Content uploaded by Kay A Wilhelm
Author content
All content in this area was uploaded by Kay A Wilhelm on Dec 17, 2013
Content may be subject to copyright.
Content uploaded by Kay A Wilhelm
Author content
All content in this area was uploaded by Kay A Wilhelm on Dec 17, 2013
Content may be subject to copyright.
RELIABILITY
OF
THE PARENTAL
BONDING INSTRUMENT AND INTIMATE
BOND MEASURE SCALES
Kay Wilhelm and Gordon Parker
The long-term reliability
of
the Parental Bonding Instrument (PBI) and
of
the
Intimate Bond Measure (IBM) are examined in a non-clinical group, with data
being examined over eleven and five years for the two respective measures.
Such reliability data are compared with reliability data on a number
of
per-
sonality measures within the same cohort. Results demonstrate considerable
stability in the PBI over an extended period and moderate stability in
IBM
scores.
Australian and New Zealand Journal of Psychiatry 1990; 24:199-202
The Parental Bonding Instrument (PBI) was
developed to measure fundamental parental dimen-
sions of care and protection (control) and to allow
quantification of any parental contribution to sub-
sequent psychiatric disorder. Reliability aspects (inter-
nal consistency and short-term test-retest reliability)
were examined in the initial paper
[
11
and its validity,
both as a measure of perceived and actual parenting,
has been examined in several studies
[2].
If
the PBI is a reliable and valid measure of subjects’
retrospective memories of their parents in their first
sixteen years, then adult subjects should return consis-
tent scores over time (ie show high test-retest
reliability).
The short-term reliability of the PBI has been ex-
amined in several studies.
In
the initial development
paper
[
11,
test-retest reliability in a non-clinical sample
was 0.76 for the care scale and 0.63 for the protection
scale over a three-week interval. Subsequently, in
a
sample of depressives initially depressed and then
Division
of
Psychiatry, Prince Henry Hospital, Little Bay,
NSW
Kay Wilhelm, FRANZCP, Staff Psychiatrist
Gordon Parker,
MD,
PhD,
FRANZCP, Professor
of
Psychiatry
Correspond with
Dr
Wilhelm
significantly improved, much higher correlation coef-
ficients (ranging from 0.87 to 0.92) were returned over
a nine-weekinterval
[2].
The higher coefficients in this
group, compared to the initial sample, were judged to
reflect the greater motivation
of
patients (in com-
parison to volunteer or importuned non-clinical
groups) to return questionnaire data conscientiously.
Subsequently, in a
US
study [3] of depressed out-
patients attending the Yale Depression Research unit,
48
depressives scored the PBI when depressed and
some four-six weeks later when significantly im-
proved. PBI scores showed
no
significant change over
time and the coefficients of agreement ranged from
0.90 to 0.96 across the four scales, slightly superior to
the Australian depressed sample. Test-retest reliability
in a group of patients with schizophrenia has been
examined [4] when the coefficients of agreement
ranged from 0.58 to 0.77. This less impressive result
was judged to be a reflection of the sample initially
scoring the PBI shortly after admission to hospital with
an exacerbation of their schizophrenia, with their
judgement and ability to complete the self-report ques-
tionnaire impaired. This interpretation is supported by
an
American study
[5]
of
26
subjects with
schizophrenia who completed the PBI form
on
two
occasions a few weeks apart, with correlation coeffi-
Aust NZ J Psychiatry Downloaded from informahealthcare.com by University New South Wales on 12/10/13
For personal use only.
200
RELIABILITY OF THE PBI
AND
IBM SCALES
cients ranging from 0.79 to 0.88, with this sample
being distinguished by the sample being selected from
those attending a community mental health centre, and
not assessed during a relapse.
Medium-term reliability data have been provided
[6], with correlation coefficients ranging from 0.79 to
0.81
on
the PBI for a non-clinical
U.S.
sample tested
seven months apart. In the only published study [7] of
the long-term reliability of the PBI, women were
studied in the post-partum period and then two to four
years (mean
30
months) later. Only the maternal PBI
form was completed by subjects, and the authors
reported that PBI scores were “remarkably stable over
time” for different sub-groups of mothers who were
either depressed on both occasions of testing, initially
depressed and then recovered, or not depressed on
either occasion.
In this paper we examine the test-retest reliability of
the PBI over a decade. The sample has been described
elsewhere [8] and
so
details only will be summarized
here. In September 1978,380 students who had under-
taken a basic Arts or Science university course, and
who were then completing a one-year training
programme at the Sydney Teachers’ College were
approached in class and invited to participate in a
longitudinal study. While our key objective was to
examine for sex differences in depressive experience
over time, the students were not informed about the
specific hypothesis, but were given details on the
range of topics (including depression) and develop-
mental issues that would be assessed longitudinally.
Those subjects completed PBI data and 170 agreed to
take part in the longitudinal study, and
so
formed the
study cohort, then having a mean age of 23.1 years.
Those taking part and those declining did not differ in
PBI scores returned for each parent. Subsequently, we
sought to interview the cohort serially, and self-report
data were obtained from 164 in 1983 and 163 in 1988.
On
each occasion the subjects were requested to com-
plete the orthodox PBI forms (assessing parenting over
the first 16 years), allowing
us
to compare PBI data
collected over extended periods.
The authors have also developed a measure of fun-
damental dimensions underlying adult intimate
relationships [9]. The test-retest reliability of that self-
report, the Intimate Bond Measure (or IBM) was as-
sessed in the initial paper, with a non-clinical sample
returning data on two occasions over a three-six week
interval, with reliability coefficients being very high
at 0.80 and 0.89. That measure was given to our
present cohort in 1983 and in 1988 and we now report
the test-retest reliability over a five-year interval. It
must be kept in mind, however, that subjects would
not necessarily be scoring the same “intimate” on both
occasions,
so
that we report consistency data for the
whole sample and for a sub-sample of those who were
married in 1988 and had rated the same “intimate” in
1983
-
a fairer test of the measure’s reliability. On each
occasion, subjects were asked to score characteristics
of the intimate “in recent times”, the IBM being more
a measure of state or current characteristics.
We also take the opportunity to report test-retest
reliability over the same extended period for a number
of other measures. We do that for several reasons.
Firstly, such reports are rare and, more importantly,
those data provide some basis for comparison against
the PBI and IBM. It is generally suggested that per-
sonality is constant and we might therefore expect that
high test-retest reliability would be demonstrated for
personality measures and give a base quantitative es-
timate of reliability against which we could judge PBI
data, in particular. That is, if personality
is
immutable,
then self-report measures of personality should show
a high level of constancy, being weakened only by
response biases and state effects (eg depression) which
are generally accepted to influence self-report scoring.
Thus, we would expect that if the PBI is a reliable
measure, reliability coefficients should be similar to
those returned on personality measures. The per-
sonality measures considered were the Eysenck Per-
sonality Inventory neuroticism scale
[
101, the
Rosenberg self-esteem scale,
[
111 the dependency
scale from the Depressive Experiences Questionnaire
[
121, the Costello-Comrey trait depression scale
[
131
which was designed to measure a “person’s tendency
to experience a depressive mood”, and the Bem sex
role inventory
[
141 (with masculinity, femininity and
social desirability sub-scales), the last being ad-
ministered in 1983 and 1988 only. Finally, we report
data on the Wilson-Lovibond state measure of depres-
sion
[
151 to again allow comparison against the “trait”
measures, anticipating that much lower levels of
agreement should be demonstrated on a state measure
over time.
Results
Table 1 reports the mean data returned for the
several measures, and the level of constancy over time,
with three intervals (1978-1983, 1983-1988, and
Aust NZ J Psychiatry Downloaded from informahealthcare.com by University New South Wales on 12/10/13
For personal use only.
KAY
WILHELM
AND GORDON PARKER
20
1
Table
I.
Consistency in scores examined
over time
Mean score Consistency
coefficients (r)
1978 1983 1988
A
0
A
with with with
(A)
(6)
(C)
c
c
PBI
Aaternal care
26.3 26.2 26.3 0.72 0.82 0.6:
inaternal protection
14.8 13.8 13.8 0.74 0.76 0.61
’aternal care
21.9 21.7 21.4
0.80
0.82 0.7;
'sternal
protection
13.0 12.1 11.9 0.69 0.67
05
BM
ntimate care
wholegroup)
N/A* 30.6 29.5 N/A
0.50
N/A
ntimate control
wholegroup)
N/A 6.7 6.9 N/A 0.49 N/A
ntirnate care
sub-group)”
N/A 31.5 30.2 N/A 0.48 N/A
ntimate control
sub-group)
N/A 6.5 7.0 N/A 0.49 N/A
Personality
qeuroticism
9.0 8.7 8.7 0.54 0.68
0.5
Self-esteem’”*
1.6 1.0 0.9 0.43 0.61 0.4
lependency
52.6 52.8 53.7 0.64 0.64 0.5
rraitdepression
31.3 31.2 29.4 0.64 0.65 0.4
Sex role Inventory
Masculinity
N/A 4.6 4.7 N/A 0.56 NII
Femininity
N/A 4.7 4.8 N/A 0.62 N/I
Socialdesirability
N/A 5.2 5.3 N/A
0.57
N/I
Statedepression
57.0
55.2
55.3 0.25 0.23 0.1
‘
N/A
=
not assessed
’*
’**
Higher scores indicate lower self-esteem
Those nominating the same intimate in
1983
and
1988
(N=96)
1978- 1988) being examined. Mean scale scores were
generally stable for most measures apart from self-es-
teem, which showed a distinct improvement from
1978 to 1983 (t
=
4.80, P<O.OOI) and from 1978 to
1988 (t
=
5.82,
P<
0.001)
but
no
change from 1983 to
1988 (t
=
0.87,
ns).
In
a similar, but less distinct
fashion, depression scores decreased from 1978 to
1988, both
on
the trait (t
=
2.03, PcO.05) and state
(t
=
2.21, P<0.05) measures.
Presumably because of the large sample size (for
mean scores were very similar), two significant dif-
ferences were established for the PBI and the IBM
measures. Thus paternal protection scores dropped
from 1978 to 1988 (t
=
2.12, p<0.05) while IBM care
scores decreased from 1983 to 1988 (t=2.86,
p<0.05).
Scores were generally more consistent in the interval
1983-1988, when the cohort had left university and
most were in full-time employment. The data for the
1
I-year interval (1978-1988) show slightly less con-
sistency, presumably reflecting changes in attitudes
over the lengthier assessment period. The test-retest
consistency in PBI scores is extremely impressive for
all three test intervals. Thus, the mean correlation
coefficients were 0.74 (1978-1983), 0.77 (1983-1988)
and
0.65
(1978-1988), contrasting with the four per-
sonality measures which returned mean correlations of
0.56,
0.64
and
0.50
respectively. Additionally, the
correlation coefficients for the PBI were superior to
each individual personality test. Stability in IBM
scores (both for the whole sample and for the sub-
group scoring the same partner) was moderate from
1983 to 1988 and clearly less stable than PBI scores,
as might be anticipated for a state measure. Finally, the
test-retest reliability of the state depression measure
was low, as anticipated.
Discussion
The sample allows a “best estimate”
of
reliability,
in that it involved a non-clinical group, comprising
volunteers who were prepared to take part in a
lon-
gitudinal study and who were, at each review period,
judged to be generous in giving their time for extended
interviews and open in their discussion with the inter-
viewers. Thus, we judge that sample members were
likely to have completed self-report measures con-
scientiously and as accurately as possible. Additional-
ly, being a non-clinical sample, it is unlikely that mood
disturbance
or
related factors weakening reliability
estimates would have been over-represented in the
sample. The data set therefore offers a “best estimate”
potential for any examination of reliability and
it
would be unlikely that non-volunteer or certain clini-
cal groups would return such high levels of agreement.
Thus, we acknowledge the unique characteristics of
Aust NZ J Psychiatry Downloaded from informahealthcare.com by University New South Wales on 12/10/13
For personal use only.
202 RELIABILITY OF THE PBI AND IBM SCALES
the cohort but suggest that such a group is the ap-
propriate one for such an examination.
The test-retest reliability of the PBI is clearly im-
pressive, both intrinsically when the correlation coef-
ficients are examined and, secondly, in comparison to
the “personality” tests which we used as our compara-
tive base.
Jorm
[
161 considered the test-retest consis-
tency of trait anxiety/neuroticism measures, and
referenced work giving a correlation
of
0.54
over the
four-six years and
0.40
over
30
years, with lower
levels for state measures of anxiety. Those data are
compatible with our neuroticism score data, and there-
fore support the likely accuracy of the latter.
The test-retest data for the IBM are somewhat less
impressive and, as we undertook a separate analysis
on
those in stable relationships with similar results,
findings cannot reflect a sub-sample scoring different
intimates over time.
As
the IBM is a measure of current
intimate relationships, some change in
the
perception
of the spouse or intimate would be anticipated over
time, particularly in such a sample of young adults.
While change in parenting over the years might also
be theoretically anticipated, the PBI, by contrast, is
designed deliberately to obtain an overall gestalt of the
parent or “product moment of innumerable experien-
ces”
so
that the instructions effectively force some
overall judgement in an attempt to minimise variation
at different developmental stages.
Each
of
the “personality” measures returned similar
levels
of
consistency over time despite some of them
(eg trait depression) conceivably being more likely to
be influenced by mood state and therefore potentially
unstable over an extended period. The extent to which
they necessarily reflect intrinsic personality, however,
cannot be addressed by such a study design.
We conclude then
that
the PBI is a highly reliable
measure over an extended period, supporting its claim
to be accurate measure of perceived parenting, and
so
useful in quantifying any parental risk to subsequent
psychiatric disorder in adulthood.
We thank our colleagues at the Mood Disorders Unit
for comments on this paper, the
NH
&
MRC
and NSW
Institute of Psychiatry for funding the cohort study,
and
Mrs
Sandra Evans for typing of the manuscript.
I.
Parker
G.
A parental bonding instrument. British Journal of Medi-
cal Psychology 1979;52:1-10.
2. Parker
G.
Parental overprotection: a risk factor in psychosocial
development. New York: Grune
&
Stratton, 1983.
3. Plantes
MM,
Prusoff BA, Brennan J, and Parker
G.
Parental repre-
sentations
of
depressed outpatients from a U.S. sample. Journal of
Affective Disorders
1988;
15:
149-155.
4. Parker G, Fairley M, Greenwood J, Jurd
S,
and Silove D. Parental
representations of schizophrenics and their association with
onset
and course of schizophrenia. British Journal
of
Psychiatry
1982; 141573-581.
schizophrenic patients’ perceptions of their parents and the course
of their illness. British Journal of Psychiatry 1988; 153:344-353.
depressive mood over time: Effects of internalized childhood at-
tachments. Journal of Nervous and Mental Disease 1987; 175:703-
712.
7. Gotlib
IH,
Mount
JH,
Cordy NI and Whiffen VE. Depression and
perceptions of early parenting: A longitudinal investigation.
British Journal
of
Psychiatry 1988; 152:24-27.
8.
Wilhelm
K
and Parker G. Are
sex
differences in depression really
necessary? Psychological Medicine 1989;19:401-414..
9. Wilhelm
K
and Parker G. The development of a measure of in-
timate bonds. Psychological Medicine 1988; 18:225-234.
10.
Eysenck HJ and Eysenck SB. Manual of the Eysenck Personality
Inventory. London: Hodder and Stoughton, 1964.
11. Rosenberg M.
The
measurement of self-esteem. Society and the
adolescent self-image. New Jersey: Princeton University Press,
1965.
12. Blatt SJ, D’Affliti JP and Quinlan DM. Depressive Experiences
Questionnaire. Unpublished report, Department of Psychology,
Yale University, 1975.
13.
Costello
CG
and Comrey AL. Scales
for
measuring depression
and anxiety. Journal of Psychology 1967;66:303-3 13.
14. Bern
SL.
The measurement
of
psychological androgyny. Journal
of Consulting and Clinical Psychology 1974;42: 155-162.
15. Wilson PH. Behavioural and pharmacological treatment of
depression. Unpublished PhD thesis. University of New South
Wales, 1979.
16. Jorm AF. Modifiability of trait anxiety and neuroticism: A meta-
analysis of the literature. Australian and
New
Zealand Journal
of
Psychiatry 1989;23:21-29.
5.
Warner R, and Atkinson
M.
The relationship between
6.
Richman, JA, and Flaherty JA. Adult psychosocial assets and
Aust NZ J Psychiatry Downloaded from informahealthcare.com by University New South Wales on 12/10/13
For personal use only.
