310 Attitudes in Social Behavior
consistency), or (c) the favorability implied by the individual’s
behavioral experience with the object (evaluative-behavioral
consistency). Most past research has examined evaluative-
cognitive consistency (see Eagly & Chaiken, 1993, 1998).
Attitudes that are high in evaluative consistency are more
stable (e.g., Rosenberg, 1968), more resistant to change (e.g.,
Chaiken & Baldwin, 1981), more likely to influence informa-
tion processing (e.g., Chaiken & Yates, 1985), and more likely
to predict behavior (e.g., Chaiken, Pomerantz, & Giner-Sorolla,
1995) than are attitudes that are low in evaluative consistency.
These effects of evaluative consistency probably reflect—at
least in part—that consistent attitudes yield similar evaluative
reactions to the object regardless of the situational salience of
attitude components. Consistent attitudes might also be held
more confidently and be more accessible than are inconsistent
attitudes (see Chaiken et al., 1995).
Ambivalence
Ambivalence refers to the simultaneous presence of conflict-
ing positive and negative elements within an attitude (Bell
et al., 1996; Kaplan, 1972; I. Katz & Hass, 1988; M. M.
Thompson et al., 1995). Ambivalence can occur between ele-
ments of the same component of an attitude, such as when
people possess both positive and negative feelings about a
minority group (intracomponent ambivalence), or between
two components of an attitude, such as when people possess
negative beliefs but positive feelings about junk food (inter-
component ambivalence). Attitudes that are ambivalent are
likely also to be low in evaluative consistency, but the con-
structs are distinct: Low consistency refers to discrepancies
between the overall evaluation and one component, whereas
ambivalence refers to discrepancies between elements of a
component or between components (Maio et al., 2000). Am-
bivalent attitudes have been shown (compared to nonambiva-
lent attitudes) to be easier to change (e.g., Armitage &
Conner, 2000) and to be less predictive of behavior (e.g.,
Lavine et al., 1998). Ambivalent attitudes have also been
shown to polarize judgments when one of the conflicting
elements is made more salient than another. For example,
MacDonald and Zanna (1998) showed that individuals with
ambivalent attitudes toward feminists made either more fa-
vorable or more unfavorable judgments about a feminist job
applicant, depending on whether positive or negative infor-
mation was made salient, whereas individuals with nonam-
bivalent but equally extreme attitudes were not affected by
the salience of positive or negative information. Ambivalent
attitudes are hypothesized to have these polarizing effects
because such attitudes contain both positive and negative
information; priming can make available one or the other
category of information, which then influences judgments.
There is also some evidence that ambivalent attitudes are less
accessible than are nonambivalent attitudes (Bargh et al.,
1992), which might explain in part why the former are more
pliable and less predictive of behavior (Armitage & Conner,
2000).Strength: An Integrative Concept?The characteristics of attitudes discussed to this point overlap
in several ways. First, they all tend to influence the degree to
which attitudes are stable, resist change, affect the perception
of attitude-relevant stimuli, and influence behavior. Also, the
characteristics tend to be interrelated. For example, attitudes
based on direct experience tend to be more extreme, less am-
bivalent, and more accessible; evaluatively consistent atti-
tudes tend to be more accessible and less ambivalent;
ambivalent attitudes tend to be less extreme and less accessi-
ble; and so on.
Intuitively, all of these characteristics reflect the extent to
which attitudes are importantto individuals (Krosnick, 1989)
or are held with conviction (Abelson, 1988). The term
attitude strengthhas become a common label for this quality
(e.g., Petty & Krosnick 1995). Theorists have linked many
attitude characteristics with strength, including extremity,
intensity, certainty, importance, embeddedness, direct experi-
ence, accessibility, conviction, evaluative consistency, am-
bivalence, and vested interest (see Bassili, 1996; Krosnick &
Abelson, 1992; Raden, 1985). Each of these characteris-
tics incorporates aspects of subjective certainty, personal
importance, and significant psychological and behavioral
consequences.
Given the conceptual overlap among these various charac-
teristics, theorists have wondered whether the variables rep-
resent more-or-less-interchangeable terms for attitude
strength—in other words, whether the characteristics form a
single dimension ranging from weak to strong attitudes. The
most common way to investigate this issue has been to mea-
sure numerous characteristics and conduct a factor analysis
of the data. If a single factor emerged, the unidimensional at-
titude strength notion would be supported, whereas if multi-
ple factors emerged, a more complex framework would be
suggested. Such studies have generally supported the multi-
dimensional view (e.g., Abelson, 1988; Krosnick, Boninger,
Chuang, Berent, & Carnot, 1993), although the precise struc-
tures of the factors emerging from the analyses have been in-
consistent. Based on these data, the most common conclusion
has been that the various characteristics should be viewed as
distinct but related constructs (e.g., Krosnick et al., 1993;
Raden, 1985).