Psychological Bulletin

1990, Vol. 107, No. 3, 367-374

Copyright 1990 by the American Psychological Association, Inc.

0033-2909/90/$00.75

Involvement and Persuasion: Tradition Versus Integration

R i c h a r d E. P e t t y a n d J o h n T. C a c i o p p o
Ohio State University
In a recent meta-analysis, Johnson and Eagly (1989) questioned our conceptualization of and evidence for the effects of involvement on persuasion (Petty & Cacioppo, 1979, 1986). In particular,
they concluded that (a) what we had termed issue involvement represented two distinct types of
involvement (outcome- versus value-relevant), (b) each type of involvement had unique effects on
persuasion, and (c) outcome involvement effects may be obtained only by 1 group of researchers.
We argue that although 2 distinct research traditions of involvement have emerged, our original
position that the 2 categories of involvement induce similar processes in persuasion situations remains viable. Evidence from Johnson and Eagly's meta-anatysis shows that as both types of involvement increase, argument quality becomes a more important determinant of attitudes. The greater
message rejection found with involvement in value as compared with outcome studies can be explained in terms of confounding factors. Finally, we note that the outcome involvement effects that
we reported initially have been replicated by other investigators, including Johnson and Eagly.

In a recent article in this journal, Johnson and Eagly (1989)

reported a meta-analysis of the accumulated research on involvement and persuasion in which they concluded that it was
useful to distinguish between value- and outcome-relevant involvement.l In addition, they suggested that although effects for
the first type of involvement were robust, effects for the latter
type of involvement may be obtained only by one group of researchers. In this article, we question both of these conclusions
and provide a brief critique of their meta-analysis.

ment" (p. 310). That is, the VRI research consists largely of
studies conducted before 1975 that were inspired by social judgment theory (e.g., Sherif & Hovland, 1961), whereas the ORI
research consists largely of studies conducted in the past decade
that were instigated by cognitive response theory (Greenwald,
1968; Petty, Ostrom, & Brock, 1981 ) or the elaboration likelihood model of persuasion (Petty & Cacioppo, 1981, 1986). Second, Johnson and Eagly argued that the two research traditions
have uncovered "distinctively different effects [of involvement]
on persuasion" (p. 290). That is, VRI studies have tended to
show that increasing involvement is associated with reduced
persuasion (a main effect for involvement as predicted by social
judgment theory; Sherif & Sberif, 1967), but ORI studies have
tended to show that involvement interactswith argument quality such that a manipulation of argument cogency has a greater
impact under high-involvement than low-involvement conditions. In some studies the interaction shows more specifically
that increasing involvement is associated with reduced persuasion when the message arguments are weak but with increased
persuasion when the arguments are strong (Petty & Cacioppo,
1979).

Utility of the Value Versus O u t c o m e

Involvement Distinction
In Johnson and Eagly's (1989) view, value-relevant involvement (VRI) occurs when the topic of a persuasive communication is "linked to important values" (p. 290) such as freedom or
equality (Rokeach, 1968), and outcome-relevant involvement
(ORI) occurs when the topic of the message is linked to the recipient's "currently important goals or outcomes" (p. 292),
such as obtaining a college degree. Johnson and Eagly criticized
those who have postulated that the two constructs have similar
effects in persuasion contexts (i.e., Petty & Cacioppo, 1979),
and they provide two primary arguments for the validity and
utility of their distinction. First, they noted that the two kinds
of involvement stem from distinct "traditions of experimentation" (p. 293) with "distinct bodies of research that r e f l e c t . . .
different ways that researchers have thought about involve-

Integrative View of Involvement

Our view is that Johnson and Eagly's (1989) categorical distinction between outcome and value involvement is premature
and that there is a more parsimonious and integrative manner
in which to view the effects of involvement on persuasion. Furthermore, we argue that the data from their own analysis support this alternative approach. The alternative view (Petty &
Cacioppo, 1979, 1986) relies on the notion of "issue involve-

Preparation of this article was facilitated by National Science Foundation Grant BNS 84-18038.
We are grateful to the following people for providing comments on a
previous draft of this article: Robert B. Cialdini, Russell Fazio, Anthony
Greenwald, Stephen Harkins, Martin Heesacker, Jon Krosnick, Diane
Mackie, Thomas Ostrom, Steven J. Sherman, Richard Sorrentino, and
Wendy Wood.
Correspondence concerning this article should be addressed to Richard E. Petty or John T. Cacioppo, Department of Psychology, Ohio State
University, 1885 Neil Avenue Mall, Columbus, Ohio 43210-1222.

Another type of involvement referred to by Johnson and Eagly as

"impression-relevant involvement" and by others as "response involvement" (e.g., Chaiken, 1980; Leippe & Elkin, 1987; Petty & Cacioppo,
1979; Zimbardo, 1960) is widely accepted as occurring when the selfpresentational consequences of one's attitude are salient (cf. Cialdini,
Levy, Herman, Kozlowski, & Petty, 1976) and is not in dispute here.
367

368

RICHARD E. PETTY AND JOHN T. CACIOPPO

ment," defined as "the extent to which the attitudinal issue under consideration is of personal importance" (Petty & Cacioppo,
1979, p. 1915; italics added. See also Apsler & Sears, 1968;
Kiesler, Collins, & Miller, 1969.) According to this view, the
topic of a persuasive message can have personal importance because it is related to a variety of self-relevant constructs such as
values, goals, people, and objects. In our alternative framework,
the most self-relevant or involving message would be one about
the self (e.g., about one's intelligence or personality). Also of
importance, however, would be messages about other people or
objects that are relevant to the message recipient, or outcomes,
ideas, values, and end states that are important to the person.
As McGuire (1989) noted, the dimension of importance is
"transcendental in that all topics can be projected on" it
whether they are "relatively concrete (such a s . . . m o t h e r ) . . .
or abstractions (such a s . . . justice)" (p. 39). The critical aspect
of issue involvement is that the topic of the message is perceived
as important to the self. That is, a message perceived as relevant
to my value of freedom, mygoal of obtaining an education, my
sister, or my car is more involving than a message about your
values, goals, siblings, or possessions. 2 Furthermore, the more
important the value, goal, sibling, or possession is to the self,
the higher the level of involvement with a message on that topic.
Is it necessary to have two or more kinds of involvement to
account for messages dealing with important values, goals, siblings, possessions, and so forth? We think not.
Instead, we proposed that where the topic of the message falls
on the personal importance continuum is more critical for understanding persuasion processes than whether the communication topic is one that deals with important values, goals, people,
or objects. In all cases, as the personal importance of the topic
increases, recipients are postulated to become more motivated
to allocate their limited cognitive resources to processing the
message. As a consequence of this, the quality of the arguments
in the message becomes a greater determinant of influence
when involvement is high rather than low (Petty & Cacioppo,
1979, 1986). Importantly, Johnson and Eagly (1989) tested this
hypothesis in their recta-analysis and found it to hold within
both the ORI and the VRI studies. The interaction of involvement and argument quality within both categories of studies is
predicted by our alternative view but cannot be explained by
social judgment theory and is awkward for the Johnson and
Eagly view that the two kinds of involvement are categorically
different (or induce qualitatively different processes). 3
The challenge for our alternative view is to account for the
main effect for involvement that occurs within the VRI studies
that does not occur in the ORI set, although this main effect is
perhaps less interesting in light of the interaction. Our view is
that involvement (personal importance) per se is one of several
variables (e.g., personal responsibility, need for cognition) that
affect the intensity of information processing, but that the direction of any Processing bias (relatively favorable, unfavorable, or
neutral) is determined by other variables. For example, possessing much attitude-congruent knowledge or prior experience
might enable people to defend their positions more effectively
than can people with low knowledge or little experience (cf.
Krosnick, 1988; Wood, 1982). This issue-relevant knowledge
or experience would be more likely to be accessed and used in
processing if the message was self-relevant than if it was not.
Similarly, because people prefer their current positions to be

correct, the discrepancy between the message and one's own

position may motivate biased information processing (i.e., people prefer their values to be upheld, their desired outcomes to
be supported, and their favorite possessions to be praised; cf.
Howard-Pitney, Borgida, & Omoto, 1986; Lord, Ross, & Lepper, 1979). This means, for example, that increasing personal
importance will enhance processing in a relatively objective
fashion when people have relatively little (or very balanced)
knowledge about a topic and a weak or relatively neutral prior
opinion, but that the enhanced processing induced by high involvement becomes more negatively biased as people's attitudecongruent knowledge and attitude strength increase (cf. Fazio,
1989). Similarly, as we noted in a previous discussion of involvement,
The joint consideration of personal relevance and m e ~ discrepancy suggests that as a message becomes more counterattitudinal,
the message arguments may have to be stronger to produce the
same degree of acceptance. Likewise, as a m ~ becomes more
proattitudinal, the m ~
arguments may have to be weaker to
produce the same degree of rejection. (Petty & Cacioppo, 1986,
p. 88)
Figure 1 depicts the expected attitude results when both the
intensity of processing and the direction of processing bias are
considered along with the quality of the arguments in a message.
Each of the three panels shows that as the intensity of processing
increases (as induced by the personal importance of a message,
for example), the quality of the arguments accounts for more
variance in attitudes. The specific form of this Intensity Argument Quality interaction varies in the three panels, however. In
the middle panel, the increased processing is relatively objective, such that more favorable attitudes result when "strong"
arguments receive increased scrutiny, but less favorable attitudes result when "weak" arguments are evaluated. In contrast
to this relatively objective processing, consider a person who
is motivated (e.g., because the message takes a very discrepant
position) and able (e.g., because the person possesses considerable attitude-consistent knowledge) to counterargue the message. This person's task is advanced to the extent that the message provides weak rather than strong arguments in support of
its position, resulting in the pattern depicted in the fight panel
of Figure 1. Similarly, in the left panel of Figure 1, the expected
results of increasing the intensity of processing are shown when
conditions foster a favorable bias. Here the person's task is advanced to the extent that the message contains strong rather

2 This view is compatible with Sherif and Cantril's (1947) view of

"ego" (1947). They wrote that "the components of the ego include the
individual's body and physical characteristics; the things he learns belong to him, such as his clothes, his t o y s . . , his sweetheart, his children; together with a whole host of social values he also learns and with
which he identifies himself--his country, his politics...." (p. 117; see
also Ostrom & Brock, 1968).
3 There are other indications that ORI and VRI produce similar
effects. For example, research has shown that whether involvement is
defined as in the social judgment theory (VRI) tradition (e.&, by using
"latitude of rejection"; Fazio & Zanna, 1978) or in the cognitive response (ORI) manner (e.g., by examining people whose personal outcomes differ; Sivacek& Crano, 1982), increasing involvement is associated with enhanced attitude-behavior consistency (see review by Ajzen,
1987).

369

INVOLVEMENT AND PERSUASION

I.

11,

T'

Favorable Bias

A W m

T'

No Bias

~ Slro~

ml

Unfavorable Bias

f
i ~
i ~

$il~mg
Aiq~n(l

li~lulllefll i

"

Er
Wellk
A iiiiimenl i

mr

,
,
Low
High
ELABORATION

l
l

i
,
Low
High
ELABORATION

l
l

Low
High
ELABORATION

Figure 1. Hypothesized effects of information processing intensity and direction of bias on attitudes in
response to strong and weak messages (adapted from Communication and PersuasiOn. Central and Peripheral Routes to Attitude Change by Richard E. Petty and John T. Cacioppo, 1986, p. 34. Copyright 1986 by
Springer-VerlagNew York, Inc. Adapted by permission.) (Panel 1: Favorable bias conditions. Panel II: No
bias [relativelyobjectiveprocessing]. Panel IIh Unfavorable bias conditions.)

than weak arguments. In sum, our point is that one must consider not only the fact that the personal importance of a message
increases the extent of information processing activity, but one
must also consider how other factors in the persuasion environment (e.g., knowledge, message position, etc.) affect the direction of processing (i.e., do they produce a favorable, unfavorable, or no bias?)
Confoundings in V R I Research
Given these theoretical considerations, it is noteworthy that
two types of confounding may have motivated or enabled a bias
toward unfavorable processing in the VRI but not in the ORI
studies included in the Johnson and Eagly (1989) review. If so,
this would explain why their meta-analysis found that ORI
studies tended to show the attitude pattern depicted in the middle panel of Figure 1, but VRI studies tended to yield the pattern depicted in the right panel. 4 The two types of confounding
that may have induced an unfavorable bias in the VRI relative
to the ORI studies are (a) confoundings with the high- versus
low-involvement distinction within the value-relevant studies
and (b) confoundings with the value-outcome distinction itself.
Regarding the first type of confounding, Johnson and Eagly
(1989, p. 306) wrote,
we suspect that involvement was sometimes correlated to some extent with variables that could be considered confounds (e.g.,
knowledgeabilityabout issues, confidence in own attitudinal position, accessibility of counterarguments), especially in the studies

that varied involvement by classifyingsubjects or varyingthe issue

of the persuasivemessage [i.e., the VRI studies].
Although they acknowledge these confoundings, they dismiss
them by arguing that "the exact nature of any confounding
would have differed across the studies, rendering less plausible
any argument that a single confound explains the effects of
value-relevant involvement" (p. 306). Despite the fact that no
single confounding may account for all of the results, it is critical to note that each of the confoundings that Johnson and
Eagly identified are likely to work in the same direction. That
is, each of the confoundings in the VRI studies is plausibly tied
to increased resistance to persuasion. 5 Furthermore, in some
of the VRI studies, the resistance bias in the high-involvement
conditions is quite blatant. For example, in the research by Miller (1965) included in the VRI set, subjects in the high-involvement conditions were explicitly instructed to list all of the reasons that favored their initial attitudes before receiving the communication and were further told that the sponsors of the study
agreed with their position and that much evidence supported
it. These aspects of the high-involvement manipulation (absent

4 Johnson and Eagly (1989) were unlikely to uncover the favorable

bias pattern depicted in the left panel of Figure 1 because they excluded
"studies or conditions within studies in which subjects received proattitudinal messages" (p. 295).
5 The confounds exist in the VRI studies because these studies tend to
be correlational, whereas the ORI research uses experimental designs.

370

RICHARD E. PETTY AND JOHN T. CACIOPPO

under low involvement) appear sufficient to account for the reduced attitude change it engendered without the need to appeal
to the concept of value-relevant involvement.
Even if we ignored all confoundings with involvement in the
VRI studies, however, there are still confounds between the sets
of studies placed in the VRI and ORI categories that can account for the different pattern of results. For example, Johnson
and Eagiy (1989) reported that people generally had more
knowledge about the topics used in the value than in the outcome studies (p. 297). If high involvement increases access to
knowledge in both the value and outcome studies, but subjects
in the value studies have available more attitude-consistent
knowledge, then involvement would produce a greater resistance to influence in the value than in the outcome studies because of this confound (cf. Wood, 1982). 6 Similarly, Johnson
and Eagly reported that VRI studies tended to rely on change
scores (e.g., pretest-posttest designs), whereas ORI studies
tended to use posttest-only designs (p. 297). In the VRI studies
using a pretest, initial attitudes might be more highly accessible
and available to bias processing than in the posttest-only ORI
studies, especially when motivation to process was high (cf.
Fazio, 1989). As a third example, we note that the relatively
large difference in the average publication date for the VRI studies (i.e., 1970) versus the ORI studies (i.e., 1984) may pose problems of interpretation, especially for effects involving argument
quality. One possible problem is that Johnson and Eagiy's contemporary raters, judging from the perspective of the 1980s,
may have overestimated the strength of the arguments used in
the older VRI studies. This might occur if arguments that were
controversial in past decades regarding various social and political issues (e.g., birth control) were more acceptable today. For
example, the arguments in favor of U.S. policy during the Vietnam War (Gant, 1970) might have seemed weaker to undergraduates during the turbulent 1960s than they do to today's
more politically conservative college students. If the "strong"
arguments used in some of the VRI studies were weaker or more
controversial for the subjects at the time of the research than
they appear to raters today, this could account for the greater
apparent rejection of the strong arguments in the VRI than in
the ORI studies. 7

Summary
In sum, although we acknowledge the view that two historical
traditions of involvement research can be identified, we question Johnson and Eagly's (1989) conclusion that two categories
of involvement are needed to account for persuasion effects because it is not clear if the processes relevant to persuasion induced by self-relevance per se in these traditions differ. Instead,
the Johnson and Eagly meta-analysis shows that when involvement is high (whether value relevant or outcome relevant), the
quality of the arguments in a message accounts for more variance in attitudes than when involvement is low. This, of course,
directly supports our view that self-relevance increases the extent of message processing. The increased resistance found in
VRI studies as compared with ORI studies can be accounted
for completely by the confoundings (e.g., attitude-congruent
knowledge, attitude extremity) that are either documented or
plausibly present in the VRI research.

Replicability a n d Generality o f the

O u t c o m e I n v o l v e m e n t Effect
Although at some level Johnson and Eagly (1989) apparently
accepted the pattern of results for outcome involvement manipulations (as they relied on these results to differentiate outcome
from value involvement), they nevertheless question whether
the outcome effects are obtained only by "Ohio State researchers." The Ohio State results are defined by papers authored or
co-authored by investigators "who obtained the PhD from
Ohio State University in the late 1970s" (Johnson & Eagly,
1989, p. 304; e.g., Leippe & Elkin, 1987; Petty, Cacioppo, &
Heesacker, 1981). Given the attention that Johnson and Eagly
give to the Ohio State group and to possible explanations as to
why this group but not others obtained the effect, readers may
be surprised to learn that Johnson and Eagly replicated the
Ohio State results themselves before final acceptance of their
article for PsychologicalBulletin.

The Johnson-Eagly and Other Replications

In a 1988 dissertation by Johnson that was supervised by
Eagly, two attempts were made to replicate our findings by "reproducing as closely as possible the procedures and materials of
the prototypical Petty and Cacioppo involvement and attitude
change experiment" (Johnson, 1988, p. 9). Panel I in Figure 2
presents our initial attitude results. The next two panels present
the results from the two experiments by Johnson (1988). In his
first replication attempt, the Involvement x Argument Quality
interaction was not statistically significant; however, consistent
with our interaction pattern, individual cell comparisons indicated that argument quality had a significant effect on attitudes
under high but not low involvement (see Panel II). In Johnson's
second replication attempt, the interaction pattern appeared
again and this time proved to be statistically significant (see
Panel III).s
6 Johnson and Eagly (1989) argued that involvement and knowledge
may be confounded in many natural settings (see also Abelson, 1988).
We have also noted this (e.g., Petty & Cacioppo, 1986) but do not feel
that this provides a compellingjustification against examining the separate contributions to the persuasion process.
7 Likewise, it is possible that current subjects underestimated the
"weakness" of certain arguments as they appeared to subjects in past
decades. The important point is that because of the time confounding,
it is difficult to directly compare the VRI and ORI studies for effects
dependent on the ratings of contemporary judges. Another issue relating to judges' ratings of argument quality in the Johnson and Eagly
(1989) review is that their judges were specifically instructed to ignore
their own views when rating argument quality, whereas in developing
the strong and weak argument manipulations used in our research, we
allow subjects to consider their own attitudes when responding. Furthermore, in our procedure, subjects do not simply rate the arguments as
strong or weak but are asked to provide a cognitive reaction to the argument to ensure that counterargning the strong arguments is unlikely
even when the subject considers his or her own position (for further
detail on these procedures, see Petty & Cacioppo, 1986).
s We only graph the conditionsfrom Johnson's (1988) secondstudy that
were conducted as an explicit replication of our work. In some additional
conditions of his experiment, subjects were provided with some background information about the topic before message exposure. The effect
of this background information was to increase protx~ng of the message
overall so that only a main effectfor argument quality was observed.

INVOLVEMENT AND PERSUASION

371

It

Ii

Petty & Cacioppo(1979)

- - Strong
An~Uments

+0.7.
0.0.
+0.$.
+014,

+0.3.
+OZ.

Johnson(I 988, ExperimentI )

+07I

0"6/"
05~+04~+0"3T
+02T

.0,]-

+I.t"

~ =s~mng

Afgumefla

0.0-

<

-0.1 -0.2-03-

k
Argumo~=

- 0 . 4 - I-

-05T
-0 IL~-0 7T

I
Low

-05"i-06T
"Q7T

High
INVOLVEMENT

+0.7
+0.6.
+O.S,
0.4+0.3.
HI *0.2+0,10.0-01< -02--

III.

IV.

Johnson (1988, Experiment 2)

Burnkrant& Unnava(I 989)

Strong

Al'~glTtnt$

5e I

/
54

Strong
Af~umont~J

5O

40
38--

-05T
-06T
-07T

Low
High
INVOLVEMENT

Weak

Arguments

:'"T
3,,~I
Low

High
INVOLVEMENT

I
Low

High
INVOLVEMENT

Figure 2. The interaction of involvement and argument quality in four experiments. (Panel I: Data from
Petty & Cacioppo [ 1979]. Panel II: Data from Johnson [ 1988, Experiment 1]. Panel III: Data from Johnson
[1988, Experiment 2]. Panel IV: Data from Burnkrant & Unnava [in press]. Panels I, II, and III report
attitude data in standard scores and Panel IV reports attitude data on a 7-point scale. In each case, higher
numbers indicate more favorable postmessage attitudes toward the topic.)
Johnson and Eagly (1989) were certainly aware of their successful replication of the Ohio State results before acceptance
of their article, but they chose not to disclose this to readers. It
is not clear why Johnson and Eagly would want to leave readers
of Psychological Bulletin with the impression that only Ohio
State researchers can obtain the effect when they have obtained
it themselves. It is also interesting to note that the Method section of their report indicates that they stopped searching standard data bases (e.g., Dissertation Abstracts International) in
mid-1987 or earlier; however, their analysis includes a 1988
master's thesis and two 1988 dissertations but not Johnson's
1988 dissertation, previously described.

Although the Johnson (1988) replication may be the most

persuasive one possible since it comes from the chief critics of
the effect, it is important to note that other studies cited in the
Johnson and Eagly (1989) article contain replications of the initial Petty and Cacioppo (1979) result as well (i.e., Huddleston,
1986; Sorrentino, Bobocel, Gitta, Olson, & Hewitt, 1988). 9
9 Johnson and Eagly (1989) eliminated the Huddleston (1986) study
from their meta-analysis because he failed to include sufficient information in his text so that they could calculate effect sizes, and they collapsed across a critical individual difference variable in the Sorrentino
et al. (1988) study, masking the fact that the effect was replicated by one

372

RICHARD E. PETTY AND JOHN T. CACIOPPO

Nevertheless, why have some investigators attempted to replicate the result and failed? We believe that this has occurred because although many of the studies have modeled the form of
our investigations, several have modified (or not included) some
critical substantive features. As one example, consider the dissertation by Homer (1987) cited by Johnson and Eagly as a failure to replicate. In this study, Homer modeled her involvement
manipulation after one previously used by us. In our study
(Petty, Cacioppo, & Schumann, 1983), involvement with a target consumer product was manipulated before exposure to an
advertisement for it. Subjects were told that the product would
be available in their local area soon (high involvement) or that
it would not (low involvement), and that after ad exposure they
would be asked to select a gift from among alternative brands
of the product (high involvement) or from among brands of
a nontarget product category. Homer (1987) manipulated the
perceived availability of the product (as we did) but enhanced
the manipulation by telling high-involvement subjects that they
would be entered in a lottery in which "the grand prize was
identified as the product shown in the experimental ad." In the
low-involvement conditions, the lottery prize was identified as
consisting of "donations from local retail establishments" (p.
30). In Homer's research, then, high-involvement subjects were
led to believe that they might actually win the product that appeared in the critical ad (a stereo), whereas our high-involvement subjects were led to believe that they would have to make
a decision about which brand among several in the product category (disposable razor) they should select for themselves. Our
subjects would presumably be motivated to evaluate the brand
featured in the ad so that they could make an informed choice.
Homer's subjects, having no influence over which brand they
would get, could only hope that the brand featured in the ad
(that which they might win) was a good one. As might be expected by the positivity bias set up by this manipulation, highinvolvement subjects evaluated the product as more desirable
than did low-involvement subjects.~

The Outcome Involvement Manipulation

Another concern Johnson and Eagly (1989) raised about the
ORI research is the "lack of variety of manipulations" used (p.
308). That is, in the accumulated ORI experiments, only three
manipulations of involvement have been used, all of which were
plausibly tied directly to outcomes that were or were not relevant to the subjects (e.g., instituting comprehensive exams before [high involvement] or after [low involvement] the subjects

personality type ("uncertainty oriented") but not the other ("certainty

oriented"). To resolve the inconsistency with past work, Sorrentino et
al. speculated and provided some evidence consistent with the idea that
the Canadian students in their study were more certainty oriented than
the American students used in previous research (p. 368). We also think
that it is important to note that, in accord with our view of the effects
of involvement, (a) in addition to Johnson ( 1988, Experiment 1), other
studies have obtained a statistically significant attitude difference between the strong and weak argument conditions under high but not low
(outcome) involvement (e.g., Axsom, Yates, & Chaiken, 1987); and (b)
nonpersuasion studies that have used the OR1 manipulation have shown
that it is effective in increasing information processing activity (e.g.,
Brickner, Harkins, & Ostrom, 1986).

graduated from college). If the same effects could be shown for

a manipulation that clearly invoked self-relevance but that did
not directly manipulate the outcome-relevance of the advocated position, then we could be more confident that the effect
is not confined to one class of manipulations. Interestingly, a
recent study by Burnkrant and Unnava (in press) addressed this
issue. In their study, the manipulation of involvement concerned whether the pronouns used in the message were chosen
to invoke self-referencing (e.g., "when your razor showed signs
of rust") and self-reflection ("you may remember f e e l i n g . . . " )
or not (e.g., "when one's r a z o r . . . ," "'one might have felt";
italics added). This manipulation was combined with one of
argument quality. The attitude results, presented in Panel IV of
Figure 2, replicated our initial finding (Petty & Cacioppo, 1979)
with a very different manipulation of self-relevance.

Summary
In sum, the ORI effect appears to be obtained by more than
the "Ohio State researchers" referred to by Johnson and Eagly
( 1989; e.g., Leippe & Elkin, 1987; Petty, Cacioppo, & Goldman,
1981). Rather, it is now clear that the effect has been obtained
by other investigators (including Johnson at Purdue) and is not
limited to just one class of manipulations.
Critique o f Meta-Analysis Procedures
Meta-analyses have been championed for making literature
reviews more objective. The Johnson and Eagly (1989) metaanalysis, however, is illustrative of the subjective judgments and
decisions that can be involved. For example, although some authors have argued that "commitment" manipulations are relevant to "involvement" (e.g., Pallak, Mueiler, Dollar, & PaUak,
1972), Johnson and Eagly decided to exclude this research from
their review because "it is not at all clear that the commitment
manipulations used in persuasion studies influenced the extent
to which the persuasive message activated subjects' values" (p.
291). Yet studies using "width of the latitude of rejection" as an
indication of involvement were included in their analysis even
though they report no evidence that latitudes influence the extent of value activation.
A second inconsistency concerns the authors' claim to have
eliminated studies with "obviously confounded manipulations" (p. 294). It is not clear what they mean by this. For example, Table 1 in their report reveals that three of the VRI studies
~0Perhaps the most important aspect of our research that has not
been followedclosely,we believe, concerns the argument quality manipulation. In our research, the arguments must meet certain criteria in
pilot testing before use in the research, and we have had to modify these
arguments over time and for subjects at different institutions. For example, the "weak" arguments used in the initial Petty and Cacioppo (1979)
study proved insufficiently weak in a subsequent experiment (i.e., they
elicited a relatively even mixture of favorable and unfavorable thoughts
rather than primarily unfavorable responses). Nevertheless, the creation
of a different set of even weaker arguments (labeled very weak) was
sufficient to produce the desired effect (Petty, Harkins, & Williams,
1980, see also Cacioppo, Petty, & Morris, 1983). Other researchers have
either used the arguments from one of our studies without examining
whether they met the criteria for their populations or have selected arguments according to criteria different from ours.

INVOLVEMENT AND PERSUASION

were explicitly coded into a "knowledge varied with involvement" category, suggesting an "obvious" confound. Also, studies that used different topics for the high- and low-involvement
manipulations introduced likely confounds. For example, in
the Rhine and Severance (1970) study, students would clearly
be more opposed to the high-involvement advocacy (increasing
tuition at their school) than to the low-involvement topic (increasing park size in a distant city). Furthermore, it is unlikely
that the arguments would have been equated for strength across
the two topics. One might also wonder why this study is classified as one of VRI rather than ORI as tuition clearly affects
students' outcomes, whereas distant park sizes do not. The authors explain this apparent anomaly by noting that "tuition was
being actively discussed by the [community and a] march on the
State Capitol had been held to protest suggestions for increased
tuition" (p. 296). In our view, this "justification" could easily
have been offered to explain the obvious outcome-relevance of
the tuition increase and certainly suggests the high level of opposition to this topic as compared with the one on park acreage.
Johnson and Eagly (1989) deliberately excluded other studies in
which "investigators compared groups differing in their initial
stands on issues" (p. 294), and they even eliminated two consumer studies in which "a difference in attitudes toward the
products probably existed prior to the experimental session" (p.
295; italics added). Clearly, different standards were used for
different studies. Some studies were excluded only when there
was "proof from data" (p. 294) of confoundings, but others
were not. Some studies were eliminated when confoundings
"probably existed" but other studies with probable confoundings were retained in the analysis.~l Although these decisions
may not have had a major impact on the overall results of their
meta-analysis, our discussion suggests that meta-analyses are
not necessarily immune to the subjective factors involved in
interpreting individual experiments (Petty, Kasmer, Haugtvedt,
& Cacioppo, 1987).
Conclusions
Our view is that it is premature to differentiate "value" from
"outcome" involvement on the basis of the two "traditions" of
research that have developed or on the seemingly different pattern of results observed for studies classified into each category.
The more remarkable finding in our view is the similarity of the
results for the two clusters of studies. That is, in each case as
involvement increased, the quality of the arguments presented
became a more important determinant of persuasion. This pattern is expected if both traditions of involvement manipulations affect persuasion by influencingthe extent of message processing. The increased resistance induced by involvement
found in the VRI but not the ORI studies can be attributed to
factors confounded with VRI, or factors confounded with the
value-outcome categories of research. Of course, we are in
agreement with Johnson and Eagly that final settlement of this
issue awaits relevant primary research.
t~ Just as it seems clear that some of the studies that Johnson and
Eagly (1989) classified as VRI used topics with obvious outcome implications (i.e., restricting teenage driving, draft deferments for college
students), it also seems likely that some of the topics of ORI studies
related to central values (e.g., the topic of "restricting collegedormitory
visits" relates to the value of "freedom").

373
References

Abelson, R. P. (1988). Conviction. American Psychologist, 43, 267-275.

Ajzen, I. (1987). Attitudes, traits, and actions: Dispositional prediction
of behavior of personality and social psychology. In L. Berkowitz
(Ed.), Advances in experimentalsocialpsychology(Vol. 20, pp. 1-63).
New York: Academic Press.
Apsler, R., & Sears, D. O. (1968). Warning, personal involvement, and
attitude change.Journal of Personality and Social Psychology, 9, 162166.
Axsom, D., Yates, S. M., & Chaiken, S. (1987). Audienceresponse as a
heuristic cue in persuasion. Journal of Personality and Social Psychology, 53, 30-40.
Brickner, M. A., Harkins, S. G., & Ostrom, T. M. (1986). Effects of
personal involvement: Thought-provoking implications for social
loafing. Journal of Personality and Social Psychology, 51, 763-769.
Burnkrant, R., & Unnava, R. (in press). Self-referencing:A strategy for
increasing processingof message content. Personality and Social Psychology Bulletin.
Cacioppo, J. T., Petty, R. E., & Morris, K. (1983). Effects of need for
cognition on message evaluation, argument recall, and persuasion.
Journal of Personality and Social Psychology, 45, 805-818.
Chaiken, S. (1980). Heuristic versus systematic information processing
and the use of source versus message cues in persuasion. Journal of
Personality and Social Psychology, 39, 752-766.
Cialdini, R. B., Levy, A., Herman, C. P., Kozlowski, L. T., & Petty,
R. E. (1976). Elastic shifts of opinion: Determinants of direction and
durability. Journal of Personality and Social Psychology, 34, 663-672.
Fazio, R. H. (1989). On the power and functionality of attitudes: The
role of attitude accessibility.In A. Pratkanis, S. Breckler, & A. Greenwald (Eds.), Attitude structure and function (pp. 153-179). Hillsdale,
NJ: Erlbaum.
Fazio, R. H., & Zanna, M. P. (1978). Attitudinal qualities relating to
the strength of the attitude-behavior relationship. Journal of Experimental Social Psychology, 14, 398-408.
Gant, V. W. (1970). Attitude change as a function of source credibility
and levels of involvement (Doctoral dissertation, Ohio University,
1970). Dissertation Abstracts International, 31, 3074A.
Greenwald, A. G. (1968). Cognitivelearning, cognitiveresponse to persuasion, and attitude change. In A. Greenwald, T. Brock, & T. Ostrom
(Eds.), Psychological foundations of attitudes (pp. 147-170). New
York: Academic Press.
Homer, P. M. (1987). Surrealism and advertising: A test of alternative
explanations (Doctoral dissertation, University of Oregon, 1986).
Dissertation Abstracts International, 47, 2657A.
Howard-Pitney, B., Borgida, E., & Omoto, A. M. (1986). Personal involvement: An examination of processingdifferences. Social Cognition, 4, 39-57.
Huddleston, B. M. (1986). An experimental investigation of the influences deceptive nonverbal cues exert on persuasive processes (Doctoral dissertation, Universityof Missouri--Columbia, 1985). Dissertation Abstracts International, 47, 709A.
Johnson, B. (1988). The effects of involvement, argument strength, and
topic knowledge on persuasion. Unpublished doctoral dissertation,
Purdue University.
Johnson, B., & Eagly, A. H. (1989). Effects of involvement on persuasion: A meta-analysis.PsychologicalBulletin, 106, 290-314.
Kiesler, C. A., Collins, B. E., & Miller, N. (1969). Attitude change. A
critical analysis"of theoretical approaches. New York: Wiley.
Krosnick, J. A. (1988). Attitude importance and attitude change. Journal of Experimental Social Psychology, 24, 240-255.
Leippe, M. R., & Elkin, R. A. (1987). When motives clash: Issue involvement and response involvement as determinants of persuasion.
Journal of Personality and Social Psychology, 52, 269-278.
Lord, C., Ross, L., & Lepper, M. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently con-

374

RICHARD E. PETTY AND JOHN T. CACIOPPO

sidered evidence. Journal of Personality and Social Psychology, 37,

2098-2109.
McGuire, W. J. (1989). The structure of individual attitudes and attitude systems. In A. Pratkanis, S. Breckler, & A. Greenwald (Eds.),
Attitude structure andfunction (pp. 37-69). Hillsdale, NJ" Erlbaum.
Miller, N. (1965). Involvement and dogmatism as inhibitors of attitude
change. Journal of Experimental Social Psychology,, 1, 121-132.
Ostrom, T. M., & Brock, T. C. (1968). A cognitive model of attitudinal
involvement. In R. Abelson, E. Aronson, W. McGuire, T. Newcomb,
M. Rosenberg, & P. Tannenbaum (Eds.), Theories of cognitive consistency:A sourcebook (pp. 373-383). Chicago: Rand McNally.
Pallak, M. S., Mueller, M., Dollar, K., & Pallak, J. (1972). Effect of
commitment on responsiveness to an extreme consonant communication. Journal of Personality and Social Psychology, 23, 429--436.
Petty, R. E., & Cacioppo, J. T. (1979). Issue involvement can increase
or decrease persuasion by enhancing message-relevant cognitive responses. Journal of Personality and Social Psychology, 37, 19151926.
Petty, R. E., & Cacioppo, J. T. ( 1981). Attitudes and persuasion: Classic
and contemporary approaches. Dubuque, IA: Wm. C. Brown.
Petty, R. E., & Cacioppo, J. T. (1986). Communication and persuasion:
Central and peripheral routes to attitude change. New York: SpringerVerlag.
Petty, R. E., Cacioppo, J. T., & Goldman, R. ( 1981). Personal involvement as a determinant of argument-based persuasion. Journal of Personality and Social Psychology, 41, 847-855.
Petty, R. E., Cacioppo, J. T., & Heesacker, M. (1981 ). Effects of rbetorical questions on persuasion: A cognitive response analysis. Journal of
Personality and Social Psychology, 40, 432-440.
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of
involvement.Journal of Consumer Research, 10, 135- !46.
Petty, R. E., Harkins, S. G., & Williams, K. D. (1980). The effects of
group diffusion of cognitive effort on attitudes: An information processing view. Journal of Personality and Social Psychology, 38, 8192.

Petty, R. E., Kasmer, J. A., Haugtvedt, C. P., & Cacioppo, J. T. (1987).

Source and message factors in persuasion: A reply to Stiff's critique
of the Elaboration Likelihood Model. Communication Monographs,
54, 233-249.
Petty, R. E., Ostrom, T. M., & Brock, T. C. (Eds.). (1981). Cognitive
responses in persuasion. Hillsdale, NJ: Erlbaum.
Rhine, R. J., & Severance, L. J. (1970). Ego-involvement, discrepancy,
source credibility, and attitude change. Journal ofPersonality and SociaI Psychology, 16, 175-190.
Rokeach, M. (1968). Beliefs, attitudes, and values:A theory of organization and change. San Francisco: Jossey-Bass.
Sherif, M., & Cantril, H. (1947). The psychology of ego-involvements:
Social attitudes and identifications. New York: Wiley.
Sherif, M., & Hovland, C. I. (1961). Social judgment: Assimilation and
contrast effects in communication and attitude change. New Haven,
CT: Yale University Press.
Sherif, M., & Sherif, C. W. (1967). Attitude as the individual'sown categories: The social judgment-involvement approach to attitude and
attitude change. In C. W. Sherif & M. Sherif (Eds.), Attitude, egoinvolvement, and change (pp. 105-139). New York: Wiley.
Sivacek, J., & Crano, W. D. (1982). Vested interest as a moderator of
attitude-behavior consistency. Journal of Personality and Social Psychology, 43, 210-221.
Sorrentino, R. M., Bobocel, D. R., Gitta, M. Z., Olson, J. M., & Hewitt,
E. C. (1988). Uncertainty orientation and persuasion: Individual
differences in the effects of personal relevance on social judgments.

Journal of Personality and Social Psychology, 55, 357- 371.

Wood, W. (1982). Retrieval of attitude-relevantinformation from memory: Effects on susceptibility to persuasion and on intrinsic motivation. Journal of Personality and Social Psychology, 42, 798-810.
Zimbardo, P. (1960). Involvement and communication discrepancy as
determinants of opinion conformity. Journal ofAbnormal and Social
Psychology, 60, 86-94.
Received J u n e 10, 1989
Revision received July 9, 1989
Accepted July 14, 1989