actual range in the population (e.g. if the sample drink more alcohol than the sampled
population) then the observed R^2 will underestimate the real cognition–behaviour
relationship. Thus it is methodologically unrealistic to expect predictive models to
explain 100 per cent of the variance in measures of behaviour.
We would claim that explaining 21 per cent of the variance in objectively measured
behaviours is ‘impressive’ (rather than ‘damning’) because of the potential for behaviour
change intervention that this figure represents. Rosenthal and Rubin (1982) translate
percentages of explained variance into expected increases in outcome or success rates
using their ‘binomial effect size display’. This approach indicates that even when 19
per cent of the variance in behaviour is explained we would expect an increase in that
behaviour from 28 per cent in a control group to 72 per cent in an intervention group
(who had adopted the cognitions that explained the 19 per cent). This would indeed
be an impressive finding for any evaluation of a behaviour change intervention (see
Godin and Conner, 2008, for an examination of different indices of the intention–
behaviour relationship for physical activity). Of course, changing the cognitions
specified by the SCMs is a challenging endeavour (see Chapter 9) but the predictive
success of SCMs strongly indicates that models such as the TPB can specify change
targets that (if successfully changed) could make important differences to the prevalence
of health behaviours in the population and, thereby, public health. Consequently, after
reviewing available evidence and providing guidance for behaviour change inter -
ventions in the UK National Health Service, the National Institute of Health and
Clinical Excellence (2007: 10–11) noted that ‘a number of concepts drawn from the
psychological literature are helpful when planning... behaviour change with indi -
viduals’. This list included ‘positive attitude’, ‘subjective norms’, ‘descriptive norms’,
‘personal and moral norms’, ‘self-efficacy’, ‘intention formation’ and ‘concrete plans’.
Finally, returning to Ogden’s critique, she suggested that measuring such cognitions
as the TPB suggests, prompts their creation rather than simply recording pre-existing
thoughts and perceptions. As Ajzen and Fishbein (2004) point out this is a common
concern in questionnaire and interview studies. Recent research has in fact supported
this concern. The effect has been referred to as the ‘question-behaviour effect’ mean -
ing that measurement by itself prompts behaviour change. The strongest effects appear
to be associated with the measurement of intentions. In Sherman’s (1980) original
demonstration of the effect, one group of participants was asked to predict how
likely they would be to perform a socially desirable or socially undesirable behaviour
(volunteering for the American Cancer Society or singing the Star Spangled Banner
down the phone, respectively), while a second group made no prediction about
their behaviour. The results indicated that participants asked to predict their behaviour
were more likely to perform the socially desirable behaviour (31 per cent versus 4 per
cent) and were less likely to perform the undesirable behaviour (40 per cent versus 68
per cent) compared to control participants making no prediction. Recent research has
shown that completing a TPB questionnaire about blood donation led to a 6–9 per
cent increase in attendance of blood donation 6–12 months later compared to groups
who did not complete such a questionnaire (Godin et al., 2008). Other research has
shown the question-behaviour effect can be used to change behaviours such as health
screening attendance and influenza vaccination (Conner et al., 2011). Reviews of the
question-behaviour effect indicate a significant but small effect on subsequent behaviour
(Rodrigues et al., 2015; Wood et al., in press). However, rather than invalidating the
158 MOTIVATION AND BEHAVIOUR