Alterations in Brain and Immune Function Produced by Mindfulness
Meditation
RICHARD J. DAVIDSON,PHD, JON KABAT-ZINN,PHD, JESSICA SCHUMACHER, MS, MELISSA ROSENKRANZ, BA,
DANIEL MULLER, MD, PHD, SAKI F. SANTORELLI,EDD, FERRIS URBANOWSKI, MA, ANNE HARRINGTON,PHD,
KATHERINE BONUS,MA,AND JOHN F. SHERIDAN,PHD
Objective: The underlying changes in biological processes that are associated with reported changes in mental and physical health
in response to meditation have not been systematically explored. We performed a randomized, controlled study on the effects on
brain and immune function of a well-known and widely used 8-week clinical training program in mindfulness meditation applied
in a work environment with healthy employees. Methods: We measured brain electrical activity before and immediately after, and
then 4 months after an 8-week training program in mindfulness meditation. Twenty-five subjects were tested in the meditation
group. A wait-list control group (N ⫽ 16) was tested at the same points in time as the meditators. At the end of the 8-week period,
subjects in both groups were vaccinated with influenza vaccine. Results: We report for the first time significant increases in
left-sided anterior activation, a pattern previously associated with positive affect, in the meditators compared with the nonmedi-
tators. We also found significant increases in antibody titers to influenza vaccine among subjects in the meditation compared with
those in the wait-list control group. Finally, the magnitude of increase in left-sided activation predicted the magnitude of antibody
titer rise to the vaccine. Conclusions: These findings demonstrate that a short program in mindfulness meditation produces
demonstrable effects on brain and immune function. These findings suggest that meditation may change brain and immune function
in positive ways and underscore the need for additional research. Key words: meditation, mindfulness, EEG, immune function,
brain asymmetry, influenza vaccine
HIV ⫽ human immunodeficiency virus; NK ⫽ natural killer cell;
EEG ⫽ electroencephalography; EOG ⫽ electrooculography; PA-
NAS ⫽ Positive and Negative Affective Scale; MBSR ⫽ mindful-
ness-based stress reduction; MANOVA ⫽ multivariate analysis of
variance.
INTRODUCTION
W
ith the widespread and growing use of meditative prac-
tices in hospitals and academic medical centers for
outpatients presenting with a range of chronic stress and
pain-related disorders and chronic diseases, under the um-
brella of what has come to be called mind/body or integrative
medicine, the question of possible biological mechanisms by
which meditation may affect somatic, cognitive, and affective
processes becomes increasingly important. Research on the
biological concomitants of meditation practice is sparse and
has mostly focused on changes that occur during a period of
meditation compared with a resting control condition in a
single experimental session (1–3). Whereas these studies have
been informative, they tell us little about changes that are
potentially more enduring. Moreover, virtually all forms of
meditation profess to alter everyday behavior, effects that are
by definition not restricted to the times during which formal
meditation itself is practiced. Thus, in the current report, we
focus not on the period of meditation itself, but rather on the
more enduring changes that can be detected in baseline brain
function as well as brain activity in response to specific
emotional challenges.
We focus on emotion-related brain activity because medi-
tation has been found in numerous studies to reduce anxiety
and increase positive affect (4 – 8). In an extensive corpus of
work on the functional neuroanatomical substrates of emotion
and affective style, we have established that the frontal regions
of the brain exhibit a specialization for certain forms of
positive and negative emotion (9, 10). Left-sided activation in
several anterior regions is observed during certain forms of
positive emotion and in subjects with more dispositional pos-
itive affect (10, 11). We therefore hypothesized that because
meditation decreases anxiety and increases positive affect,
subjects who were practicing meditation should show in-
creased left-sided activation in these territories compared with
those in a wait-list control group.
Recent studies have established that greater relative left-
sided anterior activation at baseline is associated with en-
hanced immune function using measures of NK activity (12,
13). There has been a paucity of serious research attention to
possible immune alterations that might be produced by med-
itation (14). This is somewhat surprising in light of the fact
that negative psychosocial influences on immunity have now
been well established (15–17). Recent research indicates that
relaxation and stress management procedures increase T-cy-
totoxic/suppressor (CD3⫹CD) lymphocytes in HIV-infected
men (18). On the basis of recent research demonstrating the
negative impact of stressful life events on antibody titers in
response to influenza vaccine (19), we vaccinated all subjects
at the end of the 8-week meditation program (in mid Novem-
ber), along with the subjects in wait-list control group at the
same time. We hypothesized that the meditators would show
greater antibody titers in response to the vaccine compared
with the subjects in the wait-list control group. On the basis of
From Laboratory for Affective Neuroscience (R.J.D., J.S., M.R.), Depart-
ment of Psychology, University of Wisconsin, Madison, Wisconsin; Stress
Reduction Clinic, Division of Preventive and Behavioral Medicine (J.K.-Z.,
S.F.S., F.U.), Department of Medicine, University of Massachusetts Medical
School, Worcester, Massachusetts; Departments of Medicine and Microbiol-
ogy (D.M.), University of Wisconsin Medical School; Department of the
History of Science (A.H.), Harvard University, Cambridge, Massachusetts;
Departments of Preventive Cardiology and Sports Medicine (K.B.), Univer-
sity of Wisconsin-Madison Hospitals and Clinics Center for Mindfulness,
Madison, Wisconsin; and Department of Oral Biology (J.F.S.), College of
Dentistry, Ohio State University, Columbus, Ohio.
Address reprint requests to: Richard J. Davidson, PhD, Laboratory for
Affective Neuroscience, University of Wisconsin, 1202 W. Johnson St.,
Madison, WI 53706. Email: rjdavids@facstaff.wisc.edu
Received for publication April 4, 2002; revision received December 27,
2002.
DOI: 10.1097/01.PSY.0000077505.67574.E3
564 Psychosomatic Medicine 65:564 –570 (2003)
0033-3174/03/6504-0564
Copyright © 2003 by the American Psychosomatic Society
the association we have previously reported between anterior
activation asymmetry and NK activity, we also predicted that
the magnitude of change toward greater relative left-sided
activation would be associated with a larger increase in anti-
body titers in response to the vaccine.
METHODS
Measures of brain electrical activity were recorded before random assign-
ment to each of the two groups (Time 1) and then again immediately after
(Time 2) and four months after (Time 3) the training period ended. Brain
electrical activity, or EEG, and EOG (for correcting EEG for eye movements)
was recorded during both baseline conditions and in response to a positive and
negative emotion induction using methods that have been extensively de-
scribed in previous research (20, 21). EEG was recorded from 27 sites
distributed across the scalp and referenced to linked ears during 8 1-minute
baseline trials, four with eyes open and four with eyes closed, presented in
counterbalanced order according to our established procedures (22). EEG was
also recorded during a 1-minute period before and a 3-minute period after
subjects wrote about one of three of the most positive and negative experi-
ences in their life. These events were listed on a questionnaire administered to
subjects before the start of the entire protocol. For this task, the EEG was
aggregated across the 1-minute period before and the 3-minute period after the
writing itself. Data were not collected during writing because of movement
artifact. The EEG was parsed into 1.024-second epochs, overlapped by 50%
and then processed with the use of a fast Hartley transform method to derive
measures of spectral power density in the
␣
-band (8 –13 Hz), which is
inversely related to activation (20, 22). Asymmetric activation was indexed
using an asymmetry score that is computed by subtracting log-transformed
left hemisphere
␣
-power densities from the comparable measure derived from
homologous right-sided electrodes.
After each of the writing periods, subjects were given the PANAS (23) in
state form. In addition, at each assessment, they were administered the
PANAS in trait form, along with the Spielberger State-Trait Anxiety Inven-
tory (24) in trait form. In addition, subjects in the meditation group were asked
to provide daily reports of the frequency and number of minutes and tech-
niques of formal meditation practice.
Blood draws were then obtained at 3 to 5 weeks and then again at 8 to 9
weeks after vaccination to examine antibody titers in response to the vaccine
using the hemagglutination inhibition assay (19).
A total of 48 right-handed subjects who were employees of a biotechnol-
ogy corporation in Madison, Wisconsin, were recruited to participate. Of
these, 41 subjects completed some of the measures for at least two of the
assessments. The initial laboratory evaluation was conducted before random
group assignment. Subjects were then randomly assigned to the meditation
group (N ⫽ 25; 19 female) and the wait-list control group (N ⫽ 16; 10 female)
at a ratio of approximately 3:2. There were no differences between groups in
the number of subjects who failed to complete the study. Average age of
subjects was 36 years and did not differ between group (range ⫽ 23 to 56
years). All but two subjects were white (one Asian-American in the treatment
group; one South Asian Indian in the control group). Subjects in the wait-list
control group were evaluated at each assessment period along with subjects in
the meditation group. After completion of the last assessment, the wait-list
control subjects were provided with an 8-week training program comparable
to that provided to the subjects in the meditation group.
The meditation training (known as MBSR) was delivered by J.K.-Z., and
was directly modeled on the MBSR intervention originally developed at the
University of Massachusetts Medical Center (25, 26). The effects of MBSR
have been reported in numerous clinical studies with diverse populations, as
well as in medical students (27, 28). One study demonstrated significant
effects of mindfulness on the rate of skin clearing in patients with moderate
to severe psoriasis (29) Two recent reviews of MBSR research called for
studies to elucidate potential mechanisms of action (30, 31).
The training consisted of a class that met weekly for 2.5 to 3 hours per
class, along with a silent seven-hour retreat that was held during week 6 of the
course. In addition, subjects were assigned home practice that consisted of
formal and informal meditative practices that they were instructed to perform
for 1 hour per day, 6 days per week, with the aid of guided audiotapes.
The statistical analysis of the data focused on the interactions between
group (Meditation/Wait-list control) and time (Times 1–3, with the first
assessment occurring before the intervention, Time 2 occurring immediately
after the 8-week intervention and Time 3 occurring four months after the
training period ended. MANOVAs were computed for each of the four
anterior asymmetry measures. In addition to examining main effects and the
interaction, linear trends were also tested. Follow-up ANOVAs on the sepa-
rate time periods were performed.
RESULTS
Affect and Anxiety Measures
We evaluated self-report measures of positive and negative
affect and anxiety before and after the training. There was a
significant Group ⫻ Time interaction [F(1,31) ⫽ 5.45, p ⬍
.05] on a measure of trait anxiety, the Spielberger State-Trait
Anxiety Inventory (24), accounted for by a reduction in anx-
iety for subjects in the meditation group from Time 1 to Time
2[t (20) ⫽ 2.86, p ⬍ .01; see Figure 1). There was no
significant Group ⫻ Time interaction on the Positive and
Negative Affect Scale (23). However, in light of the clear a
priori predictions for the meditators to show significant de-
creases in negative affect with treatment, we tested change
over time within each group. There was a significant decrease
in trait negative affect with the mediators showing less nega-
tive affect at Times 2 and 3 compared with their negative
affect at Time 1 [t (20) ⫽ 2.27 and t (21) ⫽ 2.45, respectively,
p ⬍ .05 for both; not shown]. Subjects in the control group
showed no change over time in negative affect (t ⬍ 1).
Brain Electrical Activity Measures
Based on previous findings linking asymmetric anterior
activation to positive affect, we specifically examined changes
in four anterior electrode sites (F3/4, FC7/8, T3/4, and C3/4 in
the International 10/20 system) during both base-line periods
and in response to the emotion inductions. We computed
MANOVAs with Group and Time (Times 1–3) as factors and
examined main effects and interactions, as well as linear
Fig. 1. Mean trait anxiety from the Spielberger State-Trait Anxiety Inven-
tory (24) measured separately by group and time. Error bars reflect
means ⫾ SE.
BRAIN AND IMMUNE FUNCTION IN MEDITATION
565Psychosomatic Medicine 65:564 –570 (2003)
trends. For the baseline period assessments, there was a mar-
ginally significant Group ⫻ Time linear trend (F(1,33) ⫽
3.73, p ⫽ .06) and a significant main effect for Group (across
time periods; F(1,33) ⫽ 4.57, p ⫽ .04). When the comparison
of change from baseline for each time period was examined,
there was a significant Group ⫻ Time interaction [F(1,37) ⫽
5.14, p ⬍ .05] for the Time 1–3 comparison and a marginally
significant Group ⫻ Time interaction [F(1,33) ⫽ 2.82, p ⫽
.10] for the Time 1-Time 2 comparison for the central leads
(C3/4). At Time 1, no group differences were present at
baseline for any region. At both Time 2 and Time 3, medita-
tors showed significantly greater relative left-sided activation
at the central sites (C3/4) compared with the wait-list control
group (p ⬍ .05 for each; see Figure 2).
The omnibus MANOVA performed on the positive emo-
tion induction condition revealed a marginally significant
overall Group ⫻ Time interaction [F(2,26) ⫽ 2.52, p ⫽ .10]
for the anterior temporal (T3/4) electrode leads. When the
comparison of change from baseline for each time period was
examined, there was a significant Group ⫻ Time interaction
[F(1,30) ⫽ 4.82, p ⬍ .05] for the Time 1-Time 2 comparison.
This same interaction for the Time 1-Time 3 comparison was
marginally significant and in the same direction [F(1,29) ⫽
3.46, p ⫽ .07]. In response to the positive emotion induction
at Time 1, no group differences were present in any region.
However, meditators showed a significant increase in left-
sided anterior temporal activation from Time 1 to Time 2 (p ⬍
.05), whereas controls showed no change (Figure 3). There
were no other significant Group ⫻ Time interactions for any
other electrode site for the positive emotion induction.
In response to the negative affect induction, the omnibus
MANOVA revealed a marginally significant linear trend
for the Group ⫻ Time interaction [F(1,27) ⫽ 2.94, p ⬍ .10]
for the anterior temporal leads. The Group ⫻ Time inter-
action for the Time 1-Time 2 comparison for the anterior
temporal region (T3/4) was again in the same direction as
the other interactions, but not significant [F(1,31) ⫽ 3.16,
p ⫽ .08].
In response to the negative emotion induction for the cen-
tral leads, an omnibus MANOVA revealed a marginally sig-
nificant Group ⫻ Time interaction [F(2,32) ⫽ 2.78, p ⬍ .08],
along with a marginally significant linear trend for this inter-
action [F(1,33) ⫽ 3.45, p ⫽ .07]. In addition, there was a
significant main effect for Group [F(1,33) ⫽ 6.78, p ⫽ .01].
For the central leads, the Group ⫻ Time interaction for the
Time 1-Time 2 comparison was F(1,33) ⫽ 3.62, p ⫽ .07, and
for the Time 1-Time 3 comparison it was F(1,37) ⫽ 5.41, p ⬍
.05. Again, there were no group differences in any region at
Time 1. At Times 2 and 3, subjects in the meditation group
showed significantly greater left-sided activation (C3/C4)
compared with subjects in the control group (for Time 2: p ⬍
.05; for Time 3: p ⬍ .01). The meditators evinced a significant
increase in left-sided activation in this region from Time 1 to
Time 2 (p ⬍ .05; not shown) and Time 3 (p ⬍ .05: Figure 4).
There were no group differences present for any of the
posterior electrodes sites for any of the conditions.
Influenza Vaccine Antibody Titers
In response to the influenza vaccine, the meditators dis-
played a significantly greater rise in antibody titers from the 4
to the 8 week blood draw compared with the controls [t(33) ⫽
2.05, p ⬍ .05; Figure 5].
Fig. 2. Means ⫾ SE of asymmetric activation during baseline for subjects in the Meditation group and Control group during Time 1 (before random assignment,
before treatment began) and Time 3. The ordinate is an asymmetric metric that represents right minus left log-transformed
␣
power density from the
C4/C3 electrode sites. This is a standard index of asymmetric activation (20). Higher numbers on this indicate greater left-sided activation.
R. J. DAVIDSON et al.
566 Psychosomatic Medicine 65:564 –570 (2003)
Relations Among Measures
To examine the relation between the magnitude of increase
in left anterior activation and the magnitude of antibody titer
rise in response to the influenza vaccine from the 4- to 8-week
blood draw, we computed a change score for each subject to
express the change in activation asymmetry from Time 1 to
Times 2 and 3 and correlated the change in activation asym-
metry with the rise in antibody titers, separately for each
group. Among subjects in the meditation group, those who
showed a greater increase in left-sided activation from Time 1
Fig. 3. Means ⫾ SE asymmetric activation (in the T3/T4 electrode sites) in response to the positive emotion induction in the Meditation group and Control
group during Times 1 and 2. The ordinate is the same metric of asymmetric activation displayed in Figure 2.
Fig. 4. Means ⫾ SE asymmetric activation in response to the negative emotion induction in the Meditation group and Control group during Times 1 and 3.
The ordinate is the same metric of asymmetric activation displayed in Figure 2 (C3/C4).
BRAIN AND IMMUNE FUNCTION IN MEDITATION
567Psychosomatic Medicine 65:564 –570 (2003)
to Time 2 displayed a larger rise in antibody titers (r ⫽ .53, p
⬍ .05; see Figure 6) while there was no significant relation
between these variables for subjects in the control group (r ⫽
.26). These correlations were not significantly different.
We also examined correlations between the frequency and
duration of reported practice and changes in the self-report
and EEG measures that showed significant Group ⫻ Time
interactions, as well as antibody titers to influenza vaccine.
There were no significant associations between the measures
of practice and any of the biological or self-report measures.
Descriptive statistics on these measures of daily practice are
provided in Table 1.
Fig. 5. Means ⫾ SE antibody rise from the 3- to 5-week to the 8- to 9-week blood draw in the Meditation and Control groups. The ordinate displays the
difference in the log-transformed antibody rise between the 3- to 5- and the 8- to 9-week blood draws derived from the hemagglutination inhibition
assay.
Fig. 6. Scatter plot for the meditation group only showing the relation between the change in asymmetric anterior activation at baseline from Time 1 to Time
2 in C3/C4 and the magnitude of rise in antibody titers to the influenza vaccine from the week 3 to 5 to the week 8 to 9 blood draw. The meditators
that showed the largest magnitude increase in left-sided anterior activation from Time 1 to Time 2 also showed the largest rise in antibody titers from
the 3- to 5- to 8- to 9-week blood draws. There was no significant relation between these variables in the control group.
R. J. DAVIDSON et al.
568 Psychosomatic Medicine 65:564 –570 (2003)
