This paper was presented at the Association for Education in Journalism and
Mass Communication in San Antonio, Texas August 2005.
If you have questions about this paper, please contact the author
directly. If you have questions about the archives, email
rakyat [ at ] eparker.org. For an explanation of the subject line,
send email to
[log in to unmask] with just the four words, "get help info aejmc," in the
body (drop the "").
Comparison of Indirect Sources of Efficacy Information in Pretesting
Messages to Prevent Drunken Driving
Ronald B. Anderson
Department of Advertising
Program in Public Relations
1 University Station
University of Texas at Austin
Austin, Texas 78712
[log in to unmask]
This experiment tested the impact of two forms of symbolic modeling
and verbal persuasion on self-efficacy beliefs and intentions to
prevent a friend from driving drunk. Three efficacy-information
public service announcements were produced to raise participants'
beliefs in their abilities to intervene successfully: a
behavioral-modeling message, which demonstrated the prevention
skills; a verbal-modeling message, which described the skills; and a
persuasive message, which only encouraged intervention. The controls
watched a public service announcement on drunken driving that did not
contain efficacy information. As predicted, both forms of symbolic
modeling engendered greater perceived self-efficacy and behavioral
intentions than did verbal persuasion, with behavioral modeling
registering the greatest effects. All efficacy-information
conditions surpassed the controls on most dependent measures.
Implications for designing public communication, or social marketing,
campaigns to prevent drunken driving are discussed, as well as
possible directions for further research.
Comparison of Indirect Sources of Efficacy Information in Pretesting
Messages to Prevent Drunken Driving
In 1991, the nation's 12 million college students spent more money
on alcoholic beverages than on textbooks, imbibing, on the average,
34 gallons apiece at a total cost of $4.2 billion. By the end of the
decade, that number had risen to $5.5 billion (Had Enough Campaign,
2001). Beer is their alcoholic beverage of choice, but not of
moderation. College students drink nearly 4 billion cans of beer
annually. To put this into perspective, if these cans were stacked
end-to-end upon each other, they would reach the moon and extend
70,000 miles beyond. It is estimated that 240,000 to 360,000 of
these students will eventually die of alcohol-related causes, such as
Binge drinking—defined as five or more drinks consecutively for
males and four for females—is the worst substance-abuse problem among
college students (Masters, 1994; Wechsler, Dowdall, Maenner,
Gledhill-Hoyt, & Lee, 1998). White males are the greatest offenders,
but White females are closing the gap. The percentage of college
women who binge drink has more than tripled in the past 20 years—up
from 10% in 1977, to 39% in 1997—trailing college men by only 11%
(Wechsler et al., 1998). The consequences of alcohol abuse among
college students are not limited to drunken-driving arrests,
accidents, and fatalities. Most campus rapes and violent campus
crime are alcohol-related. Sixty percent of college women who have a
sexually transmitted disease were drunk when they probably were infected.
Unfortunately, the public service campaigns that promoted moderation
and the designated-driver strategy so successfully during the 1980s
and the 1990s have not discouraged excessive drinking on college
campuses. This is not surprising since information alone seldom
changes risky behavior (Bandura, 1990; Flay, 1981; Maccoby &
Alexander, 1979; Maibach & Flora, 1993). People need to know why
they should cease harmful health habits, but to do so, they must
possess the necessary skills and believe they are capable of applying
them under stressful conditions.
This study investigated the impact of different forms of symbolic
modeling and verbal persuasion—two sources of efficacy information—on
self-efficacy beliefs and intentions to dissuade a friend from
driving drunk. It is part of a larger body of research on the role
of self-referent thought in the reduction of avoidance behavior
(Bandura, 1977), referred to as constraint recognition (Grunig &
Hunt, 1984, chap. 7) in the public relations literature. The overall
goal is to help health-campaign planners identify the psychological
constraints (see Anderson, 1995, 2000; Anderson & McMillion, 1995;
Grunig, 1989; Grunig & Ipes, 1983; Hertog, Finnegan, Rooney,
Viswanath, & Potter, 1993; Maibach & Murphy, 1995) that prevent
target publics from leading healthier lifestyles. Message strategies
are guided by principles of behavior modification, derived primarily
from the self-efficacy component of Bandura's (1977, 1986, 1997)
social cognitive theory. The messages tested in this study are based
on the results of several formative evaluations (Atkin & Freimuth,
2001; Kotler, Roberto, & Lee, 2002) of the drunken-driving problem
(see Anderson, 1989, 1995) that found young adult moderate drinkers
hold negative attitudes toward their friends' drunken driving but are
reluctant to express their concerns to them because they lack
confidence in their abilities to handle the situation properly and
because they have not observed this behavior among their age
group. The messages seek to remove this constraint by increasing
responsible drinkers' confidence, or self-efficacy, to manage this
sensitive interpersonal relationship. This approach, then, is
consistent with the "Friends Don't Let Friends Drive Drunk" campaign theme.
Origins of Campaigns of Self-Directed Change
Campaigns that use interpersonal communication to enhance the effects
of mass media are based on Lazarsfeld and Merton's (1948) early
notion of supplementation, which explains the importance of personal
influence, or change agents, in mobilizing public opinion on social
issues. Empirical research in the 1940s on the effects of mass
communication in presidential campaigns confirmed the mediating
influence of social networks, opinion leadership, and audience
selectivity factors on voter behavior (for reviews, see Katz &
Lazarsfeld, 1954; Lazarsfeld, Berelson, & Gaudet, 1944), laying to
rest assumptions about the omnipotent mass media and the malleable
mass audience that guided campaign planning in the early and
mid-1900s.1 Cartwright's (1949) analysis of the government's
persuasive campaigns to sell savings bonds during World War II
revealed that mass media supplemented by personal solicitation were
more effective in achieving sales than were mass media alone. He
concluded that audiences would be less likely to act upon mediated
behavioral recommendations unless they were given specific
instructions face-to-face on what to do and how to do it.
Cartwright's formulation served as one of the theoretical bases of
the Stanford Heart Disease Prevention Program (Maccoby, & Farquhar,
1975; Flora, Maccoby, & Farquhar, 1989) in the 1970s and 1980s. The
other primary theoretical orientation was Bandura's (1969, 1977,
1986) social learning theory, later renamed social cognitive theory
to differentiate it from other social learning approaches. This
conceptual scheme explains how principles of behavior modification
can be applied to modeled portrayals of risk-reducing behaviors
delivered through the mass media and interpersonally to heighten
expectations of successful performance, referred to as self-efficacy.
This long-term public health campaign compared the effects of a mass
media-only intervention with that of a mass media and
intensive-instruction intervention (what is called participant
modeling in self-efficacy theory) for a subsample of high-risk
residents in two northern California communities. A third town served
as a no-treatment control. After the first year, the media plus
supplemental personal-instruction town exhibited significantly
greater changes in cardiovascular risk and related knowledge and
behavior than did the media-only town, which outperformed the control
town. However, following the second year, differences between the two
treatment towns were not as pronounced, leading the Stanford
researchers to conclude that mass media alone can alter some
health-impairing habits when messages are designed to model
self-protective skills and raise perceived self-efficacy through
expectations of their successful performance in a variety of situations.
Constraint recognition is a key segmentation variable in Grunig's
(1989) situational theory of publics. Publics that believe they are
constrained from exercising control over situations seldom act on
issues that affect them because of a weak sense of self-efficacy.
Although constraint recognition and self-efficacy are conceptually
similar, they are inversely related; that is, when constraint
recognition is high, self-efficacy will be low, and vice versa.
Members of what Grunig and Hunt (1984, chap. 7) called a
constrained-behavior type of public often are capable of removing the
barriers that prevent them from leading a healthier lifestyle, but
convince themselves otherwise. This is because self-referent thought
intervenes between knowledge and behavior. Such self-doubt portends
failure, causing people to believe their health is beyond their
control and leaving them with a sense of fatalism or hopelessness.
However, a strong sense of personal efficacy motivates information
seeking and discussion of health issues, skills acquisition and
self-regulation, and enlistment of the social supports needed to
maintain behavior change.
Campaigns designed to remove the constraints that prevent optimal
health functioning require an understanding of the psychosocial
origins of avoidance behavior (Anderson, 1989, 1995, 2000; Bandura,
1977, 1990). According to social cognitive theory, behavioral
changes—whether instated behaviorally, vicariously, persuasively, or
emotionally—are mediated by a common cognitive mechanism, what
Bandura (1977, 1986, 1997) called self-efficacy, defined as one's
expectations of exercising control over troublesome situations. This
conceptual scheme proposes that avoidance behavior can be reduced or
eliminated by sources of efficacy information, whether behaviorally
or symbolically based. Each source of information increases
confidence to cope with subjectively threatening situations by
instilling expectations of successfully managing them, such as
overcoming the doubt that one cannot intervene successfully to
prevent a friend from driving drunk, or the fear of performing
monthly breast self-examination.
There are three types of self-efficacy beliefs: level, or magnitude,
strength, and generality. Level refers to judgments of task
performance under increasingly difficult circumstances. Strength
refers to the degree of certainty that each task can be executed
properly. Less-efficacious persons restrict themselves to the
easiest, most simple tasks, while highly efficacious persons perform
the most arduous and persevere despite dissuading
experiences. Generality refers to the extension of capability to the
performance of similar and dissimilar tasks within the same
behavioral domain or across related domains.
Maibach and Murphy (1995) have recommended that level of
self-efficacy should be indicated as a graded series of situational
demands in studies of health communication campaigns. Each successive
demand, or task, should be more difficult to achieve, or stressful to
manage. As noted, this requires formative research to identify the
behavioral scenarios to be ordered according to level of difficulty.
Strength of self-efficacy is then measured for each increasingly
difficult scenario. Such a measure captures judgments of the level
and strength of self-efficacy in a composite score. Measures of the
level dimension are needed only when threshold effects should be
determined, such as the lowest level of perceived self-efficacy
required to perform a set of fear-arousing tasks.
Depending on the situation and circumstances, people acquire
information about their capabilities from the following sources of
self-efficacy: (a) performance accomplishments, the most influential
source because it is based on direct mastery experience; (b)
vicarious experience, which uses live or symbolic modeling to instill
expectations of successful performance through observation; (c)
verbal persuasion, which relies on suggestion and exhortation to
raise perceived capability; and (d) physiological and affective
states, which provides somatic information about personal efficacy in
terms of level of anxiety and vulnerability to stress.
As a type of vicarious experience, symbolic modeling refers to
observing a model indirectly, as on television, through verbal
description, by reading, or in a picture (Maibach, 1995; Maibach &
Flora, 1993). Unlike the circumscribed effects of live modeling, the
potential effects of symbolic modeling are as pervasive as the mass
media and other mechanisms of symbolic communication. Because people
cannot retain all they observe, only the salient features of certain
modeled activities are stored for memory representation. Modeled
information is coded as symbols that subsequently guide behaviors
that are functionally valued. Rehearsal aids retention of newly
acquired behavior patterns and, if successful, enhances self-efficacy
(see Bandura, 1986, chap. 2, for a discussion of the processes
governing observational learning).
How modeling information is cognitively processed depends on the
developmental level of the observer. For example, preverbal children
rely more on modeled demonstrations than on verbal descriptions to
learn new behaviors because of their limited linguistic
competencies. As they grow cognitively, verbal modeling is used to
draw attention to important aspects of modeled activities, thereby
facilitating the acquisition of more complex and diverse ways of
thinking and behaving. In a study of the comparative effects of live
and verbal modeling, Bandura and Mischel (1965) found that while the
former had the greater impact, both forms of modeling were equally
effective at changing elementary school children's delay-of-reward
behavior, defined as their preferences for an immediate, less-valued
reward, or a delayed, more-valued reward. Similarly, the combination
of modeling and oral instruction produces better learning of
difficult subject matter by young children than does oral instruction
alone, and creates more positive attitudes toward learning the
material (White & Rosenthal, 1974).
Summary and Hypothesis
The theoretical formulation presented here presumes that observation
of the successful performance of threatening activities without
adverse consequences eliminates avoidance behavior by instating
perceived self-efficacy. Symbolic demonstrations of how to cope with
stressful situations are more convincing than are verbal descriptions
of the coping behavior, because they convey more efficacy information
and more accurately portray the conditions under which behavior will
occur and its consequences. These different forms of symbolic
modeling were employed to raise young-adult responsible drinkers'
confidence to dissuade their heavy-drinker friends from driving
drunk. Verbal persuasion was used to achieve the same result through
suggestion and exhortation. Efficacy expectations induced
persuasively are typically weaker than those instated vicariously
because they are not based on observation, or indirect experience
with the situation, but rather on the assurance of a persuader. The
impact of these treatment modalities was assessed by producing three
televised public service announcements—a behavioral-modeling
announcement, which demonstrated the intervention behavior; a verbal-
modeling announcement, which described the intervention behavior; and
a verbal-persuasion announcement, which only suggested viewers were
capable of performing the advocated behavior.
Based on the review of the literature, the following rank ordering
for sources of efficacy information on self-efficacy beliefs and
behavioral intentions was hypothesized: behavioral modeling, verbal
modeling, verbal persuasion, and control. This rank ordering also
was predicted for the most threatening of a series of hierarchically
arranged tasks involving a drunken friend.
Participants were 173 females and 68 males (N = 241) enrolled in
several undergraduate communication courses at a large southwestern
university. They volunteered to participate in the experiment
outside of class and received extra credit. One hundred eight
(44.8%) were seniors, 64 (26.6%) were juniors, 33 (13.7%) were
sophomores, 28 (11.6%) were freshmen, and 8 (3.3%) were graduate
students. Racially, 178 (73.9%) were White, 21 (8.7%) were Asian, 20
(8.3%) were African American, 19 (7.9%) were Hispanic, and 2 were
Arabic. Ninety-nine participants (41.1%) described themselves as
moderate drinkers, 76 (31.5%) as light drinkers, 50 (20.7%) as
nondrinkers, and 13 (5.4%) as heavy.2 On the average, they consumed
18.3 drinks per month and ranged in age from 17 to 43 years, with a
mean age of 21.
Preparation of Stimulus Materials
The independent variable—source of efficacy information—was
operationalized by creating three televised public service
announcements (PSAs), one for each efficacy-information condition
(i.e., behavioral modeling, verbal modeling, and verbal
persuasion). These PSAs were produced as animatics and were
approximately 60 sec in length (see Schultz & Barnes, 1999, chap.
9). An animatic is an artist's rendering, scene-by-scene, of a
television commercial, which is video taped and dubbed for
sound. Motion is simulated by using different camera movements. The
resulting spot announcement resembles a cartoon version of a
live-action public service message. Because animatics are relatively
inexpensive to produce, they are used by health-care
practitioners—particularly those who plan social marketing
campaigns—to pretest the potential effectiveness of print and
broadcast messages prior to final production (see Atkin & Freimuth,
2001). The diagnostic information from this type of formative
research is used to revise weak message strategies and executions or
approve production of those that indicate they will accomplish their
objectives, such as enhance self-efficacy and behavioral intentions.
The decision to use animatics, instead of live actors or finished
PSAs, was dictated by production, availability, and research design
considerations. Animatics are quicker and cheaper to produce than
are live-action PSAs. They are quicker because they require only an
artist, voice-over for the audio track, and a small production team,
whereas live action requires auditions, rehearsals, props, selection
of the location site, and hours of editing time. Animatics also were
chosen because the use of untrained actors could have resulted in a
loss of control over the design of the message stimuli, making the
PSAs appear unrealistic and contrived. On the other hand, close
supervision of the artist's work ensured all message formats were
similar except for the manipulation of the independent variable.
A video for pretesting health public service announcements (National
Cancer Institute, 1984) was obtained from the Department of Health
and Human Service's Office of Cancer Communications. The tape
contains a 15-min program on wildlife conservation and a series of
four product and service commercials, which appear between program
segments. The first three messages are finished commercials and the
fourth message is an animatic of a Dial Soap commercial. Between the
second and third commercials, and directly following the program, are
blank spaces for inserting the test public service announcement.
It was feared that mixing the animatics and finished commercials
within the same exposure sequence would threaten the internal
validity of the experiment, because participants, having never seen
an animatic, might be attracted to the test PSA for its novelty,
rather than for experimental purposes. To control for this potential
threat to treatment equivalency, the artist drew four storyboards of
the finished commercials so they could be videotaped as
animatics. This was accomplished by freezing each scene in the
commercial, allowing the artist to make sketches and notations of the
visual content. The product and test announcement storyboards were
videotaped and the audio portions were mixed with the video. Three
videos were produced, one for each different execution of the
independent variable (i.e., source of efficacy information).
Design and Procedure
Participants were randomly assigned to one of four conditions:
behavioral modeling (n = 66), verbal modeling (n = 60), verbal
persuasion (n = 60), or control (n = 55). Preexisting self-efficacy
was not assessed prior to random-condition assignments. Participants
read a cover story explaining that the researchers were interested in
their opinions of the wildlife program produced by a power
company. They then viewed the program titled "A Second Chance," and
were exposed twice to the distractor commercials and test PSAs—first,
about halfway through the program and again at its
conclusion. Following exposure, participants completed a
questionnaire containing measures of the dependent and demographic
variables, as well as distractor items about the program and standard
pretesting questions (see Atkin & Freimuth, 2001, for a discussion of
these measures).3 Participants were asked not to discuss the
experiment with their classmates and were thanked for their participation.
Three efficacy-enhancing PSAs were designed to encourage
participants to dissuade their heavy-drinker friends from driving
drunk. As mentioned, participants assigned to these conditions
received either the behavioral modeling, verbal modeling, or verbal
persuasion modes of efficacy induction. Each stimulus message
contained the same efficacy information (i.e., information on how
easy it is to dissuade a friend from driving drunk as long as you
believe you are capable of doing it without offending him or
her). All PSAs emphasized ease of performance and the benefits of
helping a friend, that is, contained the same reassuring
information. In each efficacy-information condition, the
spokesperson—a male college student—relates how he overcame his
reluctance to discuss with his friends their drinking and driving and
how easy this is now that he knows he can do it successfully. It was
believed that a spokesperson typical of the target public would
enhance the personal relevance of the PSAs. Indeed, attribute
similarity is often a key source factor in the message-design stage
of public health campaign planning (Devine & Hirt, 1989; McGuire,
1989; Pfau & Parrott, 1993). The critical difference between the two
symbolic modeling treatments is that the spokesperson demonstrated
the intervention behaviors, or social skills, in the behavioral
modeling condition, but only described them in the verbal modeling
condition. In the verbal persuasion condition, the spokesperson only
exhorts participants to intervene and reassures them they will
succeed. The differences among these treatments are explained in
greater detail next.
Behavioral modeling. In this condition, the intervention skills
were demonstrated by the spokesperson, who identifies himself as a
moderate drinker. He tells the audience how they can dissuade their
friends from driving drunk if they use the proper approach behavior
and persuasive arguments. He mentions how he used to worry that he
would offend his friends if he talked to them about their drunken
driving, because he never knew exactly what to say or how to say
it. The next few scenes depict how the spokesperson overcame his
reluctance. For example, he is shown casually approaching a male and
female friend and engaging in conversation with them about the legal,
financial, social, and psychological consequences of arrest for
drunken driving, such as the humiliation of having to tell your
parents, or a prospective employer. His friends are surprised to
learn of the extent and severity of these consequences and thank him
for his concern. The model then faces the camera and assures the
audience that they too can succeed, as long as they know what to say
and how to say it. The last scene is a close-up of the PSA's slogan:
"A Friend's Drunken Driving Is Your Business."
Verbal modeling. This stimulus message was identical to the
behavioral-modeling announcement, except the spokesperson modeled
verbally, rather than behaviorally, how to perform the intervention
tasks, using the familiar "talking head" PSA-message format. The
spot announcement concluded with the slogan.
Verbal persuasion. Participants in this condition were exposed only
to persuasive-efficacy information. The spokesperson exhorts
participants to prevent their friends from driving drunk and
reassures them that they will succeed if they use discretion and show
concern. Some financial consequences appear on screen as they are
mentioned. The PSA closes with the slogan.
Control. Participants in the control condition viewed a PSA
produced for the San Antonio Alcohol Safety Action Project that
relates the story of "D. W. Ier," a man who is arrested for driving
while under the influence. This spot announcement was chosen because
it was produced as an animated cartoon, and therefore resembled the
animatics. It was approximately the same length as the test PSAs and
did not contain any efficacy information, but did mentioned the
consequences of arrest for drunken driving. The PSA concludes with
the announcer asking, "Say, was that good time really worth all of
this? Don't you be a "D. W. Ier."
Efficacy expectations. As mentioned, behavioral scenarios were
created to indicate the level of perceived drunken-driving prevention
self-efficacy. The graded series consisted of four increasingly
threatening tasks involving a drunken friend, ranging from low to
high task, or constraint, recognition. These intervention behaviors
were based on the results of focus groups conducted as part of the
formative evaluations cited earlier that identified the range of
situational demands. For each hierarchical task, participants
recorded the strength of their self-efficacy on 11-point Likert-type
scales ranging from 0 (not at all comfortable) to 10 (extremely
comfortable). They judged how comfortable it would be for them right
now to ask a host/hostess or bartender not to serve a drunken friend
(least-threatening task); how comfortable it would be for them right
now to ask someone they know to help them dissuade a friend not to
drive drunk; how comfortable it would be for them right now to
express their concern by themselves; and how persistent they would be
right now if their friends counterargued with them (most-threatening
task). Cronbach's coefficient alpha for this index was .77.
To provide within-domain indices of the generality of
drunken-driving prevention self-efficacy, participants rated the
strength of their expectations in coping successfully with similar
and dissimilar threats. Participants rated on 0 to 10 scales how
comfortable they would feel expressing their concern to a casual
acquaintance (similar threat) and to a stranger (dissimilar threat)
that they do not drive drunk. They also rated the degree of
certainty for each of these nonhierarchical threats on 0 to 10
scales, where 0 indicated participants were not at all certain they
would feel comfortable and 10 indicated they were totally certain
they would feel comfortable. These scores were averaged to construct
the strength indices. The coefficient alpha reliabilities for the
similar and dissimilar threats were .77 and .64, respectively.
Behavioral intentions. Participants recorded their behavioral
intentions for each of the four tasks in the graded series on
11-point scales ranging from 0 (not at all likely) to 10 (extremely
likely). The mean of these ratings constituted the behavioral
intentions score. The alpha for this index was .67. Participants
also rated their behavioral intentions for the two generality
measures on 11-point scales. Because each item measured behavioral
intentions toward a different type of threat than presented in the
behavioral hierarchy (i.e., toward a similar or dissimilar threat)
and because they correlated only moderately (r = .57), the two
measures were treated separately in the analysis.
Planned comparisons and the chi-square test of independence were
used to test the hypothesis. A modified Bonferroni test (Keppel,
1982, chap. 8) was computed to correct the familywise error rate for
the number of analytical comparisons, resulting in an adjusted, more
conservative significance level of .025.
Manipulation check. To check whether different modes of efficacy
induction manipulated perceived self-confidence to dissuade a friend
from driving drunk, participants were asked two self-efficacy
items. The first asked how confident they were right now of their
ability to approach and express their concern to a friend about his
or her drinking and driving without offending him or her on an
11-point scale ranging from 0 (not at all confident) to 10 (extremely
confident). The second item inquired about how easy it would be to
talk this over with a friend on an 11-point scale ranging from 0 (not
at all easy) to 10 (extremely easy). The coefficient alpha of this
two-item index was .82. Planned comparisons confirmed that the
perceived self-confidence of participants in the behavioral modeling
condition (M = 7.54, SD = 2.27) was significantly greater than that
of participants in the verbal modeling (M = 7.07, SD = 1.66; t =
2.41, p <.01), verbal persuasion (M = 6.15, SD = 1.27; t = 5.39,
p <.001), and control conditions (M = 5.60, SD = 1.94; t = 6.96,
p <.001). The perceived self-confidence of participants in the
verbal modeling condition was significantly greater than that of
participants in the verbal persuasion (t = 2.90, p = .002) and
control (t = 4.50, p <.001) conditions. Participants in the
verbal persuasion condition did not differ significantly from the
controls on perceived self-confidence, although this contrast was
significant at the more liberal .05 level (t = 1.66, p =
.049). These results support the predicted ordered effects for
sources of efficacy information, thereby validating the experimental
Strength. The predicted rank ordering for the strength of
self-efficacy was largely confirmed. As presented in Table 1,
results revealed a marginally significant difference between the
behavioral modeling (M = 7.51, SD = 1.96) and verbal modeling (M =
7.07, SD = 1.17; t = 1.85, p = .03) conditions, and highly
significant differences between behavioral modeling and the verbal
persuasion (M = 6.50, SD = .94; t = 4.15, p <.001, eta2 = .07)
and control (M = 5.44, SD = .98; t = 8.34, p <.001, eta2 = .23)
conditions on perceptions of ability to perform the tasks in the
behavioral hierarchy. Participants in the verbal modeling condition
differed significantly on the strength of their expectations from
those in the verbal persuasion (t = 2.25, p = .01, eta2 = .02)
and control (t = 6.41, p <.001, eta2 = .15)
conditions. Participants in the verbal persuasion condition were
significantly more confident of their abilities to perform the
hierarchial tasks than were the controls (t = 4.20, p <.001,
eta2 = .07).
Because self-efficacy theory predicts that only people with a strong
sense of perceived self-judged efficacy will persevere in their
efforts to overcome highly aversive situations, planned comparisons
were conducted to determine which treatment mode engendered the
greatest effect on the most-threatening task in the behavioral
hierarchy—the counterargument task—which asked participants how
persistent they would be in their efforts to dissuade a friend from
driving drunk if that friend counterargued with him or her. As
expected, results indicated that participants in the behavioral
modeling condition (M = 7.72, SD = 2.35) would persist significantly
more in their efforts than would those in the verbal modeling (M =
6.63, SD = 1.98; t = 3.18, p <.001, eta2 = .04), verbal
persuasion (M = 5.88, SD = 1.61; t = 5.37, p <.001, eta2 = .11),
and control (M = 4.81, SD = 1.52; (t = 8.28, p <.001, eta2 =
.22) conditions. Participants in the verbal modeling condition
differed significantly from those in the verbal persuasion (t =
2.15, p = .015, eta2 = .02) and control conditions (t = 5.07, p
<.001, eta2 = .09) on this task. Participants in the verbal
persuasion condition would persist significantly more in their
efforts to resist a friend's counterarguments than would the controls
(t = 2.98, p = .<.001, eta2 = .04).
Insert Table 1 about here
To examine the differences between strong and weak levels of
self-efficacy, strength of efficacy expectations was further analyzed
by creating strong and weak categories, where 7-10 equaled strong and
0-6 equaled weak to moderate. A chi-square test revealed that
significantly more participants in the behavioral modeling condition
(81.7%) than in the verbal modeling (73.9%), verbal persuasion
(54.8%), and control (14.9%) conditions exhibited a strong sense of
perceived self-efficacy (X2[3, N = 195] = 54.59, p <.001, Cramer's V = .53).
Generalized self-efficacy. Although participants in the behavioral
modeling condition (M = 5.61, SD = 2.20) did not differ significantly
from those in the verbal modeling condition (M = 5.31, SD = 1.94;
t = .89, p = .18) on the generality of their efficacy
expectations for the similar threat (i.e., confidence to express
concern to a casual acquaintance), they did differ significantly from
participants in the verbal persuasion (M = 4.80, SD = 1.41; t =
2.44, p <.01, eta2 = .02) and control (M = 4.20, SD = 1.75; t =
4.20, p <.001, eta2 = .07) conditions (see Table 1). Participants in
the verbal modeling condition did not differ significantly from those
in the verbal persuasion condition on the similar-threat scale
(t = 1.52, p = .065), but they did differ significantly from the
controls (t = 3.26, p <.001, eta2 = .04). Participants in the
verbal persuasion condition did not differ significantly from the
controls (t = 1.78, p = .035), although this contrast approached
significance. As summarized in Table 1, the only contrast to reach
significance on the dissimilar-threat scale (i.e., confidence to
express concern to a stranger) was that between the behavioral
modeling condition (M = 3.67, SD = 2.54) and the controls (M = 2.56,
SD = 1.99; t = 2.73, p <.01, eta2 = .03), although there was a
trend toward significance between the verbal modeling and control
conditions (t = 1.73, p = .04).
Participants in the behavioral modeling condition (M = 7.35, SD =
1.82) were significantly more likely to perform the tasks in the
behavioral hierarchy than were those in the verbal modeling (M =
6.80, SD = 1.17; t = 2.29, p = .01, eta2 = .02), verbal
persuasion (M = 6.32, SD = 1.02; t = 4.30, p <.001, eta2 = .07),
and control conditions (M = 5.50, SD = 1.09; t = 7.56, p <.001,
eta2 = .19). Participants in the verbal modeling condition were
significantly more likely to perform the graded hierarchy of tasks
than were those in the verbal persuasion (t = 2.50, p <.01, eta2
= .02) and control (t = 5.19, p <.001, eta2 = .10)
conditions. The behavioral intentions of participants in the verbal
persuasion condition exceeded those of the controls (t = 3.28, p
<.001, eta2 = .04).
Regarding intentions to perform the most threatening task in the
behavioral hierarchy, participants in the behavioral modeling
condition (M = 7.84, SD = 2.14) were significantly more likely to
persist in their efforts to resist a friend's counterarguments than
were those in the verbal modeling (M = 6.45, SD = 2.16; t =
4.08, p <.001, eta2 = .06), verbal persuasion (M = 6.01, SD = 1.52;
t = 5.35, p <.001, eta2 = .11), and the control (M = 5.18, SD =
1.73; t = 7.60, p <.001, eta2 = .20) conditions. However, the
intentions of participants in the verbal modeling condition did not
differ significantly from those of participants in the verbal
persuasion condition (t = 1.23, p = .10), although they did
differ from the controls (t = 3.54, p <.001, eta2 =
.05). Participants in the verbal persuasion condition were
significantly more likely to resist their friends' counterarguments
than were those in the control condition (t = 2.33, p = .01,
eta2 = .02).
Generalized intentions. Participants in the behavioral modeling
condition (M = 5.80, SD = 2.55) were significantly more likely to
express their concern to a casual acquaintance that he or she not
drive drunk than were those in the verbal modeling (M = 4.41, SD =
2.30; t = 3.48, p <.001, eta2 = .05), verbal persuasion (M =
4.48, SD = 1.81; t = 3.16, p <.001, eta2 = .04), and control (M
= 4.98, SD = 2.14; t = 2.01, p = .02, eta2 = .02)
conditions. These were the only contrasts to reach significance on
the similar-threat scale for behavioral intentions. As summarized in
Table 1, the only contrasts to reach statistical significance on the
dissimilar-threat scale (i.e., intentions to express concern to a
stranger) were those between the behavioral modeling (M = 3.35, SD =
2.38) and control conditions (M= 2.40, SD = 1.94; t = 2.51, p
<.01, eta2 = .03) and between the verbal modeling (M = 3.16, SD =
2.29) and control conditions (t = 1.98, p = .02, eta2 = .02).
This study compared the relative effectiveness of two forms of
symbolic modeling and verbal persuasion on efficacy expectations and
intentions to prevent a friend from driving drunk. As predicted, both
types of symbolic modeling instated stronger self-efficacy beliefs
than did persuasive-efficacy information, although this treatment
modality impacted confidence to intervene. Regardless of the source
of self-efficacy, efficacy information produced stronger expectations
and behavioral intentions than did the no-efficacy-information
control message, which suggests that these sources can raise
confidence to intervene successfully among this age group by
overcoming beliefs of self-doubt, or by bolstering
self-efficacy. Indeed, a similar study by author (1995) found that
behavioral modeling and social persuasion affected actual
intervention behavior among this sample of drinkers more than did a
traditional persuasive message based on a fear appeal that is common
to this type of campaign.
The three-component behavioral-modeling condition that included
demonstration, verbal description, and reassuring information
engendered the greatest effects. Observation of the modeled behavior
raised expectations of successful performance and behavioral
intentions. While the difference between the two modeling conditions
was not significant at the adjusted significance level of .025, it
was at the more liberal and traditional level of .05. A stronger
modeling stimulus may have yielded the predicted result for strength
of self-efficacy. The findings for both forms of symbolic modeling
are encouraging because participants have no past experiences, or
behavioral data, from which to estimate their capabilities. These
data provide further empirical support for the postulated
As expected, participants in the symbolic modeling conditions would
persist more in their efforts to resist a friend's counterarguments
than would those in the social persuasion condition. All experimental
conditions registered stronger efficacy expectations on this measure
than did the controls. Behavioral demonstration and verbal
description of modeled activities appears to have raised
self-efficacy in these conditions to the point that a friend's
disapproval of the intervention attempt did not weaken confidence to
persist. A similar pattern of results was found for the most
difficult task in the hierarchy on behavioral intentions, although
there was no difference between verbal modeling and verbal
persuasion. Verbal description does not appear strong enough to
affect this task more than does persuasive-efficacy information
alone. This suggests a ceiling effect for the verbally modeled
behaviors in hierarchy of task demands. However, the results
indicate that symbolic verbal modeling does impact self-efficacy
beliefs about successfully intervening, and therefore should be
explored in future research. It would be interesting to compare the
efficacy expectations of verbally modeled behaviors produced in
television and radio formats, especially since this public spends
much time attending to this medium.
The inconsistent findings for within-domain generality are not
surprising, because intervening to prevent a casual acquaintance and
stranger from driving drunk is inherently more risky than engaging a
friend. The performance requirements and situational circumstances
associated with these behaviors were not addressed by the
efficacy-information messages. That differences were found between
behavioral modeling and verbal persuasion and between behavioral
modeling and the controls is encouraging. This suggests that
professionally produced PSA's aired consistently over a given time
period could encourage intervention behavior in these high-risk situations.
The results of this study suggest strategies for designing campaigns
to prevent drunken driving among this target public. Such
intervention efforts should attempt to facilitate self-directed
change by providing young adult responsible drinkers, as well as
nondrinkers, with the information, motivation, skills, self-efficacy,
and social support to engage a friend under variable conditions to
prevent him or her from driving drunk. Persuasive-efficacy messages
are well suited for satisfying the informational and
motivational campaign requirements that create the preconditions for
behavior change. The strategic importance of persuasive-efficacy
information during this stage of the campaign cannot be
overemphasized because if target publics reject the fundamental
proposition on which the campaign is built, behavioral
recommendations to intervene will not be followed.
The second stage should focus on the development of the social and
self-regulatory skills needed to translate knowledge and motivation
into efficacious actions. Newly acquired skills must be
practiced, preferably under as realistic conditions as possible to
instill a resilient sense of self-efficacy. This is important because
people will occasionally experience setbacks. Typically, those with a
weak sense of personal efficacy will abandon their efforts, while
those with a strong sense will persevere until they succeed, as the
results of this study show. Finally, campaign planners should reward
early adopters with messages that depict them as opinion leaders
whose willingness to express their concern to a friend will
influence others to do the same.
Anderson, R. B. (1989). Reassessing the odds against finding
meaningful behavioral change in mass media health promotion
campaigns. In C. H. Botan & V. Hazelton, Jr. (Eds.), Public relations
theory (pp. 309-321). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Anderson, R. B. (1995). Cognitive appraisal of performance
capability in the prevention of drunken driving: A test of
self-efficacy theory. Journal of Public Relations Research, 7, 205-229.
Anderson, R. B. (2000). Vicarious and persuasive influences on
efficacy expectations and intentions to perform breast
self-examination. Public Relations Review, 26, 107-124.
Anderson, R. B., & McMillion, P. Y. (1995). Effects of similar and
diversified modeling on African American women's efficacy
expectations and intentions to perform breast self-examination.
Health Communication, 7, 327-343.
Atkin, C. K., & Freimuth, V. S. (2001). Formative evaluation
research in campaign design. In R. E. Rice & C. K. Atkin (Eds.),
Public communication campaigns (3rd ed., pp. 125-145). Thousand Oaks, CA: Sage.
Ayres, B. D., Jr. (1994, May 22). Drunken driving slowing down.
Austin American-Statesman, pp. 1, 12.
Bandura, A. (1969). Principles of behavior modification. New York:
Holt, Rinehart & Winston.
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of
behavioral change. Psychological Review, 84, 191-215.
Bandura, A. (1986). Social foundations of thought & action: A social
cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1990). Perceived self-efficacy in the exercise of
control over AIDS infection. Evaluation and Program Planning, 13, 9-17.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Bandura, A., & Mischel, W. (1965). Modification of self-imposed
delay of reward through exposure to live and symbolic modeling.
Journal of Personality and Social Psychology, 5, 698-705.
Beck, K. H., & Frankel, A. (1981). A conceptualization of threat
communications and protective health behavior. Social Psychology
Quarterly, 44, 204-217.
Bernays, E. L. (1923). Crystallizing public opinion. New York: Boni
Blanchard, Edward B. (1970). Relative contributions of modeling,
informational influences, and physical contact in extinction of
phobic behavior. Journal of Abnormal Psychology, 76, 55-61.
Cartwright, D. (1949). Some principles of mass persuasion. Human
Relations, 2, 253-267.
Devine, P. G., & Hirt, E. R. (1989). Message strategies for
information campaigns: A social psychological analysis. In C. T.
Salmon (Ed.), Information campaigns: Balancing social values and
social change (pp. 229-258). Newbury Park, CA: Sage.
Flay, B. R. (1981). On improving the chances of mass media health
promotion programs causing meaningful changes in behavior. In M.
Meyer (Ed.), Health education by television and radio (pp. 55-91).
Flora, J. A., Maccoby, N., & Farquhar, J. W. (1989) Communication
campaigns to prevent cardiovascular disease: The Stanford community
studies. In R. E. Rice & C. K. Atkin (Eds.), Public communication
campaigns (2nd ed., pp. 233-252). Newbury Park, CA: Sage.
Grunig, J. E. (1989). Publics, audiences and market segments:
Segmentation principles for campaigns. In C. T. Salmon (Ed.),
Information campaigns: Balancing social values and social change (pp.
199-228). Newbury Park, CA: Sage.
Grunig, J. E., & Hunt, T. (1984). Managing public relations. New
York: Holt, Rinehart, and Winston.
Grunig, J. E., & Ipes, D. A. (1983). The anatomy of a campaign
against drunk driving. Public Relations Review, 2, 36-52.
Hertog, J. K., Finnegan, J. R., Rooney, B., Viswanath, K., Potter,
J. (1993). Self-efficacy as a target population segmentation strategy
in a diet and cancer risk reduction campaign. Health Communication, 5, 21-40.
Katz, E., & Lazarsfeld, P. F. (1954). Personal influence. Glencoe,
IL: Free Press.
Kotler, P., Roberto, N., & Lee, N. (2002). Social marketing:
Improving the quality of life (2nd ed.). Thousand Oaks, CA: Sage.
Lazarsfeld, P. F., Berelson, B., & Gaudet, H. (1944). The people's
choice. New York: Duell, Sloan & Pearce.
Lazarsfeld, P. F., & Merton, R. K. (1948). Mass communication,
popular taste, and organized social action. In L. Bryson (Ed.), The
communication of ideas (pp. 95-118). New York: Institute for
Religious and Social Studies.
Had Enough Campaign (2001). Sobering statistics.
Maccoby, N., & Alexander, J. (1979). Use of media in lifestyle
programs. In P. O. Davidson & S. M. Davidson (Eds.), Behavioral
medicine: Changing health lifestyles (pp. 351-370). Brunner/Mazel, New York.
Maccoby, N., & Farquhar, J. W. (1975). Communication for health:
Unselling heart disease. Journal of Communication, 25, 114-126.
Maibach, E. (1995). Moving people to behavior change: A staged social
cognitive approach to message design. In E. Maibach & R. L. Parrott
(Eds.), Designing health messages: Approaches from communication
theory and public health practice (pp. 41-64). Thousand Oaks, CA: Sage.
Maibach, E., & Flora, J. A. (1993). Symbolic modeling and cognitive
rehearsal: Using video
to promote AIDS prevention self-efficacy. Communication Research, 20, 517-545.
Maibach, E., & Murphy, D. A. (1995). Self-efficacy in health
promotion research and practice: Conceptualization and measurement.
Health Education Research, 10, 37-50.
Masters, B. A. (1994). Study shows more college women drink. Austin
American-Statesman, pp. 1, 5.
McGuire, W. J. (1989). Theoretical foundations of campaigns. In R.
E. Rice & C. K. Atkin (Eds.), Public communication campaigns (2nd
ed., pp. 43-65). Newbury Park, CA: Sage.
Pfau, M., & Parrott, R. (1993). Persuasive communication campaigns.
Needham Heights, MA: Allyn and Bacon.
Schultz, D. E., & Barnes, B. E. (1999). Strategic brand
communication campaigns (5th ed.). Lincolnwood, IL: NTC Business Books.
U.S. Department of Health and Human Services. (1984). Pretesting
television PSAs: User's guide (NIH Publication No. 85-2670).
Bethesda, MD: National Cancer Institute.
Wechsler, H., Dowdall, G. W., Maenner, G., Gledhill-Hoyt, J., & Lee,
H. (1998). Changes in binge drinking and related problems among
American college students between 1993 and 1997: Results of the
Harvard School of Public Health college alcohol study. Journal of
American College Health, 47(2), 57-68. Retrieved January 29, 2000
from the World Wide Web: http://www.hsph.harvard.edu/cas/wechsler.pdf
1Early thinking about the effects of mass communication was rooted in
theories of mass society, which, in part, held that the impersonality
of such societies impeded the development of strong social ties, or
relationships, leaving individuals psychologically isolated from each
other and social institutions. Mass communication was seen as one of
the primary means of uniting the masses and providing a stable system
of social control. It was believed that mass-mediated messages
affected the mass audience, or general public, uniformly because of
inherited biological mechanisms explained by instinct psychology,
which was at its height. It was against this intellectual backdrop
that Edward L. Bernays and Carl Byoir conducted propaganda campaigns
for the Committee on Public Information during World War I, according
to what Grunig and Hunt (1984, chap. 2) have called the two-way
asymmetrical model of public relations. Bernays (1923) later
discussed his use of the principles of mass persuasion during
peacetime in his seminal Crystallizing Public Opinion.
2Although the PSAs targeted moderate drinkers, analyses were
performed on data for the entire sample of drinkers and nondrinkers,
because there is no commonly agreed on definition of moderate
drinking shared by this age group. In the absence of such a
universal definition, it is possible that light and heavy drinkers
consider themselves moderate drinkers. Indeed, this was a key
finding in the formative evaluations cited earlier. It also is
possible that nondrinkers might identify with the moderate drinker
spokesperson in the PSAs, because he espouses a drinking philosophy
similar to theirs. Hence, all types of drinkers were included in the
sample. A subsequent investigation will report findings for a
subsample of moderate drinkers.
3Pretesting questions typically include main-idea recall,
comprehension, believability, personal relevance, and strong and weak