Intern. J. Neuroscience. 1996. Vol. 86. pp. 197-205 1996 OPA (Overseas Publishers Association)
Reprints available directly from the publisher Amsterdam B.V. Published under license by
Photocopying permitted by license only. Cordon and Breach Science Publishers SA
Printed in Malaysia
MASSAGE THERAPY REDUCES ANXIETY AND
ENHANCES EEG PATTERN OF ALERTNESS AND
TIFFANY FIELD, GAIL ROBINSON, FRANK SCAFIDI, TOM NAWROCKI,
and ALEX GONCALVES
Touch Research Institute, University of Miami School of Medicine
James Madison University
University of Maryland
SAUL SCHANBERG and CYNTHIA KUHN
Duke University Medical School
(Received March 3, 1996)
Twenty-six adults were given a chair massage and 24 control group adults were asked to relax in the massage chair
for 15 minutes, two times per week for five weeks. On the first and last days of the study' they were monitored for
EEG, before, during and after the sessions. In addition, before and after the sessions they Performed math
computations, they completed POMS Depression and State Anxiety Scales and they provided a saliva sample for
cortisol. At the beginning of the sessions they completed Life Events, Job Stress and Chronic POMS Depression
Scales. Group by repeated measures and Post hoc analyses revealed the following: 1)~frontal delta power increased
for both groups, suggesting relaxation; 2) the massage group showed decreased frontal alpha and beta power
(suggesting enhanced alertness); while the control group showed increased alpha and beta power; 3) the massage
group showed increased speed and accuracy on math computations while the control group did not change; 4)
anxiety levels were lower following the massage but not the control sessions, although mood State was less
depressed following both the massage and control sessions; 5) salivary cortisol levels were lower following the
massage but not the control sessions but only on the first day; and at the end of the 5 week period depression scores
were lower for both groups but job stress scores were lower only for the massage group.
Keywords: Massage. alertness
Despite the increasing popularity of stress-management programs (Ivancevitch, Matteson, Freedman &
Phillips, 1990) very little evaluation research has been done. Most evaluations are based on "professional
opinions" and survey studies rather than empirical studies. A recent study on stress in HIV positive men
suggested that those who were most stressed gained most from a massage therapy intervention (Ironson, Field,
Kumar, Price, Kumar, Hansen & Burman, 1995). Longterm (one month) effects indicated immunological
benefits including increased natural killer cell number and natural killer cell cytotoxicity. Massage therapy has
also been noted to decrease anxiety and depression as well as cortisol and norepinephrine levels and improve
sleep patterns in adolescents with psychiatric problems (Field, Morrow, Valdeon, Larson, Kuhn & Schanberg,
1992). Thus, massage is noted to decrease anxiety and depression based on self-report, behavior observations,
salivary cortisol and urinary norepinephrine levels and to enhance immune function.
In the above studies subjects anecdotally reported enhanced alertness instead of the expected soporific effect
following massage. The purpose of the present study was to investigate the effects of massage on alertness as
measured by accuracy of performance on math computations. The only massage study in the literature that
recorded EEG showed that facial massage was accompanied by decreased alpha and beta, a pattern that is
inconsistent with drowsiness (Jodo, Yamada, Hatayama, Abe & Maruyama, 1988). The EEG alpha was
expected similarly to decrease during the chair massage in this study and the behavioral measure of alertness,
namely math computation performance, was expected to improve following massages. In addition, anxiety,
depression and cortisol levels were expected to decrease as they did in the Field et al. (1995) study on
disturbed adolescents and the Ironson et al. (1995) study on HIV men.
The subjects were 50 medical faculty and staff members (80% females, M age = 26). The subjects were well-
educated (62% college graduate, 12% graduate school, 27% graduate degree). Income was $20,OOO for 58%,
$2~30,OOO for 35% and greater than $30,OOO for 8%. Forty-six percent of the sample exercised regularly,
with moderate numbers exercising once per week (31%), to several times per week (27%), to daily (15%). Of
the sample 50% had tried relaxation techniques, and 62% had received a massage prior to the study (62%
rarely, 31% occasionally and 8% weekly). The subjects were recruited using advertising fliers at the medical
school. They were randomly assigned to the massage and the relaxation control groups. Chi square analyses
comparing the two groups on sex, education, income, and lifestyle questions (i.e., exercise and previous use of
massage and relaxation) and a t test on age yielded no group differences (see Table 1).
Means for Demographic Variables for Massage Therapy
and Relaxation Control Group Measures
Measures Massage Control p
Age 26.4 26.2 NS
Sex (% female) 79.5 80.2 NS
Graduate degree (%) 25.8 28.0 NS
[income> $30.000 (%) 7.8 9.1 NS
Regular exercise (%) 47.3 45.4 NS
Tried relaxation (%) 49.2 51.0 NS
Tried massage (%) 64.0 60.9 NS
Massage therapy. The massage therapy was given by a professional massage therapist (different therapists
each day) for 15 minutes a day, 2 days a week for 5 weeks, and the sessions were scheduled at noon each day.
The subjects were seated fully clothed in a special massage chair, and a standard Swedish massage procedure
(kneading of muscles) was used. The procedure consisted of long broad strokes with moderate pressure on the
BACK: 1) compression to the back parallel to the spine from the shoulders to the base of the spine; 2)
compression to the entire back adding some gentle rocking; 3) trapezius squeeze: 4) finger pressure around
scapula and shoulder; 5) finger pressure along the length of the spine and back; and 6) circular strokes to the
hips below the iliac crest. ARMS: 1) drop arms to the side. Knead arms from shoulder to lower arm; and 2)
press down points of upper and lower arms. HANDS; 1) massage entire hands. Traction to the fingers; 2) press
the fleshy part of the palm between the thumb and index finger for 15 to 20 seconds; and 3) traction of the
arms both in lateral and superior directions (arm in line with the body). NECK: 1) kneading area of cervical
vertebrae; 2) finger pressure along base of skull and along side of neck; 3) scalp massage; and 4) press down
on trapezius, finger pressure and squeezing continuing down the arms.
Relaxation control group. The subjects were asked to relax by tightening and relaxing the same body parts as
those that were massaged for the massage therapy group (and in the same sequence). The subjects were briefly
shown by a research assistant how to tighten and release their muscles which they were told would help them
relax. This group was included as a control for focusing on the body and for standardizing activity level during
the assessment sessions (controlling movement artifact in the EEG measure).
On the first and last day of the study the procedure was conducted in the following order:
1) the EEG cap was positioned on the subject's head; 2) a saliva sample was taken for cortisol; 3) the subjects
completed the 3 long-term measures including the Life Events, Job Stress and Chronic POMS Depression
Scales; 4) the subjects completed the session base-line measures including the POMS Depression, State
Anxiety and the math computations; and 5) immediately after the 15 minute massage/control sessions the
subjects completed another math computation, the POMS Depression and State Anxiety Scale, and about 20
minutes after the end of the massage/control sessions they provided another saliva sample for cortisol.
Pre-post therapy session measures on first and last day. The following measures were used to assess the
immediate effects of the massage on the first and last days of the study.
a) The Profile of Mood States (POMS; McNair, Lorr & Droppleman, 1971). The POMS Scale is a 5-point
adjective Likert rating scale asking the subject to describe how well an adjective describes his/her current
feelings. The 14 items that comprise the depression factor were used. The scale has adequate internal
consistency (r = .95; McNair & Lorr, 1964) and is an adequate measure of intervention effectiveness
(Pugtach, Haskell & McNair, 1969). A summary score is obtained by adding the weight of each item. This
scale was used because positive mood state would be expected to affect alertness and performance on math
computations and because massage therapy have been noted to improve mood state in stressed adolescents
(Field et al., 1992).
b) The State Anxiety 1nventory (STAI; Spielberger, Gorsuch & Lushene, 1970). This is a 20 item scale
which measures the transitory anxiety level in terms of severity (not at all to very much so). Characteristic
items include "I feel tense," "I feel nervous" and 'I feel relaxed." The STAI has adequate concurrent validity
(Spielberger, 1972) and internal consistency (r - .83; Spielberger et al., 1970). In addition, the STAI state
scores an increase in response to situational stress and a decline under relaxing conditions (Spielberger et
al., 1970). A summary score is obtained by adding the weight of each item. This measure was included
because state anxiety is known to affect alertness and performance on cognitive tasks negatively and
because state anxiety typically decreases following massage therapy (Field et al, 1992; Ironson et al.,
c) Salivary cortisol. Saliva samples were collected and assayed for cortisol as a measure of stress that
might be expected to affect alertness and performance on math computations. In addition, salivary cortisol
levels decreased in at least two previous massage therapy studies (Field et al., 1992; Ironson et al, 1995).
The samples were collected at the beginning of the therapy sessions and 20 minutes after the end of the
sessions 0fl the assessment days. Due to the 20 minute lag in cortisol change, saliva samples reflect
responses to events occurring 20 minutes prior to collection. Salivary cortisol samples were obtained by
having subjects place a cotton dental swab dipped in sugar-free lemonade crystals along their gumline for
30 seconds. The swab was placed in a syringe, and the saliva was squeezed into a microcentrifuge tube.
d) Math computations. Before the massage sessions a series of 7 numbers was given and after the massage a
different series was given, and the subject was asked to add them. The time to complete the series and the
correct/incorrect answer were recorded. This measure was used to determine the immediate effects of
massage 0fl a task that might be expected to be enhanced by alertness.
e) EEG procedure. EEG was considered the primary dependent variable in this study as the physiological
measure of alertness. Although subjects have anecdotally reported heightened alertness in previous studies,
no direct measures have been made of alertness. Although EEG alpha and beta were noted to decrease
(suggesting heightened alertness) in a previous study (Jodo et al., 1988), face massage was used and no
self report or performance measures were included. EEG was recorded in the present study for 3 minute
periods prior to, during and after the therapy sessions with the subjects' eyes closed. The EEG was recorded
using a Lycra stretchable cap (manufactured by Electro-Cap, Inc.) that was positioned on the subject using
the standard 1~20 system. Electrode gel was injected into the following sites: F3, F4, P3, P4 and Cz.
Impedances were brought below 5 K ohms, and the impedances of adjacent sites were brought within 500
ohms of each other. The EEG signals were amplified using a Grass Model 12 Neurodata Acquisition
System with amplifiers set as follows: Low frequency filter: 1 Hz; High frequency filter: 100 Hz;
Amplification: 20,000. The line frequency filter was on for all channels. The output from the amplifiers
was directed to a Dell 325D PC fitted with an Analog Devices RTI-815 A/D board. The signal was
sampled at a rate of 512 Hz and streamed to hard disk using data acquisition software (Snapstream HEM
Data Corp.). Additional electrodes were positioned on the external canthus and the supraorbital position of
one eye to record the subject's EOG, which was used to facilitate artifact scoring.
The first step involved the elimination of data which were unusable due to artifact from eye movements,
muscle activity or technical difficulty. The artifact-free data were spectrally analyzed using a discrete Fourier
transform with a Hanning window to yield power data for the following frequency bands: 1-4 Hz (Delta), 5-7
Hz (Theta), 8-12 Hz (Alpha), 13-20 Hz (Beta-low), 21-30 Hz (Beta-high). The EEG data were analyzed using
an EEG analysis software package developed by James Long Company.
First Day/Last Day Measures
a) Life Events Questionnaire. The Life Events Questionnaire comprises a list of 9 stressful events (e.g.,
death of mate or lover, major financial difficulties). The subject is asked to check which events have
occurred in the last four weeks. The subject is then asked to rate how each event has affected his/her life
from not at all to very stressful on a 4 point scale. This measure was included to ensure that the results
of this study were not negatively affected by significant life events. The Cronbach's alpha (.71) for
internal consistency was reasonable as was the test-retest reliability (.89).
b) Job Stress Yesterday Questionnaire. This questionnaire measures job stress experienced yesterday and
consists of 31 words or phrases requiring two responses each. The first response is a word or phrase
describing the job (e.g., hectic, hassled, comfortable, too little time to think or plan). Possible answers
are YES, NO or ? (cannot decide). If the phrase does describe the job yesterday, the subject is then
asked to rate on a four point scale how much it bothers him/her. This questionnaire was included as a
self-report measure on job stress. Cronbach's alpha (.73) for internal consistency was reasonable as was
the test-retest reliability (.87).
Analyses of baseline measures yielded no group differences except that the control group had less job stress at
baseline higher scores are optimal). Data were subjected to repeated measures by group (massage/control)
ANOVAS with session (presession/postsession) and phase (pretreatment/posttreatment) as the repeated
measures. Post hoc Bonferroni corrected t tests were performed to assess the group by repeated measures
interaction effects. A priori nonorthogonal contrasts were made because based on previous anecdotal reports by
subjects and the facial massage EEG data (Jodo et al., 1988) we expected that the analyses would reveal
enhanced alertness and performance. We also expected reduced anxiety, depression and cortisol levels
following massage therapy based on previous massage therapy studies (Field et al., 1992; Ironson et al.,
1995). The analyses revealed the following (see Table 2): 1) a repeated measures effect revealed that the
massage and relaxation control groups had significantly lower POMS depressed mood state scores following
the first and last day sessions; 2) a repeated measures by group interaction effect revealed that the massage
therapy group had significantly lower state anxiety scores after the first and the last day sessions than the
relaxation control group; 3) No group differences or time changes were noted for the Life Events Scale; 4) for
the Job Stress Scale, a significant repeated measures by group interaction effect was noted, suggesting a
decrease in job stress (higher scores are optimal) but only for the massage therapy group; and 5) a repeated
measures effect suggested both groups showed a decrease in chronic depressed mood state.
Saliva Cortisol Data
A repeated measures by group interaction effect revealed a decrease in salivary cortisol levels on the first day
for the massage group and an increase on the last day for the relaxation control group.
Massage also facilitated performance on math computation tasks (see Table 2). Group by repeated measures
interaction effects suggested that the massage therapy group performed better following the sessions on both
the first and last days. The decreased time required to complete the math computation task was significantly
greater for the massage therapy group, and the decrease in the number of errors was significantly greater for the
massage therapy versus the relaxation control group.
Repeated measures ANOVAs with group as the between subjects factor and pre-, during and postsession
values as the repeated measures were conducted on the frontal alpha, beta, delta and theta values recorded
before, during and after the massage/control sessions. These group (massage/control) by trial (0efore, during,
after massage/control session) repeated measures ANOVAS revealed the following (see Table 3 for repeated
measures and repeated measures by group interaction effects and their F values): 1) a repeated measures effect
revealed that delta increased for both groups from pre- to during the session, suggesting enhanced relaxation
(see Table 3); 2) theta did not change for either group; 3) group by repeated measures interaction effects
suggested that: a) alpha significantly decreased from pre- to during the massage and from pre- to postmassage
while alpha significantly increased for the relaxation group from pre- to postmassage and; b) beta significantly
decreased for the massage group from pre- to during and from pre- to postmassage, and the control group
significantly increased from pre- to postmassage.
Relations between EEG and Math Computation Measures
Correlation analyses were performed for each of the groups to determine whether there was a relationship
between the alpha and beta EEG measures and the math computation (speed and accuracy) variables. For the
massage therapy group the number of problems correctly solved was correlated with the decrease in the natural
log of alpha from the period pre to during the massage (r,-, .43,p <.05), and the amount of time required to
complete the computations was inversely related to the decrease in the natural log of beta from pre- to
postmassage (r = -.59, p < .001).
Means for Massage Therapy and Relaxation Control Group Measures (S.D.s under means)
Day 1 Day 10 Day 1 Day 10
Measures Pre Post Pre Post Pre Post Pre Post
.4 .2 .4 .3 .6 .3 .5 .3
11.3 9.6 12.7 10.5 13.2 12.9 13.9 11.4
20.9 28.3 29.7 31.0 24.0 20.7 26.3 25.9
85.1 72.3 75.9 64.0 82.6 71.8 65.0 62.3
Salivary cortisol (ng)
.5 .5 .4 .5 .7 .6 .4 .6
Day 1 Day 10 Day 1 Day 10 Effects
b 47.0b 46.0b PxG
Chronic POMS depression
5.1a 3.6b1 p
*G = group (massage/control)
S = session (pre-session/post-session)
P = Phase (pre-treatment/post-treatment)
**higher score is optimal
Natura1 Log of Raw Power for Delta, Theta, Alpha. and Beta
Delta (1~4 Hz)
Pre Dur Post Pre Dur Post
Mean 4.29 4.55 4.16 3.93 4.66 3.99
SD .93 1.14 .81 .96 1.34 1.22
Pre v. Dur v. Post: Tria1 F = 6.32 p = .004
Pre v. Dur: Trial F = 8.07 p = .009
Theta (5-7 Hz)
Pre Post Pre Dur Post
Mean 3.02 2.71 3.02 2.92 3.25 3.00
SD 1.38 1.20 1.32 158 1.26 1.45
Alpha (8-l2 Hz)
Pre Dur Post Pre Dur Post
Mean 5.50 4.64 5.20 4.99 4.87 5.61
SD 158 1.30 1.67 1.01 1.41 1.32
Pre v. Dur: Trial F = 6.51 p = .003
Group by Trial F = 3.40 p = .04
Pre v. Dur: Trial F = 6.86 p = .02
Pre v. Post: Group by Trial F = 6.36 p = .02
Beta (1~30 Hz)
Pre Dur Post Pre Dur Post
Mean 1.71 1.45 1.41 2.03 3.04 2.84
SD 1.81 2.14 1.95 150 1.44 1.49
Pre v. Dur v. Post: TriaI F = 8.23 p = .001
Group by Trial F = 12.37 p = .000
Pre v. Dur: Trial F = 7.32 p = .0l
Group by Trial F = 21.29 p = .000
Pre v. Post: Trial F = 13.68 p = .001
Group by Trial F = 5.20 p = .03
These data, like those of other studies on massage therapy showed decreases in anxiety and stress hormones
(cortisol) immediately after the sessions (Field et al., 1992; Ironson et al., 1995). And, both the massage
therapy and relaxation therapy groups showed increased delta activity, suggesting that both therapies had a
relaxation effect and temporary and more chronic shifts in mood State which may have related to their
relaxation. The decrease in self-reported depression is consistent with other massage studies (Field et al, 1992;
Ironson et al., 1995) as well as other relaxation studies (Platania-Solazzo, Field, Blank, Seligman, Kuhn,
Schanberg & Saab, 1992).
Heightened alertness and enhanced performance on math computations occurred in the massage therapy
group. The massage sessions were characterized by an EEG pattern of alertness. Although delta increased for
both groups of subjects, suggesting relaxation, the pattern of enhanced alertness (decreased alpha and decreased
beta) occurred in the massage therapy group while a pattern of drowsiness (increased alpha and increased beta)
occurred in the relaxation control group. The decreased alpha and decreased beta were not surprising since at
least one other study documented EEG alpha decreases associated with facial massage (Jodo Yamada,
Hatayama, Abe & Maruyama, 1988). The correlation analysis further suggested that the accuracy of the math
computations and the decrease in pre- to during massage EEG alpha were related. Although the alpha decrease
occurred during the massages it could have affected the state of alertness for enhancing accuracy after the
massage. Further, the speed of performing the calculations and the decrease in EEG beta pre- to postmassage
were related. This more contemporaneous relationship suggests that performance speed may have been related
to decreased beta.
The superior performance of the massage therapy group might relate to the tactile and pressure stimulation.
Tactile and pressure stimulation, in addition to enhancing the EEG patterns of alertness and math
computations in this study, have been noted to enhance parasympathetic activity (elevated vagal tone) which is
characteristic of a more relaxed, alert state during which cognitive performance improves (Field et al., 1992).
Future research might add other measures such as vagal activity and catecholamines further to understand the
underlying mechanism for the massage therapy-enhanced alertness relationship.
In addition, a longer term follow-up would be important to assess the persistence of the effects.
Presumably, like exercise, a steady dose of massage may be required. Larger doses may also be more effective
and result in more clinically meaningful changes in mood state and cortisol than occurred in this study.
Finally, the cost effectiveness of massage therapy would need to be documented for more widespread
acceptance and adoption of the treatment.
Field, T., Morrow, C., Valdeon, C., Larson, S., Kuhn, C. & Schanberg, S. (1992). Massage therapy reduces
anxiety in child and adolescent psychiatric patients. Journal of the American Academy of Child and
Adolescent Psychiatry, 31. 12~131.
Ironson. G., Field, T., Kumar, A., Price, A., Kumar, M.. Hansen, K. & Burman, I. (In Press). Relaxation
through massage is associated with decreased distress and increased serotonin levels. International
Journal of Neuroscience.
Ivancevitch, J. M., Matteson, M. T., Freedman, S. M. & Phillips, J. S. (1990). Worksite stress management
interventions. American Psychologist, 45, 252-261.
Jodo, Yamada, Hatayama, Abe & Maruyama (1988). Effects of facial massage on the spontaneous EEG.
Tohoku Psycologica Folia, 47, 8-15.
McNair. D. M. & Lorr. M. (1964). An analysis of mood in neurotics. Journal of Abnormal Social
McNair. D. M. Lorr. M. & Droppleman, L. F. (1971). Profile of Mood States. San Diego: Educational and
Industrial Testing Service.
Platania-Solazzo, A., Field, T., Blank, J., Seligman, F., Kuhn, C., Schanberg, S. & Saab. P. (1992).
Relaxation therapy reduces anxiety in child/adolescent psychiatry patients. Acta Paedopsychiatrica, 55,
Pugtac, D., Haskell, D. & McNair, D. M. (1969). Predictors and patterns of change associated with the course
of time-limited psychotherapy. Mimeo Report.
Spielberger. C. D., Gorsuch, R. L. & Lushene, R. E. (1970). The State Trait Anxiety Inventory. Palo Alto,
Consulting Psychologists Press.
Spielberger, C. D. (1972>. Anxiety as an emotional State. In C. D. Spielberger (Ed.), Anxiety: Current trends
(n theory and research. New York, NY: Academic Press.