




 STEPHEN SUTTON
6 





 STAGE THEORIES OF HEALTH


 BEHAVIOUR







1 General background

That a whole chapter is dedicated to stage theories in this edition of the book
is indicative of the enormous amount of interest and research activity that
such theories have generated in recent years. This chapter discusses three
stage theories: the transtheoretical model (TTM; Prochaska and DiClemente
1983; Prochaska et al. 1992, 2002; Prochaska and Velicer 1997), the pre-
caution adoption process model (PAPM; Weinstein and Sandman 1992,
2002a, 2002b), and the health action process approach (HAPA; Schwarzer
1992, 1999, 2001, 2004; Schwarzer and Fuchs 1995a, 1995b), although we
argue that the last is not a genuine stage theory but a continuum theory like
the theory of planned behaviour (TPB; Ajzen 1991, 2002; Conner and
Sparks, Chapter 5 in this volume). Other stage theories that should be
mentioned, but are not discussed in this chapter, are the health behaviour
goal model (Gebhardt 1997; Maes and Gebhardt 2000), the Rubicon model,
or model of action phases (Heckhausen 1991; Gollwitzer 1996), a four-stage
model that forms the theoretical background to the work on implementation
intentions (see Sheeran et al., Chapter 7 in this volume), the AIDS risk
reduction model (Catania et al. 1990), and theories of delay in seeking
health care (Safer et al. 1979; Andersen et al. 1995).
We start by presenting a hypothetical three-stage theory to explain the
assumptions of stage theories and how they differ from continuum theories,
and we then discuss the TTM, the PAPM and the HAPA in turn. The
decision to discuss three theories necessitated some modications to the
recommended chapter format. In particular, because the PAPM and the
HAPA are relatively new and have limited evidence bases compared with
the TTM, sections on developments and application of the model are not
included for these two theories.
224 Stephen Sutton

1.1 A hypothetical three-stage theory

Figure 6.1 shows a hypothetical three-stage theory, in which the stages are
assumed to be discrete. According to the theory, a person can move from
Stage I to Stage III only via Stage II. The lower case letters a–e are causal
factors that are hypothesized to inuence the stage transitions. Increases in
factors a–c are assumed to increase the likelihood that the person will move
from Stage I to Stage II; similarly, increases in factors c–e are held to
increase the likelihood that a person in Stage II will move to Stage III. Thus,
variables a–e are the independent variables and the transitions from one
stage to the next are the dependent variables. In the simplest case, the latter
can be treated as dichotomous: a person either stays in the same stage or
moves to the next.

Figure 6.1 Hypothetical three-stage model

A key assumption of stage theories is that different factors are important

at different stages. In this example, the set of factors that inuence the
transition from Stage I to Stage II, {a, b, c}, differs from the set of factors
that inuence the transition from Stage II to Stage III, {c, d, e}. Note that
factor c inuences both transitions. This is allowable, even if c has the same
effect size for each of the two transitions, because the causal factors still
differ as a set.
A more fully specied version of the theory would also specify the causal
relationships among the explanatory factors that inuence each transition.
For example, for the rst transition, one might specify that factors a, b and
c each have direct effects on the probability of stage movement but that a
also has an indirect effect via b. This amounts to specifying a separate
causal model for each transition.
This is a very simple stage theory. In principle, a stage theory could
include only two stages, but in this case there would be only one forward
stage transition, so the assumption that different factors inuence different
transitions would not apply.
A stage theory may be made more complex by incorporating additional
stages and additional explanatory variables, and by allowing backward
transitions and transitions to non-adjacent stages. However, even the
 Stage Theories of Health Behaviour 225

simple three-stage theory outlined above has a more complex structure than
most of the theories discussed in this book. A corollary of this more com-
plex structure is that stage theories are also more difcult to test than other
kinds of theories. More specically, stage theories should be contrasted
with continuum theories. A classic example of a continuum theory is the
theory of reasoned action (TRA; Ajzen and Fishbein 1980; Conner and
Sparks, Chapter 5 in this volume). According to the TRA, the likelihood of
performing the target behaviour is a linear function of the strength of
intention to do so, which is treated as a continuous variable. A person may
move to action from any point on the intention continuum, though their
probability of doing so is assumed to be higher the further along the con-
tinuum they are. Two pseudostages, I and II, could be created by arbitrarily
dividing the intention continuum into two segments. People in these two
pseudostages would be expected to differ on variables that are assumed to
inuence intention strength (i.e. attitude and/or subjective norm). However,
this would not be a genuine stage theory for the following reasons:
1 the ‘stages’ have been arbitrarily created by dividing a continuum;
2 there is no assumption that people in Pseudostage I (low intention) have
to move into Pseudostage II (high intention) before they can move to
action;
3 everyone in the target population is assumed to have an intention with
respect to performing the target behaviour; similarly, everyone is
assumed to have an attitude and a subjective norm with respect to the
target behaviour;
4 the factors that inuence movement along the continuum (i.e. attitude
and/or subjective norm), and that therefore increase the likelihood of
action, are assumed to be the same at every point on the continuum and
to have the same effect sizes at every point on the continuum; and
5 the same intervention (i.e. one designed to increase attitude and/or
subjective norm) would be used regardless of the recipient’s position on
the continuum.

2 The transtheoretical model (TTM)

2.1 Description of the model
The TTM is the dominant stage model in health psychology and health
promotion. It was developed in the 1980s by a group of researchers at the
University of Rhode Island (hereafter referred to as the Rhode Island
group). The model derived partly from an analysis of systems of psy-
chotherapy but some of the rst empirical applications were to smoking
cessation (e.g. DiClemente and Prochaska 1982; Prochaska and DiCle-
mente 1983), and smoking remains the most popular application of the
model. Although it is often referred to simply as the stages of change model,
the TTM includes several different constructs: the stages of change, the pros
and cons of changing (together known as decisional balance), condence
and temptation, and the processes of change (Table 6.1). The TTM was an
226 Stephen Sutton

Table 6.1 The TTM constructs, adapted from Prochaska et al. (2002)

Construct Description

Stages of change
Precontemplation Has no intention to take action within the next six
months
Contemplation Intends to take action within the next six months
Preparation Intends to take action within the next 30 days and has
taken some behavioural steps in this direction
Action Has changed overt behaviour for less than six months
Maintenance Has changed overt behaviour for more than six months

Decisional balance
Pros The benets of changing
Cons The costs of changing

Self-efcacy
Condence Condence that one can engage in the healthy
behaviour across different challenging situations
Temptation Temptation to engage in the unhealthy behaviour across
different challenging situations

Processes of change
Experiential processes
Consciousness raising Finding and learning new facts, ideas, and tips that
support the healthy behaviour change
Dramatic relief Experiencing the negative emotions (fear, anxiety,
worry) that go along with unhealthy behavioural risks
Self-reevaluation Realizing that the behaviour change is an important
part of one’s identity as a person
Environmental Realizing the negative impact of the unhealthy
reevaluation behaviour or the positive impact of the healthy
behaviour on one’s proximal social and physical
environment
Self-liberation Making a rm commitment to change
Behavioural processes
Helping relationships Seeking and using social support for the healthy
behaviour change
Counterconditioning Substituting healthier alternative behaviours and
cognitions for the unhealthy behaviour
Reinforcement Increasing the rewards for the positive behaviour
management change and decreasing the rewards of the unhealthy
behaviour
Stimulus control Removing reminders or cues to engage in the unhealthy
behaviour and adding cues and reminders to engage in
the healthy behaviour
Social liberation Realizing that the social norms are changing in the
direction of supporting the healthy behaviour change
 Stage Theories of Health Behaviour 227

attempt to integrate these different constructs drawn from different theories

of behaviour change and systems of psychotherapy into a single coherent
model; hence the name transtheoretical.
The stages of change provide the basic organizing principle. The most
widely used version of the model species ve stages: pre-contemplation,
contemplation, preparation, action and maintenance. The rst three stages
are pre-action stages and the last two stages are post-action stages
(although preparation is sometimes dened partly in terms of behaviour
change). People are assumed to move through the stages in order, but they
may relapse from action or maintenance to an earlier stage. People may
cycle through the stages several times before achieving long-term behaviour
change.
The pros and cons are the perceived advantages and disadvantages of
changing one’s behaviour. They were originally derived from Janis and
Mann’s (1977) model of decision making, though similar constructs occur
in most theories of health behaviour. Note that applications to smoking
cessation usually assess the pros and cons of smoking which are assumed to
be equivalent to the cons and pros of changing (quitting) respectively.
Condence is similar to Bandura’s (1986) construct of self-efcacy (see
Luszczynska and Schwarzer, Chapter 4 in this volume). It refers to the
condence that one can carry out the recommended behaviour across a
range of potentially difcult situations. The related construct of temptation
refers to the temptation to engage in the unhealthy behaviour across a range
of difcult situations.
Finally, the processes of change are the covert and overt activities that
people engage in to progress through the stages. The Rhode Island group
has identied 10 such processes that appear to be common to a number of
different behaviours: ve experiential (or cognitive-affective) processes and
ve behavioural processes (Table 6.1).
In stage theories, the transitions between adjacent stages are the depen-
dent variables, and the other constructs are variables that are assumed to
inuence these transitions – the independent variables. The processes of
change, the pros and cons of changing, and condence and temptation are
all independent variables in this sense. Descriptions of the TTM to date
have not specied the causal relationships among these variables. It is not
clear, for example, whether the processes of change inuence pros, cons,
condence and temptation, which in turn inuence stage transitions;
whether these variables have independent effects on stage transitions; or
whether some other causal model is assumed to hold. It would be helpful if
the Rhode Island group specied causal models for each of the four forward
stage transitions.
The TTM has been applied to a wide range of different health behaviours
(Table 6.2). Because the stronger research designs have been used mainly in
applications of the model to smoking cessation and to adoption and
maintenance of physical exercise, the remainder of this section focuses on
these behaviours.
228 Stephen Sutton

Table 6.2 Illustrative applications of the TTM

Behaviour Authors

Smoking DiClemente et al. (1991); Prochaska et al. (1993)a;

Kraft et al. (1999); Borland et al. (2000); Aveyard et
al. (2003)a
Drinking Budd and Rollnick (1996); Migneault et al. (1999)
Drug use Isenhart (1994); Belding et al. (1996)
Exercise Marcus and Simkin (1993); Courneya et al. (2001);
Blissmer and McAuley (2002)a
Healthy eating Domel et al. (1996); Steptoe et al. (1996)
Condom use Evers et al. (1998); Brown-Peterside et al. (2000)a
Mammography screening Rakowski et al. (1992); Clark et al. (2002)a
Sun protection Rossi et al. (1994)a; Weinstock et al. (2002)a
a
Intervention studies.

2.2 Summary of research

This section is organized by the four research designs that can be used to
test predictions from stage theories (Weinstein et al. 1998c). These are:
cross-sectional studies comparing people in different stages; examination of
stage sequences; longitudinal prediction of stage transitions; and experi-
mental studies of matched and mismatched interventions.

2.2.1 Cross-sectional studies

A very large number of studies of the TTM have used cross-sectional
designs in which participants are classied into stages and compared on
theoretically relevant variables (i.e. processes of change, pros and cons,
condence and temptation). Stage theories predict discontinuity patterns
(Weinstein et al. 1998c; Kraft et al. 1999; Sutton 2000b). In our three-stage
example, variable b would be predicted to increase between Stage I and
Stage II but to show no difference between Stage II and Stage III, whereas
variable d would be predicted to show no difference between Stage I and
Stage II but to increase between Stage II and Stage III (Figure 6.2). This
section focuses on two important meta-analyses of cross-sectional studies
on the TTM (Rosen 2000; Marshall and Biddle 2001).
Rosen (2000) identied 34 studies, most of which were unpublished
dissertations, that reported cross-sectional data on use of change processes
by stage and included the action stage (because this is the stage in which
behavioural processes are predicted to peak and cognitive-affective pro-
cesses are predicted to decline). Although Rosen did not formally test for
linearity and departure from linearity, he noted that
For most health problems, use of behavioral processes increased fairly
linearly from precontemplation through action . . . [and] was typically
constant or increased slightly between action and maintenance. Only
 Stage Theories of Health Behaviour 229

for smoking did use of behavioral processes decline substantially

between action and maintenance . . . Behavioral processes peaked
during action or maintenance in 85% of all studies.
(Rosen 2000: 596–7)

Figure 6.2 Two discontinuity patterns in a cross-sectional comparison of people in

three different stages

For smoking in particular, this is clear evidence of a discontinuity at the

action stage. However, the decline in use of behavioural processes between
action and maintenance is not informative about factors that facilitate this
transition because it is implausible that less frequent use of behavioural
processes would increase the likelihood of the transition. A more plausible
explanation is that people in the maintenance stage need to use behavioural
processes less frequently, that is, that the change in use is a consequence of
the transition.
Rosen (2000) found that in less than half (41 per cent) of studies did
experiential processes peak in contemplation or preparation, as predicted
by the TTM. This proportion varied by behaviour. Use of experiential
processes peaked in contemplation or preparation in four out of ve studies
of smoking. By contrast, use of these processes increased fairly linearly with
stage of exercise adoption, peaking during action or maintenance in 11 out
of 12 studies. For other health behaviours, experiential processes were not
consistently associated with any particular stage.
Rosen (2000) also noted that the steepest increase in use of all change
processes typically occurred between the pre-contemplation and con-
templation stages of change, particularly in the case of cognitive-affective
processes. This could be interpreted as evidence for a discontinuity at the
contemplation stage, though Rosen (p.603) highlights the difculty of
interpreting this change in process use: ‘Does this indicate that engagement
in these processes motivates precontemplators to change their intentions?
Or only that people who are already considering change are more likely to
use cognitive-affective and behavioral processes of change?’
Rosen (2000) also reported some interesting ndings for use of specic
230 Stephen Sutton

processes. For example, consciousness raising was used most in the con-
templation or preparation stages in 80 per cent of studies on smoking and
psychotherapy but was used most in the action or maintenance stages in 88
per cent of studies on substance abuse, exercise and diet change; and
reinforcement management was used most in the action or maintenance
stage in nearly all studies of exercise, smoking and psychotherapy but was
used most during contemplation or preparation in two-thirds of the studies
on substance abuse and diet change.
The mainly linear patterns found by Rosen (2000), particularly for
behavioural processes, do not provide strong support for a stage model. If
differences between process use between adjacent stages are interpreted as
causal effects of process use on stage transition, Rosen’s ndings suggest
that interventions should encourage the use of behavioural processes
throughout the process of change from pre-contemplation through action.
As noted, the ndings for experiential processes were more variable.
One problem with Rosen’s (2000) analysis is that he combined studies
that used different staging methods. Given the differences between the
different methods, it would be preferable to combine only studies of a
particular behaviour that used the same staging method.

Table 6.3 Mean sample-weighted corrected effect sizes (d+) for differences
between adjacent stages from the meta-analysis by Marshall and Biddle (2001)

Variable k PC vs C C vs PR PR vs A A vs M

Self-efcacy 15–19 70.59* 70.36* 70.60* 70.72*

Pros 11–13 70.97* 70.01 70.24* 70.23*
Cons 11–13 70.46* 70.28* 70.37* 70.24*
Behavioural processes 5
Counter-conditioning 70.74* 70.62* 70.62* 70.37*
Helping relationships 70.55* 70.10 70.44* 70.05
Reinforcement management 70.97* 70.34 70.58* 70.03
Self-liberation 71.18* 70.41* 70.72* 70.04
Stimulus control 70.83* 70.15 70.49* 70.14
Experiential processes 5
Consciousness raising 70.93* 70.10 70.47* 70.04
Dramatic relief 70.65* 70.18 70.27* 70.07
Environmental re-evaluation 70.74* 70.01 70.36* 70.13
Social liberation 70.63* 70.19 70.32* 70.07
Self-re-evaluation 70.98* 70.01 70.57* 70.15*

Note: k = number of independent samples; PC = precontemplation; C = contemplation; PR =

preparation; A = action; M = maintenance.
* p < 0.05

Marshall and Biddle (2001) conducted a meta-analysis of applications of

the TTM to physical activity and exercise. Unlike Rosen (2000), they
excluded dissertations but they did include published conference abstracts.
 Stage Theories of Health Behaviour 231

Effect sizes for comparisons of adjacent stages are shown in Table 6.3. All
the effect sizes for self-efcacy were positive and signicant; the effect size
differed for different comparisons, though this was not tested formally. The
effect sizes for the pros of changing were all positive and signicant except
for contemplation to preparation. The cons of changing showed signicant
decreases across successive stages. Effect sizes for the processes of change
were based on fewer studies (k = 5) than for the other variables. For each of
the behavioural processes, the largest effect was for the transition from pre-
contemplation to contemplation and the smallest effect was for the tran-
sition from action to maintenance; the difference between action and
maintenance was non-signicant for four out of ve processes. For the
experiential processes, the largest effect again occurred between pre-con-
templation and contemplation. Differences between action and main-
tenance were non-signicant for four processes and signicantly negative
(i.e. showed a decrease) for the fth (self-re-evaluation).
Do these results support a stage model? For pros and experiential pro-
cesses, there is clear evidence for a discontinuity pattern. There is a steep
increase between pre-contemplation and contemplation, little or no
increase between contemplation and preparation, and an increase between
preparation and action. For both behavioural and experiential processes,
there is further evidence of discontinuity in that preparation to action is
associated with an increase whereas action to maintenance is not.
Marshall and Biddle (2001) interpret their ndings as mainly supportive
of the TTM predictions. However, our interpretation of the ndings in
terms of discontinuity patterns leads to somewhat different conclusions. If
we assume that a difference in process use between two adjacent stages
reects a causal effect of process use on the likelihood of making the
transition, then Marshall and Biddle’s ndings suggest, for example, that
pre-contemplators who use behavioural processes relatively frequently
(compared with others in that stage) are more likely to move to the con-
templation stage but that people in the action stage who use behavioural
processes relatively frequently (compared with others in that stage) are not
more likely to move to the maintenance stage (with the possible exception
of counter-conditioning). It seems unlikely that the TTM would make these
predictions. Similarly, the ndings suggest that contemplators who make
more frequent use of experiential processes are not more likely than others
in the same stage to move to the preparation stage.
This highlights an important difference in the way in which the Rhode
Island group interprets cross-sectional data on stage differences and the
interpretation suggested by Weinstein et al. (1998c) and Sutton (2000b).
Consider Figure 6.3, which shows a hypothetical pattern of means across
stages; assume that this represents the ndings for behavioural processes.
The Rhode Island group would interpret the relatively frequent use of
behavioural processes among people in the action and maintenance stages
as indicating that use of these processes is particularly important at these
stages and therefore needs to be encouraged. The alternative interpretation
focuses on the differences between adjacent stages rather than the absolute
232 Stephen Sutton

Figure 6.3 Hypothetical pattern of means across the ve TTM stages. [PC = pre-
contemplation; C = contemplation; PR = preparation; A = action; M =
maintenance.]

levels. The steepest increase occurs between pre-contemplation and con-

templation, suggesting that relatively frequent use of behavioural processes
among those in the pre-contemplation stage may increase the likelihood
that they move to the contemplation stage. Similarly, the lack of a differ-
ence in behavioural process use between action and maintenance could be
interpreted as suggesting that relatively frequent use of behavioural pro-
cesses is not benecial in moving people to the maintenance stage. Of
course, this alternative interpretation assumes a specic causal model in
which behavioural process use is treated as a potential cause but not a
consequence of the stage transition. Clearly, these two interpretations may
have very different implications for intervention. The interpretation advo-
cated here is consistent with the way that cross-sectional data on other
theories of health behaviour are usually interpreted: the analysis focuses on
the association between differences between individuals on one variable
and differences between individuals on a second variable.
Following the practice of the Rhode Island group, many cross-sectional
studies of the TTM report the results in terms of T-scores. T-scores are
standardized scores with a mean of 50 and a standard deviation of 10. This
practice creates a problem when an investigator wishes to compare absolute
levels of a variable across studies or to combine them in a meta-analysis.
Consider two studies that use the same staging algorithm. Even if the means
and standard deviations for each stage on a theoretically relevant variable
such as the pros of changing based on the raw (unstandardized) scores are
identical in the two studies, the mean T-scores and standard deviations will
differ if the distributions of individuals across stages differ between the two
studies. Similarly, consider two studies, again using the same staging
algorithm, but one reports data on all ve stages and the other reports data
on only the rst four stages (there is no one in the maintenance stage in the
second study). Even if the means and standard deviations based on raw
scores for the rst four stages on a variable such as the pros of changing are
 Stage Theories of Health Behaviour 233

identical in the two studies, the mean T-scores and standard deviations will
in general differ between the studies. Primary studies of the TTM should
therefore always report stage means and standard deviations based on the
raw scores as well as, or instead of, means and standard deviations based on
T-scores. Meta-analyses that combine data across studies in the way that
Rosen (2000) did should use the means based on the raw scores. The
technique used by Marshall and Biddle (2001), namely to compute stan-
dardized effect sizes for each pair of adjacent stages, is not affected by this
problem.1

2.2.2 Examination of stage sequences

Longitudinal data can be used to examine sequences of transitions through
the stages. Several studies have reported the full set of transition prob-
abilities for the ve-stage version of the TTM: Carbonari et al. (1999) for
smoking cessation; Cardinal and Sachs (1995), Peterson and Aldana
(1999), Plotnikoff et al. (2001) and Cardinal et al. (2002) for exercise/
physical activity; and Evers et al. (1998) for condom use. Some of these
studies used latent transition analysis (LTA2; Collins and Wugalter 1992) to
test particular models; the others used a less formal approach. Most of these
studies claimed support for the TTM, though it is not clear exactly what
predictions the TTM would make.
It should be emphasized that stage models predict discontinuities in the
transition probabilities. A pattern in which the transition probabilities for a
given stage declined steadily with increasing distance in both directions
would be consistent with a pseudostage model (Weinstein et al. 1998c). It
should also be noted that the analysis of transition probabilities assumes
that stage is measured not only validly but also reliably; in other words, that
observed changes in stage reect true changes and not simply random
measurement error.
In most of the transition matrices reported in the studies listed above,
initial pre-action stage of change predicted being in action or maintenance
at follow-up: those in the preparation stage at baseline were more likely to
be in action or maintenance at follow-up than those in contemplation, and
those in contemplation were more likely to be in action or maintenance
than those in pre-contemplation. This is what the Rhode Island group calls
a stage effect (e.g. Prochaska et al. 2004). It is a highly consistent nding in
the literature on the TTM. However, on their own, stage effects do not
provide strong evidence for a stage model because pseudostage models may
yield similar effects. For example, continuous measures of intention predict
future behaviour, and if such an intention measure is categorized into, say,
three categories, one would expect to nd a (pseudo)stage effect. Never-
theless, stage effects mean that stage measures may be of practical value; for
example, in measuring progress towards smoking cessation. However, they
may not be the best measures for this purpose (Farkas et al. 1996; Pierce et
al. 1998; Abrams et al. 2000; Sutton 2000a).
234 Stephen Sutton

2.2.3 Longitudinal prediction of stage transitions

As well as examining stage sequences, longitudinal data can be used to test
whether different theoretically relevant variables predict stage transitions
among people in different baseline stages. The assumption is that such
predictors represent causal factors that inuence stage movement. Analyses
of longitudinal data should be stratied by stage and should compare
people who move to the next stage in the sequence with those who remain
in a given stage with respect to baseline characteristics.
No prospective studies in the domain of exercise/physical activity have
used the TTM variables to predict stage transitions. By contrast, there
have been a number of such studies in the domain of smoking cessation.
Two of these (DiClemente et al. 1985; Prochaska et al. 1985) used an old
staging algorithm and an early version of the TTM. They were reviewed
by Sutton (2000a). Nine more recent studies were reviewed by Sutton
(2005): De Vries and Mudde (1998); Hansen (1999); Herzog et al. (1999);
Velicer et al. (1999); Dijkstra and De Vries (2001); Segan et al. (2002,
2004a, 2004b); Dijkstra et al. (2003).3 The ndings are briey summar-
ized here. The study by Segan et al. (2004a) is discussed in detail in Section
2.5.
These nine studies found some evidence that different predictors are
associated with different stage transitions. For example, Segan et al.
(2004b) found that the pros and cons of smoking did not predict movement
out of the pre-contemplation or contemplation stages. (In fact, none of the
TTM measures predicted movement out of the pre-contemplation stage.)
Pros and cons seemed to be important only for movement out of the pre-
paration stage, for which lower pros of smoking and lower cons of smoking
predicted forward movement.
However, there were few consistent ndings across the nine studies,
providing little support for the TTM. Most of the studies used relatively
long follow-up periods (at least six months). Future studies should use
shorter follow-up periods to minimize the likelihood of missing stage
transitions (with the proviso that at least six months is required to detect
the transition from action to maintenance).

2.2.4 Experimental studies

The strongest evidence for a stage theory would be to show consistently in
randomized experimental studies that stage-matched interventions are more
effective than stage-mismatched interventions in moving people to the next
stage in the sequence. In our three-stage example, an intervention that was
designed to increase variables a and b would be predicted to be more
effective in moving people in Stage I to Stage II than an intervention
designed to increase variables d and e; conversely, the second intervention
should be more effective than the rst for people in Stage II. Such evidence
would be strengthened by showing that the interventions do indeed inu-
ence the target variables and by mediation analyses yielding results con-
sistent with the hypothesis that this was the mechanism through which the
interventions had their effects on stage movement.
 Stage Theories of Health Behaviour 235

Only three studies to date have compared matched and mismatched

interventions within the framework of the TTM or closely related models
(Dijkstra et al. 1998a; Quinlan and McCaul 2000; Blissmer and McAuley
2002). The rst and second of these, which were on smoking cessation, are
considered rst.
Dijkstra et al. (1998a) compared the effectiveness of individually tailored
letters designed either to increase the pros of quitting and reduce the cons of
quitting (outcome information) or to enhance self-efcacy, or both. Smo-
kers were categorized into four stages of change: preparers (planning to quit
within the next month); contemplators (planning to quit within the next six
months); pre-contemplators (planning to quit within the next year or in the
next ve years); and immotives (planning to quit sometime in the future but
not in the next ve years, to smoke indenitely but cut down, or to smoke
indenitely without cutting down). The sample size for the main analyses
was 1100.
On the basis of two earlier cross-sectional studies (De Vries and Backbier
1994; Dijkstra et al. 1996), it was hypothesized that immotives would
benet most from outcome information only, preparers from self-efcacy
enhancing information only, and the other two groups from both types of
information. Thus, counter-intuitively, pre-contemplators and con-
templators were predicted to benet from the same kind of information. A
close examination of the cross-sectional studies reveals only partial
empirical support for these hypotheses (Sutton 2000a). In the event, the
Dijkstra et al. (1998a) study showed only weak evidence for a benecial
effect of stage-matched information. With respect to the likelihood of
making a forward stage transition, assessed at 10-week follow-up, there
were no signicant differences between the three types of information
among smokers in any of the four stages. However, preparers who received
the self-efcacy-enhancing information only were signicantly more likely
to have quit smoking for seven days at follow-up than preparers in the
outcome information only condition. Combining immotives and pre-
contemplators, the percentage of smokers who made a forward stage
transition did not differ signicantly between those who received stage-
matched and stage-mismatched information. Among contemplators and
preparers combined, the percentage who made a forward stage transition
and the percentage who quit for seven days were higher among those who
received the stage-matched information than among those who received the
stage-mismatched information, but these comparisons were only marginally
signicant (p < 0.10). It is not clear why the researchers combined the stages
in this way (immotives and pre-contemplators; contemplators and pre-
parers), given the hypothesis of the study.
Quinlan and McCaul (2000) compared a stage-matched intervention, a
stage-mismatched intervention, and an assessment-only condition in a
sample of 92 college-age smokers in the pre-contemplation stage. The stage-
matched intervention consisted of activities designed to encourage smokers
to think more about quitting smoking. The stage-mismatched intervention
consisted of action-oriented information and activities intended for smokers
236 Stephen Sutton

who are ready to quit smoking. At one month, 30 participants had pro-
gressed to contemplation, one participant had progressed to preparation,
and ve participants had progressed to action. Contrary to the hypothesis, a
greater percentage of participants in the stage-mismatched condition (54
per cent) progressed than in the stage-matched (30 per cent) or assessment-
only (35 per cent) conditions; however, this difference was not signicant.
Signicantly more smokers in the stage-mismatched condition tried to quit
smoking than in the stage-matched condition.
Quinlan and McCaul (2000) suggest that a mismatched intervention may
have different effects depending on whether it is matched to a later stage in
the sequence (as in their own study) or to an earlier stage. For example,
although it may not be detrimental for smokers in the pre-contemplation
stage to receive an intervention designed for those in the preparation stage, it
may be counterproductive to give preparers an intervention designed for pre-
contemplators. The Dijkstra et al. (1998a) study provided very weak support
for this hypothesis. Nevertheless, it may be worth testing in future studies.
In the most recent study, Blissmer and McAuley (2002) studied physical
activity. 288 university staff were randomly assigned to four conditions,
including: (a) stage-matched materials (personalized, stage-appropriate
covering letter plus stage-matched manuals) delivered via campus mail on a
monthly basis; and (b) stage-mismatched materials delivered in the same
way. After 16 weeks, 40.4 per cent of the matched group had progressed
one or more stages compared with 31.8 per cent of the mismatched group.
This difference was in the predicted direction. The authors did not report a
signicance test, but secondary analysis showed that it did not approach
signicance at the 0.05 level: 2(2) = 0.91, p = 0.634. A limitation of the
study, which the authors acknowledge, is that 57 per cent of participants
were in the action or maintenance stage at baseline, and the short follow-up
period would have prevented those who had recently entered the action
stage from progressing to maintenance.
Considered together, these three experimental studies of matched and
mismatched interventions found little or no evidence for the stage model
predictions. Intervention studies that have compared TTM-based stage-
matched interventions with generic, non-matched interventions or no-
intervention control conditions are considered in Section 2.6.

2.3 Developments
This section outlines several variants of the TTM. First, a group of
researchers in the Netherlands has developed a version of the TTM and
applied it in a number of studies of smoking cessation (e.g. Dijkstra et al.
1996, 1997, 1998a, 2003; De Vries and Mudde 1998). The stage denitions
in the Dutch version of the model differ from the most widely used TTM
denitions in that the pre-action stages are dened purely in terms of
intention: preparation is dened as planning to quit in the next month and
contemplation as planning to quit in the next six months but not in the next
month. In some studies, the group has subdivided the pre-contemplation
 Stage Theories of Health Behaviour 237

stage. For example, Dijkstra et al. (1998b) dened immotives as smokers

who are not planning to quit in the next ve years or who may be planning
never to quit and pre-contemplators as smokers who are planning to quit in
the next ve years but not in the next six months. Dijkstra and De Vries
(2001) relabelled the latter group postponers. However, this distinction was
not made in a recent study of stage transitions (Dijkstra et al. 2003).
In the Dutch version of the TTM, the main factors hypothesized to
inuence stage transitions are self-efcacy and positive and negative out-
come expectancies (the pros and cons of quitting), drawn from Bandura’s
(1986) social cognitive theory (see Luszczynska and Schwarzer, Chapter 4
in this volume). These correspond respectively to condence and the cons
and pros of smoking in the TTM, although the latter are operationalized
differently in the Dutch version. Processes of change are not emphasized in
the Dutch version. In some studies, the set of independent variables has
been expanded to include social inuence, based on the attitude–social
inuence–efcacy (ASE) model (De Vries and Mudde 1998; De Vries et al.
1998). Research on the Dutch version of the TTM has included cross-
sectional comparisons of people in different stages (e.g. Dijkstra et al.
1996), longitudinal studies of predictors of stage transitions (e.g. Dijkstra et
al. 2003), an experimental match–mismatch study (Dijkstra et al. 1998a,
discussed in Section 2.2), and evaluations of individually tailored inter-
ventions (e.g. Dijkstra et al. 1998b).
In a similar development, several studies in the domains of healthy eating
and physical exercise have used variables from the TPB (Ajzen 1991, 2002;
Conner and Sparks, Chapter 5 in this volume) in conjunction with the
stages of change from the TTM (e.g. Courneya 1995; Courneya et al. 1998,
2001; Armitage and Arden 2002; Armitage et al. 2003, 2004). For exam-
ple, in a longitudinal study, Courneya et al. (2001) used the TPB variables
as predictors of stage transitions in the exercise domain, although they
compared stage progression, regression and staying in the same stage rather
than stage-to-stage transitions. There was some evidence for differential
prediction. Subjective norm, for example, only predicted progression from
the pre-contemplation stage. However, a single-item measure of intention
emerged as a strong and consistent predictor across stages. Courneya and
colleagues suggest that the stages of change for exercise should incorporate
what they call intention choice (i.e. what the person intends to do) in the
post-action stages as well as the pre-action stages and that intention
strength should be included in the model as an independent predictor of
stage transitions.
In an earlier study, Courneya et al. (1998) treated stage of change, coded
as a continuous variable, as a potential mediator of the intention–behaviour
relationship. Their model specied that intention inuences stage which in
turn inuences behaviour. The theoretical basis for these proposed rela-
tionships seems dubious, rst because it does not respect the stage theory
assumptions and second because stage of change is dened in terms of
behaviour and so cannot be a cause of behaviour.
None of these studies included the TTM independent variables, so it was
238 Stephen Sutton

not possible to show that the TPB variables were more strongly associated
with stage or were better predictors of stage transitions than the TTM
variables or whether they contributed additional predictive power. There
are some similarities between the two sets of variables. For example,
behavioural beliefs in the TPB are similar in some respects to the pros and
cons in the TTM. However, they differ from pros and cons in that they are
based theoretically on the expectancy-value principle and they distinguish
between expectancy (belief strength) and value (outcome evaluation).

2.4 Operationalization of the model

2.4.1 Stages of change
Two main methods have been used to measure stages of change: multi-
dimensional questionnaires and staging algorithms. In multidimensional
questionnaires such as the University of Rhode Island Change Assessment
(URICA; McConnaughy et al. 1983, 1989), each stage is measured by a set
of questionnaire items, and scores are derived for each individual repre-
senting their position on each dimension. This approach has a number of
problems, the most serious of which is that it allows people to score highly
on more than one ‘stage’ (and many people do), which is inconsistent with
the assumption of discrete stages (Sutton 2001). By contrast, a staging
algorithm uses a small number of questionnaire items to allocate partici-
pants to stages in such a way that no individual can be in more than one
stage. This approach has a number of advantages over multidimensional
questionnaires: it is much simpler and the stages are clearly dened and
mutually exclusive. Perhaps not surprisingly, the few studies that have
compared the two approaches have found low concordance between them
(e.g. Belding et al. 1996; Skaki 2001). The staging algorithm approach has
been used in the vast majority of studies that have applied the TTM to
smoking and exercise.
Table 6.4 shows a staging algorithm for smoking that has been used in a
large number of studies since it was rst introduced by DiClemente et al.
(1991). Pre-contemplation, contemplation and preparation are dened in
terms of current behaviour, intentions and past behaviour (whether or not
the person has made a 24-hour quit attempt in the past year), whereas
action and maintenance are dened purely in terms of behaviour; ex-
smokers’ intentions are not taken into account.
Critics have pointed out a number of serious problems with this algo-
rithm, some of which stem from the way that contemplation and pre-
paration are dened (Pierce et al. 1996; Sutton 2000a; Etter and Sutton
2002; Borland et al. 2003). For example, according to this algorithm, a
smoker cannot be in the preparation stage unless he or she has made a
recent quit attempt. Thus, a smoker can never be ‘prepared’ for his or her
rst quit attempt (Sutton 1996b).
Farkas et al. (1996) tabulated some of the different denitions used in the
studies of smoking by the Rhode Island group between 1983 and 1991.
They note that the different classications have never been compared
 Stage Theories of Health Behaviour 239

Table 6.4 TTM measures for adult smoking, from http://www.uri.edu/research/

cprc/measures.htm

Stages of change
Are you currently a smoker?
* Yes, I currently smoke

* No, I quit within the last 6 months (ACTION STAGE)

* No, I quit more than 6 months ago (MAINTENANCE STAGE)

* No, I have never smoked (NON-SMOKER)

(For smokers only) In the last year, how many times have you quit smoking for at
least 24 hours?
(For smokers only) Are you seriously thinking of quitting smoking?
* Yes, within the next 30 days (PREPARATION STAGE if they have one 24-hour

quit attempt in the past year – refer to previous question . . . if no quit attempt
then CONTEMPLATION STAGE)
* Yes, within the next 6 months (CONTEMPLATION STAGE)

* No, not thinking of quitting (PRE-CONTEMPLATION STAGE)

Processes of change (short form)

The following experiences can affect the smoking habit of some people. Think of
any similar experiences you may be currently having or have had in the last month.
Then rate the FREQUENCY of this event on the following ve-point scale.

1 = Never, 2 = Seldom, 3 = Occasionally, 4 = Often, 5 = Repeatedly

1 When I am tempted to smoke I think about something else. &

2 I tell myself I can quit if I want to. &
3 I notice that non smokers are asserting their rights. &
4 I recall information people have given me on the benets of
quitting smoking. &
5 I can expect to be rewarded by others if I don’t smoke. &
6 I stop to think that smoking is polluting the environment. &
7 Warnings about the health hazards of smoking move me
emotionally. &
8 I get upset when I think about my smoking. &
9 I remove things from my home or place of work that remind me of
smoking. &
10 I have someone who listens when I need to talk about my smoking. &
11 I think about information from articles and ads about how to stop
smoking. &
12 I consider the view that smoking can be harmful to the
environment. &
13 I tell myself that if I try hard enough I can keep from smoking. &
14 I nd society changing in ways that makes it easier for non-
smokers. &
15 My need for cigarettes makes me feel disappointed in myself. &
16 I have someone I can count on when I’m having problems with
smoking. &
17 I do something else instead of smoking when I need to relax. &
240 Stephen Sutton

18 I react emotionally to warnings about smoking cigarettes. &

19 I keep things around my home or place of work that remind me
not to smoke. &
20 I am rewarded by others if I don’t smoke. &

Scoring:
Experiential processes
Consciousness raising 4, 11
Environmental re-evaluation 6, 12
Self-re-evaluation 8, 15
Social liberation 3, 14
Dramatic relief 7, 18
Behavioral processes
Helping relationships 10, 16
Self-liberation 2, 13
Counterconditioning 1, 17
Reinforcement management 5, 20
Stimulus control 9, 19

Self-efcacy/temptation (short form)

Listed below are situations that lead some people to smoke. We would like to know
HOW TEMPTED you may be to smoke in each situation. Please answer the
following questions using the following ve-point scale.

1 = Not at all tempted, 2 = Not very tempted, 3 = Moderately tempted, 4 = Very

tempted, 5 = Extremely tempted

1 With friends at a party. &

2 When I rst get up in the morning. &
3 When I am very anxious and stressed. &
4 Over coffee while talking and relaxing. &
5 When I feel I need a lift. &
6 When I am very angry about something or someone. &
7 With my spouse or close friend who is smoking. &
8 When I realize that I haven’t smoked for a while. &
9 When things are not going my way and I am frustrated. &

Scoring:
Positive affect/social situation 1, 4, 7
Negative affect situations 3, 6, 9
Habitual/craving situations 2, 5, 8

Decisional balance (short form)

The following statements represent different opinions about smoking. Please rate
HOW IMPORTANT each statement is to your decision to smoke according to the
following ve-point scale.
 Stage Theories of Health Behaviour 241

Table 6.4 cont’d

1 = Not important, 2 = Slightly important, 3 = Moderately important, 4 = Very
important, 5 = Extremely important

1 Smoking cigarettes relieves tension. &

2 I’m embarrassed to have to smoke. &
3 Smoking helps me concentrate and do better work. &
4 My cigarette smoking bothers other people. &
5 I am relaxed and therefore more pleasant when smoking. &
6 People think I’m foolish for ignoring the warnings about cigarette &
smoking.

Scoring:
PROS 1, 3, 5
CONS 2, 4, 6

Note: It states on the website that ‘All measures are copyright Cancer Prevention Research
Center, 1991. Dr James O. Prochaska, Director of the CPRC, is pleased to extend his
permission for you to use the Transtheoretical Model-based measures available on this website
for research purposes only, provided that the appropriate citation is referenced.’

empirically. This lack of standardization makes it difcult to compare

results from different studies and to accumulate the research ndings into a
coherent body of knowledge. Using data from a large sample of smokers
from the California Tobacco Survey, Farkas et al. compared the DiCle-
mente et al. (1991) staging algorithm with an earlier algorithm used by the
Rhode Island group that classied smokers into pre-contemplation, con-
templation and relapse stages. The two algorithms produced markedly
different stage distributions. For example, the earlier algorithm classied
almost half the sample in the most advanced stage (relapse) whereas the
revised scheme placed only 16 per cent in the most advanced stage (pre-
paration). The two algorithms would lead to very different conclusions
concerning the proportion of smokers for whom action-oriented pro-
grammes are appropriate. Farkas and colleagues also showed that the
earlier stage measure provided better prediction of cessation and quit
attempts assessed at 1 to 2-year follow-up than the revised algorithm and
that both schemes allocated smokers with very different probabilities of
quitting to the same stage (see also Pierce et al. 1996).
A variety of different staging algorithms have been used in the domain of
exercise/physical activity. Marshall and Biddle (2001) recommend the one
proposed by Marcus and Simkin (1993; see also Reed et al. 1997), which is
shown in Table 6.5. Although this scheme does not suffer from the logical
problems of the DiClemente et al. (1991) smoking algorithm, it seems
somewhat implausible to treat irregular exercise (preparation) as a discrete
stage between contemplation and action, implying that people move from
no exercise to irregular exercise to regular exercise and that irregular
exercise is qualitatively different from regular exercise.
A problem with most staging algorithms is that the time periods are
242 Stephen Sutton

Table 6.5 Staging algorithm for exercise, from Marcus and Simkin (1993)

Items
1 I currently do not exercise
2 I intend to exercise in the next 6 months
3 I currently exercise regularly
4 I have exercised regularly for the past 6 months

Scoring
Pre-contemplation: Item 1 = true and Item 2 = false.
Contemplation: Item 1 = true and Item 2 = true.
Preparation: Item 1 = false and Item 3 = false.
Action: Item 3 = true and Item 4 = false.
Maintenance: Item 3 = true and Item 4 = true.

arbitrary. For instance, action and maintenance are usually distinguished by

whether or not the duration of behaviour change exceeds six months.
Changing the time periods would lead to different stage distributions. The
use of arbitrary time periods casts doubt on the assumption that the stages
are qualitatively distinct, that is, that they are true stages rather than
pseudostages (Sutton 1996a; Bandura 1997, 1998).
The staging algorithms listed on the Rhode Island group’s website show
inconsistencies across different health behaviours. For example, in the
algorithm for adoption of mammography (Rakowski et al. 1992), action
and maintenance are dened partly in terms of intentions (planning to have
a mammogram in the coming year). Like the DiClemente et al. (1991)
algorithm, this algorithm has logical aws. For instance, it is possible for a
woman to move directly from contemplation to maintenance simply by
forming an intention, without passing through the action stage and without
changing her behaviour.

2.4.2 TTM independent variables

Table 6.4 shows the measures of the other TTM variables for adult
smoking as listed on the Rhode Island group’s website. These are all the
short-form measures; the long forms are also listed on the website.
Descriptions of the development of the long forms of the measures can be
found in Velicer et al. (1985) for decisional balance, DiClemente (1981),
DiClemente et al. (1985) and Velicer et al. (1990) for condence and
temptation, and Prochaska et al. (1988) for the processes of change. Fava et
al. (1995) outline the development of the short forms of these measures,
except for condence. In its studies of smoking cessation, the Rhode Island
group has favoured the temptation measure over the condence measure,
because the scores tend to be highly (negatively) correlated and the temp-
tation measure ‘is more easily responded to by subjects in some of the
 Stage Theories of Health Behaviour 243

stages’ (Velicer et al. 1990: 273). However, the assumption that the two
measures are interchangeable has been challenged by Segan et al. (2004a).
Compared with the long forms, the short-form measures are more sui-
table for use in studies that use telephone interviewing and in intervention
studies involving repeated assessment. Using the short forms, all the con-
structs in the TTM can be measured with a total of 35 items. However, it is
likely that the reliability of the short-form measures is lower than that of
the long forms, and content validity may also be compromised because a
construct may not be adequately represented by two or three items. For
example, the short-form decisional balance scale does not include items
about the health consequences of smoking, the nancial costs, or the belief
that smoking helps keep weight down.
In the measures of the TTM independent variables, scale scores are
created by computing the sum or the mean of the item scores. The short-
form processes of change measure consists of 10 two-item scales assessing
recent frequency of use; aggregate scores can also be created for the
experiential and behavioural processes respectively. Borland et al. (2000)
improved the wording of several of the items and also discussed some
remaining problems. The short-form temptation measure has three sub-
scales: positive affect or social situations; negative affect situations; and
habitual/craving situations. An aggregate temptation score can also be
computed. Finally, the short-form decisional balance measure comprises
two subscales representing the pros and cons of smoking, respectively. The
usual practice is to standardize these separately and then compare them (see
Section 2.2).
The Rhode Island group’s website lists measures of the TTM independent
variables for exercise and some other health-related behaviours, though the
full set of measures is not given for all the behaviours listed. Note that, for
many behaviours, the condence measure may be more appropriate than
the temptation measure.

2.5 Application of the model: smoking relapse

2.5.1 Introduction
The study by Segan et al. (2004a) on predictors of relapse to smoking
cessation was selected as the example application of the TTM. In their
rationale for the study, the authors point out that most quit attempts end in
failure and that a better understanding of the factors involved in relapse is
needed to improve success rates. They also note that most studies that have
applied the TTM to smoking have focused on current smokers and that
surprisingly little attention has been paid to the post-cessation stages. The
key questions addressed by this study are: do the post-cessation stages help
us understand the process of staying abstinent, and can TTM measures
predict relapse?
From the perspective of the TTM, relapse is a transition from the action
or maintenance stage to one of the pre-action stages (pre-contemplation,
contemplation or preparation). The model predicts that use of four of the
244 Stephen Sutton

ve behavioural change processes (helping relationships, counter-

conditioning, reinforcement management and stimulus control) inuences
progression from action to maintenance (Prochaska et al. 1992), the
implication being that, having quit, the smoker needs to use these processes
frequently in order to stay abstinent for six months and thus move into the
maintenance stage.

2.5.2 Method
Participants were 325 cigarette smokers who called the Quitline telephone
counselling and information service in Victoria, Australia, and were quit at
either the three-month follow-up and/or the six-month follow-up. The
mean age was 39 years (range 17 to 78) and 55 per cent were women; mean
cigarette consumption at baseline was 21 cigarettes a day.
Participants were recruited into the study after their reason for calling the
Quitline had been dealt with. Of the callers asked to participate in the study,
77 per cent did so. They completed a telephone interview (time 1) and were
posted the same questionnaire at three months (time 2) and six months
(time 3), and a shorter version at 12 months (time 4). Response rates to the
follow-ups were 76 per cent, 74 per cent and 68 per cent respectively.
The sample was part of a larger study that had several intervention
groups. Segan et al. (2004a) argue that the interventions should not be
regarded as an interfering factor because they are presumed to have their
effect by inuencing the TTM independent variables, not by producing
change in fundamentally different ways.
The predictor variables consisted of the short-form measures of the
processes of change, the pros and cons of smoking, and temptation to
smoke (see Section 2.4).4 Two of the experiential change processes, dra-
matic relief and social liberation, were not measured. Minor modications
were made to the wording of some of the change process items (see Borland
et al. 2000). At follow-up, smoking status was assessed by the question ‘Are
you currently a cigarette smoker?’ (Yes/No), and length of abstinence by the
question ‘How long ago did you quit?’, answered in days, weeks or months.
In the analysis, Segan et al. (2004a) compared ex-smokers who had quit
for less than one month and ex-smokers who had quit for more than one
month. Although the TTM does not make this distinction, the authors
present both theoretical and empirical reasons for using this time point. For
example, relapse rates are likely to differ signicantly between these groups.
In terms of the TTM stages, those who were quit at time 2 were in the
action stage; those who were still abstinent at time 3 were either still in
the action stage or had moved to maintenance. Ex-smokers who had quit for
less than one month at time 3 were in the action stage; those who were still
off smoking at time 4 had moved to the maintenance stage. Ex-smokers who
had quit for more than one month at time 3 were in action or maintenance;
those who were still abstinent at time 4 had moved to the maintenance stage.
Predictors of relapse were examined by conducting a series of logistic
regression analyses (one for each predictor measure), controlling in each
case for intervention condition and length of abstinence.
 Stage Theories of Health Behaviour 245

2.5.3 Results
Relapse between time 2 and time 3 Thirty per cent of the 247 participants
who were quit at time 2 relapsed by time 3. Relapse at time 3 was predicted
by higher aggregate temptations (p = 0.001), higher positive/social temp-
tations (p < 0.001), higher habit/addictive temptations (p = 0.03), higher
negative/affective temptations (p = 0.004), higher aggregate behavioural
change processes (p = 0.01), higher reinforcement management (p = 0.02)
and higher helping relationships (p = 0.03). There were signicant inter-
actions for the aggregate behavioural processes and reinforcement man-
agement (p = 0.03 and 0.02 respectively). For those who had quit for less
than one month higher levels of behavioural process use and reinforcement
management predicted relapse, whereas for those who had quit for more
than one month the levels of these variables were similar for both relapsers
and quitters.
Relapse between time 3 and time 4 Thirty-ve per cent of the 204 par-
ticipants who were quit at time 3 relapsed by time 4. Relapse was predicted
by higher aggregate temptations (p = 0.005), higher positive/social temp-
tations (p = 0.001) and higher habit/addictive temptations (p = 0.007).

2.5.4 Discussion
The authors note a number of potential limitations of the study. First, there
may be predictors of relapse that are specic to very recent ex-smokers in
the acute withdrawal phase (i.e. quit for less than a week), but it was not
possible to examine this because of the relatively small sample size in this
group. Second, the sample consisted of smokers who had sought help by
phoning a quitline; it is possible that the predictors of relapse would be
different in smokers who try to quit without help. Third, the short-form
measures may not adequately assess the TTM constructs (see Section 2.4).
Use of behavioural processes predicted relapse between time 2 and time 3
(though not between time 3 and time 4). However, the ndings were
contrary to the TTM predictions, with more frequent use of behavioural
processes predicting relapse between time 2 and time 3. The authors suggest
that higher use of behavioural processes may indicate greater difculties
with staying quit and hence greater likelihood of relapse. The only TTM
variables that predicted relapse in both time periods were higher levels of
temptations to smoke, a nding that is consistent with Marlatt and Gor-
don’s (1985) relapse prevention model.
There was some evidence that predictors of relapse differed between
those who had quit for less than one month and those who had quit for
more than one month. Based on these ndings and other evidence, Segan et
al. (2004a) suggest that by dividing ex-smokers into those who have quit
for less than six months (actors) and those who have quit for more than six
months (maintainers), the TTM provides an overly simplistic account of the
post-cessation phase.
This research group has recently developed a new model that species
seven ‘perspectives’ (stages) in the process of smoking cessation, including
246 Stephen Sutton

four post-cessation stages (Borland 2000; Borland et al. 2004; Segan et al.
2004a). Based on a detailed and insightful analysis of the task of quitting
smoking, the model is a promising alternative to the TTM and is potentially
applicable to other health behaviours.

2.6 Intervention studies

The TTM implies that interventions should be matched to the participant’s
stage by targeting the variables that are assumed to inuence the transition
from that stage to the next. Such interventions should be more effective than
generic interventions in which all participants are treated the same irrespec-
tive of their stage of change. TTM-based interventions have been developed
for a range of different target behaviours, including condom use (Brown-
Peterside et al. 2000) and sun protective behaviours (Weinstock et al. 2002),
as well as smoking cessation (Prochaska et al. 1993). Some TTM-based
interventions not only match materials to the participant’s stage but also
individually tailor the information on the basis of the other TTM variables.
A number of studies have compared TTM-based stage-matched inter-
ventions with generic, non-matched interventions or no-intervention con-
trol conditions. Four reviews have summarized the evidence on
effectiveness. Riemsma et al. (2003) identied 23 randomized controlled
trials of TTM-based interventions for smoking cessation and concluded
that ‘limited evidence exists for the effectiveness of stage based interven-
tions in changing smoking behaviour’ (p.1175). Using more lenient selec-
tion criteria, Spencer et al. (2002) identied 22 intervention studies on
smoking and reached a more positive conclusion. Bridle et al. (in press)
found 37 randomized controlled trials of TTM-based interventions tar-
geting seven health-related behaviours (including 13 studies on smoking
cessation and seven on physical activity). They concluded that ‘Overall . . .
there was limited evidence for the effectiveness of stage-based interventions
as a basis for behavior change or for facilitating stage progression . . .’.
Finally, Van Sluijs et al. (2004) identied 29 trials of TTM-based lifestyle
interventions in primary care (including 14 studies on smoking cessation
and 13 on physical activity) and came to a similar conclusion.
All these reviews included studies that were not proper applications of
the TTM. For an intervention to be labelled as TTM-based, it should (a)
stratify participants by stage and (b) target the theory’s independent vari-
ables (pros and cons, condence and temptation, processes of change),
focusing on different variables at different stages. Many of the studies
included in the reviews did not meet this requirement. For example, the
Newcastle exercise project involved an intervention based on motivational
interviewing and apparently did not stratify participants by stage of change
or target the TTM’s independent variables (Harland et al. 1999).
Not surprisingly, the interventions that come closest to a strict applica-
tion of the TTM are those developed by the Rhode Island group. The
group’s studies of TTM-based smoking cessation interventions have yielded
mainly positive ndings (e.g. Prochaska et al. 1993, 2001a, 2001b;
 Stage Theories of Health Behaviour 247

Pallonen et al. 1998). By contrast, adaptations of these interventions

evaluated by other research groups in the UK and Australia have yielded
mainly negative results (Aveyard et al. 1999, 2001, 2003; Borland et al.
2003; Lawrence et al. 2003).
None of these studies speaks directly to the validity or otherwise of the
TTM. There have been no process analyses published to date demonstrat-
ing that TTM-based interventions do indeed inuence the variables they
target in particular stages and that forward stage movement can be
explained by these variables.

2.7 Future directions

The TTM has been very inuential and has popularized the idea that
behaviour change involves movement through a series of discrete stages. It
has also stimulated the development of innovative interventions. However,
the model cannot be recommended in its present form. Fundamental pro-
blems with the denition and measurement of the stages need to be
resolved. Although a cursory glance at the huge literature on the TTM gives
the impression of a large body of mainly positive ndings, a closer exam-
ination reveals that there is remarkably little supportive evidence. It would
be helpful if the Rhode Island group presented a fuller specication of the
model that (a) stated which variables inuence which stage transitions and
(b) specied the causal relationships among the pros and cons, condence
and temptation, and processes of change. Predictions from the model
should be tested using strong research designs: longitudinal studies of stage
transitions with short time intervals and experimental studies of matched
and mismatched interventions (Weinstein et al. 1998c). Studies of stage-
matched interventions should examine whether the interventions inuence
the variables targeted in particular stages and whether forward stage
transitions can be explained by these variables.
It would also be helpful if the Rhode Island group addressed the detailed
critiques of the TTM by, among others, Sutton (1996a, 2000a, 2001),
Carey et al. (1999), Joseph et al. (1999), Rosen (2000) and Littell and
Girvin (2002), and responded to Weinstein et al.’s (1998c) exposition of the
conceptual and methodological issues surrounding stage theories.

3 The precaution adoption process model (PAPM)

3.1 Description of the model

The PAPM was originally developed to describe the process by which
people come to adopt the precaution of testing their homes for radon
(a naturally occurring carcinogenic gas). The model was rst described
by Weinstein (1988) but was subsequently revised. This section focuses
on the revised version, which was rst presented by Weinstein and
Sandman (1992). The theory species seven discrete stages in the process
248 Stephen Sutton

of precaution adoption (Figure 6.4). In Stage 1, people are unaware of the

health issue. People in Stage 2 are aware of the issue but they have never
thought about adopting the precaution; they are not personally engaged
by the issue. People who reach Stage 3 are undecided about whether or
not to adopt the precaution. If they decide against adopting the precau-
tion, they move into Stage 4. If they decide in favour, they move into Stage
5. Having reached Stage 5, people who act on their decision move to
Stage 6. Finally, for some behaviours, a seventh stage (maintenance) may
be appropriate.

Figure 6.4 The PAPM compared with the TTM, from Weinstein and Sandman
(2002a). [Only the acting/action and maintenance stages can be regarded as
equivalent across the two models.]

The PAPM differs from the TTM in a number of ways (Figure 6.4). It has
more stages: seven instead of ve. Unlike the TTM, there is a stage (decided
not to act) that is a side-path from the main sequence (although a person
who reaches this stage may of course return to Stage 3 at some point and
 Stage Theories of Health Behaviour 249

continue moving towards action). The decided to act stage is similar to the
preparation stage in the TTM (at least when preparation is dened purely in
terms of intentions or plans and not in terms of past behaviour). At rst
glance, deciding about acting appears to be similar to contemplation in the
TTM. However, being undecided about doing something may not be the
same as seriously thinking about doing something in the next six months.
Weinstein and Sandman (1992) suggest that the contemplation stage may
include both individuals who are undecided about action and those who
have already decided to act. Note that, unlike the TTM, none of the pre-
action stages in the PAPM refers to specic time periods, which means that
they are less arbitrary and perhaps more likely to represent genuine stages.
Finally, the PAPM in effect splits the TTM pre-contemplation stage into
three stages (1, 2 and 4), which seem to represent important distinctions; in
particular, it seems important to distinguish between (a) having never
thought about adopting a precaution and (b) having thought about it and
decided not to act.

Table 6.6 Stage transitions allowable under the PAPM

1 2 3 4 5 6

1 P P
2 P P
3 P P P
4 P P
5 P P P
6 P

Table 6.6 shows a transition matrix for the six-stage version of the PAPM
(without the maintenance stage). Allowable transitions are indicated by
ticks. The diagonal consists entirely of ticks, meaning that a person can stay
in any of the stages indenitely. For example, one person may remain
blissfully unaware of the health threat while another person may be con-
stantly trying to decide what to do. Transitions above the diagonal repre-
sent forward movements. The ticks in this part of the matrix indicate the
transitions illustrated in Figure 6.4, for example moving from being una-
ware about the issue to being aware but unengaged. Transitions in the
upper diagonal that do not have a tick represent forward skips. Such skips
may sometimes occur. For example, a person may make a decision to do
something on the spur of the moment without having thought about it.
Thus, they may move directly from Stage 2 to Stage 5, skipping Stage 3. It is
possible to interpret this example in terms of the person moving rapidly
through the intervening stage rather than skipping it completely. Con-
ceptually, it is neater to proscribe skips and to assume that change follows
the sequence postulated in Figure 6.4. In practice, it is difcult or impossible
to distinguish between the two interpretations.
250 Stephen Sutton

Transitions below the diagonal represent backward movements. Wein-

stein and Sandman (2002a: 71) state that ‘Movement backwards towards
an earlier stage can . . . occur, without necessarily going through all the
intermediate stages, though obviously it is not possible to go from later
stages to Stages 1 or 2.’
Table 6.7 shows the factors that are likely to inuence key transitions in
the PAPM. Weinstein and Sandman (2002b) emphasize the importance of
media messages in shifting people from Stage 1 to Stage 2. They also state
that the factors that inuence stage transitions may differ for different

Table 6.7 Issues likely to determine progress between stages, from Weinstein and
Sandman (2002b)

Stage transition Important issues

Stage 1 to Stage 2 Media messages about the hazard and precaution

Stage 2 to Stage 3 Communications from signicant others
Personal experience with hazard
Stage 3 to Stage 4 or Beliefs about hazard likelihood and severity
Stage 5 Beliefs about personal susceptibility
Beliefs about precaution effectiveness and difculty
Behaviours and recommendations of others
Perceived social norms
Fear and worry
Stage 5 to Stage 6 Time, effort and resources needed to act
Detailed ‘how-to’ information
Reminders and other cues to action
Assistance in carrying out action

behaviours. Although the factors listed in Table 6.7 seem plausible, and there
is a lot of indirect supporting evidence, there is as yet little direct evidence
from the few studies of the PAPM that have been conducted to date.

3.2 Summary of research

To date, the PAPM has been applied to only a limited number of behaviours
(Table 6.8). This review focuses on the longitudinal studies and an
important experimental match–mismatch study; all these studies were
applications to radon testing. An intervention study on osteoporosis pre-
vention (Blalock et al. 2002), which did not involve a comparison of
matched and mismatched interventions, is discussed in Section 3.4. First,
however, we discuss the Clemow et al. (2000) cross-sectional study on
mammography screening because it raises an important issue about the role
of past behaviour.
Clemow et al. (2000) applied the PAPM in a large sample (n = 2507)
of women aged 50–80 in Massachusetts whom they describe as
 Stage Theories of Health Behaviour 251

Table 6.8 Applications of the PAPM

Behaviour Authors

Home radon testing Weinstein and Sandman (1992); Weinstein

et al. (1998a)a
Hepatitis B vaccine acceptance Hammer (1998)
Osteoporosis prevention Blalock et al. (1996, 2002a)
Mammography screening Clemow et al. (2000)
a
Intervention studies.

‘underutilizers’ of mammography, that is women who had never had a

mammogram or who had not had one in the 24 months prior to the survey
or who had had a mammogram in the previous 24 months but had not had
one in the 24 months prior to the last mammogram. The staging algorithm
they used differed from the recommended one (see Section 3.3). Participants
were rst classied into three groups with respect to their intention to have
a mammogram in the next year or two: (a) denitely planning (Stage 5); (b)
thinking about (Stage 3); and (c) not planning. (No respondent stated that
they had never heard of a mammogram.) A second question was used to
divide the not planning group into three stages: (a) never seriously con-
sidered getting a mammogram (Stage 2); (b) considered getting a mam-
mogram, but decided against it (Stage 4); and (c) have thought about it but
still undecided (Stage 3b). Clemow et al. do not report a full comparison of
adjacent stages, but their data show some evidence for discontinuity pat-
terns. For example, compared with women in Stage 2, those in Stage 3 were
signicantly more likely to say that they worried ‘a little’ or ‘a lot’ about
breast cancer. However, the two groups also differed with respect to the
percentage that had had a prior mammogram (53.8 per cent in Stage 3 vs
11.6 per cent in Stage 2). This is a potentially important confounding
factor. Ideally, such stage comparisons should control for past behaviour,
for example by dividing the sample into those who had and those who had
not had a previous mammogram. Such an analysis could also address the
question of whether past behaviour is a moderator of stage transitions: the
factors that inuence a particular transition may differ depending on
whether or not participants have prior experience of the behaviour.
Weinstein and Sandman (1992) briey report results from three pro-
spective studies of home radon testing that examined movement from the
pre-action stages to the action stage (ordering a test). The staging algorithm
used in these studies dened Stage 4 as ‘test not needed’, which may not be
quite the same as ‘decided not to test’ in the recommended algorithm (see
Section 3.3). The ndings, which are summarized in Table 6.9, show that
the percentage who subsequently ordered a test was much higher among
those in the plan-to-test stage than among those in the other pre-action
stages; differences between these other pre-action stages were relatively
252 Stephen Sutton

small. (Weinstein and Sandman note that the higher rate of testing by
undecided participants in Study II may be a consequence of the rather
aggressive intervention that occurred after their stage of testing had been
assessed.) Although not tested formally, this is evidence not simply for what
the Rhode Island group calls a stage effect (Prochaska et al. 2004) but for
the predicted discontinuity pattern. No studies of the PAPM have examined
other stage transitions over time or have investigated predictors of stage
transitions in longitudinal studies.

Table 6.9 Stages of testing adoption and subsequent test orders (per cent ordering
a test), from Weinstein and Sandman (1992)

Prior stage Study I Study II Study III

n = 263 647 453

Never thought about it —a 2.0 5.3

Not needed 3.6 4.2 4.8
Undecided 3.3 12.9 3.5
Plan to test 26.2 23.6 28.2
a
‘Never thought about it’ was not given as a response option in this study.

Weinstein et al. (1998a) reported an experimental study of the PAPM

that compared matched and mismatched interventions. Participants (resi-
dents of Columbus, Ohio) rst viewed a general informational video in
their homes and were then staged by asking ‘What are your thoughts about
testing your home for radon?’ in a telephone interview. The statements used
for classifying people into Stages 2–5 were very similar to those in the
recommended algorithm (see Section 3.3). The statement for the action
stage was ‘I have already completed a test, have a test in progress, or have
purchased a test.’ Stage 1 people, who had never heard about radon testing,
had already been screened out of the study. Those people who were in
either the undecided or the decided to test stage were randomly assigned to
one of four experimental conditions and were sent the appropriate inter-
vention materials and a questionnaire. A follow-up telephone interview was
conducted nine to ten weeks after participants returned the questionnaire to
nd out whether they had purchased a radon test kit and, if not, to ascertain
their nal stage. (Weinstein et al. (1998a) note that buying a test kit is not
equivalent to testing, but that they chose to use test kit purchase as the main
outcome to avoid lengthening the follow-up period.)
The four experimental conditions were:
1 High likelihood. Participants in this condition received a ve-minute
video designed to convince them that they had a moderate to high
chance of nding high radon levels in their homes. The covering letter
mentioned that test kits could be ordered from the American Lung
Association (ALA) but did not include an order form.
 Stage Theories of Health Behaviour 253

2 Low effort. The ve-minute video sent to participants in this condition

described how to select a kit type (including a specic recommenda-
tion), nd and purchase a kit, and conduct a test. The procedure was
described as simple and inexpensive. They were also sent a form to
order test kits from the ALA.
3 Combination. Participants in this condition received a 10-minute video
that simply combined the high-likelihood and low-effort videos and the
same letter and order form as people in the low-effort condition.
4 Control. Participants in the control condition received a letter stating
that their assistance in viewing a second video was not needed.

Manipulation checks showed that the high-likelihood intervention

increased perceived radon risk and the low-effort intervention increased
perceived ease of testing, as intended. The outcome results are shown in
Table 6.10. The main outcome was the percentage of people who pro-
gressed one or more stages towards testing. This criterion (rather than
forward movement of only a single stage) was chosen because people in the
undecided stage who moved to the decided to test stage may have already
possessed the information or skills required to progress further to the action
stage. As predicted, both the stage by high-likelihood treatment interaction
and the stage by low-effort treatment interaction were signicant. The high-
likelihood treatment was much more effective among undecided partici-
pants than among decided-to-act participants, and the low-effort treatment
was more effective among the decided-to-act participants than among the
undecided participants. This is clear evidence for the greater effectiveness of
stage-matched over stage-mismatched interventions.

Table 6.10 Percentage of participants who progressed one or more stages toward
testing, from Weinstein et al. (1998a)

Condition

Pre-intervention stage Control High-likelihood Low-effort Combination

Undecided 18.8 (138) 41.7 (144) 36.4 (130) 54.5 (139)

Decided-to-test 8.0 (339) 10.4 (338) 32.5 (329) 35.8 (345)

Note: The group size in each cell is shown in parentheses.

There was no evidence that the stage-mismatched interventions were

counter-productive: they were still more effective than the control condi-
tion. The combination treatment was the most effective but, as Weinstein et
al. (1998a) point out, it was approximately twice as long as each of its two
components and therefore more expensive. However, it is possible that the
high-likelihood video with an accompanying letter that included instruc-
tions on how to buy a test kit and an order form would be as effective as the
combination treatment.
254 Stephen Sutton

Table 6.11 PAPM stage classication algorithm, from Weinstein and Sandman
(2002b)

1 Have you ever heard about {home radon testing}?

No Stage 1
Yes [go to 2]
2 Have you {tested your own house for radon}?
Yes Stage 6
No [go to 3]
3 Which of the following best describes your thoughts about
{testing your home}?
I’ve never thought about {testing} Stage 2
I’m undecided about {testing} Stage 3
I’ve decided I don’t want to {test} Stage 4
I’ve decided I do want to {test} Stage 5

Note: The words in curly brackets could be replaced with other precautions to develop a
staging algorithm for these precautions.

The Weinstein et al. (1998a) study is an exemplary study that provides a

model for how stage theories can be tested experimentally.

3.3 Operationalization of the model

Table 6.11, from Weinstein and Sandman (2002b), gives a stage classi-
cation algorithm that would be suitable for any behaviour for which a
maintenance stage is not applicable. These include behaviours that, if they
are performed at all, are usually performed only once, for example having a
predictive genetic test for inherited breast/ovarian cancer. Of course, vir-
tually any behaviour can be repeated: a person may test their home for
radon, then move house and test their new home for radon. If a signicant
proportion of people in the sample have adopted the precaution before,
then it may be necessary to take past behaviour into account in the analysis
and to reword the staging algorithm. Consider, for example, applying the
model to participation in mammography screening. If the investigator is
interested only in rst-time attendance for screening, he or she could either
select a sample of women who have recently reached the lower age limit for
screening and use the algorithm in Table 6.11 to stage them or select a
sample of women who have never been screened and follow them over time
until some of them have their rst screen, using the algorithm to stage the
sample on a number of occasions. Women who have had one mammogram
could be allocated to stages with respect to having another mammogram.
This would require modications to the algorithm. Stage 1 would not be
applicable for these women. And the statement used to classify women in
Stage 2 could be reworded to something like ‘I haven’t thought about
whether to have another mammogram’. (An alternative approach would be
to classify women who have had repeated mammograms in accordance
 Stage Theories of Health Behaviour 255

with the recommended schedule as being in the maintenance stage. How-

ever, it would be difcult to know how to classify women who have had
more than one mammogram but whose pattern of attendance does not
conform to the recommend schedule.)
The PAPM can also be applied to deliberate changes in ongoing
behaviours such as the frequency of taking exercise or the amount of salt
consumed per day. In this case, it is necessary to dene a criterion level of
behaviour, for example doing at least 30 minutes of moderate physical
activity every day. Here it would be appropriate to specify a maintenance
stage, possibly dened in terms of duration as in the TTM, for example
having maintained at least 30 minutes of moderate physical activity a day
for at least six months. However, as noted earlier, such time periods are
arbitrary and do not have face validity as marking a transition between
discrete stages.

3.4 Intervention studies

Apart from the Weinstein et al. (1998a) study, only one PAPM-based
intervention study has been published to date. Blalock et al. (2002)
described the effects of an osteoporosis prevention programme in which
they compared a tailored education intervention, based partly on the
PAPM, with a non-tailored intervention among women aged between
40 and 56 in North Carolina. Women in the tailored education group
(n = 273) were sent two individually tailored packets and participated in a
brief telephone counselling session. The packets contained separate cards
for calcium and exercise. The messages were partly tailored on precaution
adoption stage of change. If the participant was in the action stage
(dened as being above the recommended criterion for the behaviour at
baseline), the card included a message reinforcing that behaviour. If the
participant was in the engaged stage (not obtaining an adequate amount of
calcium/exercise at baseline but thinking about or trying to increase their
level), the card included a tailored message reinforcing her interest in
change. If she was in the unengaged stage (not obtaining an adequate
amount of calcium/exercise at baseline and not thinking about trying to
increase their level), the card included a message encouraging her to think
about trying to get more calcium/exercise. Thus, rather than targeting
factors assumed to inuence stage transitions, the tailored interventions
tried to encourage forward movement (or staying in the action stage) in a
simple, direct way.
The telephone counselling session took place about three weeks after the
woman had received the rst tailored packet. If a woman was above the
criterion level, the session simply reinforced her current behaviour. If a
woman was below the criterion but wanted to change, she was guided
through a structured protocol that included goal setting, behavioural con-
tracting, identifying potential barriers to change, and relapse prevention
strategies. The second tailored packet was sent to participants immediately
after the counselling session. This listed the behavioural goals that had been
256 Stephen Sutton

set, included a copy of the behavioural contract(s), and provided tips on

overcoming the barriers that had been identied.
Women in the non-tailored education group (n = 274) received two
packets of information with similar content to those received by the tailored
education group but with no individual tailoring. In addition, women in
this group did not receive telephone counselling.
Among women in the unengaged and engaged stages at baseline, calcium
intake increased signicantly between baseline and three-month follow-up,
and these increases were maintained at the six- and twelve-month follow-
ups. Among women in the engaged stage, the tailored group showed greater
increases than the non-tailored group at each follow-up. Among women in
the action stage at baseline, for whom further increases in calcium intake
were not appropriate, the non-tailored group showed an increase but the
tailored group did not. There were no signicant differences between the
intervention conditions for exercise. Thus, this study yielded limited evi-
dence for the greater effectiveness of the tailored intervention. This was a
well-conducted study, but the tailored intervention was based loosely on a
simplied version of the PAPM, and the study does not provide direct
evidence bearing on the validity of the model.

3.5 Future directions

Although only a handful of studies using the PAPM have been conducted to
date, it is a promising approach. The stages (particularly for the six-stage
version, without the maintenance stage) have greater face validity than the
TTM stages and make important distinctions that are not made by the
TTM. Given the problems with the TTM, researchers and practitioners
who are thinking of using the TTM should seriously consider the PAPM as
an alternative.
Key tasks for future research are to specify the variables that are
important for each of the stage transitions and to test whether they predict
and inuence these transitions. The factors listed in Table 6.7 provide a
useful starting point, though they will need to be precisely operationalized
in empirical studies. As noted in Section 3.1, Weinstein and Sandman
(2002b) suggest that the factors that inuence particular stage transitions
may differ for different behaviours (in contrast to the TTM, which holds
that stage transitions for many different behaviours are inuenced by
variables from the same limited set); however, they do not make any spe-
cic predictions about this. The model should be applied to a wider range of
behaviours, including those that are relevant for novel threats where many
people will fall into the early stages as well as those that are relevant for
more established threats. It would be helpful if future studies of the PAPM
used staging algorithms that were as similar as possible to the recom-
mended version (Table 6.11) and sufciently large sample sizes to avoid the
need to collapse stages.
 Stage Theories of Health Behaviour 257

Figure 6.5 The HAPA, adapted from Schwarzer (2004)

4 The health action process approach (HAPA)

4.1 Description of the model
The third theory to be discussed in this chapter is the HAPA. A number of
different versions of the theory have been published and different names
have been used for the same constructs. The description of the theory
presented here is based on the most recent publications including Ralf
Schwarzer’s website (Schwarzer 2004), from which Figure 6.5 is taken.
The HAPA postulates at least two distinct phases or stages: a motivation
(or pre-intentional) phase and a volition phase (also called a self-regulatory
or action phase). The latter is further subdivided into a planning phase, an
initiation phase and a maintenance phase. In the motivation phase, three
variables are held to inuence intention (or goals) directly: risk perception
(sometimes referred to as risk awareness or threat), and outcome expec-
tancies and self-efcacy from Bandura’s (1986) social cognitive theory (see
Luszczynska and Schwarzer, Chapter 4 in this volume). The model suggests
a causal order among these three predictors: ‘. . . threat is specied as a
distal antecedent that helps to stimulate outcome expectancies which fur-
ther stimulate self-efcacy’ (Schwarzer 2004). Taken literally, the arrows
would be interpreted as follows: people who, for whatever reason, have
higher risk perceptions will, as a consequence, develop more favourable
outcome expectancies; and those who develop more favourable outcome
expectancies will, as a consequence, have a higher level of self-efcacy. Or,
to put it differently, an increase in risk perceptions leads to an increase in
outcome expectancies, which in turn leads to an increase in self-efcacy.
The volition phase is represented by the right-hand part of Figure 6.5.
Descriptions of this phase of the HAPA focus on the cognitions involved in
initiating and controlling the action. For example, the formation of detailed
action plans is seen as essential to translate intentions into action
258 Stephen Sutton

(Gollwitzer 1999; Sheeran et al., Chapter 7 in this volume). Self-efcacy is

regarded as having a key role in all phases. This leads to the notion of
phase-specic self-efcacy. Maintenance self-efcacy, for example, refers to
optimistic beliefs about one’s ability to deal with barriers that arise during
the maintenance period.
Although the motivation phase of the HAPA is well specied as a causal
model, the volition phase is a framework (or ‘heuristic’, to use Schwarzer’s

Table 6.12 Applications of the HAPA

Behaviour Authors

Single-occasion drinking Murgraff et al. (2003)

Smoking Schwarzer and Fuchs (1995a)
Exercise Schwarzer and Fuchs (1995a); Lippke et al.
(2004a, 2004b,a in press); Ziegelmann, et al.
(2004a),a Sniehotta et al. (in press-a,a in press-b)
Healthy eating Schwarzer and Fuchs (1995a, 1995b); Schwarzer
and Renner (2000); Renner and Schwarzer (in
press)
Resisting dieting Garcia and Mann (2003)
Condom use Schwarzer and Fuchs (1995a)
Cancer screening Schwarzer and Fuchs (1995a)
Breast self-examination Garcia and Mann (2003); Luszczynska and
Schwarzer (2003); Luszczynska (2004)a
Testicular self-examination Barling and Lehmann (1999)
a
Intervention studies.

(2001) term) that needs further specication before the theory can be fully
operationalized and tested. Empirical applications of the model have only
recently started to try to represent the volition phase by including measures
of planning and phase-specic self-efcacy as well as behaviour (e.g. Snie-
hotta et al. in press-b).

4.2 Summary of research

Applications of the HAPA are listed in Table 6.12. Several cross-sectional
studies of the HAPA have investigated predictors of intentions (Schwarzer
and Fuchs 1995a; Garcia and Mann 2003) or current or recent behaviour
(Barling and Lehmann 1999; Renner and Schwarzer in press). Other HAPA
studies have used longitudinal designs to examine the predictors of beha-
viour as well as intentions (Schwarzer and Fuchs 1995b; Schwarzer and
Renner 2000; Luszczynska and Schwarzer 2003; Murgraff et al. 2003;
Lippke et al. 2004a, in press; Sniehotta et al. in press-b). The target
behaviours in these studies included exercise, healthy eating, single-
occasion drinking and breast self-examination. In some cases, the studies