Full text of DOI:10.1177/135910530100600407

Journal of Health Psychology
Copyright © 2001 SAGE Publications
London, Thousand Oaks and New Delhi,
[1359–1053(200107)6:4]
A Structural Evaluation
of the Expanded
Vol 6(4) 447–455; 017544
Multidim ensional
Health Locus of
Control Scale with a
Diverse Sam ple of
Caucasian/European,
Native, and Black
Canadian Wom en
Ab st ra ct
The Multidimensional Health
Locus of Control (MHLC) scale is
a widely used measure of the
factors that people believe exert
some control over their health.
The purpose of the research
reported here is to provide an
independent evaluation of the
structural properties of the
recently expanded MHLC.
Specifically we sought to address
the issue of the interdependence
of the external locus of control
scales (Powerful Others, God, and
Chance) and the legitimacy of the
addition of a measure of God
control as a separate subscale in
the expanded MHLC. The study is
based on an ethnically diverse
sample of 371 community women,
recruited through a random digit
dialing procedure, who responded
to all the items on the expanded
MHLC. Although, the three
external factors (God, Powerful
Others, and Chance) are
W I L L I A M F. C H A P L I N, K A R I NA
D AV I D SO N, V I R G I N I A SPA R R O W,
J U D I T H ST U H R , E R I C A VA N
R O O SM A L E N, & K E N N E T H A .
WA L L STO N
University of Alabama, USA
substantially correlated, structural
analyses indicate that a four-factor
model consisting of the three
external scales and the Internal
control scale provides the best fit
to the observed covariances
among the items. We view this
result as supporting the addition of
the God subscale as a separate
dimension of external health locus
of control. We also find support in
these results for the continued
scoring of four subscales on the
MHLC rather than combining the
three external subscales on one
dimension as some have
A C K N O W L E D G E M E N T S. The research reported here was supported
by the Maritime Center for Excellence in Women’s Health. The
analyses of the data on which this manuscript is based were conducted
while the first author was a visiting exchange professor at Phillips
University in Marburg, Germany
suggested.
C O M P E T I N G I N T E R E ST S: None declared.
Ke ywo rd s
A D D R E SS. Correspondence should be directed to:
W I L L I A M F. C H A P L I N , Department of Psychology, University of
Alabama, Box 870348, Tuscaloosa, AL 35487-0348, USA [email:
wchaplin@bama.ua.edu]
confirmatory factor analysis,
culturally diverse sample, locus
of control scale
447
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

JO U R NA L O F H E A LTH PSYCH O LO G Y 6(4)
T H E M U LT I D I M E N SI O N A L Health Locus of Hinrichsen, & Pollack, 1989) and non-clinical
Control Scale (MHLC) was developed in 1978 populations (e.g. Casey, Kingery, Bowden, &
to assess beliefs about the source of reinforce- Corbett, 1993; Marshall, Collins, & Crook, 1990;
ments for health-related behaviors (Wallston, Robinson-Whelen & Storandt, 1992). However,
Wallston, & DeVellis, 1978). The original a number of other studies failed to clearly
MHLC contained three subscales: Internal, support a three-factor structure of the original
Chance, and Powerful Others. Recently a fourth MHLC (e.g. Boyle & Harrison, 1981; Cooper &
subscale that focuses on the belief that God is a Fraboni, 1990; Gutkin, Robbins, & Andrews,
source of reinforcement for health was added to 1985; Meyers, Donham, & Ludenia, 1982;
the measure (Wallston et al., 1999). The purpose O’Looney & Barrett, 1983; Rogers, 1995). In
of the research reported here is to evaluate the general these studies suggest that a two-factor
new four-subscale structure of this expanded structure (internal versus external) in which
version of the MHLC on a diverse sample of Chance and Powerful Others combine into one
healthy women using confirmatory factor analy- scale is correct.
sis.
Th e e xp a n d e d fo u r-su b sca le
ve rsio n o f t h e MHLC fo r
h e a lt h y p o p u la t io n s
St ru ct u re o f t h e o rigin a l
MHLC
There are two versions of the original MHLC. Although the initial item pool for the develop-
Most of the work with the MHLC scale in ment of the MHLC contained a number of items
healthy populations has been with Form A, assessing the extent to which God was perceived
whereas Form B has been favored by those as responsible for the respondent’s health status,
studying persons with chronic illnesses. Cron- none of those items were included in Form A or
bach’s coefficient alpha reliability for the three Form B of the MHLC scale. This was primarily
subscales of Form A of the MHLC in the because Wallston et al. (1978) wanted to maxi-
development sample ranged from .67 for Power- mize the orthogonality of the Powerful Others
ful Others to .77 for Internal. The intercorrela- and Chance subscales, and the items concerning
tion matrix for the MHLC scale revealed that the role of God tended to correlate with items on
the Internal and Powerful Others subscales are both subscales. The recognition of the central
statistically independent, and the Internal and role that a belief in God plays in the lives of many
Chance subscales are only slightly negatively persons led Welton, Adkins, Ingle, and Dixon
correlated. The Powerful Others and Chance (1996) to add a six-item God subscale to Form A
subscales were positively correlated. Each of the of the original MHLC. The initial psychometric
three subscales of the MHLC significantly corre- evaluations of this scale in samples of under-
lated with its counterpart on Levenson’s (1981) graduates were promising. Wallston and his
Internal, Powerful Others, and Chance scales colleagues had also added a God subscale to a
(Wallston et al., 1978).
version of the MHLC that was tailored to
Following the publication of Forms A and B persons with rheumatoid arthritis (Wallston,
of Wallston et al.’s MHLC scale, and before the Stein, & Smith, 1994). The alpha reliability of the
addition of the God subscale, a number of inde- God Locus of Arthritis Control subscale was .91
pendent structural evaluations of the MHLC in one sample of rheumatoid arthritis patients
were conducted. For example, Lynda Anderson and .87 in a second sample (Wallston et al., 1999).
(1989) found support for the three-factor model
using LISREL VII to analyze data from Form A.
The three-factor structure was found for both
Th e p re se n t st u d y
520 persons with one or more major chronic The God subscale represents a major addition to
health conditions and for 1504 persons who had the widely used MHLC. Thus, an evaluation of
no identifiable health problems. Several other the structure of the expanded MHLC on an inde-
studies confirmed this structure in both clinical pendent sample of ‘healthy’ persons is of some
(e.g. Hartke & Kunce, 1982; Russell & Ludenia, interest. The present study was designed to
1983; Talbot, Nowen, & Gauthier, 1996; Wall, provide that evaluation. Specifically, we evaluate
448
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

CH A PLIN E T A L.: H E A LTH LO CU S O F CO NTRO L
several possible structural models of the province-wide random probability sample of 300
expanded MHLC, including an evaluation of the women stratified across the province’s 18 coun-
alternate two-factor model (external versus ties. The remaining two sample sets were
internal) that has been found in some previous designed to target specific communities in Nova
analyses of the three-subscale version of the orig- Scotia with the highest incidence of Aboriginal
inal MHLC, as well as the predicted four-factor and black women. The sample of Aboriginal
model. We also consider models in which the women was drawn from Nova Scotia reservation
factors are independent and correlated. The data communities as defined by Statistics Canada
that were available for our analyses came from a proportionate to population. The incidence of
diverse sample of healthy women (Barksdale, Aboriginals in these communities ranged from
Willis, Davidson, Loppie, & Van Roosemalen, 69 to 99 percent. The sampling frame for Black
1999) who participated in a survey of women’s women was based on census divisions with an
health concerns.
incidence of black population over 3 percent.
Table 1 shows the percentage of women for the
three ethnicity strata who were found and who
were willing to complete the survey.
Me t h o d
Sam ple design and selection
To complete surveys with 300 Nova Scotia Participants
women, 75 Aboriginal women and 75 black Table 2 provides a demographic description of
women, a private survey company conducted the participants in this study. Not surprisingly,
telephone interviews with women from each the Caucasian women reported having more
corresponding group who were aged 18 years education and a higher average household
and older. The sample for each group was drawn income than the native and black women. In
using systematic sampling procedures from a list addition, Caucasian women were more likely to
of randomly selected households compiled from have been married than the other two ethnic
listed provincial telephone numbers, drawn groups. The native women tended to be younger
from a database that is updated quarterly. than the Caucasian or black women.
Samples were selected to match the geographi-
cal distribution of the population derived from Procedure
Statistics Canada 1996 Census profiles. To mini- A telephone survey lasting approximately 20
mize the margin of sampling error across county minutes was conducted by a telephone survey
subsamples, the first sample frame comprised a company. First, the interviewer identified
Table 1. Record of random-digit dialing calls made for the Nova Scotia Women’s Health Survey
Caucasian/European
Native/Aboriginal
Black/African
Original base
Total calls
Invalid phone numbers
Did not qualify
2224 (100% )
493 (22.2% )
91 (4.1% )
1897 (100% )
386 (20.4% )
319 (16.8% )
5880 (100% )
1269 (21.6% )
760 (12.9% )
New base
Valid phone numbers
No answers/busy/answering
machines
1640 (100% )
385 (23.5% )
1192 (100% )
354 (29.7% )
3851 (100% )
1711 (44.4% )
Qualified but not available
Language problem
Quota full
Callbacks
Refused to complete
Refusals
61 (3.7% )
7 (0.4% )
17 (1.0% )
275 (16.8% )
9 (0.6% )
581 (35.4% )
3 (0.2% )
302 (18.4% )
55 (4.6% )
8 (0.7% )
6 (0.5% )
264 (22.2% )
3 (0.2% )
411 (34.5% )
10 (0.8% )
81 (6.8% )
171 (4.4% )
35 (0.9% )
46 (1.2% )
585 (15.2% )
16 (0.4% )
1197 (31.1% )
15 (0.4% )
75 (2.0% )
Duplicates
Total number completed
449
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

JO U R NA L O F H E A LTH PSYCH O LO G Y 6(4)
Table 2. Demographic characteristics of survey participants
Ethnicity
Demographic variable
Caucasian/European
Native/Aboriginal
Black
Age
45.4 (14.7)
37.7 (12.7)
45.5 (15.7)
Marital status
Never married
Married/common-law
Divorced/widowed
9.7 (% )
74.2
11.7
34.6 (% )
40.7
16.0
30.6 (% )
43.1
15.3
Education
≤ High school
42.6 (% )
26.4
8.0
59.2 (% )
4.8
16.0
57.0 (% )
15.3
15.3
Trade/tech./other school
Some university
≥ Graduated university
23.1
19.7
12.5
Household income
< $20,000
$20,000–$49,999
≥ $50,000
22.7 (% )
39.8
22.7
39.5 (% )
17.3
7.4
26.4 (% )
40.3
9.7
herself, and asked to speak with a woman who extent to which missing values might compro-
lives in the household who is 18 years of age or mise our analyses. In addition, we evaluated the
older. Then, the interviewer briefly explained the characteristics of the four a priori subscales (i.e.
purpose of the interview and attempted to obtain Internal, Chance, God, and Powerful Others).
consent. If consent was given, respondents were
asked their age and their major ethnic identity. Missing values
We used a phone interview because the partici- A substantial number of the 458 research par-
pants were selected by sampling from telephone ticipants did not respond to at least one of the 24
numbers using random digit dialing. The phone health locus of control items. The number of
administration of the MHLC is more representa- missing responses ranged from 0 (n = 371) to 10
tive of its applications as a research instrument. (n = 1). Of the 87 participants with missing
However, we have no a priori reason to expect responses, 66 had only one or two missing
that the individual administration of the instru- responses, 14 had three or four missing
ment in clinical contexts would change the struc- responses, and the remaining seven had between
tural results reported here.
five and 10 missing responses. Because of the
Following some questions concerning their potential statistical problems and conceptual
health, Form A of the MHLC scale (Wallston et issues raised by replacing missing values with an
al., 1978), with the newly developed God Locus estimate (e.g. mean substitution) we elected to
of Health Control subscale items (Wallston et exclude all participants who had any missing
al., 1999) randomly interspersed, was verbally values. This reduced our sample to 371 partici-
administered to each woman. Finally, 14 ques- pants (248 Caucasians, 67 African Americans,
tions regarding pertinent demographic infor- and 56 Native Americans).
mation were asked. These questions assessed To gain some insight into the extent to which
family status, work history, educational back- this decision might bias our results, we created a
ground, and socioeconomic status.
variable to indicate if a participant was excluded
from our analyses (1 = excluded, 0 = not
excluded) and correlated this variable with the
scale scores, item responses, and age. The corre-
lations between this index of exclusion and the
Re su lt s
Prelim inary analyses
Prior to evaluating the structure of the 24 items scale scores ranged from –.12 to .05, whereas for
of the expanded MHLC we considered the the item responses the range was from –.11 to
450
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

CH A PLIN E T A L.: H E A LTH LO CU S O F CO NTRO L
.06. Although the correlation between the Inter-
Although all of the subscale intercorrelations
nal scale and exclusion (r = –.12) was statistically in Table 4 are positive and all but one are statis-
significant (p = .01), indicating that participants tically significant, the Internal subscale is rela-
with a more internal health locus of control are tively independent from the other three
slightly less likely to be excluded, the relation is subscales. The subscales that represent different
small, accounting for slightly more than 1 forms of external control (Chance, God, and
percent of the variance. Further none of these Powerful Others) are all substantially positively
correlations was statistically significant using a correlated. This finding suggests that for com-
Bonferroni correction. There was, however, a munity populations such as the one we studied,
small but statistically significant correlation combining these three subscales into a general
between exclusion and age (r = .18) with older External subscale might be justified. This is one
participants being more likely to be excluded. of the possible structures we evaluate in our
Finally, we considered whether exclusion was structural analyses of the items.
related to the ethnicity of the participant. There
was no evidence that participants from one Structural analyses
ethnic group were more likely to be excluded The focus of this article is on the structure of the
(based upon a missing response) than those 24 health locus of control items in a diverse
from another (chi-square with 2 degrees of sample of women. In addition to the possibility
freedom = 2.36, p = .31).
of a general control factor (a one-factor model)
as the basis for participants’ responses to the
items and a four-factor model that is consistent
Analyses of the subscale scores
Prior to the structural analyses, we considered with the proposed scoring of the expanded
the basic descriptive and psychometric proper- MHLC, we also evaluate a two-factor model
ties of the four proposed subscales. In Table 3 we that combines the God, Chance, and Powerful
report the descriptive statistics for the four sub- Others into a single External factor. All the
scales and in Table 4 we report the correlations structural analyses reported here were done as
among the four subscales. As shown in Table 3, confirmatory factor analyses using EQS
participants generally scored higher on the (Bentler, 1995).
Internal Health Locus of Control subscale than
Before assessing the goodness of fit of the
they do on any of the external subscales, and various models to the observed covariances
among the three external subscales they score among the items, we evaluated the data for
highest on Powerful Others. For the overall multivariate normality. The normalized esti-
multivariate analysis of repeated measures to mate of Mardia’s coefficient was 32.8, suggesting
compare these for subscales, Wilks lambda =.42, a possible violation of the assumption of multi-
F3, 368 = 170, p < .01). For the contrast between variate normality. Thus, we used a robust
the Internal subscale and the three external sub- maximum likelihood procedure to evaluate the
scales, F1, 370 = 405, p < .01. For the contrast fit of the different models. We report both stan-
between Powerful Others and the remaining two dard chi-squares and comparative fit indices
external subscales, F1, 370 = 136, p < .01. For the (CFI) and their robust counterparts (e.g. Sat-
remaining contrast between God and Chance torra-Bentler scaled chi-squares and robust
the difference was not significant.
CFI) available in EQS. All the models evaluated
Table 3. Descriptive statistics for the four Health Locus of Control scales
Scale
M
SD
Coefficient
alpha
Range of corrected
item–total correlations
Internal
Chance
God
Powerful Others
4.35
3.26
3.28
3.78
.65
.60
.68
.88
.65
.14–.45
.35–.47
.47–.77
.10–.49
.80
1.10
.80
N =371. There are six items on each scale. Ratings were made on a six-point scale where higher values
indicate more control
451
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

JO U R NA L O F H E A LTH PSYCH O LO G Y 6(4)
Table 4. Correlations among the four Health Locus of
Control subscales
acceptable fit. Thus, this model provides some
justification for the proposed scoring of four sep-
arate subscales, with the understanding that the
scores might not be strictly independent. Inspec-
tion of the estimated correlations among the
latent factors for this model indicated that the
Internal factor is largely independent of the
other three factors (the estimated correlations
were all .09), whereas the estimated correlations
among the three external factors are substantial
(.77, .71, and .56). This finding is consistent with
the correlations shown in Table 4 and suggests
that a more parsimonious scoring scheme might
be to reduce the God, Chance, and Powerful
Others subscales to a single External subscale.
However, as can be seen in Table 5, neither a
two-independent nor a two-correlated factor
model provide an acceptable fit to the correla-
tions among our participants’ responses to the
24 health locus of control items. Therefore, col-
lapsing the three external subscales into a single
External subscale is not supported by these
structural analyses. The last model considered in
Table 5 maintains the distinction among the four
factors but adds the constraint that the Internal
factor be independent of the three external
factors. Adding this constraint to the four corre-
lated factor model did not reduce the fit of the
model substantially. (The difference between
the chi-squares of the two models is 2.4, which,
with 3 degrees of freedom is not significant.)
Thus, the model that provides the most com-
pelling characterization of the structure of the
expanded MHLC in this sample of Canadian
women is a four-factor model with the three
Subscale
1
2
3
4
1. Internal
2. Chance
3. God
–
.10
–-
.14
.54
–-
.14
.50
.43
–
4. Powerful Others
N = 371. Correlations larger than .11 are significant at
the .05 level, two-tailed test
were fit after fewer than 10 iterations and none
of the models resulted in unusual parameter
estimates.
Evaluation of the m odels Table 5 summa-
rizes the goodness of fit of seven different
models to the observed covariance matrix. The
independence model represents the ‘null’
hypothesis that all the items are uncorrelated;
that is, there is no structure underlying the 24
health locus of control items. Not surprisingly,
this model yields a chi-square that is much larger
than any of the alternative models and can be
rejected. A one-factor model, implying that par-
ticipants’ responses to the items are influenced
by a general control factor, provides a better fit
to the data, but the fit of this model is still not
acceptable.
We next considered the four-factor model that
represents the proposed scoring of the scale
when administered to healthy samples.
Although treating the four factors as totally
independent did not provide a good fit to the
data, relaxing the constraint that the four factors
external
dimensions
correlated.
The
be independent resulted in
a marginally
Table 5. Summary of the goodness of fit indices for different structural models of the Health Locus of Control
items
Model
Degrees of Chi-square S-B scaled
CFI
Robust
CFI
Standardized
RMSR
freedom
Chi-square
Independence
1 factor
276
252
252
246
252
2679.7
963.4
763.3
490.4
764.3
–
–
–
–
763.4
625.3
399.6
610.8
.704
.787
.898
.787
.722
.797
.916
.805
.089
.156
.067
.084
4 independent factors
4 correlated factors
2 independent factors
(internal vs external)
2 correlated factors
4 factors with three external
factors correlated
251
249
762.0
492.8
609.6
401.2
.787
.899
.805
.917
.082
.069
N = 371. S-B = Sattora-Bentler; CFI = comparative fit index; RMSR = root mean squared residual
452
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

CH A PLIN E T A L.: H E A LTH LO CU S O F CO NTRO L
standardized estimates of the path coefficients analyses support treating it as a separate source
for this model are presented in Table 6.
of women’s beliefs about factors that control
their health. In addition, the God subscale has a
higher degree of internal consistency than the
other three subscales. One possible reason for
the higher internal consistency of the God sub-
scale is the greater variance of scores on this sub-
scale. Although the average degree to which
Discu ssio n
The addition of the God Locus
of Health Control (GLHC)
subscale
Although the God subscale is substantially cor- women attribute their health outcomes to God is
related with the Chance and Powerful Others significantly lower than the degree to which they
subscales in this population, our structural attribute health outcomes to internal factors or
Table 6. Standardized estimates of the path coefficients for the four-factor model with the ‘external’ factors
correlated for the HLCS
Factor
Item
Internal
Chance
God
Powerful
Others
Error
I1
I2
I3
I4
I5
I6
.36
.35
.14*
.58
.58
.73
.93
.94
.99
.81
.81
.68
C1
C2
C3
C4
C5
C6
.46
.41
.53
.55
.47
.60
.89
.91
.85
.84
.88
.80
G1
G2
G3
G4
G5
G6
.73
.49
.78
.78
.84
.83
.68
.87
.63
.62
.54
.56
P1
P2
P3
P4
P5
P6
.47
.50
.10*
.61
.60
.72
.89
.87
.99
.79
.80
.70
Correlations among the factors
Chance
–
God
Powerful Others
Chance
God
.71
–
Powerful Others
.77
.56
–
N = 371. I = Internal, C = Chance, G = God, P = Powerful Others. Items are grouped by factor. The number of
the item corresponds to the order it appears on the MHLC. All paths are statistically significant except for
those marked with an asterisk(*)
453
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

JO U R NA L O F H E A LTH PSYCH O LO G Y 6(4)
powerful others, the greater variance on the
God subscale indicates that, for some women,
t
t
810 = 9.56) and Powerful Others (2.81 vs 3.78,
810 = 16.99) scales, and slightly higher scores on
God is viewed as an important influence on their the Internal (4.46 vs 4.35, t810 = 2.28) and God
health. Thus, the addition of the God subscale to (3.49 vs 3.28, t810 = 2.40) scales than did our com-
the MHLC is justified psychometrically and it munity sample. (All above t-values are statisti-
appears to assess a factor that for some women cally significant at the .05 level, two-tailed test.)
is a major source of control beliefs of health. Of
Of greater importance for the structural
course, the predictive and construct validity of results reported here is that the correlations
the God subscale must be further assessed in among the four subscales suggest more scale
studies that focus on the relation of this scale to homogeneity for the community sample than for
various psychological and health outcomes. the college sample. For example, in the com-
Thus far, evaluations of the validity and utility of munity sample the correlations shown in Table 4
this new subscale appear promising (Wallston et are all positive, whereas in the above college
al., 1999; see also Welton et al., 1996).
sample the correlations between the Internal
scale and the three external sales are negative.
Moreover, the correlations among the external
Tw o factors versus four factors
The results of these analyses provide support for scales, although still positive, are lower in the
a four-factor structure of the 24 items on the college sample (averaging .26) than in the com-
expanded MHLC. Although the three factors munity sample (averaging .49). The homogene-
that represent external causes (Chance, God, ity of the scales in the community sample would
and Powerful Others) are substantially corre- generally serve to increase the goodness of fit of
lated in these data, our structural analyses clearly a one- or two-factor solution. Yet despite this
support the continued separation of these factors homogeneity we still found the best support for
rather than their aggregation into a single Exter- the four-factor solution in our community
nal factor as has been suggested by some analy- sample. This underscores the robustness of the
ses of the original MHLC (e.g. Boyle & Harrison, solution across non-college student samples.
1981; Gutkin et al., 1985; O’Looney & Barrett,
1983). Thus, the current scoring of the expanded Lim itations
MHLC on four separate subscales remains our Although our sample is diverse and based on a
recommendation. However, the substantial sophisticated sampling scheme, it has some
correlation among the three external subscales obvious limitations. First, it consists only of
must be considered when interpreting results women. The reason for this limitation was that
obtained from these measures. In particular, if the study in which the data presented here were
these subscales are used together as predictor collected focused on women’s concerns about
variables in a multiple regression analysis the their health (Barksdale et al., 1999). Restricting
multicollinearity of these subscales may pose the psychometric and structural analyses to a
some statistical and interpretive problems.
sample of women is not unreasonable because
there is some evidence that the factor structure
of the MHLC differs for men and women
Sam ple diversity
Although our sample suffers from the same (O’Looney & Barrett, 1983). However, an
volunteer bias as nearly all studies in psychol- evaluation of the expanded MHLC on a sample
ogy, we view the ethnic diversity and com- of men is an obvious target for future research.
munity recruitment of our sample as one of the The focus and sampling scheme of the study also
unique strengths of the study we report here. precluded the collection of criterion information
The difference between our sample and the against which the predictive validity of the
more usual college student samples used in MHLC scales could be evaluated.
psychological research can be appreciated by
A second limitation is that this study was con-
comparing the responses of our sample to those ducted in Canada. Health care systems differ
of a sample of 441 college women recently col- substantially across different countries, and the
lected by Willis and Wallston (1999). The availability and national philosophy of health
college students in this sample had substantially care could plausibly influence women’s beliefs
lower mean scores on the Chance (2.73 vs 3.26, about health locus of control. The extent of this
454
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015

CH A PLIN E T A L.: H E A LTH LO CU S O F CO NTRO L
influence will limit the cross-cultural generality
internality, powerful others, and chance. In H. M.
Lefcourt (Ed.), Research with the locus of control
construct (Vol 1). (pp. 15–63) New York: Academic
Press.
Marshall, G. N., Collins, B. E., & Crooks, V. C. (1990).
A comparison of two multidimensional health locus
of control instruments. Journal of Personality
Assessment, 54, 181–190.
Meyers, R., Donham, G. W., & Ludenia, K. (1982).
The psychometric properties of the Health Locus of
Control scale with medical and surgical patients.
Journal of Clinical Psychology, 38, 783–787.
O’Looney, B. A., & Barrett, P. T. (1983). A psycho-
metric investigation of the Multidimensional Health
Locus of Control questionnaire. British Journal of
Clinical Psychology, 22, 217–218.
Robinson-Whelan, S., & Storandt, M. (1992). Factor-
ial structure of two health belief measures among
older adults. Psychology and Aging, 7, 209–213.
Rogers, W. S. (1995). Deconstructing the Health Locus
of Control scale. Studia Psychologica, 37, 165–167.
Russell, S. F., & Ludenia, K. (1983). The psychometric
properties of the Multidimensional Health Locus of
Control scales in an alcoholic population. Journal of
Clinical Psychology, 39, 453–459.
Talbot, F., Nouwen, A., & Gauthier, J. (1996). Is health
locus of control a 3-factor or a 2-factor construct?
Journal of Clinical Psychology, 52, 559–568.
Wall, R. E., Hinrichsen, G. A., & Pollack, S. (1989). Psy-
chometric characteristics of the Multidimensional
Health Locus of Control scales among psychiatric
patients. Journal of Clinical Psychology, 45, 94–98.
Wallston, K. A., Malcarne, V. L., Flores, L., Hansdot-
tir, I., Smith, C. A., Stein, M. J., Weisman, M. H., &
Clements, P. J. (1999). Does God determine your
health? The God Locus of Health Control scale.
Cognitive Therapy and Research, 23, 131–142.
Wallston, K. A., Stein, M. J., & Smith, C. A. (1994).
Form C of the MHLC scales: A condition-specific
measure of locus of control. Journal of Personality
Assessment, 63, 534–553.
Wallston, K. A., Wallston, B. S., & DeVellis, R. (1978).
Development of the Multidimensional Health
Locus of Control (MHLC) scales. Health Education
Monographs, 6, 160–170.
Welton, G. L., Adkins, A. G., Ingle, S., & Dixon, W. A.
(1996). God control: The fourth dimension. Journal
of Psychology and Theology, 24, 13–25.
of these findings. Another target for future
research would be the evaluation of the
expanded MHLC in different countries.
Co n clu sio n
Despite the limitations of the research reported
here, we believe that the support we found for
including God as an additional factor in
women’s beliefs about the locus of control for
their health is impressive. Likewise, our results
endorse the continued scoring of the three
‘external’ subscales as separate, but related, con-
structs rather than combining them into a single
External subscale.
Re fe re n ce s
Anderson, L. (1989). Structure of the MHLC scales:
Healthy versus chronically ill populations. Unpub-
lished manuscript, Department of Psychology, Uni-
versity of Michigan, Ann Arbor, MI.
Barksdale, C., Willis, S., Davidson, K., Loppie, C., &
Van Roosemalen, E. (1999). Women’s health con-
cerns. Unpublished manuscript, Department of Psy-
chology, University of Alabama, Tuscaloosa,
Alabama.
Bentler, P. M. (1995). EQS structural equations program
manual. Encino, CA: Multivariate Software.
Boyle, E. S., & Harrison, B. E. (1981). Factor structure
of the Health Locus of Control scale. Journal of
Clinical Psychology, 37, 819–824.
Casey, T. A., Kingery, P. M., Bowden, R. G., Corbett,
B. S. (1993). An investigation of the factor structure
of the Multidimensional Health Locus of Control
scales in a health promotion program. Educational
and Psychological Measurement, 53, 491–498.
Cooper, D., & Fraboni, M. (1990). Psychometric study
of Forms A and B of the Multidimensional Health
Locus of Control scale. Psychological Reports, 66,
859–864.
Gutkin, T. B., Robbins, J. R., & Andrews, L. (1985).
The Health Locus of Control scale: Psychometric
properties. Educational and Psychological Measure-
ment, 45, 407–409.
Hartke, R. J., & Kunce, J. T. (1982). Multidimension-
ality of health related locus of control items. Journal
of Consulting and Clinical Psychology, 50, 594–595.
Levenson, H. (1981). Differentiating among
Willis, A. S., & Wallston, K. A. (1999). The college
students’ health beliefs study. Unpublished data.
Samford University, Birmingham, AL.
455
Downloaded from hpq.sagepub.com at University of Ulster Library on April 12, 2015