302 Attitudes in Social Behavior
Measurement The expectancy-value model prescribes
a method for measuring attitudes: Participants must first
consider a list of potential attributes of an attitude object and
then for each attribute rate (a) the probability that the object
possesses the attribute, and (b) the desirability of the attribute.
In most research, the probability ratings are made using scales
from3(very improbable)to3(very probable) or from
0(not at all)to1(definitely). The evaluative ratings are
made using evaluative scales from3 (e.g.,very bad)to 3
(e.g.,very good). To derive the overall attitude, the product
of the probability and evaluative ratings is computed for
each attribute, and the products are summed across all of the
attributes.
The expectancy-value model is also compatible with an
open-ended thought-listing procedure for measuring atti-
tudes. In this procedure, participants list their beliefs about
the attributes of the attitude object, and they rate the desir-
ability of each attribute. An overall index of attitude is then
obtained by summing the desirability ratings. The thought-
listing procedure does not require probability ratings because
it elicits attributes that participants perceive as being highly
associated with the attitude object (Esses, Haddock, & Zanna,
1993; Esses & Zanna, 1995).
Evidence Research has examined the utility of the
expectancy-value model by testing whether people’s reports
of their own attitudes are correlated with the summed products
of the attitude-relevant expectancies and values. Results indi-
cate that there are at least moderate correlations between atti-
tudes and the expectancy-value products (e.g., Budd, 1986;
van der Pligt & de Vries, 1998), although there have been sta-
tistical and methodological criticisms of these findings (e.g.,
Bagozzi, 1984; Sparks, Hedderley, & Shepherd, 1991).
To test the expectancy-value model directly, it is necessary
to examine experimentally the causal impact of beliefs and
evaluations on attitudes. Fortunately, studies of persuasion
have yielded some support for the idea that persuasive mes-
sages influence evaluative beliefs, which influence attitudes
(e.g., Maio, Bell, et al., 1996), although this effect may occur
only when people are motivated and able to process persua-
sive messages in a systematic manner (Chaiken, Liberman, &
Eagly, 1989; Petty & Cacioppo, 1986; E. P. Thompson,
Kruglanski, & Spiegel, 2000).
Reconciling the Three-Component
and Belief-Based Models
The comprehensiveness of the expectancy-value model is
challenged by the findings noted earlier—that affect and past
behavior predict attitudes independently of beliefs. This
challenge can be met by the argument that affective reactions
and past behaviors are simply different types of beliefs about
the attitude objects. For example, the three-component model
suggests that people may form a positive attitude toward an
object that makes them feel happy. The expectancy-value
model can account for this process by suggesting that people
believethat the object makes them happy and valuetheir
own happiness—the effect of happiness is reduced to an
expectancy-value product. Nevertheless, affective beliefs
and behavioral beliefs are made salient only by considering
the three-component model. Thus, at the very least, the three-
component model spurs the expectancy-value formulation to
consider different types of beliefs.
On the other hand, the three-component model would be
more compelling if the relations between the three attitude
components and attitudes fell into a discernible pattern that
could be explained by prior theory. Discovering such a pat-
tern is difficult, partly because there is conflicting evidence
for some attitude objects. For example, Esses et al. (1993)
found that affect played the greatest unique role in predicting
attitudes toward two groups (French Canadian and native
people), whereas beliefs about out-group values played the
greatest unique role in predicting attitudes toward two other
groups (Pakistani and homosexual people). These researchers
also obtained evidence that the relative dominance of emo-
tions and cognitions depended on individual differences and
situational factors.
These findings indicate that there is a need for theory de-
scribing when one component should be more influential than
another. Such a theory would need to consider evidence that
the roles of affect and cognition may depend on the psycho-
logical functions fulfilled by attitudes (see Maio & Olson,
2000). For example, affective reactions may be stronger pre-
dictors of attitude when the attitude object has a hedonic pur-
pose than when the object has a utilitarian purpose (Kempf,
1999). In addition, attitudes toward social partners become
more imbued with affect as people get older and when they are
diagnosed with a critical illness—conditions that presumably
increase the importance of close affective ties with others
(Carstensen, Isaacowitz, & Charles, 1999).
It is also important to consider that affect and cognition
may have different processing requirements. For example,
affective associations may be more accessible (Verplanken,
Hofstee, & Janssen, 1998) and they may be processed more
easily (Reeder & Pryor, 2000). Perhaps the ease of affective
processing explains why (a) affective reactions exert a
stronger influence on attitudes when there is a conflict between
affect and cognition (Lavine, Thomsen, Zanna, & Borgida,
1998), (b) affect has a stronger influence on mental represen-
tations of others in general (Jussim, Nelson, Manis, & Soffin,