Subjects and Meditation Protocols
Two specific meditative techniques were studied: (i) Chinese Chi (or Qigong) meditation (as taught by Xin Yan) and (ii) Kundalini Yoga meditation (as taught by Yogi Bhajan).
The Chi meditators were all graduate and post-doctoral students. They were also relative novices in their practice of Chi meditation, most of them having begun their meditation practice about 1-3 months before this study. The Kundalini Yoga subjects were considered to be at an advanced level of meditation training. The subjects of both meditation groups were in good general health and did not follow any specific exercise routines. All subjects provided informed consent in accord with a protocol approved by the Beth Israel Deaconess Medical Center Institutional Review Board.
The eight Chi meditators, 5 women and 3 men (age range 26-35, mean 29 yrs), wore a Holter recorder for approximately 10 hours during which time they went about their ordinary daily activities. At approximately 5 hours into the recording they each practiced one hour of meditation. Meditation beginning and ending times were delineated with event marks.
During these sessions, the Chi meditators sat quietly, listening to the taped guidance of the Master. The meditators were instructed to breath spontaneously while visualizing the opening and closing of a perfect lotus in the stomach. The meditation session lasted about one hour.
The four Kundalini Yoga meditators, 2 women and 2 men (age range 20-52, mean 33 yrs), wore a Holter monitor for approximately one and half hours. 15 minutes of baseline quiet breathing were recorded before the 1 hour of meditation. The meditation protocol consisted of a sequence of breathing and chanting exercises, performed while seated in a cross-legged posture. The beginning and ending of the various meditation sub-phases were delineated with event marks.
In addition to comparing the pre-meditation and meditation states, we also made comparisons to three healthy, non-meditating control groups from a database of retrospective electrocardiogram (ECG) signals: (i) A spontaneously breathing group of 11 healthy subjects (8 women and 3 men; age range 20-35, mean 29) during sleeping hours. (ii) A healthy group of 14 subjects (9 women and 5 men; age range 20-35, mean 25) during supine metronomic breathing at 0.25 Hz. (iii) A group of 9 elite triathlon athletes in their pre-race period (3 women and 6 men; age range 21-55, mean 39) during sleeping hours. Except for exercise training in the triathlon athlete group, the overall general health conditions for the meditation groups and control groups were comparable.
Signal Processing and Data Analysis
The Holter tapes were scanned and annotated using a Marquette Electronics Model 8000T Holter scanner and annotations manually verified. The resulting annotation files were then transferred to a Sparc workstation for further analysis. A small fraction (< 1%) of the instantaneous RR interval heart rate time series for each recording was identified as outliers and deleted. Instantaneous heart rate time series were then derived by taking the inverse of each successive interbeat interval.
We applied an ECG-derived respiration algorithm [4,5] to obtain information about the frequency and relative amplitude of respiration. Briefly, this technique is based on the observation that the body surface ECG is influenced by electrode motion relative to the heart and by changes in thoracic electrical impedance as the lungs fill and empty. Measurement of axis shifts at each normal QRS interval provide a continuous ECG-derived respiration signal. The relation between this signal and respiration has been confirmed by comparing the changes in axis direction with simultaneous measurements of chest circumference taken with a mercury strain gauge or pneumatic respiration transducer . We cross-correlated the heart rate time series with the ECG-derived respiration signal. In particular, we uniformly resampled both the instantaneous heart rate and the ECG-derived respiration time series at 2 Hz, then calculated the coherence  of these two signals.
To quantify the amplitudes of heart rate oscillations during meditation and to compare them with those under usual basal conditions, two independent quantitative algorithms have been applied:
(1) We calculated Fourier spectral power by applying the Lomb periodogram method for unevenly sampled data . Spectral power was measured in the frequency range 0.025-0.35 Hz to ensure that all respiration related heart rate oscillations would be included.
(2) We also used a Hilbert transform-based algorithm [7,8]. The advantages of using the Hilbert transform are two-fold: (i) it does not require stationarity of the signal; and (ii) it measures the amplitude and frequency of the dominant oscillation in the signal at each moment. However, since the Hilbert transform can be applied only on narrow band signals, the heart rate time series has to be pre-processed. First, the heart rate time series signal was bandpass filtered over the same frequency range (0.025 to 0.35 Hz) studied in the Fourier analysis. Next, a Hilbert transformation was performed on the filtered signal. Thus for each subject's heart rate time series, we obtained a sequence of amplitudes describing the time-dependent magnitude of the oscillations.
We then calculated the median value, Am, of the oscillation amplitude obtained by the Hilbert transform for each subject. The median value is a robust measurement even when a substantial number of outliers are present in the data. The Hilbert transform and median amplitude procedure described here can be applied to time series with an arbitrary number of data points. Therefore, the results can be compared among subjects with data sets of different lengths.
To determine the effect of meditation on oscillation amplitude and spectral power in the meditators, values measured before and during meditation were compared using a paired t-test. The Student t-test was used to compare values obtained during meditation to those obtained from each control group. A p value of less than 0.01 (two-sided) was used as the level of significance for rejecting the null hypothesis that values measured during meditation were similar to those obtained outside of meditation. Since the number of subjects in each meditation group was small, we pooled these subjects (n=12) for the comparisons with the control groups. Statistical analysis was performed using SAS software release 6.12 (Cary, North Carolina). Results are reported as mean standard deviation.