Journal o/Psychosomatic Research, Vol. 33, No. 5, pp. 549-559. 1989. 0022-3999189 $3.00 + .

OO
Printed in Great Britain. C 1989 Pergamon Press plc

THE AUTONOMIC NERVOUS SYSTEM RESPONSE

INVENTORY (ANSRI): DISCRIMINATION AMONG
MIGRAINE HEADACHE SUFFERERS, MUSCLE
CONTRACTION HEADACHE SUFFERERS,
AND NORMAL CONTROLS

WILLIAM F. WATERS, DAVID P. MCANULTY and RAY R. Buss

(Received 14 April 1988; accepted in revised ,jmn 23 March 1989)

Abstract-Twenty-three migraine headache sufferers, sixteen muscle contraction headache sufferers, and
thirteen no headache control subjects were selected to test the discriminant validity of the ANSRI.
Significant Chi-Squares indicated reliable differences among the groups in ANSRI F scale scores under
the Anger condition and in P scale scores based on the means of items across four emotions (All E).
Discriminant analyses yielded 69% correct classifications for All E P scales and 58% correct
classifications for Anger F scales. All E P scale analyses revealed that the Muscle Tension scale separated
the headache groups from the control group. A second function separated the three groups from each
other, with the Cardiac, Respiration, and Gastrointestinal scales most responsible. Anger F scale analysis
showed the Peripheral Vasoconstriction, Cardiac, and Pattern 2 scales separating the headache groups
from each other and from the control group. Results demonstrated discriminant validity for the ANSRI,
and were consistent with muscle tension and vasoconstriction as variables in muscle contraction and
migraine headache, respectively.

INTRODUCTION

THE AUTONOMIC Nervous System Response Inventory, or ANSRI developed by

Waters, Cohen, Bernard, Buco, and Dreger [l] was designed to define a subject’s
individual-response stereotypy [2], within and across emotions, such that it can
be described as a profile of peripheral physiological responses. Responses which
tend to be unusually large, enduring, and/or frequent within or across emotions
theoretically are of clinical interest, i.e., may pertain to the etiology, course, or
response to treatment of a patient’s psychophysiological disorder. It also was
designed to be a research tool in the investigation of the roles of individual-response
stereotypy and stimulus-response specificity in the etiology and/or exacerbation of
psychophysiological disorders in general.
The ANSRI elicits an individual’s self-reported patterns of peripheral
physiological response to emotion-provoking stimuli. It consists of 51 items, with
each item representing a peripheral somatic or autonomic nervous system response
which often occurs during emotional distress, pleasure, or physical activity. Each
item is rated as being present (in varying degree) or absent during a particular
emotional state/situation that is reconstructed from memory and then imagined.
An essential step in determining the experimental validity (and, to some degree,
the clinical utility) of the ANSRI is to ascertain the ability of the ANSRI to
discriminate among patients with various psychophysiological disorders, and

Address correspondence to: William F. Waters, Ph.D., Department of Psychology, Audubon Hall,
Louisiana State University, Baton Rouge, LA 70803.

549
550 WILLIAM F. WATERS et al.

between such patients and healthy people. Assuming that there is some
reasonable degree of validity to the assertion that migraine headaches involve a
predisposition to excessive vasoactivity and that muscle contraction headaches
involve a predisposition to excessive scalp and neck muscle activity [3], by whatever
etiological mechanisms, it would be reasonable to hypothesize that migraine
headache sufferers and muscle contraction headache sufferers might generate
differential patterns of actual physiological response to some or all emotional
stressors. Similarly, they should produce different self-reported patterns of
physiological response on the ANSRI, and both groups would be hypothesized to
so differ from a contrast group of healthy (no headache) individuals.
The data pertaining to patterns of actual psychophysiological response (individual-
response stereotypy) in muscle contraction headache sufferers are mixed. Several
investigators have found that groups of muscle contraction headache sufferers show
higher levels of resting muscle activity (frontalis, neck, temporalis or trapezius EMG)
than no headache control subjects [4-S]. Other investigators have not found such
differences [9-131. Complicating matters is the finding that migraine headache
sufferers sometimes show equivalent or greater resting EMG levels than controls
[5-7, 11, 141.
Similar results have been obtained with respect to EMG reactivity to neutral,
painful and stressful stimuli. EMG responses to neutral stimuli have not
discriminated between muscle contraction headache sufferers and normal controls
[5, 11, 14, 151, nor have studies using physical stressors [ 10, 16, 171 or some studies
using mental/emotional stressors [12, 15, 17, 181. There are, however, some
investigations in which EMG reactivity differences between muscle contraction
headache sufferers and controls did emerge in the hypothesized direction. All of these
employed mental/emotional stressors [5, 14, 191.
Maber, Kuczmierczyk and Adams [20], and McAnulty [21] have found evidence
that there exist, within groups of subjects presenting with complaints of muscle
contraction headache, at least two subgroups: a muscle contraction headache
subgroup and a psychogenic head pain subgroup. The former tend to show the
hypothesized EMG response levels and reactivity while the latter do not. Thus, the
highly mixed individual-response stereotypy data gathered to date may well reflect
the combined responses of two different subgroups of subjects complaining about
muscle contraction headache symptoms, only one of which demonstrates any
anatomical basis (pathophysiology) for the complaints.
Migraine headache sufferers have been differentiated physiologically from no
headache control subjects in a few studies, each showing greater cephalic vaso-
constriction during or after laboratory mental/emotional stressors for the former
than the latter [5, 11, 161. On the other hand, other studies [7, 10, 171 did not find
such differences. At least four other studies, however, have documented cerebral
blood flow patterns reflecting vasoconstriction prior to migraine headache and/or
vasodilation during migraine headache [22-251.
Because so many psychophysiological headache investigations have generated
such conflicting results, inquiry into the self-reported patterns of physiological
response to emotional stressors of migraine headache sufferers and muscle
contraction headache sufferers would be worthwhile as an additional tool for
defining individual-response stereotypy and as a discriminating diagnostic method
 ANSRI: Discrimination of headache types 551

in its own right. The ANSRI would be a proper instrument for carrying out
such inquiries.

METHOD

Sulrjects
Participants in the present study were 39 chronic headache sufferers (23 classic and common nugraine
headache sufferers, the Migraine Headache group; and 16 muscle contraction headache sufferers, the
Muscle Contraction Headache group) and 13 no headache control subjects, the No Headache Control
group. Headache patients were recruited via newspaper articles requesting volunteers and by physician
referrals to the Psychology Consultation Service at the Earl K. Long Memorial Hospital in Baton Rouge,
Louisiana. Control subjects were volunteers recruited by the headache subjects, who were non-blood
relatives or friends of the same sex and approximate age. Control subjects had fewer than six headaches
per year and no history of diagnosed mental disorder or treatment for same. The Migraine Headache
group consisted of 21 females and two males, with a mean of 38.5 years (range 2658). The Muscle
Contraction Headache group included 12 females and four males, with a mean age of 38.9 years (range
23-65). The No Headache Control group included 1I females and 2 males. with a mean age of 33.9 years
(range 15-54).

Volunteers and referred patients were interviewed independently by a Board certified neurologist and
an advanced doctoral student in medical/clinical psychology from the Louisiana State University’s clinical
psychology training program, each interview generating an independent diagnosis. Discrepancies in
diagnostic classification of headache were discussed by the two interviewers, and inclusion of a subject
in the study was allowed only when agreement was reached.
Specific inclusion and exclusion criteria were used in making the diagnostic decisions [26,27]. Migraine
headache was characterized by at least three of the following six sets of descriptors: (a) nausea, vomiting,
diarrhea or constipation during headache; (b) unilateral onset of pain; (c) visual prodromata; (d) throbbing
or pulsating pain; (e) relief of pain from vasoconstrictive medications; (f) hypersensitivity to light or noise.
Further, there could be little report of any of the muscle contraction headache symptoms listed below
(most migraine headache sufferers report some associated muscle tension).
Muscle contraction headache was characterized by the presence of the following typical complaints:
(a) frequent headaches, i.e. at least three per week; (b) pain described as ‘aching’, ‘tightness’, ‘tenseness’.
‘caplike or bandlike pressure’; (c) bilateral pain, frequently beginning in the neck and shoulders. In
addition, the exclusion of the following vascular headache symptoms was established: visual prodromata,
unilateral pain and vomiting during headache. Subjects reporting symptoms of mixed headache were
excluded from both categories and from the present study.
Finally, a number of conditions automatically excluded a potential subject from the study. These
included: temperomandibular joint syndrome (based on a confirmed dental diagnosis); structural damage
or physical trauma (such as concussion or pinched nerve); severe sinus headache (with confirmed medical
diagnosis); past or preseni major mental disorder (i.e., psychosis of any type); absence of a negative
neurological screening including at least a CAT scan, brain scan or electroencephalogram (EEG).
Following the initial diagnostic interviews, subjects who were accepted for the study were scheduled for
the administration of the ANSRI, which was completed in one session of about 25 minutes, with an
examiner available to answer inquiries about the test instuctions.

The Autommic Nervous S>ssstem Response Inventory (ANSRI)

The ANSRI [l], created to describe self-reported patterns of peripheral physiological response to
emotion-provoking stimuli, consists of 51 items representing peripheral somatic or autonomic nervous
system responses often occurring in emotional distress. pleasure or physical activity. The items were
selected from a list derived from the literature on the psychophysiology of emotion and from inquiries
put to colleagues in Departments of Psychiatry and Psychology. Each item is rated as being present (in
varying degree) or absent during a particular emotional state/situation that is reconstructed from memory
and them imagined. Subjects initially recollect a prototypical fear situation and then reconstruct it in
imagery, taking note of physiological responses remembered from that situation and of physiological
responses occurring during the imagery. The 51 inventory items are then reviewed and rated as having
been present from I = not at all, to 5 = very strong. The same procedure is then performed with
prototypical anger, sadness, joy and physical activity situations and imagery.
Twelve physiological scales (P scales) were developed on the basis of the physiological coherence of
items, each scale representing a physiological response system. Factor scales (F scales) were derived from
factor analyses of the 51 items as rated under each of the conditions (fear, anger, sadness, joy, and physical
552 WILLIAM F. WOERS et a[.

activity) and as averaged across all five conditions (All C) and the four emotions (All E). The
principal-components method was used on a matrix of Pearson product-moment correlations of the
ratings of the ANSRI items under a given condition. The principal components were rotated orthogonally
using the Varimax procedure. Only factors with an eigenvalue of 1.00 or greater were accepted, and the
Scree Test was used to determine the optimal number of factors.

TABLE I.-ANSRI P SCALES AND F SCALES FOR FEAR, ANGER, SADNESS, JOY,
ACTIV1TY.AL.Lc ANL! ALL E

Condition Scale no Scale lable No. items

- -
Fear Pl Cardiac* 4
P2 Cardiovascular* 9
P3 Skin* 5
P4 Sudomotor
P5 Piloerection 2
P6 Muscle tension II
P7 Respiration* 6
P8 Gastrointestinal 9
P9 Excretory* 4
PlO Energy activation* 2
PI1 Eyes 3
PI2 Thermoregulation* 10
Fl Gastrointestinal 9
F2 Muscle tension
F3 Headache I
F4 Sudomotor
F5 Energy activation*
F6 Pattern I
F7 Respiration (amplitude)*
F8 Vasoconstriction (cephalic)
F9 Vasoconstriction (peripheral) 2

Anger PI-PI2 As above

FI Vasoconstriction (peripheral)
F2 Muscle tension
F3 Headache 2
F4 Vasoconstriction (cephalic)
F5 Cardiac
F6 Respiration (amplitude)*
Fl Pattern 2
F8 Gastrointestinal

Sadness PI-P12 As above

Fl Pattern 3 9
F2 Energy activ-ation*
F3 Muscle tension
F4 Gastrointestinal (gas)
F5 Gastrointestinal (excretory)
F6 Gastrointestinal (stomach)
Fl Vasoconstriction (cephalic)
F8 Respiration (rate)

JOY PI- PI2 As above

Fl Gastrointestinal
F2 Cardiac*
F3 Pattern 4
F4 Sudomotor
F5 Vasoconstriction
F6 Headache 3
F7 Vasodilation
F8 Muscle tension (limbs)
Continued on facing page
 ANSRI: Discrimination of headache types 553

TABLE I.--continued

Condition Scale no. Scale lable No. items

Activity PlLPI2 As above
Fl Headache 4 12
F2 Pattern 5 10
F3 Muscle tension 7
F4 Respiration 5
FS Gastrointestinal 6
F6 Energy activation 4
Fl Vasoconstriction (cephalic) 4

All C PI-P12 As above

Fl All C pattern 1 9
F2 Gastrointestinal 10
F3 Muscle tension 7
F4 Reduced function 3
F5 All C pattern 2 9
F6 Energy activation* 4
F7 Vasoconstriction (cephalic) 7

All E PI-PI2 As above

Fl Muscle tension 7
F2 Gastrointestinal 11
F3 All E pattern 9
F4 Reduced function 4
F5 Vasoconstriction (cephalic) 8
F6 Energy activation* 3
F7 All E headache 9
*Bipolar scale.

Nine clusters of items (factors) emerged when subjects responded to the 51 items under the fear
condition, eight factors emerged under the anger, the sadness, and the joy conditions, and seven factors
emerged under the activity condition. All C and All E generated seven factors each. An item from a factor
was included on a factor scale if its factor weight was 0.40 or greater. A cross-validation factor analysis
was performed on a separate sample on a separate occasion, the cross-validation factors correlating well
with the original factors (74% of the factors exceeded the correlation criterion of 0.65). and the
cross-validation factor scale content corresponding well to the original as estimated by item overlap (57%
of factor scales had 50% or greater overlap; 74% of factor scales had 33% or greater overlap). Further
details are provided in Waters ef al. [l].
The scales, developed to describe the relative activity of different physiological response systems under
different emotional states. are listed in Table I for each condition, All C and All E. Considering the 51
items under each condition as a separate inventory, there are from 19 to 21 scales for each condition (the
12 P scales plus from 7 to 9 F scales). Waters et al., found that I1 of the 12 P scales met test-retest reliability
and internal consistency criteria, and 37 of the 54 F scales were sufficiently reliable and internally
consistent, and were also replicable upon cross-validation. Pursuant to those findings, some necessary
wording changes were made on the unreliable P scale (P9), and the unreliable or unreplicable F scales
were treated as tentative and experimental.

Statistical Anal.vses
Each condition (fear, anger, sadness, joy, activity) is a separate administration of the ANSRI, a separate
inventory, since the stimulus conditions (fear imagery, anger imagery, etc.) are different each time a subject
responds to the 51 items. Thus the scale scores for each condition were analyzed apart from each other
condition. The activity condition was not included among the analyses in order to help limit the number
of analyses: it was selected for elimination because it had the least theoretical bearing on the etiological
issue at hand, and because it alone did not reflect an emotional state. All E data were analyzed, but All
C data was not, in order to limit the number of analyses and because All C differs from All E only in
the inclusion of the activity condition data (already excluded).
All of the ANSRI scales were included in the data analyses, except for the F scales previously found
to be most unreliable in our earlier research [l]. Even though Cronbach, Gleser, Nanda and Rajaratnam
[28] recommend the use of unreliable scales mixed with reliable scales in preference to the use of too few
scales and an excessive restriction of the domains sampled, the unreliable F scales were deleted in order
554 WILLIAM F. WATERS et al.

to reduce the number of variables in the F scale analyses (it was possible to do so without neglecting any
of the domains sampled since there is redundancy among the scales). The omitted scales were: Fear F3.
Anger F6, Sadness F2 and F6, Joy F5, Activity Fl and F6, and All E F6.
Discriminant Analvses were oerformed on the ANSRI data in order to test for the ability of the ANSRI
to discriminate among the groups, and to identify which scales contributed the most to the-discrimination.
Because the P scales and the F scales were independently derived, and because there is considerable overlap
between some P scales and F scales, separate Discriminant Analyses of the P scales and F scales for each
condition and All E were required. A Discriminant Analysis was,considered statistically significant if the
Chi-Square was non-chance (p = or ~0.025) for any Canonical Discriminant Function. A dependent
variable (P scale or F scale) was considered contributory if its Standardized Canonical Function
Coefficient exceeded +0.40 or -0.40. The contribution of such a dependent variable was considered to
be important if the pooled within groups correlation between the dependent variable and the Canonical
Discriminant Function equalled or exceeded +0.30 or - 0.30.
To establish the validity of some scales and the invalidity of others called for the simultaneous
comparison of all relevant scales on the same samples at the same time. Thus. the extensive statistical
treatments applied to these data. Despite the fact that the administration of the ANSRI under each
condition is a separate inventory, this creating independence among the statistical analyses for each
condition and All E, it was necessary to further counter the potential Type I error effects of performing
a large number of statistical tests on the data. This was accomplished by: (1) eliminating the Activity P
and F scale analyses and the All C P and F scale analyses, thus reducing the number of Discriminanr
Analyses to ten; and (2) by setting the required level of significance to p = 0.025.

RESULTS

Statistically significant Chi-Squares were obtained for the All E P scale

Discriminant Analysis (Chi-Square = 37.99, df = 22, p < 0.019) and the Anger F
scale Discriminant Analysis (Chi-Square = 26.62, df = 14, p = 0.021).
All E P Scales. P4 and P&P1 1 all had sufficiently large Standardized Canonical
Discriminant Function Coefficients, but only P6 (Muscle Tension) was adequately
correlated with the significant first Canonical Discriminant Function, r = 0.3 1.
Discriminant functions evaluated at the group means for the first function yielded
values of 0.28, 0.65 and - 1.29 for the Migraine Headache, Muscle Contraction
Headache and No Headache Control groups respectively. The first function, thus,
separated the two headache groups from the control group. It should be noted that
function 2 in this analysis produced a Chi-Square = 16.85 (df = 10) with a probability
level of 0.077, and that three P scales correlated well with this function (Pl. Cardiac,
r = 0.47; P7, Respiration, r = 0.39; and P8, Gastrointestinal, r = 0.30). This second
function separated the two headache groups from each other and from the control
group: Discriminant functions evaluated at the group means produced values of
0.70 (Migraine Headache), -0.82 (Muscle Contraction Headache) and -0.23 (No
Headache Control). The relevant group means are presented in Table II.
The All E P scales correctly classified 69 % of all subjects into the three categories.
There was a 74% rate of correct predictions for the Migraine Headache group, 69%
for the Muscle Contraction Headache group, and 62% for the No Headache Control
group.
Anger F Scales. Anger F scales Fl. F3. F5 and F7 each had Standardized
Canonical Discriminant Function Coefficients greater than +0.40 or -0.40. Three
of these F scales also had sufficiently large correlations with the Canonical
Discriminant Function to be interpretable: Fl (Vasoconstriction, Peripheral;
r = 0.36), F5 (Cardiac; r = 0.34) and F7 (Pattern 2; r = 0.47). Thus, the function
is primarily a cardiovascular one, although respiration and gastrointestinal
components are evident in the pattern scale (F7). Discriminant functions evaluated
 ANSRI: Discrimination of headache types 555

TABLE II.-MEAN ANSRI SCALE SCORES ON SCALES THAT CORRELATED WITH THE CANONICAL
DISCRIMINANT FUNCTION COEFFICIENTS

Migraine headache Muscle contraction No headache

ANSRI scale group headache group control group
All E function I
P6 muscle tension 22.29 22.63 19.29
All E function 2
PI cardiac 12.30 11.00 11.48
P7 respiration 14.80 13.73 13.02
P8 gastrointestinal 16.07 14.16 14.61
Anger function 1
Fl vasoconstriction 9.96 1.63 8.61
F5 cardiac 12.21 9.69 11.15
F7 pattern 2 16.30 12.13 13.31

at the group means resulted in values of 0.88 for the Migraine Headache group,
- 0.98 for the Muscle Contraction Headache group and - 0.36 for the No Headache
Control group. This measure clearly separates the headache groups from each
other with the control group in between. The relevant group means are presented
in Table II.
The total correct classification rate was 58%, with 65% of the Migraine Headache
group correctly classified, 69% of the Muscle Contraction Headache group correctly
classified and 3 1% of the No Headache Control group correctly classified.
When all of the Anger F scales were included in these analyses, the same scales
emerged as significant, the same separation was found, and the total correct
classification rate improved to 64% (with the Migraine Headache group at 65%, the
Muscle Contraction Headache group at 75%, and the No Headache Control group
at 42%).
The obtained estimates of successful classification may be considered somewhat
optimistic since the sample size was insufficient to select a holdout sample on which
the original estimates could be cross-validated. Thus, a jackknife procedure
(BMDP7M) was used on the data from the entire sample to provide a conservative
estimate of successful classification. The jackknife procedure removes one subject
from the sample and generates classification estimates for the N - 1 sample,
obtaining an overall estimate of classification accuracy after having run the
procedure with each subject removed once. The results were, as might be anticipated,
notably less impressive: All E P scales dropped from 69% to 50%; Anger F scales
(reliable scales) from 58% to 42%, and; Anger F scales (all scales) from 64% to 40%.

DISCUSSION

The present data provide good evidence for the experimental discriminant validity
of the ANSRI. Not only did the P scale scores, when averaged across all four
emotions (All E) correctly identify 69% of the carefully diagnosed migraine and
muscle contraction headache patients and control subjects, but the major variable
contributing to the successful All E discriminations between the headache groups and
the control group was P scale 6, Muscle Tension. Although no cardiovascular scales
were isolated as major (statistically significant first function) variables in the All E
discrimination, the All E P scale Discriminant Analysis did reveal a second, weaker
556 WILLIAM F. WATERS et al

function in which three P scales correlated well (above +0.30) with the Canonical
Discriminant Function, foremost among these P scale 1, Cardiac. That function also
discriminated the two headache groups from one another. The ability of the ANSRI
to discriminate between types of headache sufferers and between headache sufferers
and no headache subjects, however, must be considered to be limited by the results
of the jackknife procedure. Regardless, replications on other, similar samples must
be carried out before the actual discriminant abilities of the ANSRI can be known.
The cardiovascular contribution to the ANSRI’s discrimination among the groups
became more evident and more certain in the statistically significant Anger F scale
Discriminant Analysis, where three scales correlated well with the Canonical
Discriminant Function: Anger F scale 1 (Vasoconstriction, Peripheral), Anger F scale
5 (Cardiac) and Anger F scale 7 (Pattern 2, which contains primarily gastrointestinal
and respiration items). In this case, the effective discriminations were between the two
headache groups and between the headache groups and the control group. It is of
interest that in the All E analysis, the second function’s major variables were also
cardiac, respiration and gastrointestinal scales, and there too, the scales discriminated
between headache groups.
It thus appears that the self-report of five physiological response systems’ activity
during imaginal emotional stress contributed significantly to the ANSRI’s ability to
discriminate migraine headache sufferers from muscle contraction headache sufferers
and headache sufferers from normal controls: muscle tension, peripheral vascular
activity, cardiac activity, respiration, and gastrointestinal activity. The muscle
tension scale contributed most to the discrimination of headache subjects from
no headache controls, while the cardiovascular, cardiac, respiratory, and gastro-
intestinal scales contributed most to the discrimination between migraine and muscle
contraction headache sufferers.
Although the present study did not address etiological issues directly via the
correlation of self-reports of physiological phenomena and their objective measures
assessed at the time of headache (an issue that should be addressed in future research
despite the typically small correlations between such measures), there are limited
implications in the results for the etiology of headache. It is likely that the
above-mentioned peripheral physiological systems, in different ways and degrees,
contribute to or reflect possible pathophysiologies of migraine and muscle con-
traction headache. The etiological connections between vascular activity and
migraine headache, and between muscle tension and muscle contraction headache,
have already been noted, but the possible roles of cardiac, gastrointestinal and
respiratory activity in these disorders have not. Some admittedly speculative
possibilities are: (a) gastrointestinal disruption may be produced by a general
vasoconstriction preceding and/or causing migraine headache, or it may be a
response to the severe pain produced by that type of headache; (b) increased cardiac
activity (greater output via increased rate and stroke volume) may reflect a reflexive
compensatory sympathetic nervous system response to the vasodilation commonly
present during migraine headache; and (c) changes in respiratory activity may be
responsive to, parallel to, or even may drive to some degree the cardiac activity noted
above.
It may be meaningful that anger was the only emotion to generate a statistically
significant discrimination between the headache groups, in view of the hypothesis
 ANSRI: Discriminatioli of headache types 557

that inability to deal with hostility generates migraine headache [3,29, 301. Further,
only in the anger condition did vasoconstriction emerge as a major variable in
the discrimination among the groups. This has direct implications for the
anger-migraine headache hypothesis because anger has more extensive vascular
effects than fear, as noted, for example in studies by Ax [31] and Weerts and Roberts
[32]. Both of these studies indicated greater increases in diastolic blood pressure
during anger than fear, and increases in diastolic blood pressure require greater and
more general vasoconstriction than increases in systolic blood pressure. Moreover,
Mason’s review [33] of studies on peripheral adrenergic and noradrenergic activity
in anger and fear pointed out that noradrenergic activity with its greater and more
widespread vasoconstrictive effects, is more typically associated with anger, and
adrenergic activity is more closely associated with fear.
Although the present data do not directly confirm the association between muscle
tension and muscle contraction headache, or that between vasoconstriction and
migraine headache, they do clearly implicate those two physiological responses in
the discrimination between subjects having the two types of headache and in the
discrimination between headache and normal control subjects. The data thus
strengthen such associations.
Perhaps of even greater significance is that the ANSRI has been shown to
discriminate statistically between two samples having different psychophysiological
disorders with different pathophysiologies and between those samples and a sample
of subjects with no history of either disorder. Further, the ANSRI scales that were
the most important in determining the group discriminations were scales reflecting
the physiological activity most commonly associated with the etiology and course of
each of the two disorders. This may be seen as a demonstration of empirical
discriminant validity for the ANSRI, although the actual correct classification rate
could not be considered clinically useful. Classification errors were primarily false
positives in the anger analysis, but were more evenly distributed between false
negatives and positives in the more predictive All E data.
The present results probably do not represent the maximum discriminatory power
of the ANSRI, for two reasons. First, there are not available, as yet, adequate norms
for the conversion of ANSRI scale scores to standard scores which would enable
comparisons among different scales and comparisons of patterns of scale scores.
Second, the muscle contraction headache group may have been contaminated by the
inclusion of subjects presenting with psychogenic head pain rather than muscle
contraction headache [20,21], a factor that would operate to decrease discriminabil-
ity based on physiological self-report.
A reasonable suggestion might be made that much of the ANSRI’s predictive
power derives from a correspondence between patient responses to the ANSRI items
and patients responses to the diagnostician’s questions. It is true that the ANSRI
would be of little clinical help in predicting the relatively obvious, as in the present
study. Rather, the ANSRI’s usefulness would derive from the empirical association
of a scale pattern (profile) with the later development of a psychophysiological
disorder, or the association of a scale pattern with a positive response to a form of
treatment, or the association of a scale pattern with a particular subset of individuals
presenting with a given psychophysiological disorder (implying an etiological role for
a particular pattern of stress response in such patients).
558 WILLIAM F. WATERS et al.

The ANSRI is still a long way from being the clinically useful instrument it
now appears it may become, even though data from our laboratory show that
in regression analyses, ANSRI scales predicted cardiovascular. electrodermal,
respiratory and other electrophysiological measures at non-chance levels, accounting
for 27% of the variance on average [34]. There remains the need for other
demonstrations of discriminant validity with samples of patients having other
psychophysiological disorders, and for full standardization of the instrument such
that scores for individuals can be evaluated against the norm and different scales and
profiles can be compared (T scores).

REFERENCES

I. WATERS WF, COHEN BA, BERNARDGA, Buco SM. DHEGER RM. An Autonomic Nervous System
Response Inventory (ANSRI): Scaling, reliability and cross-validation. J Behar Med 1984; 7:
315-341.
LACEY JI. Individual differences in somatic response patterns. J C‘omp fh.~?~iol /?r~‘chol 1950; 43.
338-350.
WOLFF HG. Headache and O/her Head Pain. New York: Oxford. 1963.
ANDRASIKF, HOLROYDKA. Physiologic and self-report differences between tension and non-tension
headache sufferers. J Behar .4ssrss 1980; 2: 135 141.
COHEN RA, WILLIAMSON DA. MONGIJILI.OTJE, HIJTCHINSOI‘: PC. G~TTLIEB J, WATERS WF.
Psychophysiologic response patterns in vascular and muscle contraction headaches. / Behar Med
1983; 6: 93- 108.
6 PHILIPS C. A psychological analysis of tension headache. In Contriburions /u Medical Psycholog>
(Edited by RACHMAN S), (Vol 1). Oxford: Pergamon Press, 1977.
7 POZNIAK-PATEWIECZ E. Cephalic spasms of head and neck muscles. Headache 1976; 15: 261-266.
8 VAN BOXTEL A, VAN DER VAK JR. Differential EMG activity in subjects with muscle contraction
headaches related to mental effort. Headache 1978; 17: 233-237.
9 ACOSTA F, JAMAMOTO J, WILCOX S. Application of EMG biofeedback to relaxation training.
J Consult Clin Psych 1978; 46: 383-384.
10 ANDERSON CD, FRANKS RD. Migraine and tension headache: Is there a physiological difference?
Headache 1981; 20: 63 71.
11 BAKAL DA, KACANOV JA. Muscle contraction and migraine headache: Psychophysiological
comparison. Headache 1977; 17: 208-215.
12 MARTIN PR. MATHEWS AM. Tension headaches: Psychophysiological investigation and treatment.
Jpsychosom Res 1978: 22: 389-399.
13 SUTTONEP, B~LAR CD. Tension headache patients versus controls: A study of EMG parameters.
Headache 1982: 22: 133Sl36.
14 PHILIPS C, HUNTER M. A psychophysiological investigation of tension headache. Headache 1982; 22.
173-179
15 SWRGIS ET. Lability and rracticir_y m headache acfit1ir.v. Paper presented at the annual convention
of the Association for the Advancement of Behavior Therapy. New York. 1981.
16. GANNON LR, HAYNES SN, SAFRANEKR, HAMII.TON J. A psychophysiological investigation of muscle
contraction and migraine headache. J psychosom Res 1981; 25: 271-280.
17. ANURASIK F, BLANCHAR~ EB, ARENA JG. SAUNDERSNL. BAKRON KD. Psychophysiology of recurrent
headache, Methodological issues and new empirical findings. Behac Ther 1982; 13: 407-429.
1X. FEUERSTEIN M, BUSH C, COSBISIERO R. Stress and chronic headache: A psychophysiological
investigation of muscle contraction and migraine headache. J ps_pcho.wm Res 1982; 26: 167- 182.
19. BRANTLEK PJ. Headache: The Role of Enrironmenttrl and Emotional Stwss. Unpublished doctoral
dissertation, University of Georgia, 1980.
20. HABER JD, KUCZMIERCZYK AR, ADAMS HE. Tension headache: Muscle overactivity or psychogenic
pain? Headache 1985; 25: 23-29.
21. M~ANULTY DP. The Role of EMG ill Tension Headache. Unpublished doctoral dissertation,
Louisiana State University, 1985.
22. O’BRIEN MD. Cerebral blood flow changes in the migraine headache. Headuc,he 1971; 11: 139-143.
23. SAKAI F, MEYER JS. Regional cerebral haemodynamics during migraine and cluster headaches
measured by the 133 xenon inhalation method. Headache 1978; 18: 122-132.
24. SKINHOJ E. Haemodynamic studies within the brain during migraine. Arch Neural 1973; 29: 95-98.
 ANSRI: Discrimination of headache types 559

25. TUNIS MM, WOLFF HG. Analysis of cranial artery pulse waves in patients with vascular headache
of the migraine type. Amer J Med Sci 1952; 224: 565--568.
26. AD Hoc COMMITTEE ON CLASSIFICATIONOF HEADACHE. J Amer Med Assoc 1962; 179: 717-718.
27. DIAMONDS. DALFSSIODJ. The Practicing Physician’s Approach to Headache 2nd, Edn. Baltimore:
Williams and Wilkins, 1978.
28. CRONBACH LJ, GLESER GC, NANUA H, RAJARATNAM N. The Dependahilily of Behavioral
Measurements. New York: Wiley, 1972.
29. FROMM-REICHMANNF. Contributions to the psychogenesis of migraine. Psychoanal Rev 1937; 24: 26.
30. WOLFF HG. Personality factors and reactions of subjects with migraine. Arch Neural and Psychiaf
1937; 37: 895.
31. Ax AF. The physiological differentiation between fear and anger in humans. Psychosom Med 1953;
15: 433442.
32. WEERTS TC, ROBERTS R. The physiological effects of imagining anger-provoking scenes.
Psychophysiology 1976; 13: 174 (abstr.).
33. MASON JW. Organization of psychoendocrine mechanisms: A review and reconsideration of
research. In Handbook of Psychophysiology (Edited by GREENFIELD NS, STERNBACH RA). New York:
Holt. Rinehart and Winston, 1972: 3-121.
34. WATERS WF, BERNARD BA, Buco SM. The Autonomic Nervous System Response Inventory
(ANSRI): Prediction of psychophysiological response. J psychosom Res. 1989; 33: 347-362.