not particularly good predictors of behaviour per se which has generated work exploring
the intention behaviour gap (Sutton 1998a; Gollwitzer 1993). Therefore, studies have also
used the TRA and the TPB to explore the cognitive predictors of actual behaviour. For
example, Shepherd and Stockley (1985) used the TRA to predict fat intake and reported
that attitude was a better predictor than subjective norms. Similarly, attitudes have also
been found to be the best predictor of table salt use (Shepherd and Fairleigh 1986),
eating in fast food restaurants (Axelson et al. 1983), the frequency of consuming low fat
milk (Shepherd 1988) and healthy eating conceptualized as high levels of fibre and fruit
and vegetables and low levels of fat (Povey et al. 2000). Research has also pointed to the
role of perceived behavioural control in predicting behaviour particularly in relation to
weight loss (Schifter and Ajzen 1985) and healthy eating (Povey et al. 2000). The social
norms component of these models has consistently failed to predict eating behaviour.Adding extra variables
Some studies have explored the impact of adding extra variables to the standard frame-
work described within the social cognition models. For example, Shepherd and Stockley
(1987) examined the predictors of fat intake and included a measure of nutritional
knowledge but found that this was not associated with either their measure of attitudes
or their participants’ behaviour. Povey et al. (2000) included additional measures of
descriptive norms (e.g. ‘To what extent do you think the following groups eat a healthy
diet’), and perceived social support (e.g. ‘to what extent do you think the following
groups would be supportive if you tried to eat a healthy diet’) but found that these
variables did not add anything to the core cognitions of the TPB. Research has also
examined the impact of accessing the individual’s hedonic responses to food with a focus
on beliefs about the sensory properties of the food concerned. The results, however, in
this area have been contradictory. For example, Tuorila-Ollikainen et al. (1986) asked
participants to rate their beliefs about low salt bread both before and after tasting some
bread and reported that the post-tasting hedonic ratings predicted eating behaviour
above their measure of attitudes. In contrast, Tuorila (1987) asked participants to rate
milk which varied in its fat content for its hedonic properties and reported that these
ratings of the sensory aspects of the food did not add anything to the basic cognitive
model. Shepherd (1989) provided a review of these studies and suggested that the
hedonic responses to food may be more important if the food is novel than if it is familiar.
The attitudinal research described so far conceptualizes individuals as holding either
positive or negative views towards a given object. In terms of eating behaviour it is
assumed that people either like or dislike certain foods and that this value-laden attitude
predicts food intake. Recent studies, however, have also explored the role of ambivalence
in predicting behaviour (Thompson et al. 1995) and this has been applied to eating
behaviour (Sparks et al. 2001). Ambivalence has been defined in a variety of different
ways. For example, Breckler (1994) defined it as ‘a conflict aroused by competing evalu-
ative predispositions’ and Emmons (1996) defined it as ‘an approach – avoidance conflict- wanting but at the same time not wanting the same goal object’. Central to all
definitions of ambivalence is the simultaneous presence of both positive and negative
values which seems particularly pertinent to eating behaviour as individuals may hold
144 HEALTH PSYCHOLOGY