM and post-resting periods, respectively. To correct for multiple testing the Bonferroni correction was applied. A p-value < 0.05 was considered statistically significant for all tests.

Both for the respiratory cycle and the RR interval, the Friedman test showed considerable variation across pre-resting, OM and post-resting (p = 0.001 and p < 0.01, respectively). In the left panel of Figure 3, the respiratory cycle during pre-resting, OM and post-resting is shown; a large and significant difference was found between OM and both resting periods (>14.0 s), but not between both resting periods. For the RR interval, which is shown in the right panel of Figure 3, no significant differences were found among the three pairs after the Bonferroni correction.

Since SBP was not calibrated, we investigated differences in SBP. The difference in SBP between OM and pre-resting for a particular subject is ΔSBP = SBP_OM—SBP_PRE. The differences in SBP between OM and post-resting, and between pre-resting and post-resting are defined analogously. In the left panel of Figure 4, all three ΔSBP across all subjects are depicted. A distinct and significant difference in SBP was found between OM and both resting periods (>21.0 mmHg). In the two resting periods SBP was similar.

BRS showed significant variation across pre-resting, OM and post-resting (p < 0.01). BRS during pre-resting was significantly higher than during OM. Between OM and post-resting and between pre- and post-resting, no significant differences were found (right panel of Figure 4).

The Friedman test showed considerable variation across pre-resting, OM and post-resting for synchronisation γ of RR-RESP (p < 0.001), RR-SBP (p < 0.01) and RESP-SBP (p < 0.001). Synchronisation in all subjects during OM chanting and both resting periods is shown in Figure 5. Synchronisation during OM was always significantly stronger than during both pre-resting and post-resting periods (median values for pre-resting, OM and post-resting during RR-RESP: 0.41, 0.88, 0.39; RR-SBP: 0.59, 0.93, 0.57; RESP-SBP: 0.41, 0.87, 0.40). Synchronisation during pre-resting and post-resting was similar, which indicates that there was no carry-over effect of OM chanting to the post-resting period with respect to synchronisation between signals.

The time order of the signals was always RESP—SBP—RR. The median (IQR) phase differences amounted to ɸ_RESP-SBP = 0.26π (0.18π to 0.46 π) and ɸ_RESP-RR = 0.52π (0.38π to 0.59 π), which corresponded with a delay of 2.33 (1.47–3.27) seconds for SBP and of 3.84 (3.33–4.80) seconds for RR intervals.

During one OM chant, a clearly visible non-respiratory second oscillation appeared in the time signals of SBP and RR intervals (c.f. example in Figure 2), which seemed to depend on the breathing frequency. To quantify the amplitude of the second oscillation, its amplitude was calculated relative to the amplitude of its preceding oscillation. A relative amplitude close to 1 implies that the amplitude of the second oscillation was as high as its preceding oscillation. If the relative amplitude is close to 0 then the second oscillation is not visible, i.e. a second oscillation does not appear. Figure 6 shows the average relative amplitude of the second oscillation in the RR tachogram and SBP as a function of the first harmonic of the RR intervals, f_0,RR, and the first harmonic of the SBP, f_0,SBP, respectively (the first harmonics of the RR intervals, SBP and RESP were very similar during OM). A second oscillation with an amplitude almost as high as in the preceding oscillation (relative amplitude >0.7) appeared in the RR tachogram of three subjects. These oscillations occurred at frequencies of the first harmonics in the range 0.046–0.057 Hz. There were also three subjects that hardly showed a second oscillation in the RR tachogram. They showed relative amplitudes <0.1 and had frequencies of the first harmonic between 0.057 and 0.092 Hz. Furthermore, three subjects showed intermediate levels of the amplitude of the second oscillation (frequency range 0.05–0.073 Hz). For the SBP time series the amplitude of the second oscillation was in the intermediate range (relative amplitude >0.2 and <0.6, frequency range 0.048–0.075 Hz) for all subjects except one (at 0.094 Hz).

As in our preliminary study (von Bonin et al., 2001), a second oscillation appeared in the time signals of RR intervals and SBP during one OM chant, i.e., one respiratory-related oscillation, which was initiated by the inspiration (Penaz and Burianek, 1963), and another oscillation that obviously appeared without any trigger. If the breathing frequency elicited during OM chanting is at 0.05 Hz (i.e., 3 cycles per minute) the second oscillation in SBP and RR intervals is most distinct, indicating an oscillation at 0.1 Hz (Figure 6).

Which mechanisms have been reported to explain the behaviour of the RR and SBP oscillations observed? Slow breathing at 0.1 Hz in a resting condition (sitting or supine) leads to a “baroreflex resonance” called Traube-Hering waves (Vaschillo et al., 2002; Vaschillo et al., 2006; Billman, 2011; Sevoz-Couche and Laborde, 2022). I.e. a slow inspiration decreasing heart period will cause a surge in blood pressure due to the increasing amount of circulating blood. This response shows a time delay of approx. 5 seconds, caused by the inertia of the blood flow through the vascular system (Sevoz-Couche and Laborde, 2022). Hence, at a respiratory frequency of 0.1 Hz, a complete cycle of Traube Hering wave lasts approx. 10 s. As described, both oscillations are out of phase (phase difference approx. π). This baroreflex resonance is vagally mediated (Kromenacker et al., 2018; Sevoz-Couche and Laborde, 2022) and seems to involve pulse pressure modulations as well (Barnett et al., 2020).

Besides the Traube Hering waves, the so-called Mayer waves occur in the form of blood pressure oscillations at 0.1 Hz independent of respiration (Julien, 2006; Julien, 2020). Function and origin of the Mayer waves, with a central tendency at 0.1 Hz in humans, appearing in sympathetic nerve discharge, blood-pressure and HRV are still under debate. They may consist of two components, one originating in the baroreflex loop and the other in a centrally generated oscillation (Julien, 2006). The latter may be located in the pons/brain stem and drive cardiac autonomic oscillations with a short delay (Pfurtscheller et al., 2020). There is also evidence for a capacity of the spinal cord to generate such oscillations independent of supraspinal inputs (Inoue et al., 1991; Montano et al., 2000), which would lead to the assumption of a spinal source for Mayer waves (Koh et al., 1994). To complete these observations, an independent oscillation in the 0.15 Hz rhythm band in cutaneous blood flow has been described (Perlitz et al., 2004a; Perlitz et al., 2004b) and confirmed by recordings of similar blood-oxygen level-dependent oscillations (Pfurtscheller et al., 2019). If respiration rate came close to 0.15-Hz, a strong synchronisation of cardiovascular and respiratory oscillations occurred. However, simultaneous cardiovascular and respiratory oscillations at about 0.1 Hz did not affect the appearance of the 0.15-Hz rhythm in blood flow (Perlitz et al., 2004a; Perlitz et al., 2004b). Taken together, there is accumulating evidence for a neurogenic origin of cardiovascular oscillations at approx. 0.1 Hz as well as support for the hypothesis of the cardiovascular system generating these oscillations (Blinowska et al., 2020).

In the present study, Traube Hering waves are unlikely to generate the observed second non-respiratory related oscillation. The inspiration during OM chanting could have initiated Traube Hering waves. However, respiration and oscillations in heart period during OM chanting are not out of phase by π as they should be in the case of Traube Hering waves (Sevoz-Couche and Laborde, 2022). Instead, the standing position during OM chanting will elicit orthostatic stress and, hence increase sympathetic activity (Cooke et al., 1999). This will result in increasing blood pressure activity during OM chanting and may enhance the appearance of Mayer waves.

However, the present data cannot answer this question unambiguously. Previous results indicate that the cardiovascular response on respiration at 0.05 Hz is different to respiration at 0.1 Hz (Vaschillo et al., 2002). Hence, studies investigating respiration at 0.05 Hz with different inspiration to expiration ratios would be needed to further clarify the origin of the second non-respiratory related oscillation observed.

The non-respiratory oscillations at 0.1 Hz in SBP and RR interval in our experiment occur predominantly at a breathing frequency half of the Mayer waves (1:2 ratio). At slightly higher respiration frequencies (close to but still below 0.1 Hz) the second non-respiratory oscillation attenuates. Hence, the second non-respiratory oscillation does not originate from entrainment with e.g. blood pressure oscillations. Such an entrainment would lead to enhanced oscillations instead of diminished oscillations. Instead, it seems that the second oscillation only appears if there is enough time for a complete oscillation. It has been suggested that the human organism tends to establish a coordinated time structure of physiological functions such as heart period and respiration (Hildebrandt, 1986; Hildebrandt, 1991). This time structure is reflected by e.g. whole number frequency coordination of their oscillations. This endogenous time structure prefers integer number relationships in the spontaneous rhythms (Hildebrandt, 1986; Hildebrandt, 1991). In line with this assumption, most of our study subjects preferred a breathing frequency for OM chanting with an integer number ratio (1:2) to the Mayer waves (e.g. the study subject in Figure 2). Where this ratio was not maintained (faster breathing), the second oscillation disappeared. We do not expect confounding effects of varying stress levels since all speech practitioners were experienced and deliberately chose the individual frequency of OM chanting.

The emergence of Mayer waves may also have contributed to the higher statistical dependence between RR and SBP compared to RESP and both SBP and RR during resting: RESP is largely independent of both SBP and RR, whereas SBP and RR are much more correlated. Moreover, we often observed a main frequency component of SBP and RR of approximately 0.1 Hz which explains the larger value for both γ_PRE and γ_POST in RR-SBP as compared to RR-RESP and RESP-SBP (Figure 5).

The time order of the oscillations is always RESP-SBP-RR if the second oscillation in SBP and RR (Figure 1) is used to establish this time order, i.e., the single oscillation in respiration always precedes the second oscillation in both SBP and RR. Physiologically, the speech induced respiration triggers an increase in blood pressure and heart period oscillations (respiratory sinus arrhythmia) which, in the case of slow breathing at approximately 0.05 Hz, is complemented by a second oscillation of SBP and RR of approximately 0.1 Hz. In the present study, we were able to demonstrate this effect but not to establish its origin.

The subjects showed slightly varying patterns of OM chanting leading to partially different responses on cardiovascular regulation. Future studies should use a stricter OM chanting pattern to reduce the variance in the responses on the cardiovascular regulation. Furthermore, the possibility of gender differences in cardiovascular responses (Voss et al., 2015) should be considered in future studies. The lack of calibration of the blood pressure signal permitted only the analysis of SBP changes. Absolute changes in SBP would give further insight in blood pressure regulation during OM chanting. Furthermore, partial pressure of O₂ and CO₂ were not measured and, hence, it is not known whether these OM chanting patterns, i.e., low respiratory frequencies, lead to hypoxia or hypercapnia which also may affect cardiovascular regulation.