You are on page 1of 2

Safety Science Vol. 24, No. 3, pp.

229-236, 1996
Pergamon Published by Elsevier Science Ltd. Printed in the Netherlands
0925.7535/97 $17.00 + 0.00

Book Reviews
Predicting Health Behaviour: Research and Practice with Social Cogni-
tion Models, edited by Murk Conner and Paul Norman, Open University
Press, Buckingham, 1996. ISBN 0-335-19320-X. 230 pp. (pbk), price E14.99.

Tom Sa yver: Are there fish in that river? I’m hungry.


Huckleberry Finn: There must be - there are bubbles on the surface. We need a rod and then they will all be
ours. Here’s a twig, a piece of string and a bent nail. Let’s start.
Tom Sawyer: Do you think we should go home and get my rod - it’ll only take half an hour.
Hucklebery Firm: No. There are so many fish and this twig looks as though it will work fine.

2 hours later

Tom Sawyer: Well we haven’t caught any fish.


Huckleberyv Finn: No, but I’m sure that the fish are there and that’s what really counts. I’ve had a couple of tugs
on the line and that certainly proves the fish are there.

Moral 1. Common sense needs to be proved. Indications are not proof. Moral 2. If you do
not have good quality tools, you will not know for sure that the fish are there.
Throughout ‘Predicting Health Behaviour’ we are presented with the details of studies in
which researchers have failed to predict health behaviour. The models described; the Health
Belief Model (HBM), Health Locus of Control (HLC), Protection Motivation Theory (PMT),
the Theory of Planned Behaviour and Health Behaviours (TPB) and Self-Efficacy (SE) all
seem to be reasonable theoretical explanations of health behaviour. In reality, they rarely
explain more that 5% of the variance in health behaviours.
It is difficult for the reader to determine whether this is due to the ‘common sense’ theories
being inappropriate or to the poor experimental design that has been employed by those
applying the models. Research in this field uniformly employs measurement scales that lack
adequate coverage and include items written by the researchers rather than derived from
qualitative interviews with relevant respondents. The scales have inadequate consistency and
are not tested for reliability or validity. To compound these failings, the researchers choose to
apply statistical models to the data collected that cannot be justified by the properties of the
scales employed.
The editors’ introductory chapter is followed by five more, each describing a different
model. Each of these chapters consists of an introduction to the model, a review of key
research studies and recent developments, how the model has been operationalised, a sample
application relevant to health behaviour and concluding remarks. A final chapter, by the
editors, considers future directions in the field.
Sheeran and Abraham describe the HBM. They report that this model explains between 0.5
and 4% of variance in behaviour due to “poor operationalisations of the model and failure to
check the reliability and validity of constructs”. They also conclude that the model excludes
cognitions (unidentified) that have been shown to be powerful predictors of behaviour and

229
230 Book re~.icws

fails to account for behaviour under social and affective control. Unfortunately, the authors
provide a sample of ‘good’ methodology that is far from impressive.
The HLC model is described by Norman and Bennett. They conclude that the “evidence
linking internal HLC beliefs with global indices of preventive health behaviour has been
weak”. Their sample study is Heartbeat Wales, which looked at the dietary choice of 13,000
people. This study also employed inadequate scales and was found to explain less than 5% of
the variance in dietary choice.
Boer and Seydel use their own study of PMT related to mass screening for breast cancer, as
the sample study. Again, this was able to predict only 5% of the variance in participation in
screening.
TPB is covered by Connor and Sparks. The poor quality of scales employed in this field is
well illustrated in this chapter. Sample items such as “I ate fruit as part of my midday meal
today” (with seven response options ranging from “definitely did not” to “definitely did”)
and “most people who are important to me think I should/should not eat fruit as part of my
midday meal tomorrow” are included. The sample study (on healthy eating) again predicts 5%
of variance in behaviour. The authors state that there have been numerous ‘successful’
applications of TPB but do not appear to have included them in the chapter.
SE appears to be a more promising approach, as reviewed by Schwarzer and Fuchs. The
model was able to predict 20% of the variance in healthy eating behaviour, although questions
should be asked about the use of multiple regression with the data presented. This chapter
suffers from the inclusion of a great deal of theoretical information unsupported by evidence.
The final chapter of the book is remarkably frank. Norman and Conner admit that there is
little to choose between the models and that they share many similar constructs. They also
admit that little attention has been paid to the psychometric properties of the scales developed
to test the models. The authors report that “many health behaviours are determined not by
social cognitive variables... ...but rather by one’s previous behaviour”. However. past
behaviour may predict rather than explain future behaviour.
Overall, the book may appeal to researchers working in the field of social cognition. The
content should be approached warily by students. Much of the advice provided on the
development and testing of rating scales is dated and inappropriate (notably that by Sheeran
and Abraham). It is unfortunate that so few researchers understand what reliability is, how it
should be assessed and why high reliability is such a crucial requirement. The continued
reliance on Cronbach’s (Y as an estimate of reliability is hard to justify now that it has been so
clearly discredited.
On the positive side, the book does bring together a considerable amount of material not
previously available in any one volume.
Unfortunately, the attempt to include so much material has led all the authors to leave out
the most important information from the studies reported, the proportion of variance in
behaviour predicted. Too often statements such as “the constructs are useful predictors” or
“significant predictors” are used. At times lists of studies are given without any description of
the work. It is to be hoped that more scientific rigour will be applied to the study of social
cognition models in the future. Then we may end up with fish rather than a worm on the end
of our line.

Stephen P McKenna
Galen Research, Manchester, UK

You might also like