,’ Cl

 

LIBRARY
Michigan State

University

_ _.._.. -.-,_..... ~m -

 

 

This is to certify that the
dissertation entitled

CULTURAL DIFFERENCE IN THE PERSUASIVE IMPACT
OF A ROLE NORM MESSAGE

presented by
SANG-YEON KIM

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Communication

7

‘

Major Professor’s Signature
5 ~ 18 ' 200‘?

 

 

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 Kzlprq/AcwpresICIRC/Dateous.indd

 

CULTURAL DIFFERENCE IN THE PERSUASIVE IMPACT
OF A ROLE NORM MESSAGE

By

Sang-Yeon Kim

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Communication

2009

ABSTRACT

CULTURAL DIFFERENCE IN THE PERSUASIVE IMPACT
OF A ROLE NORM MESSAGE

By
Sang-Yeon Kim
This study examines the persuasive impact of role norm appeal relative to the
persuasive impact of traditional social norm strategies. Role norms constitute a particular
form of social norms, stipulating what one ought to do as the holder of a particular role.
This study hypothesizes that a role norm appeal produces greater persuasion than a social
norm appeal, because social sanctions from related others tend to be perceived as more
threatening than social sanctions from those that are unrelated .This study also predicts a
relatively greater persuasive impact of role norm appeals in holistic cultures versus
analytic cultures because holists tend to manifest a greater role-dependency compared to
their analytic counterparts.

The predictions were examined employing a 2 (U.S., Korea) x 3 (role norm,
social norm, no-norm control) independent groups design (NTo-r = 702; NUS = 412, NKOR

= 290). These three groups argue separate messages: firstly, that college students should
avoid excessive drinking as a responsible child of his/her parents; second that excessive
drinking is avoided because the person is a responsible community member; and lastly
that drinking is avoided for the subject's own good, respectively. The participant’s own
behavioral intention (BI) and the projected behavioral intention of others (PBI) served as
major dependent variables. Neither the main effect for treatment, nor the culture by
treatment interaction, were statistically significant at level or = .05. However, a

participant's culture had a significant main impact on both BI and FBI; across conditions,

Korean students’ BI and PBI scores indicated lower intentions to drink than the US.
students. Path analyses indicate that, across cultures, social norm factors exert only minor
impacts on behavioral intention; whereas participants’ perception of problem severity
explains more variance. Path models also suggest that Korean participants take a dual
cognitive processing in which BI and PBI are explained by two separate sets of variables.

This tendency remained less pronounced among the US. participants.

Dedicated to my wife, Yeol-Hee Lee, who has willingly sacrificed
six years of her youth for my doctoral study.

iv

ACKNOWLEDGEMENTS

I am truly indebted to Dr. Frank Boster, who has lent me his unfathomable knowledge
and keen insights, to Dr. Maria Lapinski, who has given me invaluable research
experience, and to Dr. Toby Teneyck, who has given me new diverse perspectives with
which cultures may be viewed. I must give my deepest thanks to Dr. Timothy Levine, the

very first person who showed me how loveable social science can be.

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................................... viii
LIST OF FIGURES ...................................................................................................... ix
LITERATURE REVIEW ............................................................................................... 1
Social Norms and Persuasion ............................................................................... 1
Cross-Cultural Difference in the Impact of Role Norm Appeals ......................... 5
METHOD .................................................................................................................... 10
Participants ......................................................................................................... 10
Experiment Design ............................................................................................. 13
Procedure ........................................................................................................... 13
Measures ............................................................................................................ 14
Induction check ........................................................................................ 15

Perceived salience of social sanction ....................................................... 15

Locus of consequence .............................................................................. 16

Personal and projected behavioral intention ............................................ 17

Other potential mediators ......................................................................... l7

Descriptive and injunctive norm .............................................................. 18

TRA measures .......................................................................................... 19

Tests of Measurement Invariance ...................................................................... 20
RESULTS ..................................................................................................................... 22
Induction Checks ............................................................................................... 22
Perception of roles ................................................................................... 22

Perceived social sanction ......................................................................... 32

Perceived argument quality ...................................................................... 33

Perceived locus of consequence ............................................................... 33

Perceived severity of drinking problem ................................................... 33

Descriptive norm of drinking ................................................................... 34

Injunctive norm of drinking ..................................................................... 34

Subjective norm of parents ...................................................................... 35

Subjective norm of community ................................................................ 35

Motivation to comply with parents .......................................................... 35

Motivation to conform to the community's norm .................................... 35

Hypothesis Testing ............................................................................................. 35
Exploratory Structural Equation Modeling ........................................................ 37
Phase 1 ..................................................................................................... 37

vi

Phase 2 ..................................................................................................... 56

DISCUSSION .............................................................................................................. 65
Check for Validity of Rationales ........................................................................ 68
The Third Person Effect ..................................................................................... 72
Koreans More Willing to Drink Less ................................................................. 73
Small Social Norm Effect .................................................................................. 75
Holists and Perception of Roles ......................................................................... 77
APPENDICES ............................................................................................................. 79
Message Stimuli (English) ................................................................................. 79
Message Stimuli (Korean) ................................................................................. 80
Measures (English) ............................................................................................ 81
Measures (Korean) ............................................................................................. 85
REFERENCES ............................................................................................................ 89

vii

LIST OF TABLES

TABLE 1: Measurement Invariance Tests ................................................................... 23
TABLE 2: Descriptive Statistics by Culture and Treatment Condition ....................... 29
TABLE 3: Full Correlation Matrix .............................................................................. 31
TABLE 4: Correlation Matrix for the US. .................................................................. 50
TABLE 5: Correlation Matrix for Korea ..................................................................... 51

TABLE 6: Mean Difference Between the US. and South Korea at Construct Level .70

viii

LIST OF FIGURES

FIGURE 1: Predicted Cultural Difference in Persuasive Impact of Role-Norm and Social

Norm Appeals .............................................................................................................. 1 1
FIGURE 2: Found Cultural Difference in Persuasive Impact of Role-Norm and Social
Norm Appeals (Behavioral Intention) .......................................................................... 38
FIGURE 3: Found Cultural Difference in Persuasive Impact of Role-Norm and Social
Norm Appeals (Projected Behavioral Intention) .......................................................... 39
FIGURE 4: Full Model (U.S.) ..................................................................................... 42
FIGURE 5: Full Model (Korea) ................................................................................... 43
FIGURE 6: A Reduced Model (U .S.) .......................................................................... 44
FIGURE 7: A Reduced Model (Korea) ........................................................................ 45
FIGURE 8: A Reduced Model for the US. (Rearranged) ........................................... 46
FIGURE 9: A Reduced Model for Korea (Rearranged) ............................................... 47
FIGURE 10: Final Model (U .S.) .................................................................................. 52
FIGURE 11: Final Model (Korea) ............................................................................... 53
FIGURE 12: US Model Fitted to Korean Sample ..................................................... 54
FIGURE 13: Korean Model Fitted to the US. Sample ............................................... 55
FIGURE 14: TRA Model (Entire Sample) .................................................................. 58
FIGURE 15: TRA Model (U.S.) .................................................................................. 59
FIGURE 16: TRA Model (Korea) ............................................................................... 60
FIGURE 17: Separate Process Model (U.S.) ............................................................... 63
FIGURE 18: Separate Process Model (Korea) ............................................................ 64
FIGURE 19: Separate Process Model (U .S.), Composite Indices ............................... 66

ix

FIGURE 20: Separate Process Model (Korea), Composite Indices ............................ 67

Literature Review
Social Norms and Persuasion

Considerable scholarly energy has been devoted across many disciplines of social
science to the understanding of the influence of social norms on human behavior.
Anthropologists have long investigated the nature of social norms as the key to
understanding the basic principles of society (e.g., Malinowski, 1926). Sociologists (e. g.,
Birenbaum & Sagarin, 1976; Nadel, 1957; Piddocke, 1968) and students of law (e.g.,
McAdams, 1997; Sustein, 1996) have produced a vast literature on aspects of social
norms, particularly social norms’ contributions to the maintenance of a social order
without resorting to legal measures. Persuasion scholars have been investigating ways to
maximize the effect of social norms in inducing pro-social behavior among the public
(e.g., Cialdini, Reno, & Kallgren, 1990; Reno, Cialdini, & Kallgren, 1993).

Across this broad spectrum of applications, the concept of “social norms” has
typically been defined as having two elements: (a) a consensus on the patterns of
behavior considered to be desirable by most members of a collective, and (b) social
enforcement, where compliers are rewarded, defectors are punished, or both (Birenbaum
& Sagarin, 1976; Malinowski, 1926; Nadel, 1957; Piddocke, 1968). Individuals raised in
the same culture come to develop a similar rule set stipulating which behaviors and value
systems correspond or contradict those desired by the majority of people. Possessing a
code of conduct presumed to be shared by most others, one can predict, albeit roughly,
whether his or her behavior will be viewed as acceptable if executed in public. This
ability to predict potential social consequences of one’s behavior fimctions as an internal

regulator that encourages pro-social behaviors and discourages anti-social behaviors,

because by so doing, one can gain social approval and avoid potential consequences for
violating rules. Consistent with these observations, this study defines social norms as a
set of socially shared guidelines to the expected behavioral patterns whereby social order
is spontaneously maintained through social enforcement.

While symbolic interactionists have paid more attention to the process in which a
social consensus evolves, and how it becomes internalized in the minds of individual
members of the society (Blurner, 1969; Mead, 1934), the concept of social sanctions has
attracted continued interest from social psychologists as they examine direct and indirect
normative influences on human behavior (Firth, 1958; Homans, 1950; Piddocke, 1968).
Firth (1958) and Piddocke (1968) in particular point out that one’s decision to carry out a

particular action is immediately determined by the actor’s projection of potential social

sanctions as well as personal predispositions or habits.l

Cialdini and colleagues (1990, 1993, 2000) are among many persuasion scholars
who have successfully demonstrated the influence of social norms in inducing pro-social
behavior. In particular, the researchers devoted a series of field experiments to test the
effectiveness of anti-littering social norms on curtailing littering in public space (Cialdini,
Reno, & Kallgren, 1990; Kallgren, Reno, & Cialdini, 2000; Reno, Cialdini, & Kallgren,
1993). In the main, the data were consistent with the prediction; targets in whom the anti-
littering social norm had been induced were less likely to litter than were the targets in

whom the anti-littering social norm had not been induced. Particularly, an increase in the

 

1 This premise holds true to the extent that social sanctions, especially punishments, are indeed perceived
as such by most members of the society; a death penalty may not be valued as punishment for a martyr and
thus would exert little influence on her religious behavior (Piddocke, 1968). Whether it be a direct physical
attack from the community, restricted freedom of action, a ruined self-image, or feelings of shame or
anxiety, a social sanction must be an external or internal measure that arouses at least minimal displeasure
in the norm violator. A social norm that tells what is desired but unsubstantiated by sanctions is what we
call an ideal (Homans, 1950).

strength of an induced anti-littering social norm produced a proportionate decrease in
littering. Cialdini and associates (1990, 1993, 2000) attributed this outcome to the
salience of the experimental stimulus. Cialdini et a1. opined that most social norms tend
to remain dormant in the human mind until becoming cognitively accessible to the target
through a salient stimulus, and posited that the probability of norm conformity enhances
as the stimulus becomes more salient. This is because a more salient stimulus can render
the pertinent social norms, and thus potential social sanctions, more noticeable to the
target (see Rimal & Lapinski, 2005).

To the extent that this premise is correct, a social norm which is made more
accessible to the target’s cognition should produce an increased norm conformity. This
study thus proposes that a stronger norm congruence maybe inducible by invoking the
target’s role norms than by inducing other types of social norms.

A social role refers to a position endowed to a person fulfilling expected duties in
a particular relationship. That is, a social role does not exist until one enters a
relationship and plays a part complementary to that of the counterpart(s) in that particular
relationship (Banton, 1965; Nadel, 1957). For example, in order for a man to be a
legitimate father, he must enter a father-child relationship and perform a set of
obligations for relationship maintenance. Demonstrably, no doctors would exist in that
role of “doctors” in the absence of patients that have to be treated.

In particular, the relational duties expected of a role holder can be referred to as
role norms, hereby defined as “a set of norms and expectations applied to the incumbent
of a particular position ” (Banton, 1965, p. 29). Role norms are a particular form of social

norms exclusively confined to the role-specific behavior domain. That is, role norms

operate only within the domains where the interactants communicate as ‘role players,’
whereas social norms cover the whole gamut of social interactions including those in
which individuals need not be identified as a role bearer. For example, social norms and
role norms provide an equal set of instructions as to what one ought to do as a parent, a
student, a manager, a romantic partner, etc. But it is only more general social norms that
remain valid beyond the relational boundaries, requiring, for example, to follow basic
table manners, not to pick nose in the presence of others, not to stare at people, and so on.

As mentioned, social norms include role norms. For convenience, however, this
study limits the scope of social norms exclusively to public domains where interactants
feel little role obligations to the other(s) due to the absence of a perceivable relationship.
Social titles as ‘community member’ or ‘fellow citizen,’ for example, are nominal in
nature, requiring minimal expectations about what one ought to do for the other(s), and
would thus fall into the domain of social norms.

A corollary of this view is that role norms comprise the anticipations from related
others whereas social norms represent the expectations of unrelated others. To the extent
that the expectations from related others are more salient in human minds than the
expectations from unrelated others, people should also perceive the sanctions from
related others as more threatening than the sanctions from unrelated others. It further
follows that norm-congruent pro-social actions are more likely to occur in the targets in
whom role norms have been invoked, opposed to the targets in whom social norms have
been induced. Following this reasoning, this study proposes its first hypothesis as follows.

H1: Persuasion messages using role-norm appeals are more effective in promoting

pro-social behaviors than the persuasion messages using more general social

9

2
norm appeals.

Cross-Cultural Difference in the Impact of Role Norm Appeals

Cross-cultural researchers have invested significant effort to understanding the
causes and consequences of cultural differences. Culture itself is a very broad concept,
and depending on which aspects of culture is highlighted, some find the cause of cultural
divides in the difference in language (Hamaguchi, 1977; Kashima & Kashima, 1998;
Lebra, 1976), in context-dependency (Hall, 1976), in history of philosophy (King, 1985;
Munro, 1985), in perception of self (Markus & Kitayama, 1991, 1998; Kashima et al.,
2004), in worldview (N isbett, 2003), or in the structure of society (Lim, 2002; Lim, Allen,
Burrell, & Kim, 2007).

One observation is that behind these diverse frames of reference lies a common
assumption that there exists a substantial cultural difference in context-dependency,
particularly between Europeans and East Asians. For example, Kashima and Kashima
(1998) documented a context-dependency of Japanese culture by the fact that, in
Japanese language, the reference to self (i.e., first person singular pronoun) takes
different forms according to varying relational contexts. Hall (1976) pointed out that
Eastemers’ communication depends more on contextual cues, whereas Westerners tend

to focus more on the message itself as the primary source of meaning. Markus and

 

Some may argue that role norms are conceptually identical to subjective norms (Ajzen, 1985; Fishbein &
Ajzen, 1975), which refer to one’s perception of norms dominant in his/her reference group (e.g., ‘what my
family would expect me to behave '). Subjective norms, however, lack social desirability that is by definition
imbedded in role norms (e.g., ‘what you ought to do as a parent ’). The norms active in one’s family, for
instance, are a combination of the social norms universally embraced by most other families and the norms
idiosyncratic to that particular family. It is determined that the former conveys social desirability while it
remains probabilistic if so does the latter. Therefore, the norm-congruent actions induced by a subjective
norm appeal might be either pro- or anti-social in nature. But the norm-conforming behaviors that a role
norm appeal produces should always be pro-social.

It should be also noted that the discussion of social norms and role norms here focuses on what Cialdini
et al. (1990, 1993, 2000) call injunctive norms.

Kitayama (1991, 1998), Kashima et al. (2004), and Lim et a1. (2007) have proposed that
East Asians’ perception of self tends to vary depending on whom they communicate with
at the moment while one’s identity tends to remain relatively unaffected by the relational
contexts in European cultures.

Recently, Nisbett (2003) and Nisbett, Peng, Choi, and Norenzayan (2001) have
used a similar approach to explain cultural differences in the way of understanding and
perceiving the world. Specifically, Nisbett (2003) proposes that Europeans, the
descendants of ancient Greek culture, tend to possess an analytic worldview. East Asians,
as the offspring of Chinese tradition, manifest a holistic tendency. This new scheme - the
distinction of analytic versus holistic cultures - posits that individuals raised in societies
where analytic world views prevail tend to see the world as an aggregate of objects
separable from one another and fiom the context that they belong to. In analytic cultures,
objects dissociated from the context receive cognitive attention in a belief that knowing
individual objects’ internal attributes provide a reliable means of understanding physical
and social phenomena. This cultural assumption implies that fundamental characteristics
of an individual object remain relatively constant, independent of other objects and
varying contexts. In contrast, Nisbett postulates that East Asians with a holistic
worldview tend, as a whole, to see things in context, or are less apt to distinguish
individual objects from the pertinent background. To holists, everything is interrelated
and the attributes of an object change depending upon where the object is situated. Hence,
the knowledge of a separated object is considered to be of little use to understand the
world. Instead, knowing how things are related and being able to see their relational

dynamics in context is regarded as important in holistic cultures.

Existing data corroborate the assertion that analysts tend to perceive things
independent of context, whereas holists’ perception of the world is more context-bound.
For example, Japanese participants tended to recall better the objects placed in the
background of a visual stimulus (e.g., water, rocks, or plants in an aquaritun) whereas
American participants were more likely to report on the objects in the foreground (e.g.,
big fish in the same aquarium; Masuda & Nisbett, 2001). Also, among Japanese
participants, the retention rate of a previously shown object was significantly lower when
the same stimulus was presented again later against a different background than when the
visual context remained unchanged. American participants, however, maintained an equal
retention rate across conditions (Masuda & Nisbett, 2001). In Morris and Peng’s (1994)
study, Chinese participants tended to view situational factors as more important causes of
a fictitious murder case, whereas American participants were more likely to evaluate the
murderer’s presumed personality traits as the primary cause of the murder (see Nisbett,
2003 for review of other related findings).

Given Nisbett’s (2003) postulate, it should follow that analysts’ perception is less

sensitive to social roles whereas holists’ perception is more sensitive to social roles. This

is because, as mentioned, social role is by definition context-bound.4 Role recognition
necessitates relational context, which changes constantly according to the object of
communication. A social role comes to exist only in the presence of a relationship
whether it be marriage, familial, fiiendship, or companion. One is a husband ‘in relation’

to his wife, a father ‘in relation to’ his child, a fi'iend ‘in relation to’ his/her friend, an

 

This reasoning implicates an association between cognitive context dependency and sensitivity to social
roles. Consistent with this conjecture, Dershowitz (1971) found out that children who had received
domestic education that spells out role relations are more likely to manifest a field dependency, an
inclination that one’s perception of objects being influenced by the background, compared to children from
normal families.

employee ‘in relation to’ his/her boss, etc. A social role cannot hold independent of the
counterpart assuming a reciprocal role in that particular relationship (Banton, 1965). This
notion implies that one’s social role keeps changing from one to another as he/she travels
through various relational contexts (see James, 1950). One assumes the role as a parent
when communicating with a child but acts out a student role at school, particularly before
a professor. Thus, staying sensitive to varying contextual cues is imperative to role
recognition.

To the extent that the preceding reasoning is correct, whether or not a role factor
is made apparent in a persuasion message should create a more pronounced perceptual
difference among people from holistic cultures than among people from analytic cultures.
This is primarily because holists are more role-sensitive and thus more likely to catch and
incorporate role-related information in forming impressions of the message whereas
analysts, who are relatively insensitive to social roles, would have a limited chance to do
so. Presurning that it is a pan-cultural tendency for people to experience the potential
sanctions from related others as more threatening than the sanctions from unrelated others,
and to the extent that a proximate social sanction is more likely to produce a pro-social
behavior than would a social sanction perceived as distal, holists should manifest the
same behavioral tendency of analysts but to a greater magnitude. That is, a role-norm
appeal should promote pro—social actions from holists more than it does from analysts,
and a social-norm appeal should demote pro-social actions from holists more than it does
from analysts.

In fact, it has been documented repeatedly that East Asians tend to discriminate

against people in whom they find little relational connection, while offering greater

generosity to people tied in any form of relationship (King, 1985; Triandis, 1995). Cross-
cultural psychologists (DeVos, 1973; Hamaguchi, 1977; Lebra, 1976; Lim, 2002; Lim,
Allen, Burrell, & Kim, 2007; Triandis, 1995) and philosophers studying culture (King,
1985; Munro, 1985) postulate that role-dependency, which characterizes the essence of
holistic culture, constitutes the primary cause of discrimination based on presence or
absence of relational connections. Specifically, Munro (1985) and Lim, Allen, Burrell,
and Kim (2007) conjecture that individuals in holistic societies are viewed as mutually

interdependent parts of a whole, fulfilling complementary roles assigned by the whole,

not as autonomous yet unique-different individuals as in analytic cultures.5 Holists are,
thus, more likely to self-identify as a role player that attains significance only when
functioning in the role within the whole but turns into a meaningless entity when
separated from the whole or dissociated from related others (DeVos, 1973; Hamaguchi,
1977; Lebra, 1976). Empirical findings indicate that the descendants of holistic cultural
heritage tend to identify themselves by referring to their social roles and often experience
difficulties describing self not using role terms (Bachnik, 1994; Cousins, 1989; Cross,
Kanagawa, Markus, & Kitayama, 1995; Ip & Bond, 1995; Lim, Allen, Burrell, & Kim,
2007; Markus, Mullally, & Kitayama, 1997; Rhee et al., 1995; Snyder & Gangestad,
1986)

The conjecture that holists’ identity originates from their social roles translates
that one’s significance as a social entity comes from the presence of related others,

namely the ones that imbue the otherwise insignificant with meanings. Fulfilling one’s

 

5 In their original works, Munro (1985) and Lim et al. (2007) maintained the traditional term
‘individualistic culture’ to refer to the cultural characteristics of East Asian nations, which is represented by
‘analytic cultures’ in Nisbett’s book (2003 ). Simply for the sake of communication with readers, this study
chose to use the term ‘analytic’ consistently throughout this manuscript.

relational duties to other(s) in a relationship may thus be commensurate with respecting
the actor him/herself as well as the counterpart and the relationship itself. In contrast,
unrelated others should remain as unimportant social objects because of their inability to
invoke an identity as a social being. Particularly, in situations where unrelated individuals
communicate, there remain few rules on appropriate treatment of the counterpart or

interaction rituals of politeness. Once released from all relational ties, East Asians are

freed from all social restrictions regulating utterances and behaviors.6 For this reason,
Chinese philosophers have warned that Chinese people, as a whole, can be cold-hearted
or extremely impolite to people in whom they find few meaningful relational connections
(King, 1985; see also Triandis, 1995). Building upon the preceding accounts of holistic
culture, this study proposes a second set of hypotheses (see Figure 1).

H2: The difference in magnitude between the persuasive impact of role-norm
appeals and the persuasive impact of social-norm appeals is more pronounced
in holistic cultures than in analytic cultures.

H2a: A role norm appeal promotes pro-social actions more among holists than
analysts.
H2b: A social norm appeal demotes pro-social actions more among holists
than analysts.
Method

Participants

Participants (NTOT = 851) were recruited from the US. (NUS = 469) and South

 

6 The basic disciplines of Confucianism, one of the most influential life principles among most East Asian
countries, have been teaching for centuries about what one ought to do in certain specific relationships but
reticent on how one should behave outside those relational contexts (King, 1985; Markus, Mullally, &
Kitayama, 1997).

10

Figure 1 Predicted Cultural Diflerence in Persuasive Impact of Role-Norm and Social-

Norm Appeals

1 _.

.i
z | '0
a 1 ' - - '0- - - Holistic Culture
< l '
E; i —I—Analytic Culture
:3. _' ,.

i

i 9'

i.z-.-,_,..,,_z ,3 , _-_ m,

Social Norm Role Norm

NORM TYPE

11

Korea (NKOR = 382) to constitute a sample of an analytic and a holistic culture,

respectively. The experiment was advertised in social science classes including
communication, sociology, and journalism, in both nations. The investigator was also
able to reach potential American participants outside these classes using an online subject
pool operated in the Communication Department at Michigan State University. Students
participated in the experiment online by accessing a designated website where the
experiment stimuli and measurement sets were posted. Research credits were awarded in
exchange for participation.

Duplicate responses were screened by examining registered IP addresses and the
time points at which a particular respondent began and finished the experiment. All
responses associated with an IP address that appears more than once in the server were
considered as duplicates and deleted (Lehmiller, 2008). However, the duplicate response
preceding all others sharing the same IP address was retained, noting that the
participant’s first response cannot be contaminated by how he/she responded in a later
session. Multiple responses were further detected by participant’s student ID numbers
and excluded from the data set following the same procedure. In total, 125 responses
were identified as duplicates, and thus eliminated from the data set. The file containing
participants’ IP addresses and student ID numbers were permanently destroyed after
cleaning the data.

Approximately 2 percent (n = 10) of the US participants came from countries out
of research interest (e.g., Latin America and Middle East) and were excluded from the
US. sample. Chinese participants constituted 5.2 percent (n = 15) and Japanese

participants accounted for .3 percent (n = 1) of the Korean sample. Responses from

12

Chinese or Japanese participants were kept considering that both countries manifest the
holistic cultural tendency as in Korea (see Nisbett, 2003). Responses from one Latin
American subject were deleted from the Korean data set. Also deleted were responses
from participants who failed to report their cultural origin (n = 13).

After eliminating duplicate responses and the responses from cultures out of

research interest, the sample size has reduced to NTOT = 702; NUS = 412 and NKOR = 290.

Sixty six percent of the US. sample were females. Mean age was 19.98 (SD = 1.56).
Females constituted 48 percent of the Korean sample, with the mean age of 21 .94 (SD =
2.43).
Experiment Design

The predictions were examined employing a 2 (US, Korea) X 3 (role norm,

social norm, no-norm control) independent groups design. Participants from each country

were randomly assigned to one of the three treatment conditions, resulting in "role = 143,

"social = 143, and "control = 126 in the U.S. sample, and "role = 94, "social = 104, and

”control = 92 in the Korean sample.

Procedure

Upon entering the experiment website, participants were asked to read the
assigned message and evaluate it using a set of Likert-type items. The role norm, social
norm, and no-norm control messages warned against the habit of excessive drinking
among college students, the potential subject pool of this study. The role norm condition
and the social norm condition were created by varying the norm type that the message

uses to dissuade excessive drinking around campus. Specifically, the role norm message

13

indicated that potential negative consequences of excessive drinking may fall on their
parents as well as themselves, emphasizing that avoiding excessive drinking is a part of
fulfilling duties as a responsible child of their parents (i.e., “Your parents have been
sacrificing themselves to raise you for almost twenty years. Now, they just want you to be
safe and healthy. Drink responsibly. Responsible children do not make their parents
sufler! ”). On the other hand, the social norm message reminded the reader that excessive
drinking may hurt other innocent community members, urging the reader to implement
the duties as a responsible community member (i.e., “Drink responsibly. Responsible
community members do not make other fellow members sufler! ”). The no-norm control
message attempted to deter excessive drinking without using norm appeals; it simply
discussed potential negative consequences for the excessive drinker him/herself (i.e.,
“Drink responsibly. Protect yourself? ”).

The quality of the argument remained strong across conditions. The arguments
themselves presented factual or statistical evidence from qualified sources. The evidence
was obtained from the website of “University Mothers Against Drunk Driving” or
UMADD (2009), a nationwide organization established to help prevent college binge
drinking at community level (see Appendices A and C for English messages and
measures, respectively). All experimental materials were administered in the primary
language spoken in the country in which the data were collected. The initial survey was
created in English, and then translated and back-translated into Korean by native Korean
speakers to ensure equivalence of meaning (Brislin, 1980; see Appendices B and D for
Korean messages and measures, respectively).

Measures

14

Induction check. Eight 5-point Likert items (1 = strongly disagree and 5 =
strongly agree) were used to assess if the treatment message has induced its
corresponding norm type as intended. The first four items checked if the message had
manipulated the role norm (i.e., the role as a responsible child), and the latter four items
tapped into the perceived salience of social norm (i.e., the role as a responsible
community member). To the extent that the message induction was successful, higher
ratings should result from the first four items and lower ratings from the latter four items
for the role norm condition. The opposite trend should indicate successful induction for
the social norm condition. Lower scores should result from all the eight items provided
that the no-norm control message has induced neither.

Included items were “This message points out that avoiding excessive drinking is
a way of becoming a responsible child of my parents,” “This message appeals to my role
obligation as the child of my parents,” “This message argues that I avoid excessive
drinking as a responsible child of my parents,” “This message indicates that excessive
drinking may prevent me from functioning as a responsible child of my parents.” “This
message points out that avoiding excessive drinking is a way of becoming a responsible
community member,” “This message appeals to my role obligation as a community
member,” “This message argues that I avoid excessive drinking as a responsible
community member,” and “This message indicates that excessive drinking may prevent
me from functioning as a responsible community member.”

Perceived salience of social sanction. To test the rationale that role norm appeals
can exert a greater conformity than social norm appeals via the enhanced perception of

social sanction, measures of perceived salience of social sanction were included in the

15

survey as a mediator. Four S-point Likert-items (1 = strongly disagree and 5 = strongly
agree) measured the perceived salience of social sanction. Included items were “I feel
like people would think less of me if I drank excessively,” “People would disapprove of
me if I drank excessively,” “Our society views heavy drinkers negatively,” “Society is
intolerant of excessive drinking.”

Locus of consequence. Powell (1965) found that the effectiveness of a fear appeal
becomes greater when the potential consequences of non-compliance were cast on the
target’s family, than when the expected ramifications were said to affect the nation or the
target himself (i.e., men with family). This finding led to an alternative explanation that
the role norm message may produce more persuasion than does the social norm message,
due to the locus of potential consequences being nearer to the subjects in the former (i.e.,
parents) than to the subjects in the latter (i.e., community) not because of the perceived
social sanctions being made more salient in the role norm condition than in the social
norm condition. This study attempted to measure and statistically control the potential
nuisance effect of the locus of consequence to document a separate main impact of
perceived norm salience in inducing pro-social behavior.

Four 5-point Likert-items (1 = strongly disagree and 5 = strongly agree)
measured the perceived locus of consequences of excessive drinking. Included items
were “This message shows that excessive drinking will have consequences for people
that are close to me,” “This message demonstrates the negative effects of excessive
drinking that people close to me are likely to face,” “This message indicates that my

excessive drinking may affect people close to me,” and “This message claims that people

16

close to me can be the potential victims of my excessive drinking.” 7

Personal and projected behavioral intention. Personal and projected behavioral
intentions were the primary dependent measures. Five 5-point Likert-items (1 = not at all
and 5 = very much) were used to assess the strength of participant’s behavioral intention
to avoid excessive drinking. Included items were “I would be willing to sign a petition
supporting regulation on excessive drinking,” “I will try to drink less in the future,” “I
will avoid excessive drinking,” “If I drink, I will drink in moderation,” “I would be
willing to ask my friends to drink less,” and “I would be willing to talk to my friends
about binge drinking issues.”

Also included were another four 5-point Likert-items (1 = not at all and 5 = very
much) that purport to measure participants’ projected behavioral intention of other
readers. Included items were “Many of those who read this message will avoid excessive
drinking to the best that they can,” “Most people who read this message will try not to
drink excessively,” “People who read this message will try dissuading their significant
others from excessive drinking,” “Most people will NOT quit excessive drinking because
of this message (R).”

Other potential mediators: A potential threat to the internal validity of this
experiment involves the treatment messages’ difference in content. That is, all three
messages claimed that excessive drinking may cause serious damages, yet each approach
uses different facts in doing so. In particular, the role norm appeal introduced, for

example, how many college students die from excessive drinking every year, and then

 

7 In these items, the locus of potential consequences, either community or parents, were left unspecified but
made to vary in relational closeness to make the resulting scores comparable between the role norm and the
social norm condition. This arrangement assumes that people feel closer to their parents than other, mostly
unknown, community members.

17

reminded that the same number of parents are losing their children to excessive drinking.
The social norm appeal, on the other hand, discussed statistics on misdemeanors often
committed by heavy drinkers in public space (e.g., picking fights with strangers,
damaging public properties, or killing innocent people while driving under the influence
of alcohol). To control for the potential nuisance effect for the messages being
confounded, two additional measures were attached to the survey; argument quality and
perceived severity of drinking problem. Six 5-point Likert items (1 = strongly disagree
and 5 = strongly agree) measured perceived argument quality. Included items were “The
argument of this message is strong,” “valid,” “flawed (R),” “high in quality,” “weak (R),”
and “invalid (R).” Four 5-point Likert items (1 = strongly disagree and 5 = strongly
agree) tapped into perceived severity of drinking problem. Included items were

9, ‘6'

“Excessive drinking among students leads to severe and undesirable consequences, rs a

’9 6"

serious problem, rs dangerous,” and “is NOT a serious problem (R).”

Descriptive and injunctive norm. 8 Descriptive norm of drinking among college
students was estimated using six 5-point Likert items (1 = strongly disagree and 5 =
strongly agree). Included items were “It is commonplace that college students drink
excessively (R),” “Most college students drink excessively now and then (R),”
“Excessive drinking is prevalent among college students (R),” “Few college students
drink excessively,” “In general, college students tend to drink too much on occasions
(R),” and “It is usual that college students consume alcohols excessively (R).”

Five 5-point Likert items measured students’ injunctive norm of drinking (1 =

strongly disagree and 5 = strongly agree). Included items were “Most college students

 

For an easrer mterpretatron of results, the responses were recoded such that higher ratings indicate a
greater anti-drinking descriptive norm or a greater anti-drinking injunctive norm.

18

consider excessive drinking to be acceptable (R),” “Most students object to excessive
drinking among college students,” “At my university, it is okay for college students to
drink excessively (R),” “Most students at my university would think that college students
should not drink excessively,” “Most students at my university disapprove of excessive

drinking among college students.”

T RA measures. 9 To explore for the potential impact of participant’s subjective
norm on the behavioral intention, measures of subjective norm of drinking and
motivation to comply were included. Both measures were sub-divided into two separate
sets according to the targeted reference group; parents and community.

Subjective norm of drinking in family was tapped into by using six 5-point Likert
items (1 = strongly disagree and 5 = strongly agree). Included items were “Excessive
drinking is acceptable in my family (R),” “My parents object to excessive drinking,” “My
parents think it is okay that I drink excessively (R),” “My parents think that I should not
drink excessively,” “My parents allow excessive drinking (R),” “My parents disapprove
of my excessive drinking.” The measures of subjective norm of drinking in community
were created simply by replacing the word ‘parents’ and ‘family’ by ‘community’ and
‘other members of my community,’ respectively.

Five 5-point Likert items measured participant’s motivation to comply with
parents (1 = strongly disagree and 5 = strongly agree). Included items were “It is
important that I obey my parents,” “It is okay for me to disobey my parents (R),” “It is
important that I do as my parents wish,” “It is important that I meet the expectations of

my parents,” and “I want to do what my parents would think is a good thing.” Another

 

As in the measures of descriptive norm and the measures of injunctive norm of drinking, the responses
were recoded for higher scores to represent a greater anti-drinking subjective norm.

l9

five 5-point Likert items were used to assess participant’s motivation to comply with the
norm of community. Included items were “It is important that I conform to the norms of
my community,” “It is okay for me to violate the norms of my community (R),” “It is
important that I do as other community members would wish,” “It is important that I
meet the expectations of other community members,” and “I want to do what other
community members would think is a good thing.”

Tests of Measurement Invariance

Before testing the current hypotheses, all scales were subjected to measurement
invariance tests for cultural differences. A set of items may produce factor loadings that
are different in sign, magnitude, or both, across cultures due to unclear wordings or
meanings that may have been lost or added in translation. Removal of such items
enhances confidence that the cultural tendencies were comparable based on latent
constructs functionally equivalent across cultures.

Exploratory Factor Analysis (EFA) preceded the measurement invariance tests
whenever the investigator was unsure about the factor structure of a given set of items.
Principal Axis Analysis was conducted and the emergent factor loadings were rotated
using Varimax Method for separate cultures. Only the items that consistently loaded on a
same factor for both cultures were kept and then put into subsequent Confirrnatory Factor
Analyses (CF A) for measurement invariance tests. Factor structures obtained by EFAs
were maintained in CF As with no fiu'ther adjustments. Missing values were imputed
using Full-Information Maximum Likelihood Method (F IML; Little & Rubin, 2002) to
produce more precise fit indices and parameter estimates.

Tests of measurement invariance were performed according to a conventional

20

procedure (Muthe’n & Muthén, 1998-2004; Steenkarnp & Baumgartner, 1998) which
comprises the following four phases: (a) separate CFAs were conducted for each culture
to see if the same factor structure holds across cultures; (b) a chosen factor structure was
compared with the data from both cultures simultaneously, this time with all the
parameters freed; (c) phase (b) was repeated with restrictions on the parameters; (d)
finally, the model was fitted to the entire means-covariance structure for both cultures
simultaneously. At each phase, items producing relatively large errors were eliminated to
enhance model-data fit. Efforts were made to preserve the reliability of resulting
estimates; when the deletion of a particular item led to a minimal improvement of the fit,
the item was kept to maintain an acceptable level of degrees of freedom. Using this
elimination rule, it was ensured that each latent construct has at least three indicators at
the final phase.

The fit of each model was estimated with Mplus (Muthén & Muthén, 1998-2004)
with Comparative Fit Index (CF I) and Root Mean Square Error of Approximation
(RMSEA) as fit indices. CF I compares the obtained covariance matrix to the predicted
model and to a baseline null model in which all the variables included in the covariance
matrix are set to be uncorrelated. As CFI approaches 1, the obtained covariance matrix
better approximates the predicted model relative to the null model, and 0 for the opposite
case. As an approximated rule of thumb, a CFI of .90 or higher represents a reasonable
model fit. RMSEA computes the overall distance between the observed and the predicted
covariance matrix, with no null models involved. Conventionally, RMSEA of less
than .05 is considered as a good model fit. Both CPI and RMSEA are among the fit

indices that are least affected by sample size (Fan, Thompson, & Wang, 1999).

21

Obtaining an acceptable model fit at the final phase (i.e., full measurement
invariance or intercept and factor loading invariance) has rarely been found to be
achievable in the discipline of cross-cultural studies (see for review Steenkamp &
Baumgartner, 1998). Larger error terms resulted as more restrictions were added to the
model, and most measurement invariance models failed to pass the conventional fit test,
particularly RMSEAS .05, at the final phase. The conventional RMSEA criterion was
thus relaxed for a mean-covariance structure to be considered as acceptable when the
lower bound of 90 percent confidence interval hovered around .08 (see Table 1 for the
final measures, fit indices, factor loadings, and reliability coefficients). Only the items
that have survived the full measurement invariance test were used for further analysis
(see Table 2 for descriptive statistics and Table 3 for full correlation matrix).

Results
Induction Checks

Perception of roles. The induction of role perception was successful. The role

norm appeal (M = 3.89, SD = 0.68, n = 237) exerted a stronger induction of participant’s

reported role as a child than did the social norm appeal (M = 3.05, SD = 1.02, n = 245) or
the no-norm appeal (M = 3.00, SD = 0.90, n = 217); F (2, 693) = 73.22, p < .001, n2

= .17. A contrast coefficient test (i.e., “role norm” = + 2, “social norm” = “no norm” = —
1) produced a similar result, t(696) = 12.41, p < .001, n2 = .18. Tukey B post-hoe
comparisons also indicated that the role norm message induced the perception of child
role more powerfully than did the social norm or the no norm control condition, which

produced equally less powerful inductions of role perception as a child. Neither the main

effect for culture nor the culture by treatment interaction were statistically significant at

22

 

2382: $5558
o_£m:o%2 a ma wfieomuoaa
Sch 2: E265 has: wet—:5.
the 3.. 023.686 35 8286:. 0330:. £5...
:uonEoE
£5588 033.8%»: a wEEooon
mo ‘83 a m_ wet—5.6 029.888
Ami Nb. m52o>a 35 5o SE2— owmmmoE £5... ow. «NOEREEQD»

:.3:an b:
.«o 230 033.5%»: a we may—28:3
Set 2: 2.26.5 .88 wet—£6
83 mm. 028088 35 330%.: owmmmoE £5...
:éeoBQ be .«o 220
Beacons. a we mix—2.6 033088

a: mo. 23m _ $5 Bowen owmmmoE £5...
2.353; be Co 220 05 we eozaflzo

Chg mm. 28 >5 3 mason—nu owammoE mat...
asap—pan.

be me 220 933553. a wEEooon
mo $3 a m_ wet—5.6 028088
A219 mm. m£Eo>a $5 So 3&3 owammoE 25... no. «3.258
ans. 2:. 30. S. v vm madc—

 

Qossvmaw SE26 mass<mmsa <55. Eu u xx a». e

 

«NE u \6 3.3K 8.8.23.5 Esfiuuameo: _ 2an

23

 

his...

33 N... 36888 go .5533. m. bosom:
.mbozumwo:
AS... ow. €8.56 .98... @503 3068 .50:
.mbozmmooxo x53.
33 a... C. as... 32.58... 2.53 289... 2:. mm .. m3. 3. v v 3.3 8. 8.85m 338.0“.

2v £03308 £5 «o 8:83 was—5.6
33 .v. season a... 52 as 282. a8...

.meo—Euv 02.6083 8o... Eofio
saga.” so... waves... 9 =3

 

2 EV no. owammoa. £5 26. 2.3 0:32. .82..
.mbozmmooos x56 8 Ho: .3 EB
:33 K. newsman... £5 30.. 2.3 0302. .82:
:50 .85 35 .mp5
05 o. mac—5.6 023088 296 53
8.3 mm. 0938:. $5 on»... 2.3 omofimo >52: cw. «hoBmxogkx
...mmo_ E
E... E. s as... as v_a s 2...? 2. :83 ...
SEQ S. :dozfloeofi 5 E E...» _ .xfifi C...
83 E. :23... as a a0. v.5... a E :5 ...
.mwsxatv
3.335 .5 coma—amp. weigh—am
63 S. 85.8 s cm... a use... 2. 2.53 ... E. Se...»
hwc. one. m3. 5. V on GEEBE 3.2223.
28. N a: w 8:85 Bessfimzy. Ems... Eu a x.» N N e

 

SE38 . as;

24

 

£88030:

 

:3 mm. m. 3:025 m:o.:0 $356 0>_mm00xm..
#8030... 032.3 a
3%.. uh. m. 3:028 w:o:... wfixfifi 02038.9.
:.m00:0:com:oo
038.002: 0:: 0.0>0m o. 300—
Amng K. 3:025. w:o:a mic—5.6 02308.9. .8. we. go. .o. v v med. 3.. b_.0>0m E0305
amid—:30 03308.0 >8
.«o 0:50... 35:08: 05 0.. :00 0:. e.
00.20 0.30: .05 05.0.0 0wsmm0E 0:2...
8.... as. 2.2.. a 08.0
038.. .00.»... .8... 3.0.5.6 02300.3
>8 .05 3.00:5. 0wemm08 £5...
.9... 8. :08.
o. .200:— 03 0:. o. 820 0.30.. .05
mic—:30 02000008 mo floor—0 0389: 3:258:00
$3 5. 0.: 00.8.0580: 0380.: 02,—... :5. V omo. 3a. Vm. v 3.: S. no 0.300..—
2. .....83
35 v». m. 09.50:. 02. Co 20:53.. 2:...
:53... a.
83 S. :0... a. oases a... .o asap...“ 2:.
9.. 2.8%..
Chg me. 0. 0wamm08 0.5 .«o .:0:.:m.0 of...
...w:obm
a... R. m. ”was... as... .55.? 2F. as. t .. Rs. 8. v o. 3 .mm as. $.30 20830.2.
28. V a: 0 8580 Essafimza Ems... E0 .. s. N 0. e

 

SEES . 2......

25

 

...m.:00:.m 09:00 w:0:.0
30.5.0 02308.0 .0 0>0.....0m.0

 

a... 2.. 0.3.02.5 .8. .0 0.000.... .82..
:2023088 .50
.0: 032.3 0.50:3 0w0=00 .0... .5...
85.. 3.. 0.:03 b.3025 b: .0 0.50:3 .303...
...m.:00:.m 0M0..00 w:0:.0 w:.0.:...0
Gm. 0m. 02308.0 0. .00....0 0.50:3 .303... m... «03.02:.9t
9.. :....02308.0 3.2.00.0 0:53:00
.2... 00. 0.80:... 000:8 .0... .03: m. ....
9.. ...m:0.m0000 :0 20:... 00. 2:..0
AN... 3.. 0. 0:0. 3:00am 0w0..00 ..0.0:0m E:
9.. :.m.:00:.m 0m0..00
8a.. 0... m:0:.0 .:0.0>0.... 3. 90.5.0 02308.9.
9.. 2:0... 0:0 30: 2023088
8.... mm. ..:...0 3:003... 0w0..00 .303...
.05 8. .m. «0230.000Q.
0w... .0... .00. .c. v cm «005 E002 .0.00m
Q09. w .3. w 5.30:0 00<§<mm§ aim—>9. EU Q \0 m R B

 

02......000 . 03....

26

 

20555.0 02308.0
.0 305.3050 5:05:00 :3...

 

. . 0.. 2.. 0.. ...0...........
02308.0 030:0 555500 :2:
25.. v... :....02308.0
V.5..0 .0: 050:3 0:0 .0... x5...
85.. 2.. 5:55.00 :5 .0 30:50:. .050:
.055...
8.... m... 02308.0 0. 3.00.30 5:55:00 :3... 0... 00.5.5500.
0...
8.... 2.. 20555.0 02308.0 30:0 0.5.0.. :2:
9.. :....02308.0
2 .0 a... .5... . .2. .8... m. .. .5... 05...... .2:
2055.0
80.. mm. 02308.0 0. .0030 0.5.0.. :3...
0.. ......§.
30.. m... :5 5 0.00.5000 3. 9.0.5.0 02308.9. mm. «0.5.3....
N00. on... :3. S. V cm hm. . m. 5.02 0>..00_...:m
0.8. 0 .3 . 8.02.0 0.0.8532: <32: .5 a x... 0. 0

N

 

00:02:00 . 0B0...

27

 

2&5... 000m 0 m. 0.55 0.003 0.00505

 

.00.. a... 5:00.500 .050 .003 00 0. E03 ...
20.00505
>..:::.:.00 .050 ..0 55.0.0098
a... a. 2.. .82 _ 0... .5085. m. =3 0.. 00.5.3500.
:05... 080 a 0 .55
8.... Nb. 0.003 050.00 >5 .05» 00 0. .503 ...
2.0.5.0.. >5 .0 55.0.0098
83 R. 20 .85 . 05 .508... 0 0..
90.;mm3
3.... mm. 3:0.00 >5 00 00 . .05 50:00.5 0. .... 0». «030.8%
an. m». 2.0.5.0.. >5 >000 . .05 50:005. 0. .... >...:.00
N00. .3. mac. 5. V 0N 0. .oh 0. :0..0>..0S.
.08. 0 0m: 0 80.26 30.8.55 $.03... .00 0 \0 N N 0

 

0.30.0000 . as...

28

 

 

 

$0": 80".. 80".. 2.0": can: 80".. 8m": .00”:
G5 a... :5 a... $3 $0 .05 .00 :5 8.. .5. 8.. .0... 0... :5 00m .0...
an: an: an: an: an: an: an: an:
0.5 on. .05 .2 80.. 00.0 85 3.0 0.5 8.0 :5 an. a... N... 80.. a... 0.0.0
3.": 3.": 3.": 8."2 3.": 3.": .5."0 8.":
a5 5.. .5. 8... .00.. 8.. .05 8.0 E. 0.... $0.. 8.. .05 3.. :5 8.. 0.5588
8": an: .8": 00".. Non: an: 8": No":
.5. 8... :5 on... .05 20 G5 0.... .3. m... a... 8.. :5 0. .m .05 3.0 .2050 3.3.
0...": 0...": 0...": 0...": 3.": 0...": 0...": :7:
a5 0.... .05 Fm 85 5.. 85 0... :5 3... .2... a... .0... $0 :5. .5 00m .00.
m3": 3.": 3.": 9.”: 8.": 0...": 9.": ME":
:5 0.... .00.. 2... G5 3.... 85 2.. a... 0.... G5 a... :5 m... a... 8.. 2:6
2.": 3.": 0...": 2...”: m3": 0:": 0...": .2”:
$5 5... :5 a... .5. on... 85 .0... 0.5 a... .05 z... :5 .0... :5. .0 5.. 0:58:50
8.": 00...... 00.": 00.": 0.2": 00.": 00.": 2.":
0.0.. a... 0.5 a... .50 0.... :5 2.. :5 3... 8.... 00m :5 0.... :5 8... .0080 0.0
002% 3:88.80 0.05 8.050 .0 8.0.3... 3.55.80. 0.20.
00500.0. 0.80.. .:0.::m.< 00500.0: .0.0.>0...0m 50000:. 55000...

 

=0.§0=0U 2.0.500... .020 0.3.3.0 >0 002.58% 00.3.2000Q m 0.00...

29

 

 

 

3N": EN": 8an 8N": 8N": 8N":
..5 ..N .8. o... .8. v... .85 8.. .8. 8N .8. ..N .90..
..u: 3n: an: an: an: «an:
..5 KN ..5 .v. .N5 a... :5 8. ..V :8. N..N ..5 SN 2E0
3.": 3.": 2...": 3.": 3.": 8.":
.8. ...N .N8. 8.. .8. m... .85 8.. .8. ..N .8. SN 0.55.80
Nan: Nan: Na": Na": Na": Nan:
..5 ..N .8. h... .8. .N. ..5 ....V .N8. .8.N .8. NN.N .380
N3": :7: 27: NS": N..N: NS":
.3... .... .8. x... .8. N... .8. .. ..V ..5 SN .8... N..N .30..
.2": .2": .2": .3": .2": .2":
..... 2.. .N8. 8.. ..5 a... .8. E... ..5 ..N .N... N..N 8....0
.3": NE": .3": .2": .2": .1":
.3... v. .. .8. 8.. ..5 3.. ..5 S e ..5 EN .8. E .N $5.858
8N. u: 8N. u: 8N. u: 8N. u: 8N. n: 8N. u:
.8. 8.. .8. 8... .2... a... .5. .5. .N5 N..N .8. SN 3250
25.2.8. .383... .383... 552 252
39:00.2 3.5.00-2 8.82% 8.52% 02.05.?— 03310.09

no.3.

.m.D

 

3%....30 N 0......

3O

.N H «23. 2 u .m.D ._ 503.0382 .5. .I. Q .05. m... M Q ... .5805ch ...—80:838. 203 .... N L was mo. N. ..

 

 

.... NN. .5. N.. N..- .N. ... ... ... .N. ... .N. ... ...- .... 8.. 3.855.880.5508...
.N. ..N. 8.- ..N.- ... N.. N.. 8.. 8. .N. 8.. NN. .N.- 88. 8... .355.......580-2.8.

... ... 8. .N. 8. 8.. N.. .... NN. ... ... N..- R. .... .b.§5580.§ozo>.58.5=8...

8.- 8..- 8.. .... 8.. .... 8. .... 8. .... .....- ... ...... .3885..§ozo>..uo..8=m.N.

... .... ....- .8.- .N. ... .... 8... .5. ... .... ...N 85282555....

...- ...- .N.- ...: 8.. 5.- 8..- 8..- NN. .8. .N.N 8528:5550...

8N. .... ... ... .... .N. .N. 8..- 8. 8... 5.55. 83.85....

.N. N.. N.. .N. ... 8.- 8.. N... 8:33.50 5 88-. ..

..N. 8.. ..N. ... ..N. 8... ... .N.. 8.30 525.82 ..

..N. NN. .... .N. NN.- 8.. N. .. 5.5.58 83.85. .8

.... .N. 8.. .N. ... 8..N ... 8285.. .8

.N. NN. ... .... 3.. .5855. 58.55.. ...

..N. 8..- .... .8. 3.55580. 5.5.5... ..

8.. 8. N... ......0. 5.5.5:. .N

885.50..

... .. N. .. ... a . . 8 8 .- . N . 08 x. 8.8.;

 

...N... u 2. 5.5:. 5.5.250 ...... . 5.88.

31

level a = .05.

Similarly, participant’s role perception as a community member was induced
more powerfully in the participants exposed to the social norm appeal (M = 3.74, SD =
0.75, n = 247) compared to the participants in the role norm condition (M = 3.46, SD =

0.86, n = 237) or the participants in the offset control condition (M = 3.32, SD = 0.85, n =

218); F(2, 696) = 15.36, p < .001, n 2 = .04. A contrast coefficient test (i.e., “social

norm” = + 2, “role norm” = “no norm” = — 1) produced a consistent result, ((699) = 5.40,
p < .001 , n 2 = .04. Tukey B post-hoc comparisons also indicated that the social norm
message induced the perception of role as a community member more strongly than did
the role norm or the no norm control condition, which produced non-different and less

powerfiil inductions of role perception as a community member. The induction strength

of perceived role as a community member, however, also varied by culture, F (1, 696) =
18.43, p < .001, n 2 = .03. Across conditions, the US. participants (M = 3.63, SD = 0.83,
n = 412) experienced a stronger sense of role obligation as a community member than did
the participants in South Korea (M = 3.36, SD = 0.83, n = 290). The interaction between

culture and norm type was within sampling error at level a = .05.

Perceived social sanction. The main effect for the treatment on perceived social

sanction was not statistically significant, F (2, 695) = 0.05, p > .09, n 2 < .001. The

perception of social sanction, however, varied by culture, F (1 , 695) = 34.03, p < .001,

n 2 = .05. Specifically, heavy drinkers were perceived as deserving greater social

sanctions in the US. (M = 3.26, SD = 0.80, n = 412) than in South Korea (M = 2.92, SD =

0.72, n = 289). The culture by treatment interaction was non-significant at level a = .05.

32

Perceived argument quality. Perception of argument quality remained relatively

invariant across the treatment conditions, F (2, 696) = 2.06, p > .13, n 2 = .005. The

impact of culture on perceived argument quality, however, was statistically significant,
F(1, 696) = 123.20,p < .001, n2 = .15. Korean participants (M= 2.89, SD = 0.65, n =

290) rated the message as lower in argument quality than did the US. participants (M =
3.47, SD = 0.70, n = 412). The interaction between culture and treatment produced non-
significant impact on perceived argument quality.

Perceived locus of consequence. Compared to the no norm appeal (M = 3.43, SD
= 0.80, n = 218), the role norm appeal (M = 3.88, SD = 0.69, n = 237) and the social norm
appeal (M = 3.82, SD = 0.65, n = 247) were associated with greater perceptions of the

potential consequence of excessive drinking for people close to the participants

themselves, F (2, 696) = 24.10, p < .01 , 11 2 = .06. Tukey B post-hoc analysis found the

mean difference between the role norm appeal and the social norm appeal as statistically
insignificant however. Cultural difference affected the perceived locus of consequence,
but was trivial in magnitude, F (1, 696) = 5.00, p < .05, n 2 = .007. The culture by
treatment interaction was not significant at a = .05.

Perceived severity of drinking problem. The treatment produced little variance in

perceived severity of drinking problem, F(2, 695) = 0.25, p > .78, n 2 < .001. A

significant culture difference existed however, F (1 , 695) = 26.37, p < .001, n 2 = .04.

The US. participants (M = 3.88, SD = 0.72, n = 412) found the problem of student
drinking as more severe than did the Korean participants (M = 3.59, SD = 0.76, n = 289).

The culture by treatment interaction failed to produce significant impact on the dependent

33

variable.

Descriptive norm of drinking. Participants exposed to the social norm appeal (M =
2.29, SD = 0.64, n = 246) found that excessive drinking in college is slightly less

predominant than the participants who read the no norm appeal (M = 2.14, SD = 0.58, n =

218); F(2, 695) = 4.61, p = .01, 11 2 = .01. Tukey B comparisons indicate that the mean of

the role norm condition (M = 2.25, SD = 0.64, n = 23 7) significantly differs from neither

the mean of the social norm condition nor the mean of the role norm condition. The main

impact for culture was significant, F(1, 695) = 33.99, p < .001, n2 = .05; Korean

participants (M = 2.39, SD = 0.67, n = 289) viewed excessive drinking among students as
less prevalent than did the US. participants (M = 2.12, SD = 0.56, n = 412). The culture
by treatment effect fell within sampling error.

Injunctive norm of drinking. The main effect for treatment was statistically
significant, F(2, 696) = 4.53, p < .05, 11 2 = .01. Specifically, the participants in the social
norm condition (M = 2.64, SD = 0.71, n = 247) perceived the norm against excessive
drinking as stronger than the participants in the role norm condition (M = 2.49, SD = 0.69,
n = 237) or the participants in the control condition (M = 2.44, SD = 0.69, n = 218).

Tukey B post-hoe test grouped the role norm and the control condition as a homogeneous

subset. Reported injunctive norm also varied by culture, F (1 , 696) = 15.40, p < .001, 11 2

= .02; perceived social disapproval of excessive drinking was stronger among Korean
participants (M = 2.65, SD = 0.63, n = 290) than among the US. participants (M = 2.44,
SD = 0.73, n = 412). The culture by treatment interaction produced non-significant

impact on the perception of injunctive norm.

34

Subjective norm of parents. Neither treatment nor culture had significant impact
on participant’s subjective norm of parents on drinking. Their interaction term also
remained within sampling error of zero at level a = .05.

Subjective norm of community. Analysis indicates a significant main effect for

culture, F(1, 694) = 10.19, p = .001, r] 2 = .02. Specifically, the community’s norm

against excessive drinking was reported as more powerful among the US. participants (M
= 3.52, SD = 0.83, n = 410) than among the participants in South Korea (M = 3.34, SD =
0.65, n = 290). Neither the main effect for treatment nor the culture by treatment
interaction passed the significance test at level a = .05.

Motivation to comply with parents. A stronger motivation to comply with parents

resulted among the US. participants (M = 3.84, SD = 0.63, n = 411) than among Korean

participants (M= 3.50, so = 0.65, n = 290);F(1, 695) = 47.05, p < .001, n2 = .06. The

main effect for treatment and the interaction between culture and treatment produced
non-significant impact on motivation to comply with parents.

Motivation to conform to the community’s norm. The main effect for culture was
statistically significant, F(1, 695) = 15.80, p < .001, n 2 = .02. The US. participants (M =

3.13, SD = 0.84, n = 412) had a greater motivation to follow the norm of community than
did their counterparts from South Korea (M = 2.89, SD = 0.74, n = 289). The p—values for
the treatment effect and the culture by treatment interaction exceeded .05.
Hypothesis Testing

H1 and H2 were tested with 2-way independent groups AN OVAs examining the

effects of norm type and culture on behavioral intention (BI) and projected behavioral

35

intention (PBI) to avoid excessive drinking. H1 predicted that a role norm appeal is to
produce a stronger intention to avoid excessive drinking in participants (BI) than would a

social norm appeal. Non-significant main effect resulted for norm type, F (2, 696) = 0.13,

p > .80, n 2 < .001. A similar finding resulted when the analysis was rerun after
excluding the control condition, F (1, 480) = 0.01 , p > .90, r] 2 < .001. Null findings
continued when the projected behavioral intention (PBI) served as the dependent
measure; F (2, 696) = 0.73, p > .40, n 2 = .002 with the control condition included, and
F(1, 480) = 0.14, p > .70, n 2 < .001 with the control condition excluded. The current

data were therefore inconsistent with H1.
H2 was also inconsistent with the data because the culture by treatment

interaction effect was not statistically significant for either dependent measures; F (1, 480)

= 0.1 l,p > .70, n2 < .001 for Bland F(1, 480) = 0.23,p > .60, n2 < .001 for PBI. An

ANCOVA was conducted to examine H2 controlling for the nuisance effect of potential

covariates; age and sex. Both age and sex explained a significant amount of variance in

BI, F(1, 470) = 5.41,p = .02, n2 = .01 and F(1, 470) = 14.39,p < .001, n2 < .03,

respectively. A subsequent regression analysis indicated that female respondents (M =

3.62, SD = .78, n = 275) had a greater intention to drink less than male respondents (M =

334,51) = .76, n = 203), [3: .16, t(473) = 3.49, p = .001,n2 = .03. The impact of age,

however, became insignificant when entered into the equation with sex presumably due

to the more powerful impact of sex suppressing the impact of age, B= - .03, t(473) = -

.68, p > .40, n 2 = .001. Regardless, only a minimal increase in F -value resulted for

36

culture by treatment interaction effect for BI, with confirming the same null finding as

above; F(l , 470) = 0.38, p > .50, n2 < .001. Entering age and sex as covariates made
little contribution in reducing the error variance for PBI; F (1 , 470) = 2.86, p > .09, n 2

= .005 and F(1, 470) = 0.28, p = .60, n2 < .001, respectively. The culture by treatment

interaction remained as statistically non-significant as a result, F (1, 470) = 0.38, p > .50,

n2<.001.

AN OVA detected a significant main effect for culture. Specifically, Korean

participants (M = 3.61, SD = 0.69, n = 198) were more willing to avoid excessive
drinking, F(1, 480) = 6.76, p = .01, n2 = .01, than American participants (M= 3.42, so
= 0.84, n = 286). Korean participants (M = 2.83, SD = 0.75, n = 198) also had a greater

expectation of other readers’ compliance, F (1 , 480) = 30.15, p < .001, n 2 = .06,

compared to their US. counterparts (M = 2.46, SD = 0.71, n = 286). Figures 2 and 3
visually represent the results from hypothesis testing for comparison with the initial
prediction depicted in Figure l.
Exploratory Structural Equation Modeling

Investigation continued to explore paths linking the induction of role obligation to
behavioral intention, and their possible cultural differences. Conj ectured path models
were compared to the data using Structural Equation Modeling technique (SEM). Mplus
(Muthén & Muthén, 1998-2004) computed path coefficients and estimated model-data fit,
enabling to assess the empirical validity of each model. Missing responses were
approximated using FIML in producing parameter estimates (Little & Rubin, 2002).

Phase I. A full path model was first created to incorporate all the variables

37

Figure 2 Found Cultural Difference in Persuasive Impact of Role-Norm and Social-Norm

Appeals (Behavioral Intention)

. o .........
""""" o

E"
u:

"-0-- Korea

[WILL
DRINKLESS
l: 9) UI
LII Ur

S”
A

._. A_.._ .2... _. _-_.._._._..__ 7 ,_._v _. _ _ ._r g. _‘ —. __

Community Merrber Child
NORM TYPE

5»
9’!»
wan

38

Figure 3 Found Cultural Difference in Persuasive Impact of Role-Norm and Social—Norm

Appeals (Projected Behavioral Intention)

 

31
I o ------------
2.8 ‘ """"" 9
..lU)
u—lm I
53 l ---o---Korea
>.. E26l
En: l -—I—US
D 1 e a
2.44‘
l
!
2.2t__,__._ — — 4 a - ~ ~ a
Commnity Member Child

NORM TYPE

39

measured for this study. The induction was dichotomized for an easier interpretation of
the results by dummy coding for the three treatment conditions, which resulted in the
induction of role norm and the induction of social norm as two separate independent
variables. All measured intervening variables were entered as potential mediators. That is,
perceived social sanction, argument quality, locus of consequence, and perceived severity
of drinking problem were specified to relay the impact of perceived role salience (i.e.,
sensitized role obligation as a child or a community member) to behavioral intention and
projected behavioral intention simultaneously. Descriptive norm, injunctive norm,
subjective norm of parents, and subjective norm of community were also considered as

mediators although they may in fact constitute antecedent variables exerting direct impact

on the dependent measures independent of the message induction. 10 These decisions
reflect the notion that the resulting path coefficients may reveal the position of a
particular variable in the equation model. Specifically, a particular variable substantially
correlated with both of the presumed antecedent(s) and the consequence(s) is likely be a
mediator, whereas non-induced antecedents can affect the dependent(s) while minimally
affected by other independent(s).

The intervening variables were entered in the model, unordered, as few theory-
based predictions were available to determine the order amongst them. A researcher,
however, can determine post-hoc the order among mediators in a causal chain through

observing the size of their respective path coefficients to the dependent variable(s).

 

10 The subjective norm in the model constitutes the product of the initially measured subjective norms and
the motivation to comply with the corresponding reference group (i.e., parents or community). This
decision followed the assumption that subjective norm alone may not exert powerful influence on
behavioral intention particularly in the absence of motivation to comply; one may believe that it is
important to do as his/her parents would wish yet still lacking in motivation to actually do so. That is, one’s
subjective norm is likely to impact his/her behavior in reality only when combined with the motivation to
conform to the norm of the reference group.

40

Specifically, a mediator more powerfully correlated with the dependent variable is likely
to rest nearer to the dependent variable than a mediator less correlated with the
consequence. This statistical decision rule, however, becomes tenable only when it
corresponds to reasonable conjectures derivable from the pertinent literature.

The full model was compared with the data separately for each culture. The data
from both cultures produced a reasonable fit to the model although the path coefficients
differed in size (see Figures 4 and 5 for the models and fit indices). The model, however,
provided too complicated a picture of cognitive process, making it difficult to
comprehend and explain in a coherent manner. One major cause of this problem involves
retaining small but statistically significant path coefficients. The small path coefficients

seem to have passed the significance test due to the relatively large sample sizes (i.e.,
NUS = 412 and NKQR = 290) rather than to exhibiting substantial associations. Keeping

minor path coefficients just for statistical significance is prone to lower the probability
that the model replicates in future studies, particularly when the model was built for
exploratory purposes under little theoretical guidance (Rakov &Marcoulides, 2000).
Following this rationale, the significance level a was lowered from .05 to .01, and path
coefficients associated with p > .01 were deemed as statistically insignificant and
eliminated from the model. Figures 6 and 7 illustrate thus reduced path models for the
US. and South Korea, respectively. The path models in Figures 8 and 9 are identical to
those in Figures 6 and 7, respectively, but simplified after excluding constructs all of
whose paths were removed for being insignificant.

The reduced models were tested with the trivial paths excluded. Results, however,

suggested a further model adjustment for both samples. For the US. sample, the

41

 

.3580
0.53 8.83508 53 280$:me
A82 .3. u e .02: “a Emu—$36

£3qu8 0.3 3.56508 53 =<...

8.82 3:835

 

355868 —

 

 

 

 

 

 

  

 

 

353$
— .502 25835 —4/
E ,
$55.59 «N 355568
8:85. m fl 2 29m 88%.:
v/ a. a.
3 .562 >\ 2
0231809
mm :s
a.
.5365
MN bE>om e T
v\ e. .N 4
genes m \ 6863628 as £38 a. £28
38035 mace..— o_o~_ cocoa-9:

 

44” 2 cm

M: $56 \ cm

omammo—z

   
      

  

5:82am
333.65—

...Qo. ... .6 $36.. aea<mm2m .meo. n «.332 ”ma. n Eu :8. v a 58 use .3: 2 u N m .34: News: 23.4 e as»;

42

 

83:80

203 $820508 53 8858me

-802 .3. n a .05. 8 885:»?
3:85:88 98 $820508 53 =<...

Erase _

8.52 0280.35

8538
882 03835
- R

85:35 m 3:28?—
./ a-

8 532 \

cesarean em

wm ~ 2-
‘ 2
cm

 

 

 

 

      
   
     

 

 

  

8288:— m
38qu

85.68.80
384

       

 

8 £25

838.8%
@023qu

  
  

33. n a 836.. $3528 £3. u <88 63. u Eu mso. v a .

 

 

 

 

 

Esseé : geeeaeov
23— 85265
£28 £28
30% BV 95:25:—

 

Ec he .22 : u N .4 ”area 3ch ER m 6.5mm

43

 

 

 

 

355.82 R 3.55.58
835:: m 1/ 23. 8:265
a.
R
a
8 Masai
to>um /
_ mm

a

89.25 m \ £28 3 _ 6.28 _

@080qu 23— 5:26:—

./

mm mm

 

  

 

.5588
323.6“—

:5. n a :33 §<mm2m ”m2. u <mmzx ”Ra. u Eu :8. v a :8 "km 56?. n N a .35» 3%: mass»: w w 2:3

44

5::qu m

    

 

Nausea m

:5. n 5 E33 §N<mm2m a3. u <mms2 58. n Eu :8. v a SM n a. “Fm: u N N ”328: 38: naéé V 5 25»;

88%on 41//

mm av

 

3:296

882 0380.35 /

mm

 

 

 

$55508
2oz

Nun—poi
bigom

       

E
8.28

Bod

   

.3130
$9302

       

  

scream
33080.n—

  

      
 

 

 

 

 

 

wvl

 

£28
accuse:—

 

 

45

  

  
    

 

 

 

  
   

:53on $E=EE¢B AbENNEEoUV
cause:— m we bto>om mm 0.0: pm :38sz
a _ mm
8:53 m \ 885m £28 828
380.85 mm 323.3 mm 23— 3 5:26:—

om

   

 

two. N U ”530; $8<mm§
”m8. u <mm2m KS. n Eu :8. v a :8 ”xx .538 u N R ..«Emzutgs VS 2: 35%.»: usage w x 25mm

46

 

 

    

 

 

 

 

  

 

 

                    

  

 

  

 

  
 
 

Amt—Pan: $55808 335E508
5::qu m / mm .502 038036 mm 0.0m _N c2830:—
nm cm
5039i \
b_u0>0m
2 mm
£25
03302 E
«v
m—
:2225 m >\ .55st £28 2228
0080.85 mm “003020; 20% av conga:—

  

 

 

 

cvo. u a $38 $3,..me
£3. n <mm2m .03 u Eu :8. v NN .2; "\n .53 u N a “Quanatgfi 33M 350.8: .32.an w o. 230E

47

association between the perceived role as a community member and injunctive norm
became statistically significant with the elimination of the minor path links. This data-
driven information was incorporated into the final model for the US. sample. The model-
data fit remained acceptable after the adjustment; CF I = .925 and RMSEA = .053 (see
Figure 10). In the sample from South Korea, the induction of role as a child has lost its
function as a mediator as it produces only minor effects on its immediate dependent
variables; perceived argument quality and perceived problem severity. The induction of
child role was thus eliminated from the final model for South Korea. The revised model
adequately fit the data; CFI = .895 and RMSEA = .057 (see Figure 11). All path
coefficients in both models are statistically significant at level a = .01.

Each final model was then fitted to the means-covariance structure of the opposite
culture in order to examine the potential moderating effect for culture. The resulting
model-data fit should deteriorate substantially if the model takes a widely different
structure across cultures. The fits were lowered, yet remained as acceptable, although the

power of some path coefficients weakened (see Figures 12 and .13). Interestingly, the

results from x 2 goodness-of-fit test indicated that the US. model may fit Korean sample,

1 2 (265) = 546.41, p < .001, slightly better than it fits the US. data set, X 2 (265) =

571.47, p < .001. An objective conclusion, however, remains unavailable because
identical models with the same degrees of freedom are incommensurable using
Likelihood Ratio Test (LRT). The other fit indices, on the other hand, indicate that the
US. model may fit the US. sample better (CFI = .925, RMSEA = .053) than it fits the
Korean sample (CFI = .894, RMSEA = .061). Still, both CF Is and RMSEAs constitute

descriptive fit indices unable to determine whether the US. model performs significantly

48

better with the US. sample or with Korean sample. Here, however, a greater focus should
be placed on the observation that both models produced an acceptable fit with the sample
from a presumably different culture, with revealing little moderating effect for culture.

Two observations common for both countries merit mention. First, perceived
severity of the drinking problem exerted a substantial positive impact on behavioral
intention across cultures. The willingness to drink less tended to increase as the

participants perceive the problem of drinking as more severe. The effect size may seem

culturally different when comparing for the strength of path coefficients; Bus = .70 and

BKQR = .60 (see Figures 10 and 11). However, it should be noted that the two models are

qualitatively different, which do not allow for direct comparison of path coefficients in
either size or magnitude. As an alternative, their respective zero-order correlations were
compared afier z-transformation (see Tables 4 and 5 for correlation matrix for the US.
and Korea, respectively). The difference between the two correlations fell within

sampling error at level a = .05; — .06 _<_. Pus — pKOR S .24. Second, the perceived social

sanction predicted projected behavioral intention positively for both countries; B us

= .17 and BKQR = .26. Participants anticipated a greater willingness to avoid excessive

drinking among others as they perceive potential social sanctions as more threatening.
The effect size remained approximately equal for both cultures at level a = .05;
-.24 S Pus —pKOR S .06.

A further observation revealed potential culture differences. The injunctive norm
exerted significant impact on projected behavioral intention in the US. (0 = .22) while

producing non-significant influence in the Korean sample. The subjective norm of

49

50200023.. .8. u Q use no. M Q “a “505%? £35308 203 2. N a can 2. N x ...

 

 

2.. 8. 8. 8. .7 2. 2. 8. 2. 8. 2. 2. 2. 8. 28 3255883382.:
8. NM. 8.. 2.- 8. 2. :H. 3. 2. a... 2. 8. 8. 32 38382358.: .2

8. 3. 8. 8. 8. 2. .m. 2. R. 8. 2. 2. 82 22855888220232.8382

8.- 2.- 8. 2. 2. 8. 8. 8. S. 8. S. 2.2. 3838 Eozozeozsmi

8. 8. 8.- s.- 8. 8. 2. 8. 3.- 8. 3.8 5.238352

8.- 8.- 2.- 8. 2. 8.- 2.- 2.- 8. 28 Eozoéacaooa

8. 8. 8. 2. a. 8. 8. Q. 8.... 2.38 8385 .w

.m. .2. E. 8. NM. 9.. 8. PM 85:32.8 8 882 s

z. 2. 8. 8. 8. 8. 2% 2.25 .5582 8

8. 8. _ _. 8. 8. 88 8858 828.8 .m

8. .N. 8. E. 38 5 2.28.8: .2.

R. 8. 88 805.5 @2530 .m

8. 2. 88 35588 88.65 .N

8.. 88 22200 8:265 ._

2 2 : 2 a a h e m 2. m N _ S. 8 2853

 

.. «m:- u 2» we: 2: L888: 53330 2. 23

50

5038083.. .5. H Q 28 3. u Q 3 2805.88 8:85:38 0.53 2. N. L can m_.N L ...

 

 

8. 2. 8.- 8. 8.- 8. 8. 8. 2. 8. 8. 8. 2. 8. 28 2252581358-: .2
2. :. 8. 8.- 8. 8. 2. 8. 2. 8. 2. 8. 8. 8.8 @538 2888-2 .2

8. 8. :. 8. 8. 8. 8. 2. 2. 2. 2. 8. 88 35888 8.0288038 .2

8.- 2.- 8. 8. 8. :. :. 8. 8. 2. 8. 8.2. Aacoaaéozozeosam.:

8. 8. 8. 2. 8. 8. 2. 2. 2. 8. 8.8 82.288522 .2

8.- 2.- 2.- 8. 8.- 2.- 2.- 2.- 8. 28. 8.8288828

2. mm. 8. 8. a. 8. 2. 8. 8.8 2:58 8288 .8

8. 8. 2. 2. 8. 8. :.. 88 88:32.8 8 88-2 s

8. 8. 8. 8. 8. 8. 28 2:25 “55:22, .8

8. 8. 8. 2. 8. 3.8 8:28 3288 .m

8. 8. 8. :. E 88%: .2.

2. 2. 8. 8.8 8822 .2288 .m

8. 8. 8.8 22.58.88 .8885 .8

a. 8.8 6.28 582.5 ._

2 2 : 2 8 a \- 8 m v m N _ a. 8 28:9

 

.. 88 u 28 9:8 88%: 8.3.38 m 28h

51

 

 

5282 352202 2288508

 

  
   

.8252: m bto>om 23— 53265 _

 

8 8 8

      

      

 

  

.8352: m
38.32

5:22am

32832 0.3. zero—.6:—

‘ 8 II.—
6.28 3. III— £28

 

mm

832

 

 

0265?:

28. n 6 833 28.822 88. n .8852 ”88. n Eu :8. v a .88 "8 2.. E u N .8 8E 888 322 2 288

52

 

m—

 

  

 

388%

8.82 03200.35 /

an em

vm

 

‘III
53:00:— m /

   

E0383—
bto>0m

  
 
 

mm

 

  

38.25808
03%

325.5588
cocoa—2:

  
  

 

2.80 >\
03802

2. 3.

    

cob—.085 m \

838.32 . 8

  

5:028
“0030202

  

  

28. u a 833 38.6.85. 88. u 88,2 88. u :5 :8. v a .88 "8 .832. n N p. .8288 888 35 2 _ 2&2

53

S. H d “a 2305an 3.005506 000 $565000 53 0050 =< :o. A Q ...

 

  
 

 

 
  

 

 

      

      

  

E0395 Aécaeeouv Abigufioov _
.8352: m . we bt0>0m _ mm 0.3— mm 5:26:— _
@_ o_ *0—
=oz=3£ m \ 835% 6.38 8.28
00000.85 A 5N 003080,— A. ...w _ 03”— av 5:265

 

%

 

 

.N

:8. u $532 0.3. n Eu :8. v 0 .SN ”xv .303 n N N 035% 5020M e 30E 3%: as S 050E

...».8. n a «033 §<mm§

54

 

hm

 

   

S. H d :0 2.00536 bfiozmsfim 000 3565000 £3 0050 =< :o. A Q ...

 

 

 

  
    

 

T ...No u Away—pan:
coup—005 m / 8.82 028035 1/
cc *o—
E0395
bt0>0m um
3:55:58 .8 32:55.08
0.0M note—.6:—
NV
0:95 >\
03302
N.— @—

      

:BEBE m \

@060on n «N

  

cozocmm
“002029—

.03. n a «$8 «53.0%:
$8. u $62: 6;. n Eu :8. v a .mmm use {2% u N N ..sasum .05 a: e 0202 38: :88: 2 050E

55

parents constituted a moderately powerful predictor of behavioral intention in South
Korea ( B = .24), but remained relatively uncorrelated with all other variables in the US.

sample. As aforementioned, however, the two models are incomparable, particularly for
the impact of a variable present in one model but absent in the other.

Phase 2. Further efforts were made to create a culturally equivalent structural
equation model, which would justify testing for potential culture difference in
interrelationships among the constructs of research interest. The two tendencies common
for both nations provided the basis for constructing the new model; the main effect for
perceived problem severity on behavioral intention and the impact of perceived social
sanction on projected behavioral intention. A conjecture arose by extending these
observations that the norm variables (i.e., injunctive norm, descriptive norm, and
subjective norm) may predict projected behavioral intention independent of the message,
whereas participants’ own behavioral intention can be better explained with the
immediate outcomes of message induction (i.e., perceived social sanction, perceived
argument quality, perceived severity of problem, and the locus of consequence). The
model explained the data adequately with CFI = .893 and RMSEA = .046. The results,
however, suggest that a second-order unidimensionality model is invalid for the norm
factor. Neither the injunctive norm nor the descriptive norm hung together with the
subjective norm constructs, with contributing little to the higher-order norm component;
2 = .04 and -.24, respectively. As an alternative, the injunctive and the descriptive norm
construct were isolated from the two subjective norm constructs, and presumed to be
second-order unidimensional on their own. Two independent second-order norm factors

evolved as a result. The model-data fit remained as acceptable after the adjustment; CFI

56

= .903 and RMSEA = .044. Still, the second-order unidimensionality assumption failed to
hold for the injunctive and the descriptive norm components, particularly due to their
factor loadings being substantially different in size; 2 = .40 and .95, respectively.
Moreover, the second-order factor for the injunctive norm and the descriptive norm
construct behaved in unpredictable ways when related with other constructs; it correlated
negatively, albeit insignificantly, with the subjective norm factor while producing a
significant negative impact on projected behavioral intention. The injunctive norm and
the descriptive norm constructs were excluded from further fit tests for instability in
factor structure.

The model after the adjustment seemed to assimilate the traditional TRA model
particularly when considering the perception of the message as a proxy of the
participant’s attitudes on drinking (see Figure 14). This time, both the attitude
components and the norm components substantially loaded onto their corresponding
second-order factor. The model-data fit was acceptable when tested with the entire
sample; CFI = .924 and RMSEA = .044. The model also fit the data separated by culture;
CFI = .918 and RMSEA = .042 for the US. and CFI = .919 and RMSEA = .039 for
Korea (see Figures 15 and 16). The model, however, seemed far fi'om ideal for either
country when examining the factor loadings and path coefficients. In the US. data, social
sanction hardly hung on the second-order factor for message perception (Q = .38) and
the second-order construct for subjective norm had a negative impact on projected

behavioral intention (B = - .24), which is inconsistent with past findings. Moreover, B: -
.24 was statistically non-significant while a smaller partial correlation (i.e., [331-1431

= .18) passed significance test at the same level alpha. This result originated from the

57

8. u a 8 2.858% 8.888: 03 38.0808 £8 .28 =< :8. A a ...

 

   
  

 

 

 

 

    
   

 

 

3:08.—
Zm
1/ X
_ m S
nos—.02: m ...oN - A§§EEouv _ mm A§§EEQUV
38.38 23— 8885
NV
3 Vm
om «
VN .
:2385 AEEUV AEEUV
_Bo_>20m we . cm 23— . mv _ 550.65
3. 2. mm 3

888028 bt0>0m 5:030 couogm
be 80°..— v0>_00._0m 093002 300m

.. :8. u a 838 §N<mm2m :8. n <82: :8. n Eu :8. v NN .28 use .88: n N .N. .888 288 N08: 3: E 2:08

58

S. .I. a 8 8005.86 b82855: 0.8 38035000 :88 8050 =< 2o. A Q ...

      
          

4/ \
cm MN.

COW-buns m I AbmczEEOUv ”N

38.3.: .8 28:

av
cc 3
2 %

8280.5 9.28

8.5383: we Alnl cm 23— Al 3
_m 2. mm mm

   
   

3:85 :cuocam
0w0mm02 300m

bt0>0m
.0202?—

888088
.«o 953

  

88. u .0 838 88:2: 68. n <82: :3. u 20 :8. v N: .98 "8 .88: u N N 8.8 382 5: 2 2.8:

 

— 98888288

 

_ 8885

 

 

— AEEUV

 

888.:

 

59

S. H a an Emu—mama b—aocmtfim 08 356508 53 850 =< :o. A A *

 

     

     

 

 

 

 
 

 

       

 

 

 

 

1/ \4
cv we
5.228:— m N_ - AbEseeoov vm AQEEEoUV
380.85 ... Box .5526:—
we
3 .3 G
.. e
8::25 €28 _ 528
_flo_>m:um EV . _m 23— : 5:265
3” we we 3

   

      

  
   

Sago
.3 300..—

£30
awammuZ

€05
3:020."—

:ouocam
38m

6 $38 xa<mm2m ”35. u «$32 as. n :6 :8. v a “as n»? 538 n N R .«SBE BE: 5: 2 2&5

6O

second-order norm factor being associated with a large standard error which is
attributable to the two subjective norm constructs failing to meet the second-order
unidimensionality assumption. Similar problems occurred with Korean data. The factor
loadings for social sanction (Q = .49) and perceived locus of consequence (2 = .34)
were relatively small compared to those of message quality (2 = .64) or perceived
problem severity (2 = .66), suggesting that the former two factors may jointly constitute
a separate cluster or independently form two respective clusters. The second-order norm

construct continued behaving erratically due to its unstable factor structure; the ample

path coefficients as BNQRM-pERCEpT = .46 or BNORM-BI = .42 were found statistically

insignificant.

Fit tests continued incorporating these results. Locus of consequence was
excluded from the model, while perceived social sanctions were maintained as a separate
norm construct, this time considering it as a measurement of perceived social pressure
external to message perceptions. The second-order norm factor further broke into two
original subjective norms (i.e., parents and community). Adopting Eagly and Chaiken’s
conjecture (1993) that subjective norms may affect behavioral intention indirectly via
attitudes, both message perception and perceived social sanction were predicted by the
subjective norms as well as the message induction. The model was further simplified by
restricting the subjective norm of parents to affect the message perception only and the
subjective norm to predict perceived social sanction alone. This last decision followed the
observation that the subjective norm of community correlates more powerfully with
perceived social sanction (r = .32) than with the two components of message perception;

r = .28 and r = .16 for perceived problem severity and perceived argument quality,

61

respectively. The reverse pattern prevailed for the subjective norm of parents, whose

correlation with problem severity (r = .35) surpasses the correlation with perceived social

sanction (r = .19). ll Perceived argument quality, however, remained relatively
uncorrelated with either the subjective norm of parents (r = .14) or the subjective norm of
community (r = .16). That perceived argument quality lacks association with the
variables of interest and, more importantly, it behaves differently from how its
companion variable (i.e., perceived problem severity) behaves suggest its elimination
from the model. As a result, perceived problem severity alone replaced the second-order
construct for message perception. Path models illustrated in Figures 17 and 18 (i.e., the
Separate Process Model) represent the above reasoning applied to the US. and Korean
sample, respectively.

Having two equivalent path models warranted a comparison of path coefficients
in magnitude and sign, which would further the investigation of potentially meaningful
cultural difference in cognition. Four observations merit mention. First, the induction of

child role exerted a larger impact on perceived social sanction in the US. ([3 = .24) than
in South Korea ([3 = .11), though the difference was only marginally significant;
- .01 s pus — p KOR S .29. The US. participants seem to believe that one may subject to

social sanctions for worrying parents with troubles he/she created while drunk. Korean
participants, on the other hand, seem to distinguish family matters from society, viewing
that society may not intervene to chastise someone for failing to fulfill duties as a child of
his/her parents.

Second, the impact of perceived role obligation as a community member

 

l .
The margin of error for 95 percent confidence was i .10 afier z-transformation (N = 702).

62

S. u d 3 ESE—Em Enouflasm 0.8 $56508 53 .550 =< :o. A & ...

W

cos—.8:— m \ somgm
38.35 38m

...No

  

 

 

All L _ $58.58 A R Again—:88

 

 

    

 

    

   

  

0— CM— £05056:—
a
8
E
access £33m £28 £28
3.6328 3 3283 . .2 23. . 3 8:265

  
  

 

+

an

@—

aSSmIZm

...?o. u a $33 §a<mm2m
55. n <55 68. u :6 :8. v a .2: use 435 u N N 5,. ..Q 38: ”seek 22.5% 2 2am;

63

 

53:8:— m
@2885

C

con—.25
_Eo_>m:om

        

   

    

S. n d “a EducEwi 3.3.53.5 2a £56508 53 550 =< :o. A m ...

 

    

 

  

 

 

 

    

  

NM
\ ... mo 6
an 385.88 mm AbmcsEEoUv

20m c0326.:
mm
L _

8:28 828 _ £28
3 @028qu . .. _ _ 20M 3 5:26:—

 

  

 

  

 

VN mm

*QMO. H .U mmBOA $om<mm§
68. u <82: :8. u Eu :8. v a. .68 nee «3.8 u N \N. .8288 38: $85 2336.... 2 28E

predicted perceived social sanction significantly better in South Korean ([3 = .28) than in
the US. (B = .11); — .33 S Pus - pKOR S —.03 . In Koreans’ perception, society may

exercise the right to punish people for causing troubles to the community, whereas
Americans tend to deny society such privilege. Americans may depend more on the legal
measures than informal social pressures to treat trouble makers to the community.

Third, perceived problem severity produced a substantially more powerful impact

on BI ([3 us = BKOR = .59) than on PBI ([3 us = .16 and BKOR = .18) for both countries;

 

.38 S p31 - ppm S .66 for the US. and .34 S p31 - ppm S .66 for Korea. Perceived

social sanction had a significantly larger impact on PBI ([3 U5 = .30 and BKOR = .36)

than on B1 ([3 us = .15 and BKOR = .04) for both cultures; C195% ranged

—.30 S pm — ppm S —.02 and —.50 S p31 — ppm S -—.18 for the US. and South Korea,

respectively. These two findings demonstrate that participants from both cultures applied
two separate mechanisms in determining BI and PBI. Participants reported that their own
intention to drink less would increase as the drinking problem was perceived as more
severe, while anticipating that others may try to avoid excessive drinking as perceived
social sanctions grow stronger.

These two models were reanalyzed using composite indices, the averaged sum of
individual items. The model-data fit deteriorated slightly due to reduction in degrees of
freedom, but the path coefficients remained similar to those estimated using individual
items (see Figures 19 and 20).

Discussion

This study predicted that a role norm appeal would produce greater persuasion

65

_

8. n a a .5888; 88:28: 93 3:22.808 5% .28 .2 :o. A a ..

 

_ 85.55848 _

_

mm

MN All! .8 @5588 T MN _ 85588
E 38.3.2 5:88 .208 23. _ 8885
N

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a w
8825 853 6.28 _ 6:28
_ 8.6322... _. «m _ 828.2. # . ...: _ 22 _ _ a. _ 8:265

mm

253215

 

 

 

 

 

 

 

 

 

38. u a 533 §<mm22
:.2. u <82: :8. u Eu :8. v N: .8 use .53 n N N .323 22358 .28 :28: 2.822 2838M. 2 252.2

66

S. H d “0 3005ch biocmsfim 000 30205000 500 0050 =< :o. A Q ...

 

 

 

 

 

 

 

 

 

 

 

.1
I

 

 

 

 

 

 

 

 

$588815
0N
0m . 2 32:25:08 . C Agassi—00v
E 080085 _ coaoqnm 300m 0.0: 0030.65
mm
...oo -
020:8:— om £63m - 2:28 A 3:58
.8030som 60300.5; .3 20x 3 00:26:—
_m
255:5

 

.08. u a 838 2.32.822
: 2 2. u $522 5.5 M £0 :8. v a .8 "\n .35 u N R .623 88258 .3338 38: .2822 22.8.4 8 ”am;

67

than a traditional social norm approach, and this pattern should be more pronounced in
holistic cultures than in analytic cultures. Neither predictions were consistent with the
data. Participants who read the role norm message (i.e. “You should avoid excessive
drinking as a responsible child of your parents ”) were no more willing to drink less than
the participants exposed to the social norm appeal (i.e., “You should avoid excessive
drinking as a community member ”). This pattern of data prevailed across the US. and
South Korea, an analytic and a holistic culture, respectively. These findings suggest that
(a) role norm‘appeals may create little perceptual difference from traditional social norm
appeals at least in the persuasion context tested here, producing little enhancement in the
amount of compliance, and (b) role norm appeals’ inability to garner a greater persuasion
than would social norm appeals may be common for both analytic and holistic cultures.
C heck for Validity of Rationales

H1 followed the rationale that a role norm appeal should produce greater
conformity than a social norm appeal because people often find social sanctions from
related others to be more threatening than the potential sanctions from unrelated others.

As evidenced in the induction check, however, perceived social sanction on excessive

drinking remained nearly invariant across the three message conditions, 11 2 < .001. The

path coefficients in the SEM models (see Figures 17 and 18) provide mixed support for

this reasoning. The induced role perception as a child ( [3 = .24) had a greater impact on

perceived social sanction than the perceived role as a community member ([3 = .11) for

the US. data; .00 s Pchild — Pcommunity S .28 .12 The direct opposite was the case for

 

12 .
The reported C195% is wider than actual due to rounding error.

68

South Korea however. Perceived social sanction was significantly higher among

participants primed with the role as a community member (B = .28) than among the

participants reminded of the role as a child ([3 = .11); .34 S Pchild — Pcommunity s —.02.

These findings suggest that there exists a stronger family-community connection
in the US. than in South Korea, or the community pressure for norm conformity remains
more powerful in US. families than in Korean families. That is, US. students seem to
believe that one who fails to fulfill his/her duties as a child may well deserve sanctions at
the community level as well as immediate punishment from parents. The term
‘institutionalization’ has been used frequently to encapsulate the typical American culture
(e. g., Hall, 1976). A problem of individual families ofien becomes a problem of
community. The police, after receiving a call from a neighbor, often engage in incidents
of domestic violence or meddle in quarrels between a father and his son. There exist
numerous support groups to help people unable to lead a normal family life for being
addicted to substances (e. g., “Alcoholics Anonymous”), losing a child (e.g.,
“Compassionate Friends”) , being infertile (e.g., “RESOLVE”), suffering from minor
disorders to serious diseases like cancer (e.g., “The Wellness Community”), or even
being in debt (e. g., “Money’s Debt Support Group), to mention a few. That the US.
participants had a significantly higher motivation to conform to the community norms
than did the Korean participants also corroborates this conjecture (see Table 6 for culture
differences at construct level). The US. correlation between the motivation to comply
with parents and the motivation to follow the norm of community, r(412) = .42, p < .01,
also surpassed that of South Korea, r(290) = .34, p < .01, although the difference

remained marginally significant at level a = .05; — .05 S Pus — pKOR S .25 (see

69

Table 6 Mean Diflerence Between the US. and South Korea at Construct Level

 

 

D = KOR — US
Behavioral Intention .34*
Projected BI .64“
Perceived Sanction -.58**
Argument Quality -l.17**
Locus of Consequence -.20*
Severity of Problem -.45**
Descriptive Norm .49“
Injunctive Norm .41**
Subjective Norm (Parents) -.13
Subjective Norm (Community) -.32*
Motivation to Comply (Parents) -.57**
Motivation to Comply (Community) -.36**

 

70

*p<.01**p<.001

Tables 4 and 5).

Korean students, on the other hand, seem to more strictly separate family matters
from those of community by reporting that violating role norms in a family entails few
reasons for the community to interrupt or impose sanctions. One would incur sufferings
to his/her parents for a wrongdoing he/she committed while drunk. Yet the problem stays
within the familial boundary and the community’s right to intervene remains restricted.
This rationale corresponds to the prediction from the analytic versus holistic cultural
distinction. Possession of a group membership, hence a social role, turns an otherwise
insignificant object into a meaningful social entity (e.g., Lim, Allen, Burrell, & Kim,
2007). Devoid of substantial roles, the ‘generalized others’ remain as insignificant beings
who can exert little normative pressure on one another.

The final path model for Korea at Phase One (see Figure l 1) captures this holistic
tendency to separate ‘our path’ from ‘their path’ based on the presence or absence of role
relationship; the path leading to the projected behavioral intention remains relatively
independent of the path to the participants’ personal intention to avoid excessive drinking.
Specifically, Korean participants anticipate that other people’s intention to drink less
would enhance in a linear function of the message’s argument quality and the Strength of
potential social sanction on heavy drinkers, while attributing their own intention to drink
less to perceived severity of drinking problem and the subjective norm of drinking in the
family.

H2 reasoned that role norm appeals should be perceived as more persuasive
among holists than among analysts because holists tend to be more sensitive to role-

bound obligations than analysts. To the extent that this postulate holds true, the role norm

71

appeal should induce a stronger induction of role perception than does the social norm
appeal among Korean participants, and the difference in induction strength between the
role norm and the social norm appeal should be significantly larger in the Korean sample
than in the US. sample in that particular direction. A stronger role perception resulted in

the role norm condition (|3= .48) than in the social norm condition ([3 = .22) as expected.
Their difference in magnitude was statistically significant at level a = .05;

.14 S Pchild — Pcommunity S .46. However, the same pattern reappeared in the US. data;

the induction strength of role norm appeal ([3 = .44) exceeded that of the social norm

appeal ([3 = .27) in a statistically significant fashion; .05 S Pchild — Pcommunity S .33.

The two C1395o/o had an overlapping region, indicating a non-significant moderating effect

for culture on the difference in induction strength. Therefore, the failure to find the
predicted culture by treatment interaction may be attributable in part to the invalidity of
the postulate that holists possess a more sensitive role perception than analysts.
The Third Person Efiect

Findings demonstrate that participants from both cultures apply two separate
mechanisms in determining BI and PBI. Participants reported that their own intention to
drink less would increase as the drinking problem be perceived as more severe, while
anticipating that others may try to avoid excessive drinking as the perceived social
sanction grows stronger. This cognitive pattern is analogous to the third person effect
hypothesis (Davison, 1983), which maintains that people tend to overestimate the
influence of media for others while perceiving themselves as more immune to media

impact. In particular, a meta analysis (Paul, Salwen, & Dupagne, 2000) indicates a

72

stronger third person effect for college-aged students, the major subject pool of this study.

The third person effect seems to hold in the domain of interpersonal
communication too. Johansson (2005) found that people tend to anticipate that others’
political attitudes be more influenced by interpersonal communication as well as mass
media, while attributing their own political attitudes to personal experiences. The current
finding is consistent with the third person effect hypothesis presuming that the perception
of social sanction originates not only from mass media coverage but also from daily
contacts with other members of society.
Koreans More Willing to Drink Less

The data from this research provide little explanation for the cultural difference in
the willingness to drink less. The willingness to avoid excessive drinking was higher
among Korean participants than in the US. participants even though the findings indicate
the opposite as more plausible. That is, compared to the US. students, Korean students
found (a) potential social sanctions as less threatening, (b) the message’s argument as less
persuasive, (c) the locus of consequence to be more distal, and (d) the problem of
drinking as less severe (see Table 6 for culture difference at construct level). Both
injunctive and descriptive anti-drinking norms were stronger among Koreans than among
Americans. These normative influences, however, had only trivial impact on behavioral
intention as evidenced in path analyses, and hence provide little rationale for the stronger
readiness to drink less among Korean participants. Therefore, the found cultural
difference may well be attributed to the factors left unmeasured for this study.

There are at least two lines of argument consistent with this finding. First, the two

student samples may differ in the amount of exposure to anti-drinking campaigns targeted

73

at college students. Michigan State University, the subject pool of the US. participants
for this study, has been making continued efforts to reduce student binge drinking around
campus since 2006. The school has initiated a wide variety of campaigns applying
descriptive (e.g., “82% of MS U students drink moderately or not at all on football
Saturdays”) and injunctive norm approaches (e.g., “Fewer than 1 in 10 MSU students
allow alcohol to interfere with their academics ” or “Most (94%) of MSU students
disapprove of pressuring others to drink”). And those campaign messages have been
advertised repeatedly around campus and on local newspapers as well (Michigan State
University, 2008). Fewer such attempts seem to have been made at colleges in South
Korea. That colleges are investing less effort to intervene may translate to students in
Korea drinking less, with incurring fewer drinking related problems than students in the
US. However, results from a recent survey contradict this conjecture (Yonhap, April 15),
in which nearly 22 percent of students at a premier Korean university were found to drink
2-4 times a week and 3 percent to drink more than 4 times a week, with more than 17
percent classified as at risks of alcoholism (N = 431).

The US. students may have turned into a more persuasion-resistant sample after
the repeated exposure to a series of anti-drinking campaigns. Reactance theory posits that
people tend to resist any attempts to threaten freedom and resistance arises in an attempt
to re-establish, and to prevent further loss of, the freedom (Brehm, 1966; Brehm &
Brehm, 1981). Resistance often takes behavioral or attitudinal opposition to the position
endorsed by the source (i.e., the ‘persuasive boomerang effect’). In particular, a repeated
freedom threat may induce anger in the target audience, producing a further resistance.

Heavy smokers often find anti-smoking campaigns irritating as they make constant

74

attempts to threaten freedom. Reportedly, angered smokers feel forced to ridicule anti-
smoking campaign messages as absurd or experience impulses that they must light up
especially right in front of the posters stating “No Smoking” (Wolburg, 2006). Similarly,
the US. participants’ lowered willingness to refrain from excessive drinking may have
resulted due to their repeated exposure to freedom threats, not to the Korean students
being more compliant.

The second conjecture stems from the potential group difference in perceived
novelty of the argument. Morley (1987) maintains that subjective judgment of
information novelty constitutes an important determinant of belief change in conjunction
with perceived plausibility and significance of the claim. Information that is already
familiar to the target suffers limited chance to receive a thoughtful attention from the
target, whereas novel information can induce a more effortful processing in the target’s
cognition. Provided that the target has perceived the argument as both plausible and
important, a greater change in belief becomes more likely in a target who also found the
information as novel rather than banal. Previous exposure to similar norm campaigns
may have rendered the treatment messages more familiar to the US. participants, while
the same stimuli might have seemed less familiar to Korean students who had remained
relatively unaffected by similar attempts at persuasion. The finding that Koreans were
more compliant than Americans may thus be attributable to the treatments being
perceived as less familiar among the former than among the latter.

Small Social Norm Eflect
Results from path analyses indicate that, for both countries, social norm factors

exert only minor influences on behavioral intention, which was largely determined by

75

participant’s personal assessment of problem severity (see Figures 17 and 18). This
finding is in fact consistent with past TRA literature: a meta analysis of 37 tests of TRA
model reveals that attitude components significantly outweigh the normative factors in
predicting behavioral intention (Farley, Lehmann, & Ryan, 1981). Specifically, the
attitude-behavioral intention correlation surpassed the subjective norm-behavioral
intention correlation and this tendency remained persistent after controlling for the
potential moderating effect for subject types (i.e., student vs. non-student sample),
method of inquiry (i.e., survey vs. experiment), or the discipline of study (i.e., social
psychology vs. consumer marketing).

Eagly and Chaiken (1993) maintain that this pattern of data makes mathematical
sense when assuming that subjective norm influences behavioral intention indirectly via
the attitude components (see also Stitt, 2004). The Separate Process Model (see Figures
17 and 18) corroborates this Eagly and Chaiken’s conjecture to the extent that perceived
problem severity and perceived social sanction can be considered as indicative of
participant’s attitudes on drinking. As illustrated in the model, the subjective norm of
one's parents had a relatively smaller impact on behavioral intention than perceived
problem severity, and similarly, the subjective norm of a community produced a minor
influence on projected behavioral intention than did the perceived social sanction. Also,
both subjective norm factors exerted a greater impact on the quasi-attitude variables,
while affecting behavioral intention substantially less.

The debate seems to still be continuing on whether to see subjective norms and
attitudes as two separate factors making distinctive contribution to behavioral intention

(Ajzen, 1985; Fishbein & Ajzen, 1975) or to view subjective norms as a determinant of

76

attitudes (Eagly & Chaiken, 1993; see for review Dillard & Pfau, 2002; O’Keefe, 2002).
Meanwhile, this study documents another evidence favoring the latter view.
Holists and Perception of Roles

This study predicted that a role norm appeal would induce greater attention from
holists than analysts for manifesting a higher context dependency (Hall, 1976; Nisbett,
2002) and a greater role dependency in self-perception (Cousins, 1989; Cross, Kanagawa,
Markus, & Kitayama, 1995: Lim, 2002; Lim, Allen, Burrell, & Kim, 2007; Markus &
Kitayama, 1991, 1998). The current data, however, were inconsistent with this conjecture.
Participant’s role perception enhanced when exposed to the role norm appeal than when
exposed to the social norm appeal. But this tendency was common for both countries (see
Figures 17 and 18). This single null finding cannot invalidate the cumulated past findings
consistent with the theorized cultural difference. Rather, this inconsistency may well be
attributed to either the characteristics particular to this study (i.e., content effect) or
potential problems with measurement.

Perhaps the presumed cultural difference in sensitivity in role perception may
remain very subtle in reality, which may become unobservable when examined with too
crude a tool. That is, both holists and analysts may possess an equally strong role
obligation as a child but equally minimal sense of duty as a community member. In
accordance to this conjecture, Lim (2002) maintained that family constitutes the most
cohesive group in both American and Korean culture. Unlike our conventional wisdom,
family does not always function as the most fundamental group in other cultures. For
instance, the strongest relational bonding may occur at corporate level as in Japan or at

community level as in many Afi‘ican societies.

77

The U.S.-Korea difference in role dependency may reside in relational domains
between the two extremes, for example, the relationship with siblings, friendship,
romantic interests, casual classmates, or a neighbor living next door. Future experiments
may be able to find the predicted cultural difference by exposing participants to more

diverse relational contexts.

78

Appendix A Message Stimuli (English)

1. Role Norm Message Condition

Binge drinking among college students is a family problem. Every year, more
than 696, 000 students are physically assaulted by another who has been drinking and no
less than 97, 000 are victims of alcohol-related sexual assault or date rape (W echsler et
al., 2002; William, 2002). About 1,700 college students die each year fiom alcohol-
related injuries (Hingson et al., 2005). Behind the students who sufler are their parents
who must go through the same consequences together. In other words, parents of 696, 000
college students experience the consequences of their children’s injuries and 97, 000
families fall victim of sexual assault. And every year, parents of l , 700 college students
are losing their children to alcohol-related accidents. Your parents have been sacrificing
themselves to raise you for almost twenty years. Now, they just want you to be safe and
healthy. Drink responsibly. Responsible children do not make their parents sufier!

11. Social Norm Message Condition

Heavy-drinking college students not only risk their own health, but also
jeopardize the well-being of other members of their community. One out of every four
college students who drink report having forgotten where they were or what they did
during the school year. The incidence of blackout was doubled (54%) among fiequent
binge drinkers (Wechsler et al., 2000). A study shows that being rude toward or picking a
fight against innocent people, damaging public properties, or committing sexual assault
were among the most common behaviors students do while blacked out (Kim, 2008;
Wechsler et al., 2002). Most importantly, binge drinking can kill others. A total of 4, 55 3
people were killed in 2001 in crashes involving 18-24-year-old drunk drivers, and as
many as 46 percent of them were innocent victims (Hingson et al., 2005). Drink
responsibly. Responsible community members do not make other fellow members sufirer.’

III. No-norm Control Condition

It is important that college students drink responsibly and avoid binge drinking.
Evidence indicates that binge drinking causes many health related problems. A recent
study (Hingson, et al., 2005) found out that, binge drinkers are eight times more likely
than non-binge drinkers to get hurt or injured. In the longer term, research shows that
binge drinking during one ’s college days substantially increases the chances of long term
alcohol abuse and dependence, which in turn, is associated with numerous health
problems such as liver disease (Hingson, et al., 2005). Moreover, binge drinking causes
education problems too. In particular, excessive consumption of alcohol is associated
with lower academic performance (Wechsler et al., 2002). Poor grades or, worse yet, a
criminal record (even for a minor oflense; Hingson, et al., 2005) make it more difi‘icult to
find a good job after graduation. Do not binge drink Protect yourself?

79

Appendix B Message Stimuli (Korean)

I. Role Norm Message Condition

CHItgi AlOIOIIHQJ mitt SZFE- OIXII 7189.! EXIIEJLJEL nH'd 633,000 %
OIQQJ EHIt’ét’EOI #83 file ttgiml 9J6H $21.18 $31.71 21204. 9r 88,000 04%94
ttgim 8258-3 gamer 111611! 213'. 229L412}. if}, 9r 1,500 l$9.! ttgim IJHIei
sear: Ami Alanna. 0W?“ S25E14 Alli nlsfll 21E #85 Halli are
138 mil 2510M 8H5 #22150! net—Int. CHI INH,633,000 04 twist IFELEIEOI
IlLial 811E agatm, 88,000 01 7l§§0| giitsl nltflxm Ell flaunt. Stall DH'rj
1,500 04%94 'FE‘EJEOI seat! A112 xlull 243'. Stet—IE}. 04345391 #2218 XIH
2015 52* 046158 9I$7l $18" 'éttli asaaeuu. r3024 Irla‘alol xla Blair? 71%
Evil 9021 mm anal—act. W15 auxin. 81%]? 2.35 XifilE $22M 13% 2171!
WI Stet—IE}.

11. Social Norm Message Condition

3:83 Efltrwi 74021 m; Eel OlL—lal :1 xl°42FE|iQJ arias anal—In.
82.5! §7IE CHIPS HI a 3 a rag #Iéotl IHJOI OICIOII 2,1225% 2 $918 new
7I9-16‘rxl iflflz Emu}. Olaltt El! 37:! {date as m5 UMIE tigi MOIOIIA-I 9!
1? HH :1 WF LlF-l‘dL—IEL a 93-? anion mag. $13 Malia gull; Stall
WHEN—l AIHII 74E 31. as Alt-ii mil, JELT'. gifiOI ttfiiOl 31-“: ENE We": 3
Mt {it 7,12; ZEMEI‘zteL—ICL :4 320 2:18, £89.; alt" El! AlEtEOI ii 4‘- E
2MB awn. XIH 2001 +1, 18-24 Idl Mole! 82587517471 ENE 4115an 301 IEDI
ass 2421204. :1 E 46%? ii DMIXI eta Ataiomeuu. W? was. 8196
Ste Male Olizfi'PJEOI 31.88 2171! am smut}.

III. No-norm Control Condition

Eflfl’éioilfll 8301. ii fililfilfl 38% IllfiE 31'8- éREl WEIL-lfil. OiE‘l 2.5M
ZIP-loll UrElH Hl2l ZIOI 5&8- EQOII amend. El fi-THHI 120295, #35 §7|E
Alibi?- Jixl 818 Mililll Clip-M éltfll 9.18 lllol BHH 7i?! Elf—l1 ElL-IEl. Iii.
EHtléfi Al’iil'FE-l 58% AI’S-itlE AliliE ’37I?JO-|l an Wii 'El-SOII—l éioil ail
HE’SOI 011-?- $204, 2 fiflli if HEP-l #13 042-1 N530“ §E|7| +11% 7122.
L-lEl‘deL—llll. C‘lfiol i818 I'L-S’Sifil Boll/HE tllfiel—lf—l. Itl, age ttm 55-13
XiflAIZIL—lf-l. SEOI Xiifi'lHl—l 82$ 4 fig 52E PJ‘Ei BEWII‘OI 22E 73?, EH‘ii
iii # :8- ‘BJIlEII :I’S‘rfl OifieL—ICI. 512:5 #43102. Xlt'fi “fit—MIR.

8O

Appendix C Measures (English)

Role salience (Manipulation Check)

1.

NP‘

This message points out that avoiding excessive drinking is a way of
becoming a responsible community member.

This message appeals to my role obligation as a community member.
This message argues that I avoid excessive drinking as a responsible
community member.

This message indicates that excessive drinking may prevent me from
functioning as a responsible community member.

This message points out that avoiding excessive drinking is a way of
becoming a responsible child of my parents.

This message appeals to my role obligation as the child of my parents.
This message argues that I avoid excessive drinking as a responsible child of
my parents.

This message indicates that excessive drinking may prevent me from
functioning as a responsible child of my parents.

Perceived social sanction

l.
2.
3.
4.

I feel like people would think less of me if I drank excessively.
People would disapprove of me if I drank excessively.

Our society views heavy drinkers negatively.

Society is intolerant of excessive drinking.

Locus of consequences

1.

2.

3.

4.

This message shows that excessive drinking will have consequences for
people that are close to me.

This message demonstrates the negative effects of excessive drinking that
people close to me are likely to face.

This message indicates that my excessive drinking may affect people close to
me.

This message claims that people close to me can be the potential victims of
my excessive drinking.

Behavioral intention (Selfl

l.

9‘95“?!”

I would be willing to sign a petition supporting regulation on excessive
drinking.

I will try to drink less in the future.

I will avoid excessive drinking.

If I drink, I will drink in moderation.

I would be willing to ask my friends to drink less.

I would be willing to talk to my friends about drinking issues.

81

Behavioral intention (Projection)

1. Many of those who read this message will avoid excessive drinking to the best
that they can.

2. Most people who read this message will try not to drink excessively.

3. Most people who read this message will try dissuading their significant others
from excessive drinking.

4. Most people will not quit excessive drinking because of this message.

Argument quality

1. The argument of this message is strong.

2. The argument of this message is valid.

3. The argument of this message is flawed.

4. The argument of this message is high in quality.

5. The argument of this message is weak.

6. The argument of this message is invalid.

Severity of drinking problem

1.

Excessive drinking among students leads to severe and undesirable
consequences.

2. Excessive drinking among students is a serious problem.
3.
4. Excessive drinking among students is not a serious problem.

Excessive drinking among students is dangerous.

Perceived descriptive norm

P‘S‘PP’NI“

It is commonplace that college students drink excessively.

Most college students drink excessively now and then.
Excessive drinking is prevalent among college students.

Few college students drink excessively.

In general, college students tend to drink too much on occasions.
It is usual that college students consume alcohols excessively.

Perceived injunctive norm

95”.“?

Most college students consider excessive drinking to be acceptable.

Most students object to excessive drinking among college students.

At my university, it is okay for college students to drink excessively.

Most students at my university would think that college students should not
drink excessively.

Most students at my university disapprove of excessive drinking among
college students.

82

Subjective norm (Parents)

Excessive drinking is acceptable in my family.

My parents object to excessive drinking.

My parents think it is okay that I drink excessively.
My parents think that I should not drink excessively.
My parents allow excessive drinking.

My parents disapprove of my excessive drinking.

99:“pr

Subjective norm (Community)

Excessive drinking is acceptable in my community.

My community objects to excessive drinking.

Other members of my community think it is okay to drink excessively.
Other members of my community think that one should not drink excessively.
My community allows excessive drinking.

My community disapproves of excessive drinking.

9999!”?

Motivation to comply (Parents)

It is important that I obey my parents.

It is okay for me to disobey my parents.

It is important that I do as my parents wish.

It is important that I meet the expectations of my parents.
I want to do what my parents would think is a good thing.

.U'PPNZ"

Motivation to comply (Community)

It is important that I conform to the norms of my community.

It is okay for me to violate the norms of my community.

It is important that I do as other community members would wish.

It is important that I meet the expectations of other community members.
I want to do what other community members would think is a good thing.

MPPNT‘

83

Demographics

1.

2.

I am Male / Female (please circle)

I am years old

. In which country were your born and raised? Please choose the continent that

it belongs to. Please specify your country name in the blank after “other” if
unsure.

North America (i.e., the US. or Canada)

Latin America (e.g., Mexico, Brazil, Argentina, etc.)
Europe (e.g., England, Germany, Italy, etc.)

Asia (e.g., China, Korea, Japan, Hong Kong, etc.)
Middle East (e.g., Iran, Iraq, India, Saudi Arabia, etc.)
Africa (e.g., Kenya, Tanzania, South Africa, etc.)
Australia

Other:

P“!° rm 9.0 9‘!»

 

84

Appendix D Measures (Korean)

Role salience (Manipulation Check)

1.

0| 3% £88 Sill Eli:- Z‘JOI ’Wélél 22E X|°—’l §§ii|2| ‘ilfiol 51E El Elfigfi
XIHEHZL

2. 0| 313- UP} 7091 E‘Siil—Ql ‘Hflgékl Xlfiol if 94$Oil iiéll 22121.

0| E13, Mm 8193 21E Xl9-1 S‘SiiI—‘ll ‘iJflOIElE ifii II3H0|= EH11 WSW-11.

4. 0| 38, Xlth' 82F; XI‘EI SSiIIBI ilflgikl SHOII fl Qiili Cl 3170 5 fl 43

9’35”.“

5 Elf—LE 30343111.

0| 318- ‘88 IIHQEH XH‘JOI ’Sfi‘élél Eli-:- Xlél‘élfi Eoii 45 “Eli! Xl’ildll.
0| 5‘8 HM 11412;)?! X|9=10l if 94510“ 13:313. 2151.

0| 313-, UN 511819 221% XlNOIP—H’J 3:83 EISHOI EH11 4533?}.

0| 38, IILH‘J 845E. ”NEE/H 8110|= it ‘fifii El 3W f I’ 4= 5 2,1111
Xl’i'lfllil.

Perceived social sanction

1.
2.
3.
4.

LH7l iii 316, 350-! Niliol LlOil EH8" i-XI 8113 2'83 2!?“ i 7L'OIE1.
AliliE LH7l 5:23P:- Z'di S-E'fiXI Eli ZI'OIEL

‘FEI Al’EIE iSfiE Alilii #21123 15131551.

0| MERE iii 833W Elia-El.

Locus of consequences

PSPN.‘

0| 5‘19. 8:SOI L12} 7l77h‘2- Affilioilfll 77W "F’S’ilfil 85'2‘53 ’5-‘MIE'E1.

0| 38, Lisa} 7177112- kliliol 2117“ ill BEE Iii—9| Ql‘éfil‘ii Eoiéfl.

01 38, L121 a:EOI Urfll 717711? Alilioilfll Qi‘is'fiii § -’:‘- Elf—131 Xl’SllEif-l.

0| 5‘18, L121 7177118 Aliliol L121 3:822 PJSH BISHI I -’:‘- 8151.7. 455111121.

Behavioral intention (Self)

99.90.03

LHE ‘83 7073181301? gllfi IIII‘BIE BEA-Ml kl‘fiil QJQ'OI 2.1121.
Ll-E 92E 3% fiolflifl 51:34? 720M.

Lh'E 3:38 mil 710M.

3% W401 SCH! H9331 UMIZi'E-l.

LlE LH fl-‘r‘ioilfll ii §0|ELT1 ilil gzlol 22111.

L—lE JEEP—l E1141?! 0|¢rioil 33H {Will 0|0|=7|il QIQ‘OI 21C}.

85

Behavioral intention (Projected)

1. EH3 Miliol 0| ii 941 if 4> 21E El #8317“ 912311 51:11? 710M.
2. 0| 38 848 EH'FEP—l Affiliol 3:85 EMF—11 i451? 710M.

3. 0| ii 948 Affilifi- Xlt'lll EEOI 921-“: Miiioilfll 3:83 Fifi? atom.
4. EH’FEQJ ”@513 0| '38 S—HEELT’. £88 éEffilXIE ‘33 310E}.

Argument quality

0| 3.9-1 ‘FEE NEHOI Q3111}.

Ol 32! $38 EEI’EIEE. ElQS‘rEl.

0| EQJ 5&8 EH71 QJXI SEC}.

0| 38- ‘23519-1 5&8 113. 53,1111.

0| 39-! $38- N‘ifiol 931121.

0| 321 3588 EEI’ELQE 5193le 811121.

P’S’IPP’N.‘

Severity of drinking problem

1. EH‘léfli-‘B-l 3:813 am Hlill‘lé'lxl 93E BID-ll ZEEHEEL
2. Efltlgifil “(ES $17-13 Salt}.

3. Ellilfiifll KEE- fillfifiltl.

4. “91’6"in iifiol £1113 ERIE- OlL—ltl.

Perceived descriptive norm

. EHQl’éfiiOI $8311? 7:13 i °'E ‘ilOIEl.

. EH’FEQ—l 121131358 §§ as 14817“ Ult'El.

. ififi- EH‘El’é‘i AlOIOIIA-l 3113515101 Silt}.

. ”fifth?- Efltl’gie 7121 22121.

. Elltlgifi‘ £3 8% Li? 5113171! DMIE @Q‘OI Elf—l.
. Elltlgiol ii El'é‘flil DMIE 728 8183421 <ilollll.

arm-th-s

Perceived injunctive norm

1. EHJPEQJ Elltlgie 3:33 Eli 4* 21E <ilOIElI’. EC}.
2. EHW‘EQJ W858 Ellil’SQ—l ifioil Elf-HEEL
3. $El ELENA-IE- tl’fiiol 3:88 SHE Emil.
4. $El Email ClL—IE EH‘FEQI ”SEE- Efltl’giol 5:35 Silk-IE El EELE QZlElE-l.
5. $El tLIPJHI ElL—IE- EH‘T‘EP-l ”€58 EHEléfii AlOIOilkl—Ql 33% 893W EIEEL

86

Subjective norm (Parents)

1. $El EJO‘IIME a(Sol 88153111.

2. $EI ‘FE‘EJE ifioil 312113113121.

3. $31 ‘i‘E‘alE LH7l 8% 31317“ 01445 Elfiltlfl gzia‘rtllll.

4. $El ‘FE‘EIE- U171 33 Elfi'fll lelklr‘f 21 Elf—ll gzltlfltl.
5. $EI $22118- ifii SElfit'C-l.

6. $31 ’FE'EJE ‘88- fixl 9.18- EEOIELT'. it'll

Subjective norm (Community)

1. $EI Xl‘l’l Siiiloilklf‘: itfiol 89.117551.

2. SF—u‘il X|°—‘l EEfiIE 3:801 EiEHEHZl.

. $31 X|°—‘l $9.55 88 3131711 01415 HSEL'F. ngEiE-l.

4. SPEI XIQ-‘l 359.158- 33 311317“ DWIHE El Elf-13. "$113111.
5. $EI X|94 Séflloilklr'f 3:88 318351.

6. SPEI X191 géiiloilkli‘f 388 fiXI BIB QEOIELE Sci.

(D

Motivation to comply (Parents)

1. #2151771! &fifiE 728 59?! 9.101313. $113121.

. $215M Eéfiixl 3015 Emilia ’eflZtEHZl.

. FIE'EIOI HlElE CHE fiSfiE 7:413 $9.3 Wolf-111 $11351.
. ‘FE'EIP—l 711211011 $8311? 728- LH7il éfifilf—l.

. Lit- ‘FE‘EIOI £513. gzlfilbli'f 713 5111131313. 5543f}.

011th

Motivation to comply (Community)

XI‘ZI 858191 REE [GEE 713 $9.3 SJOIELE $113121.

XI?! 258121 WES 01745 EST-113'. ’éflZlEHT-l.

Eh! Xl°—‘l EF'FJEOI HlilEiEi $58 We 31'3- éfifl S’JOIELE ’ngEHT—l.
El! Xl‘l’l 2511594 7|EHO|| 5183115 718 LH7-ll 39.31111.

'41? El! Xl‘l‘l ’F'PJEOI Elf-ll ngSl-‘E Edi 1115131311 13513121.

9995’?

87

Demographics
1. it‘lfilQ-I 29%: 323' / 0%; ($93115 foil 0 .E. 3113MB)
2. i‘i'B'iQ-l LlOI: El _ HI

3. H3115 01!: LlEloilkl EH01L111 anew»? Sil‘él'a‘rE foil 0 i 31¢Ml2. 01211 E71
é 311%311'5 i‘OI Bit-:- 75‘-?. 7lElEioi| :l LlElSll Olfii ’5-1’51 342MB ELIE-l.

71. Eli
L1. 551‘
Cl. ‘31!
El. 7|El:

 

88

References

Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl & J.
Beckmann (Eds.), Action control: From cognition to behavior. Berlin,
Heidelber, New York: Springer-Verlag.

Bachnik, J. M. (1994). Introduction: Uchi/soto: Challenging our conceptualizations of
self, social order, and language. In J. M. Bachnik & C. J. Quinn, Jr. (Eds),
Situated meaning: Inside and outside in Japanese self society, and language
(pp. 3-3 7). Princeton, NJ: Princeton University Press.

Banton, M. (1965). Roles: An introduction to the study of social relations. New York:
Basic Books. Inc., Publishers.

Birenbaum, A., & Sagarin, E. (1976). Norms and human behavior. New York: Praeger
Publishers, Inc.

Blumer, H. (1969). Symbolic interactionism: Perspective and method University of
California Press.

Brehm, J. W. (1966). A theory of psychological reactance. New York: Academic Press.

Brehm, S. S., & Brehm, J. W. (1981). Psychological reactance: A theory of fieedom and
control. New York: Academic Press.

Brislin, R. W. (1980). Translation and content analysis of oral and written materials. In H.
C. Triandis & J. W. Berry (Eds.), Handbook of cross-cultural psychology (Vol.
2, pp. 389-444). Boston: Allyn & Bacon.

Cialdini, R. 8., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative
conduct: Recycling the concept of norms to reduce littering in public places.
Journal of Personality and Social Psychology, 58, 1015-1026.

Cousins, S. D. (1989). Culture and self-perception in Japan and the United States.
Journal of Personality and Social Psychology, 56, 124-131.

Cross, S. E., Kanagawa, C., Markus, H. R., & Kitayama, S. (1995). Cultural variation in
self-concept. Unpublished manuscript. Ames: Iowa State University.

89

Davison, W. P. (1983). The third-person effect in communication. Public Opinion
Quarterly, 4 7, 1-15.

DeVos, G. A. (1973). Socialization for achievement: Essays on the cultural psychology of
the Japanese. Berkeley, LA: University of California Press.

Dershowitz, Z. (1971). Jewish subcultural patterns and psychological differentiation.
International Journal of Psychology, 6, 223-231.

Dillard, J. P., & Pfau, M. W. (2002). The persuasion handbook: Developments in theory
and practice. Thousand Oaks, CA: Sage.

Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX:
Harcourt Brace Jovanovich College Publishers.

Fan, X., Thompson, B., & Wang, L. (1999). Effects of sample size, estimation method,
and model specification on structural modeling fit indexes. Structural Equation
Modeling, 6, 56-83.

Farley, J. U., Lehmann, D. R., & Ryan, M. J. (1981). Generalizing from “imperfect
replication.” Journal of Business, 54, 597-610.

Firth, R. (1958). Human types. New York: New American Library.

Fishbein, M., & Ajzen, l. (1975). Belief attitude, intention, and behavior: An
introduction to theory and research. Reading, MA: Addison-Wesley.

Hall, E. T. (1976). Beyond Culture. Garden City, NY: Anchor Press.

Hamaguchi, E. (1977). 'Nihon rashisa’ no sai hakken [Rediscovery of 'Japaneseness 7.
Tokyo: Nihon Keizai Shimbunsha.

Homans, G. C. (1950). The human group. New York: Harcourt, Brace and Co.

Ip, G. W., & Bond, M. H. (1995). Culture, values, and the spontaneous self-concept.
Asian Journal of Psychology, 1 , 29—35.

90

Malinowski, B. (1960). A scientific theory of culture and other essays. New York:
Oxford University Press.

Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implications for cognition,
emotion, and motivation. Psychological Review, 98, 224-253.

Markus, H. R., & Kitayama, S. (1998). The cultural psychology of personality. Journal of
Cross-Cultural Psychology, 29, 63-87.

Markus, H. R., Mullally, P. R., & Kitayama, S. (1997). Selfways: Diversity in modes of
cultural participation. In U. Neisser & D. A. Jopling (Eds.), The conceptual self
in context: Culture, experience, self-understanding (pp. 13-61). Cambridge:
Cambridge University Press.

Masuda, T., & Nisbett, R. E. (2001). Attending holistically vs. analytically: Comparing
the context sensitivity of Japanese and Americans. Journal of Personality and
Social Psychology, 81, 922-934.

McAdams, R. H. (1997). The origin, development, and regulation of norms. Michigan
Law Review, 96, 338-433.

Mead, G. H. (1934). Mind, self and society. Chicago: University of Chicago Press.

Morley, D. (1987). Subjective message constructs: A theory of persuasion.
Communication Monographs, 54, 183-203.

Michigan State University. (2008). Michigan State University’s social norms program
model. Retrieved April 20, 2009, from http://socialnorms.msu.edu/

Morris, M. W., & Peng, K. (1994). Culture and cause: American and Chinese attributions

for social and physical events. Journal of Personality and Social Psychology,
67, 949-971.

Munro, D. J. (1985). Individualism and holism: Studies in Confircian and Taoist values.
In D. J. Munro (Ed.), Individualism and holism: Studies in Confucian and
Taoist values (pp. 1-32). Ann Arbor: Center for Chinese Studies, The
University of Michigan.

91

Muthén, L. K., & Muthén, B. 0. (1998-2004). Mplus user ’s guide (3rd ed.). Los Angeles:
Muthe’n & Muthén.

Nadel, S. F. (1957). The theory of social structure. Glencoe, IL: The Free Press.

Nisbett, R. E. (2003). The geography of thought: How Asians and Westerners think
differently...and why. New York: The Free Press.

Nisbett, R. E., Peng, K., Choi, I., & Norenzayan, A. (2001). Culture and systems of

thought: Holistic versus analytic cognition. Psychological Review, 108, 291-
310.

O’Keefe, D. J. (2002). Persuasion: Theory and research (2“d ed.). Thousand Oaks, CA:
Sage.

Piddocke, S. (1968). Social sanctions. Anthropologica, 10, 261-285.

Powell, F. A. (1965). The effect of anxiety-arousing messages when related to personal,
familial, and impersonal referents. Speech Monographs, 32,102-106.

Paul, B., Salwen, M. B., & Dupagne, M. (2000). The third-person effect: A meta-analysis
of the perceptual hypothesis. Mass Communication & Society, 3, 57-85.

Rakov, T., & Marcoulides, G. A. (2000). A first course in structural equation modeling.
Mahwah, NJ: LEA.

Reno, R. R., Cialdini, R. B., & Kallgren, C. A. (1993). The transsituational influence of
social norms. Journal of Personality and Social Psychology, 64, 104-1 12.

Rhee, E., Uleman, J ., Lee, H., & Roman, R. (1995). Spontaneous self-descriptions and
ethnic identities in individualistic and collectivistic cultures. Journal of
Personality and Social Psychology, 69, 142-152.

Rimal, R. N., & Lapinski, M. K. (2005). An explication of social norms. Communication
Theory, 15, 127-147.

Snyder, M., & Gangestad, S. (1986). On the nature of self-monitoring: Matters of
assessment, matters of validity. Journal of Personality and Social Psychology,

92

James, W. (1950). The principles of psychology. New York: Dover Publications.

Johansson, B. (2005). The third-person effect: Only a media perception? Nordicom
Review, 26, 81-94.

Kallgren, C. A., Reno, R. R., & Cialdini, R. B. (2000). A focus theory of normative
conduct: When norms do and do not affect behavior. Personality and Social
Psychology Bulletin, 26(8), 1002-1012.

Kashima, E. S., & Kashima, Y. (1998). Culture and language: The case of cultural
dimensions and personal pronoun use. Journal of Cross-Cultural Psychology,
29, 461-486.

Kashima, Y., Kashima, E., F arsides, T., Kim, U., Strack, F ., Werth, L., et al. (2004).
Culture and context-sensitive self: The amount and meaning of context-
sensitivity of phenomenal self differ across cultures. Self and Identity, 3, 125-
141.

King, A. Y. C. (1985). The individual and group in Confucianism: A relational
perspective. In D. J. Munro (Ed), Individualism and holism: Studies in
Confucian and Taoist values (pp. 57-70). Ann Arbor: Center for Chinese
Studies, The University of Michigan.

Lebra, T. S. (1976). Japanese patterns of behavior: Honolulu: University of Hawaii Press.

Lehmiller, J. J. (2008). Conducting relationships research online: A primer on getting
started and making the most of your resources. Relationship Research News, 7,
6-12.

Lim, T. (2002, July). Inside in, outside out: The dichotomous nature of Korean
interpersonal dialogues. Paper presented at the annual meeting of the
International Communication Association, Seoul, South Korea.

Lim, T., Allen, M., Burrell, N., & Kim, S. (2007). The relativity and salience of identity
across cultures. Speech and Communication, 8, 178-202.

Little, R. J ., & Rubin, D. B. (2002). Statistical analysis with missing data. New York:
Wiley.

93

51, 125-139.

Steenkamp, J ., & Baumgartner, H. (1998). Assessing measurement invariance in cross-
national consumer research. The Journal of Consumer Research, 25(1), 78-90.

Stitt, C. R. (2004, May). The influence of social norms on attitudes: Considering a
modified value-expectancy approach. Paper presented at the annual meeting of
the International Communication Association, New Orleans, LA.

Sustein, C. R. (1996). Social norms and social roles. Columbia Law Review, 96, 903-968.

Triandis, H. C. (1995). Individualism and collectivism. Boulder, CO: Westview.

UMADD. (2009). Statistics. Retrieved February 12, 2009, from
http://www.umadd.org/Home/Statistics.html

Wolburg, J. M. (2006). College students’ responses to antismoking messages: Denial,
defiance, and other boomerang effects. The Journal of Consumer Aflairs, 40,
294-323.

Yonhap News. (2009, April 15). Report shows 17pct of Seoul National University
students at risks of alcoholism. Yonhap News. Retrieved from
http://english.yonhapnews.co.kr/news/2009/ 04/ 1 5/0200000000AEN2009041 50
03400315.HTML

94