316 Attitudes in Social Behavior
behavioral control extends the model to behavior that is not
fully under volitional control; for example, individuals who
believe that they cannot easily perform a behavior might not
do it even if they have a positive attitude toward the behavior,
perceive that other people want them to perform it, or both
(see also Gollwitzer & Moskowitz, 1996). Thus, the nature of
the behavior—specifically, its controllability or difficulty—
influences the strength of the attitude-behavior relation.
Nature of the Attitude
Characteristics of the attitude also influence the strength
of the attitude-behavior relation, a point that we noted earlier
in this chapter. One of the first attitude qualities to be studied
in this regard was direct versus indirect experience: Attitudes
that are based on direct experience with the attitude object
predict behavior better than do attitudes that are based on in-
direct experience (see Fazio & Zanna, 1978, 1981). Presum-
ably, these findings reflected that attitudes based on direct
experience are stronger—more confidently held, more stable,
and so on—than are attitudes based on indirect experience.
Indeed, in a meta-analysis of the attitude-behavior consis-
tency literature, Kraus (1995) concluded that such attitudinal
qualities as direct experience, certainty, and stability pre-
dicted the strength of the attitude-behavior relation.
Fazio (1990) has proposed that the effects of direct expe-
rience operate through another manifestation of attitude
strength—namely, the accessibilityof the attitude. As noted
earlier, accessible attitudes are more likely to be associated
with biased perceptions of stimuli (e.g., Houston & Fazio,
1989; Schuette & Fazio, 1995). If accessible attitudes are
more likely to be evoked spontaneously in the presence of the
attitude object (e.g., Fazio et al., 1986) and to guide individ-
uals’ perceptions of situations, then they seem likely to serve
as the basis for action as well. In line with this reasoning,
Fazio and Williams (1986) found that voters who reported
their evaluations of candidates quickly were subsequently
more likely to vote for their preferred candidate than were
voters who reported their evaluations more slowly. In his
meta-analysis of past studies, Kraus (1995) also found that
attitude accessibility predicted attitude-behavior consistency.
Another attitude characteristic that has been related to
attitude-behavior consistency is ambivalence, which (as
described in earlier sections) refers to inconsistency within
or between the components of an attitude (e.g., affective-
cognitive ambivalence involves oppositely valenced affect
versus cognition). Ambivalent attitudes are generally less
predictive of behavior than are nonambivalent attitudes (e.g.,
Armitage & Conner, 2000), presumably because the conflict-
ing elements may become differentially salient at various
times or in various settings, thus inducing inconsistent ac-
tions. In an interesting twist on this reasoning, however,
Jonas et al. (1997) showed that encountering a new attitude
object that has both positive and negative aspects can arouse
attitudinal ambivalence, which in turn can cause individuals
to process information systematically (because of uncer-
tainty; see also Maio, Bell, et al., 1996—which in turn can
actually produce higher attitude-behavior consistency. Thus,
although ambivalence in existing attitudes may serve to re-
duce attitude-behavior consistency, ambivalence in newly
forming attitudes may have the opposite effect.
Lavine et al. (1998) showed that when attitudes were high
in affective-cognitive ambivalence, the affective component
predicted behavior better than did the cognitive component.
In contrast, for nonambivalent participants, the affective and
cognitive components of attitudes were equally predictive of
behavior. MacDonald and Zanna (1998) found that manipu-
lations of evaluative priming had a significant effect on the
behavioral intentions of attitudinally ambivalent individuals
but did not affect the intentions of nonambivalent individu-
als. Consistent with the studies described in the preceding
paragraph, these data indicate that ambivalent attitudes yield
greater behavioral variability across time and settings than do
nonambivalent attitudes.Personality VariablesFinally, some people may behave in accordance with their at-
titudes to a greater extent than do other people—that is, col-
lapsing across attitude-behavior domains (hence, ignoring
the nature of the attitude and the nature of the behavior),
personality variables might predict the strength of attitude-
behavior consistency.
The variable that has received the most attention in this re-
gard is self-monitoring (Snyder, 1987). Self-monitoring re-
flects the extent to which people base their behavioral choices
on internal versus external cues. Low self-monitors rely on
internal cues to guide their behavior, whereas high self-
monitors use external, situational cues as guides to action.
Given that attitudes are an internal construct, low self-
monitors should exhibit stronger attitude-behavior consis-
tency than do high self-monitors. Several researchers have
produced data in support of this prediction (e.g., Snyder &
Kendzierski, 1982; Zanna, Olson, & Fazio, 1980). Ajzen,
Timko, and White (1982) obtained evidence that the source
of the heightened attitude-behavior consistency for low self-
monitors was that these individuals were more likely to fol-
low through on their behavioral intentions than were high
self-monitors. Presumably, high self-monitors are easily di-
verted from their intended courses of action by unanticipated