Journal of Personality and Social Psychology Copyright 1996 by the American Psychological Association, Inc.

1996, Vot. 70, No. 5,996-1011 0022-3514/96/S3.00

More Than Words:
Li ngui sti c and Nonl i ngui s t i c Pol i teness i n Two Cul tures
Nal i ni Amba dy and Jasook Koo
Harvard University
Fi ona Lee
University of Michigan (Ann Arbor)
Rober t Rosent hal
Harvard University
This article extends the validity of politeness theory ( P. Brown & S. Levinson, 1987 ) by investigating
the nonlinguistic aspects of politeness in 2 cultures. Politeness strategies expressed through different
channels of communication (silent video, speech, full-channel video and audio, and transcripts of
speech ) were examined, and it was found that politeness strategies were communicated nonlinguis-
tically as well as linguistically and that nonlinguistic strategy usage was related to social and
contextual factors. Two studies revealed that Koreans' politeness strategies were influenced more by
relational cues, whereas Americans' strategies were influenced more by the content of the message.
This research represents a 1st attempt to explore the nonlinguistic communication of politeness
across cultures.
Consider the following scenario (familiar to some): Mary, a
graduate student getting ready to face a dismal job market, re-
ceives a letter informing her that a paper coauthored with her
advisor, a fellow graduate student, and a senior undergraduate
has just been accepted by a prestigious journal with very few
revisions. Mary rushes to share the good news with her coau-
thors. Consider another alternative: Mary reads that the article
has been rejected. She knows that the paper will need major
revisions if it is to be accepted in any journal. She has the un-
pleasant task of conveying the news to her coauthors. How will
Mary convey the good and bad news? And how will she convey
the news differently to her advisor, to her peer, and to the
undergraduate?
Pol i t eness Theor y
These are interesting questions for social psychologists be-
cause the manner in which such messages are communicated
Nalini Ambady, Jasook Koo, and Robert Rosenthal, Department of
Psychology, Harvard University; Fiona Lee, Department of Psychology,
University of Michigan (Ann Arbor).
Authors' names are listed alphabetically. Preparation of this article
was funded by the Bayer Foundation, the James McKeen Cattell Fund,
a Knox Bequest Research Grant from the Harvard University Depart-
ment of Psychology, and a Faculty of Arts and Sciences Sabbatical
Award from Harvard University.
Portions of this research were presented in a poster session at the
meeting of the American Psychological Society, June 1994, Washing-
ton, DC. We thank Joyce Kuo and Eva Ho for their assistance in col-
lecting the data.
Correspondence concerning this article should be addressed to Nalini
Ambady, Department of Psychology, Harvard University, 33 Kirkland
Street, Cambridge, Massachusetts 02138. Electronic mail may be sent
via the Internet to na@wjh.harvard.edu.
might provide information about Mary' s perceptions of her in-
terpersonal relationships and her construal of her social envi-
ronment. Politeness theory (E Brown & Levinson, 1987), a
"social psychological theory of language usage" (Holtgraves &
Yang, 1990, p. 719 ), suggests that Mary, when confronted with
the news of the acceptance or the rejection, can take five possible
courses of action. Labeled as superstrategies, these courses of
action, in increasing order of politeness, are (a) stating the news
baldly on record, with no politeness to couch the message (e.g.,
just saying "The paper has been accepted!" ); (b) displaying pos-
itive politeness by indicating solidarity with the target ("This is
a terrific paper, but they made a mistake and rejected it. Let' s
work on it and I' m sure it will get i n"); (c) displaying negative
politeness by respecting the target's freedom of action, using
restraint, and being formal and self-effacing ("I' m awfully sorry
the paper was rejected. It is my fault; I should have added the
additional data. Could you suggest what I should do next?");
(d) stating the news offthe record by hinting or using metaphor,
irony, or understatements ("Perhaps another journal would bet-
ter fit our needs"); or, (e) in the most polite case, not conveying
the message at all.
Which of these superstrategies might Mary adopt? Politeness
theory asserts that the strategies individuals use depend on the
nature and context of the social relationship between the
speaker and the target. Specifically, politeness theory suggests
that speakers will be more polite when (a) the relative power
of the target over the speaker increases, (b) the social distance
between the target and the speaker increases, and (c) the degree
of imposition on the target increases. Politeness theory predicts
that Mary will use one oft be more polite superstrategies when
she has to convey bad news (because bad news is more imposing
than good news) and when she is conveying the news to her
advisor (the person who has the greatest relative power over
Mary and probably the most social distance). Conversely, the
996
NONLINGUISTIC POLITENESS 997
t heory predicts t hat Mar y will use one of the less polite super-
strategies when she has t o convey good news t o a peer or t o an
undergraduate.
The validity of politeness t heory across different cultures has
been demonst rat ed in anthropological (Baxter, 1984; R. Brown
& Gilman, 1989; Smith-Hefner, 1988) as well as experimental
social psychological studies (Holtgraves & Yang, 1990, 1992).
For example, Holtgraves and Yang used vignettes in which a
speaker made a request t o a target. They mani pul at ed the power
of the target over the speaker, the social distance between the
target and the speaker, and the size of a request. Ameri can and
Korean participants rated the likelihood t hat different polite-
ness strategies would be used in these vignettes (1990) or indi-
cated exactly what t hey would say t o make the request (1992).
Overall, the results validated the effect of power and size of im-
position on politeness and provided mixed support for the dis-
tance variable. These results provided some evidence for the lin-
guistic universality of politeness theory: Although Ameri cans
and Koreans weighed power and distance differently in their po-
liteness strategy usage patterns, politeness t heory provided a
valid explanation of strategy usage in bot h cultures. The present
study expanded the current research literature on politeness
t heory by posing two research questions: Can politeness t heory
be extended t o nonlinguistic channels of communi cat i on? and
Is the nonlinguistic communi cat i on of politeness universal or
cul t ure specific?
Nonl i ngui s t i c As pect s o f Pol i t eness
Although the sociolinguistic research has cont ri but ed greatly
t o refining politeness theory, one potentially i mport ant aspect
of politeness t heory has been ignored in the literature. P. Brown
and Levinson (1987) speculated t hat politeness t heory extends
t o nonverbal behavior; t o our knowledge, however, no studies
have exami ned the nonlinguistic aspects of politeness theory.
This neglect of nonverbal behavior is surprising. First, much of
day-to-day communi cat i on is not purely verbal or linguistic in
nature; for example, face-to-face communi cat i on and tele-
phone conversations bot h include a mix of linguistic and non-
linguistic communi cat i on cues. Second, politeness t heory orig-
inated out of Erving Goffman' s (1967) conceptualization of
"facework," or communi cat i ve strategies used t o preserve the
self-esteem and "face" of bot h the communi cat or and the
target. Goffman suggested t hat a great deal of facework occurs
t hrough expressive behavior and nonverbal channels of com-
munication. Interpretation of polite utterances t o save face,
therefore, depends not only on the semantic cont ent of the ut-
terance but also on the accompanyi ng nonverbal cues- - vocal ,
kinesic, and f aci al - - t hat contextualize the utterance. Consider
the earlier example of Mary telling her coaut hors about a rejec-
tion. She might say, "Well, guess what? The editors said ' n o ' -
looks like it' s back t o square one!" with a confident vocal tone,
directly gaze at the listener, shrug her shoulders, and smile, or
she could say the same thing with a downcast gaze, a hesitant
t one of voice, and no smile. Even t hough the linguistic cont ent
is the same, the two scenarios will be interpreted quite differ-
ently. I n the first case, Mar y appears optimistic t hat revisions
will improve the paper and t hat the rejection and upcomi ng re-
visions are just "par t of life" and do not necessarily have much
bearing on her own or her coaut hors' perceived competence. In
the second case, Mary appears much more disappointed about
the rejection, pessimistic about the paper' s future chances of
acceptance, and even dismayed t hat more work has t o be done.
The present study moved beyond a purely linguistic t heory of
politeness to exami ne the communi cat i on of politeness t hrough
nonlinguistic channels. P. Brown and Levi nson' s (1987) work
on politeness described the linguistic strategies individuals use
t o communi cat e face threats and exami ned how power, dis-
tance, and size of imposition affect strategy usage. The present
study at t empt ed t o explore the linguistic as well as nonlinguistic
cues and strategies used t o communi cat e politeness. Specifi-
cally, we exami ned politeness strategy usage in four different
channels commonl y used in past nonverbal research: a purely
nonlinguistic channel (silent video), a purely linguistic channel
(transcripts of speech ), and two channels with mixed linguistic
and nonlinguistic cues (full channel, including video and
speech, and speech alone). The exami nat i on of different chan-
nels provided information about ( a) the types of politeness
strategies t hat are t ransmi t t ed nonlinguistically, ( b) whether
linguistic and nonlinguistic channels of communi cat i on
t ransmi t different types of politeness strategies, and (c) whether
nonlinguistic strategy usage is related t o the social and
contextual relationship between the speaker and the target.
Types of Nonlinguistic Strategies
E Brown and Levinson (1987) proposed t hat the superstrat-
egies (bald on record, positive and negative politeness, off the
record, and no communi cat i on) are the five possible ways t o
linguistically communi cat e face-threatening information, and
specific politeness substrategies are subsumed under each su-
perstrategy. Will these same types of superstrategies and sub-
strategies emerge when nonlinguistic cues are also taken i nt o
account ? There are several reasons t o suggest t hat analysis of
nonlinguistic strategies will yield a slightly different typology of
politeness strategies. First, Brown and Levinson designed their
typology of strategies exclusively for linguistic communi cat i on,
and most linguistic rules t hat govern interpersonal communi ca-
tion do not necessarily generalize t o nonlinguistic domains. Sec-
ond, although many of Brown and Levi nson' s linguistic sub-
strategies can be communi cat ed bot h linguistically and nonlin-
guistieally (e.g., being deferential or. apologetic), some of the
linguistic substrategies do not have nonlinguistic counterparts.
For example, one linguistic substrategy speakers use t o couch a
request is t o impersonalize the request by omitting first and sec-
ond person pronouns or nominalize the request by using the
passive voice. Obviously, such strategies cannot be communi -
cated nonlinguistieally. Third, there is evidence t hat the five
types of superstrategies, as described by Brown and Levinson,
are not as clearly differentiated as once assumed. For example,
recent research (Craig, Tracy, & Spisak, 1986; Li m & Bowers,
1991; Scollon & Scollon, 1983 ) suggests t hat positive and nega-
tive strategies may not be mutually exclusive categories but t hat
people often use a mix of bot h types of strategies. Furthermore,
the typology and organization of strategies may be different for
different social situations.
998 AMBADY, KOO, LEE, AND ROSENTHAL
Al l of t hi s evidence suggests t hat when nonl i ngui st i c channel s
of communi cat i on are t aken i nt o account , pol i t eness strategies
will not strictly conf or m t o t he t ypol ogy of superst rat egi es and
substrategies as pr oposed by P. Brown and Levi nson ( 1987) .
However, because no past studies have expl ored nonl i ngui st i c
politeness, we cannot propose any concret e pr edi ct i ons about
t he st r uct ur e of nonl i ngui st i c pol i t eness strategies. Therefore,
we first at t empt ed t o del i neat e t he strategies expressed t hr ough
linguistic and nonl i ngui st i c channel s of communi cat i on (i. e. ,
What t ypes of superst rat egi es will be t r ansmi t t ed when nonl i n-
guistic channel s are t aken i nt o account ?) .
How Different Channels Communicate Different
Strategies
The most pol i t e superstrategy, ot her t han omi t t i ng t he mes-
sage altogether, is communi cat i ng t he message off t he record.
Accordi ng t o E Brown and Levi nson ( 1987) , off-t he-record
strategies allow speakers t o say t hi ngs " i n such a way t hat it is
not possi bl e t o at t r i but e only one cl ear communi cat i ve i nt en-
t i on t o t he act," and t hese strategies are "essent i al l y i ndi r ect uses
of l anguage" (p. 211 ). Off-t he-record strategies are par t i cul ar l y
amenabl e t o nonverbal channel s of communi cat i on. Nonver bal
or nonl i ngui st i c behavi or is ext r emel y subt l e and is often used
t o communi cat e t hought s and feelings t hat cannot be commu-
ni cat ed verbally. Such behavi or is oft en used t o convey t he
" t r ue " meani ng of a st at ement , for exampl e, smi l i ng when mak-
ing a sharp r emar k t o show t hat t he r emar k was a j oke. Thus,
nonl i ngui st i c cues allow t he speaker t o hi nt , imply, or suggest
t he i nf or mat i on i ndi rect l y t hr ough expressi ons or t one of voice,
in effect faci l i t at i ng off-t he-record communi cat i on o f t h e " t r u e "
message. Our first hypot hesi s was t hat off-t he-record pol i t eness
strategies woul d most oft en be conveyed t hr ough nonl i ngui st i c
channels.
Relationship Between Strategy Usage and Social-
Contextual Factors
P. Brown and Levi nson ( 1987) pr oposed t hat communi ca-
t or s' choice of superst rat egi es ( i n i ncreasi ng or der of politeness:
bal d on record, positive politeness, negative politeness, off t he
record, and no communi cat i on) depends on power, di st ance,
and level of t he i mposi t i on. Specifically, as power, di st ance, and
i mposi t i on increase, i ndi vi dual s will use higher level superst rat -
egies. For exampl e, pol i t eness t heor y suggests t hat negative
strategies are mor e "pol i t e" t han positive strategies and t hat in-
di vi dual s will use mor e negative strategies t han positive st rat e-
gies when speaki ng wi t h a powerful person. Recent research,
however, has pr ovi ded equi vocal evidence suppor t i ng Brown
and Levi nson' s ( 1987) hi er ar chy of superstrategies. For exam-
ple, i ndi vi dual s do not necessari l y use mor e negative strategies
t han positive strategies under si t uat i ons in whi ch t heoret i cal l y
mor e pol i t eness is r equi r ed ( Cr ai g et al., 1986; Fi el d, 1991; Lee,
1993; Li m & Bowers, 1991; Scollon & Scollon, 1983). On t he
ot her hand, t here is consi derabl e evidence t hat communi cat or s
si mpl y use mor e el abor at e strategies t o i ndi cat e mor e pol i t eness
(R. Brown & Gi l man, 1989; Lee, 1993). For exampl e, Mar y
t he gr aduat e st udent mi ght t el l her colleagues t hat t hei r paper
was rej ect ed by saying ( a) " I ' m so sor r y t o t el l you t he paper
was r ej ect ed" or ( b) "I am so sor r y about our paper ; it was
rej ect ed and it appear s t o be my faul t because I di d not i ncl ude
t he addi t i onal dat a. I a m sure you will never be abl e t o forgive
me. But i f you will consi der giving me anot her chance, I will be
greatly i ndebt ed t o you i f you let me t r y salvaging t he manu-
scri pt ?' These are bot h negative pol i t eness strategies and mi ght
be consi dered as equal l y pol i t e on t he basi s of Brown and Lev-
i nson' s hierarchy. But, clearly, t he l at t er st rat egy seems t o be
mor e pol i t e t han t he f or mer one. Thus, i nst ead of consi deri ng
pol i t eness in t er ms of a choice of superstrategies, i t mi ght be
mor e appr opr i at e t o consi der pol i t eness in t er ms of t he ext ent
t o whi ch strategies are used.
We now consi der in mor e det ai l how t he power, distance, and
st rengt h of t he i mposi t i on affect t he ext ent t o whi ch strategies
are used. First, pol i t eness t heor y posi t s t hat , as t he power of t he
l i st ener over t he speaker increases, mor e pol i t eness is needed
t o communi cat e t he face-t hreat eni ng i nformat i on. Thus, one
woul d be mor e pol i t e t o a mor e powerful ot her t han t o a less
powerful other. In t er ms of Mar y' s exampl e, one woul d expect
Mar y t o be mor e pol i t e when communi cat i ng to her superi or
t han when communi cat i ng t o her peer or subor di nat e because
her superior, by definition, has mor e power over Mary.
Second, politeness theory posits that, as the social distance be-
tween the speaker and the listener increases, politeness increases.
In other words, one would be more polite t o someone who is more
socially distant (say, a stranger ) t han to someone who is more so-
cially close (say, a family member ) . How does social distance differ
within a hierarchical setting? Because individuals occupying the
same level within the hierarchy are likely t o have more similar
experiences, be mor e familiar with one another, and interact more
informally with one another (Kanter, 1977; Mintzberg, 1973),
distance is greater with superiors and subordinates t han with peers.
Thus, one would expect more politeness with superiors and subor-
dinates t han with peers.
I n a hi erarchi cal si t uat i on, bot h power and social di st ance
define any i nt erpersonal i nt eract i on. Thus, in organi zat i onal hi-
erarchies, power and di st ance ar e not ort hogonal , separat e di -
mensi ons but are necessari l y rel at ed and must be consi dered in
conj unct i on wi t h one anot her (Lee, 1993). How will power and
di st ance t oget her affect st rat egy usage in hi erarchi cal settings?
P. Brown and Levi nson ( 1987) di d not consi der how social and
cont ext ual cues may differentially affect different strategies; ac-
cordi ng to pol i t eness theory, i ndi vi dual s si mpl y t end t o use
mor e strategies as power and di st ance i ncrease ( R. Brown &
Gi l man, 1989; Fi el d, 1991). However, it is likely t hat some
strategies are mor e sensitive t o di st ance differences and some
are mor e sensitive t o power differences. For exampl e, one woul d
expect t he strategy of bei ng deferent i al t o be mor e sensitive t o
power differences because t he st rat egy i t sel f di rect l y refers t o t he
power rel at i onshi p between t he speaker and t he listener. Con-
versely, t he strategy of emphasi zi ng common gr ound woul d be
mor e sensitive t o di st ance differences because t he st rat egy itself
di rect l y refers t o t he si mi l ar i t y and f ami l i ar i t y between t he
speaker and t he listener. Therefore, our second hypot hesi s was
t hat some strategies (specifically power-sensitive st rat egi es)
woul d be used mor e wi t h superi ors t han wi t h subordi nat es and
NONLINGU1STIC POLITENESS 999
peers and t hat ot her st rat egi es (speci fi cal l y distance-sensitive
st rat egi es) woul d be used l east wi t h peers.
How will pol i t eness strategies differ for good news and for
bad news? Politeness t heor y posi t s t hat i ndi vi dual s will be mor e
pol i t e when t he size of t he t hr eat is larger. Here, i t is obvi ous
t hat bad news is mor e t hr eat eni ng t han good news. Bad news
t hr eat ens t he t arget ' s needs t o be appr eci at ed and acknowl edged
( r ej ect i on of one' s paper undoubt edl y under mi nes t hese needs
consi der abl y) , as well as t he t arget ' s needs t o be free of i mposi -
t i ons (rej ect i on of one' s paper necessari l y suggests addi t i onal
r ounds of revisions, agai n undoubt edl y i mpi ngi ng on t he
t arget ' s freedom). Our t hi r d hypot hesi s was t hat i ndi vi dual s
woul d be mor e pol i t e when communi cat i ng bad news t han
when communi cat i ng good news.
Cu l t u r e a n d No n l i n g u i s t i c Pol i t e ne s s
Cross-Cultural Similarities
One of t he mor e uni que charact eri st i cs of pol i t eness t heor y is
its cl ai m of universality. Accor di ng t o P. Brown and Levi nson
( 1987) , pol i t eness t heor y is not cul t ur e specific; rather, i t de-
scribes strategy usage i n all cul t ures. I n t hi s art i cl e, we compar e
t he nonl i ngui st i c communi cat i on of pol i t eness in t he two cul-
tures: Amer i ca and Korea. These t wo cul t ur es were selected be-
cause t hey were compar ed for l i ngui st i c pol i t eness st rat egi es by
Holtgraves and Yang ( 1990, 1992) and, mor e i mpor t ant , be-
cause t hey r epr esent very different t r adi t i ons i n t er ms of con-
st rual s of t he self, others, and rel at i onshi ps ( Mar kus & Ki t a-
yama, 1991; Triandis, 1989). I f pol i t eness strategies are f ound
t o be si mi l ar across these two diverse cul t ures, one can have
mor e confi dence r egar di ng t he general i t y of nonl i ngui st i c st rat -
egies in t he communi cat i on of politeness. When nonl i ngui st i c
channel s are t aken i nt o account , will pol i t eness strategies still
show universal t r ends across cul t ures? By and large, t here is
mor e uni versal i t y in nonl i ngui st i c communi cat i on t han in l i n-
gui st i c communi cat i on ( Ekman & Fri esen, 1975). Our f our t h
hypot hesi s was t hat t ypes of pol i t eness strategies, communi ca-
t i on of strategies across different channel s, and t he effects of
soci al - cont ext ual cues on pol i t eness woul d show si mi l ar pat -
t erns across t he Kor ean and Amer i can samples.
Cross-Cultural Differences in Interpretation of Social
and Contextual Cues
Hol t graves and Yang ( 1992) f ound t hat power, di st ance, and
size of i mposi t i on had t he pr edi ct ed effect on pol i t eness st rat e-
gies in bot h Kor ean and Amer i can sampl es; in ot her words, po-
liteness i ncr eased when power was high, when di st ance was
high, and when t he size of t he i mposi t i on was high. Politeness
var i ed mor e wi t h power and di st ance differences, however,
among Kor eans t han among Amer i cans. I n ot her words, differ-
ences in social or r el at i onal factors were weighed mor e heavily
by t he Kor ean par t i ci pant s. Thi s is consi st ent wi t h ot her re-
search ( Mar kus & Ki t ayama, 1991; Shweder & Bourne, 1984;
Triandis, 1989) suggesting t hat i ndi vi dual s from East ern cul -
t ures pay mor e at t ent i on t o i nt erpersonal and rel at i onal cues. In
cont rast , Amer i cans rel y less on rel at i onal cues but mor e on
t he cont ent of t he verbal exchange t o i nfer meani ng ( DeMent e,
1988; Hynson, 1990; Park, 1987). Our final hypot hesi s was t hat
Kor eans' usage of pol i t eness woul d be mor e di fferent i at ed ac-
cordi ng t o t he social rel at i onshi p and Amer i can usage of pol i t e-
ness woul d be mor e di fferent i at ed accordi ng t o t he cont ent of
t he message.
S u mma r y
On t he basis of t he tenets of pol i t eness t heor y and previ ous
research, t hi s art i cl e exami nes t he following question: What
t ypes of pol i t eness strategies will be t r ans mi t t ed when nonl i n-
guistic channel s are t aken i nt o account ? Al so expl ored ar e t he
following hypotheses:
Hypothesis 1." Mor e off-t be-record pol i t eness strategies are
conveyed vi a nonl i ngui st i c channel s.
Hypothesis 2: For some strategies, st rat egy usage will be
different when communi cat i ng t o superi ors t han t o subordi -
nat es and peers; for ot her strategies, st rat egy usage will be
different when communi cat i ng t o peers t han t o superi ors and
subordi nat es.
Hypothesis 3: I ndi vi dual s will be mor e pol i t e when commu-
ni cat i ng bad news t han when communi cat i ng good news.
Hypothesis 4: Types of pol i t eness strategies, communi cat i on
of strategies across different channel s, and t he effects of soci al -
cont ext ual cues on pol i t eness will show si mi l ar pat t er ns across
t he Kor ean and Amer i can samples.
Hypothesis 5." Kor eans' usage of pol i t eness strategies will be
mor e sensitive t o rel at i onal cues, whereas Amer i cans' usage of
pol i t eness strategies will be mor e sensitive t o t he cont ent of t he
message.
Th e Pr e s e n t St u d i e s
In summary, the goal of the present research was, first, t o exam-
i ne the nonlinguistic communi cat i on of politeness and, second, t o
exami ne whether the nonlinguistic communi cat i on of politeness
is universal or cul t ure specific. A novel methodology was intro-
duced in the present research to exami ne politeness. Most social
psychological investigations of politeness t heory (e.g., Field, 1991;
Holtgraves & Yang, 1990, 1992) have coded responses to vignettes
to exami ne politeness strategies. Such studies have cont ri but ed
greatly to extending and refining the theory. However, the met hod-
ology is somewhat artificial and is l i mi t ed t o exami ni ng only the
linguistic aspects of the theory. Keeping such limitations in mind,
we conducted a more naturalistic, real-life study exami ni ng polite-
ness conveyed through bot h verbal and nonverbal channels of
communi cat i on. Instead of using a paper-and-pencil test t o gener-
ate responses t o different scenarios, we asked part i ci pant s to com-
muni cat e t o their targets as i f they were really talking t o them.
Moreover, instead of using experimenter-concocted vignettes, we
asked part i ci pant s t o t hi nk of "real-life" targets from their work-
place and asked t hem t o convey i nformat i on relevant t o their
work. By videotaping these interactions, we were able t o conduct
fine-grained analyses of the communi cat i on of politeness via
different channels of communi cat i on.
In this article, we discuss two studies t hat investigated politeness
in the hierarchical communi cat i on of good and bad news in two
1 0 0 0 AMBADY, KOO, LEE, AND ROSENTHAL
di f f er ent cul t ur es. I n t he first study, par t i ci pant s wer e empl oyees
of a Kor ean br oker age f i r m. I n t he second study, par t i ci l xmt s wer e
Ame r i c a n gr a dua t e st udent s. Bot h set s o f par t i ci pant s were vi deo-
t a pe d c o mmu n i c a t i n g good a nd b a d news t o a superi or, subor di -
nat e, or peer f r om t hei r wor kpl ace. Because i t was i mp o r t a n t t h a t
t he news be real i st i c a nd r el evant t o t he par t i ci pant s ' real-life work,
t hese messages were f r a me d di fferent l y for t he t wo sampl es. For
t he Kor e a n br oker sampl e, t he good news a nd ba d news r el at ed t o
a n i ncr ease or decr ease i n commi s s i on. For t he Ame r i c a n gr aduat e
s t udent sampl e, t he good news a nd ba d news r el at ed t o t he accep-
t ance or r ej ect i on of a pa pe r s ubmi t t e d for publ i cat i on. Not e t hat ,
i n b o t h st udi es, t he good a nd ba d news was r el evant t o t he t ar get
as well as t o t he par t i ci pant : The c ommi s s i on i ncr eas e or decr ease
affect ed all f our peopl e ( t hr e e t ar get s a nd t he pa r t i c i pa nt ) , as di d
t he accept ance or r ej ect i on o f t he pa pe r for t he gr aduat e s t udent
sampl e. The message, t her ef or e, was pot ent i al l y face t hr e a t e ni ng
t o a ny of t he t hr e e t ar get s whe n i t was b a d news. I n s umma r y, each
par t i ci pant was vi deot aped rol e-pl ayi ng six scenar i os ( t wo t ypes
o f messages [ good news a nd ba d news] by t hr ee t ypes of t arget s
[ super i or s, s ubor di nat es , a nd peer s] ).
The r esul t s f r om t he t wo st udi es ar e pr es ent ed separ at el y be-
cause t he st udi es i nvol ved di f f er ent scenar i os a nd becaus e s uch a
pr es ent at i on allows for a mor e car ef ul cons i der at i on o f t he nonver -
bal expr essi on o f pol i t eness behavi or wi t hi n each cul t ur e. Behavi or
is e xa mi ne d separ at el y f r om t he fol l owi ng sour ces o f c o mmu n i c a -
t i on: si l ent vi deo ( n o s peech) , speech ( n o vi deo) , full c ha nne l
( a udi o a nd vi deo) , a nd t r ans cr i pt s o f speech. Finally, r esul t s f r om
t he t wo st udi es ar e c o mp a r e d met a- anal yt i caUy t o pe r mi t us t o
ma ke t ent at i ve concl usi ons a bout t he gener al i t y of t he nonve r ba l
expr essi on of pol i t eness.
Me t h o d
St udy 1: Korean Sampl e
Participants. Thirty-six stockbrokers from a large brokerage firm in
Korea participated in the study. All of the part i ci pant s were male be-
Tabl e 1
Three-Factor Solution: Factor Loadings After Rotation
Di mensi on Fact or 1 Fact or 2 Fact or 3
At t ent i ve .91 - . 1 l - . 16
Concerned .87 .08 - . 02
Seeking agreement .85 .24 .07
Encouraging .82 .14 .09
Polite .80 - . 4 0 .27
Approvi ng .76 .29 .21
Deferential .73 - . 28 .40
Positive .70 .05 - . 03
Empat hi c .66 .59 .09
Professional .63 - . 08 - . 17
Open .23 .92 - . 17
Affiliative .18 .92 .14
Joki ng - . 28 .73 .00
Uncert ai n - . 03 .08 .86
Indirect .10 - . 22 .75
Avoi dant - . 57 .05 .66
Apologetic .42 .06 .55
Note. Numbers in boldface indicate di mensi ons loading on each
factor.
cause there were no female brokers or managers i n t he firm with super-
visory responsibilities ( t he present study' s design required part i ci pant s
with real-life superiors and subordinates wi t hi n the firm). Participants
were selected by t he fi rm' s managers t o take par t in t hi s study,
Procedure. Participants were tested individually in an office wi t hi n
the brokerage firm. Participants were first asked to t hi nk of a peer, a
superior, and a subordi nat e from t hei r work. A peer was defined as
"someone at t he same level as you are, whom you work fairly closely
wi t h- - maybe a member of your group or t eam. " A superior was defined
as "someone you report t o who supervises you," and a subordi nat e was
defined as "someone who reports t o you, whose work you supervise."
Participants were asked t o t hi nk of a specific person i n each of t he t hree
categories. To make t hei r targets more salient, part i ci pant s were asked
to write down t he target' s name and to rat e each target on ( a) the extent
to which the target had power over t he part i ci pant and ( b) t he extent to
which t he part i ci pant felt close to the target. Participants were t hen
shown two messages, one of which was good news and one of which was
bad news. The good news involved an increase in commission t o the
group because of good work, and the bad news involved a decrease in
commission because of poor work. A senior employee at the firm was
consulted before t he messages were framed so as t o make t hem realistic
and relevant to the brokers.
Next, part i ci pant s were asked t o imagine how they would communi -
cate t he good news and t he bad news to each target. Participants t hen
role-played how they would actually tell each target about t he good news
and t he bad news. These role-play scenarios were videotaped. Overall,
each part i ci pant role-played six different scenarios: t ransmi t t i ng good
news to a specific superior, subordinate, and peer and t ransmi t t i ng bad
news to a specific superior, subordinate, and peer. The sequence in
which part i ci pant s role-played the scenarios was counterbalanced. After
the part i ci pant s had completed the role-play for all six scenarios, they
were debriefed and t he experi ment was t ermi nat ed. Because there were
36 part i ci pant s each generating six scenarios, there was a t ot al of 216
clips.
Coding. Several types of dat a were collected from each of t he clips
generated from the experimental procedure: ( a) full-channel video-
tapes of t he role-play scenarios wi t h bot h video and audio data, ( b)
videotapes with video dat a only (i.e., silent video and no audio dat a) ,
(c) audiotapes generated from the videotapes (i.e., speech only, no
video dat a) , and ( d) t ranscri pt i ons of t he verbal cont ent of part i ci pant s'
role-play. We t ri ed t o obt ai n pure tone-of-voice dat a by cont ent filtering
the speech, but background noise such as traffic made it impossible to
obt ai n usable samples.
Coding categories were first drawn from E Brown and Levinson' s
( 1987 ) lists of substrategies for positive (p. 102) and negative politeness
(p. 131 ). These categories were carefully selected to represent different
politeness strategies t hrough a "bot t om- up" approach.
There are two different coding methodologies from past research on
politeness. One met hod was used by Holtgraves and Yang (1992), who
asked coders t o categorize part i ci pant s' responses i nt o one of t he five
superstrategies described by P. Brown and Levinson ( 1987 ). The sec-
ond met hod was used by Lee (1993), who asked coders t o rate sub-
strategies from politeness theory on a 9-poi nt scale (substrategies are
simply more specific strategies t hat could be subsumed under t he
broader superstrategies) and t hen created larger composite strategies by
analyzing how the substrategies related t o one another. For t hi s study,
Lee' s coding methodology was more appropri at e for several reasons.
First, as mentioned, we do not expect nonlinguistic politeness strategies
t o assume t he same st ruct ure as linguistic strategies, so imposing Brown
and Levi nson' s st ruct ure on the nonlinguistic dat a would not he appro-
priate. Second, in exami nat i ons of specific situations (e.g., communi -
cating bad news in a hi erarchy), a more specific st ruct ure is appropri-
ate. She found t hat although participants di d not use all of t he strategies
NONLINGUISTIC POLITENESS
Tabl e 2
Between-Factor and Within-Factor Correlations
Korean sample Ameri can sample
Ot her Ot her
Variable ori ent ed Attiliative Circumspect oriented Affiliative Circumspect
Ot her ori ent ed .64 - . 05 - . 0 4 .50 .44 .16
Al~liative .83 - . 02 .51 - . 05
Ci rcumspect .21 .58
Note. All correlations along t he diagonals are mean correlations among individual variables defining a
composite variable. All correlations off t he diagonals are mean correlations between the individual variables
of one composite variable and t he individual variables of a different composite variable. In a well-defined
set of composites, t he mean correlations offt he diagonal should be substantially lower t han the mean corre-
lations on t he diagonal. For the present composites, t he medi an correlation on t he diagonal is .55, and the
medi an correl at i on offt he diagonal is - . 03.
1001
Only strategies t hat could be communi cat ed nonverbally, such as
seeking agreement and being apologetic, were included. Thus, five sub-
strategies t hat were purely linguistic (e.g., going on record as i ncurri ng a
debt by saying "Do me a favor") were excluded. All i n all, 20 categories
reflecting different substrategies were generated.
Categories t hat capt ured positive politeness strategies included being
affiliative, attentive, concerned, approving, encouraging, supportive,
empat hi c, emotional, and positive; seeking agreement; and joking.
These categories correspond t o P. Brown and Levi nson' s ( 1987 ) polite-
ness strategies of noticing or at t endi ng to t he target; showing interest,
approval, and sympat hy for t he target; seeking agreement; asserting
common ground; joking; asserting knowledge of t he target' s wants and
concerns; being optimistic; and asserting reciprocity. No nonverbal par-
allel could be found for t he strategies of giving gifts t o the target or asking
and giving reasons.
Categories t hat capt ured negative politeness included being indirect,
not being open, being apologetic, mi ni mi zi ng t he imposition, and being
uncert ai n, professional, and deferential. These categories correspond to
Brown and Levi nson' s ( 1987 ) strategies of bei ng indirect, hedging, be-
ing pessimistic, mi ni mi zi ng t he imposition, being deferential, and being
apologetic. No nonverbal parallel could be found for the strategies of
impersonalizing t he target and t he speaker, nominalizing, or going on
record as i ncurri ng a debt. Two ot her categories were included: being
avoi dant and being critical.
Four different sets of coders who were all native speakers of Korean
(four coders for the full-channel, video, and audi o and two coders for
the t ranscri pt s) rat ed t he full-channel video and audio, video only, au-
di o only, and t ranscri pt data. Wi t hi n each set, hal f of the coders rated
the clips or t ranscri pt s i n t he order in whi ch the dat a were generated,
and the ot her hal f rat ed t he clips or t ranscri pt s i n t he reverse order. Each
clip or t ranscri pt was rated on the 20 categories on a 9-poi nt scale. Thus,
each coder rat ed each clip on every category on a 9-poi nt scale ranging
described by Brown and Levinson, some of t be strategies were more fine
grained t han what t he t heory posits. For example, mi ni mi zi ng t he
t hreat was one of t he strategies in Brown and Levi nson' s list, but partic-
i pant s used a variety of met hods to mi ni mi ze t he t hreat ; some discred-
ited t he source of t he i nformat i on, some disagreed explicitly with the
bad news, and some decreased t he i mport ance and significance of the
bad news. Thi s more bot t om- up approach (formi ng strategies from the
dat a) , rat her t han a top-down approach (formi ng strategies from the
t heory), was bet t er able to extract a specific typology of strategies for
t he specific situation being studied.
from not at all ( 1 ) to extremely ( 9) . All of t he coding was conducted in
Korean. Each coder rated 216 clips on 20 different categories.
Study 2: American Sample
Participants. Thi rt y graduat e students (15 women and 15 men)
from Harvard University part i ci pat ed i n t he study. Only graduat e stu-
dents were recrui t ed because they are likely t o work with superiors (e.g.,
advisors), peers (ot her graduat e st udent s), and subordi nat es (e.g., un-
dergraduate research assistants) in t hei r research. Students were re-
cruited from a posted sign-up sheet. They were pai d for t hei r
participation.
Procedure. As in Study 1, participants were asked t o role-play de-
livering good and bad news to t hree targets. In Study 1, t he bad news
was a reduction in commission, and t he good news was an increase in
commission. Although t hi s type of good news and bad news would be
realistic in the organizational life of Korean brokers, t he same piece
of i nformat i on would have much less relevance t o graduat e students.
Therefore, t he cont ent of t he good news and bad news was adjusted for
t he present sample: The good news was the acceptance of a paper by a
maj or j ournal , and the bad news was t he rejection of a paper by a maj or
j ournal . The remai nder of t he procedure and t he coding system used
were t he same as for Study 1.
Coders were native speakers of English. Four sets of coders with four
coders each rated t he full-channel video and audi o clips, video-only
clips, audio-only clips, and transcripts. The same 20 di mensi ons coded
in Study 1 were used t o code the clips and t ranscri pt s in Study 2. All of
the coding was conduct ed in English.
Re s ul t s
Prel i mi nary Considerations: Pri nci pal -Component s
Anal ysi s and Creation of Composi t e Variables
Ea c h o f t he 36 Ko r e a n p a r t i c i p a n t s p r o d u c e d 6 cl i ps
( c o mmu n i c a t i n g t wo t ypes o f news t o t h r e e l evel s i n t he hi e r a r -
chy; a t ot a l o f 216 cl i ps ) . Ea c h o f t he 30 Ame r i c a n p a r t i c i p a n t s
al so p r o d u c e d 6 cl i ps ( a t ot al of 180) . Th e me a n o f t he j udge s '
r at i ngs o f e a c h of t he 20 d i me n s i o n s was c o mp u t e d acr os s t he
216 cl i ps of t he Ko r e a n s a mp l e a n d t he ! 80 cl i ps o f t he Ame r i -
c a n s a mp l e separ at el y. Ea c h of t he 20 r a t i ngs was s t a n d a r d i z e d
t o e l i mi n a t e c or r e l a t i ons r e s ul t i ng f r om a ny s ys t e ma t i c g r o u p
1002 AMBADY, KOO, LEE, AND ROSENTHAL
Table 3
Reliabilities oJTudges' (Js) Ratings of Behaviors
Korean sample
Video and Video and
audio Video Audio Transcript audio
Variable 4Js a IJ b 4Js IJ 4Js 1J 2Js IJ 4Js a I J b
American sample
Video Audio Transcript
4Js IJ 4Js 1J 4Js 1J
Other oriented .72 .39 .57 .25 .72 .39 .60 .43 .70 .37 .64 .31 .62 .29 .71 .38
Affiliative .75 .42 .74 .42 .84 .57 .74 .59 .69 .36 .80 .50 .59 .27 .71 .38
Circumspect .41 .15 .41 .15 .73 .40 .35 .21 .68 .35 .55 .23 .48 .19 .68 .35
a Reliability ofthe mean of these judges' ratings, bReliability of a typical single judge.
differences between the Korean and American samples. This
was done for all four channels of communication.
Principal-components analyses were conducted on the stan-
dardized ratings across all four channels to reduce the number of
strategies into a few global ones. To use the same global strategies
for both samples for all channels, we conducted principal-compo-
nents analyses on the standardized data from both samples over
all four channels. A step-up-by-one principal-components analysis
with varimax rotation yielded a satisfactory, theoretically inter-
pretable three-factor solution. This solution was supported by an
examination of the intercorrelations between the factors and
within the factors. Table 1 shows the factor loadings of this three-
factor solution. Three dimensions (emotional, trivializing, and
critical) did not load on any of the three factors, and raters re-
ported confusion about the interpretation of these ratings. These
three dimensions were dropped from all subsequent analyses. Ta-
ble 2 shows the average between-composite and within-composite
variable correlations; as can be seen, the three-factor solution was
satisfactory for both the Korean and American samples.
On the basis of the principal-components analysis, three
composite variables were created: affiliative strategies (open,
affiliative, and joking), circumspect strategies (uncertain, indi-
rect, avoidant, and apologetic), and other-oriented strategies
(attentive, concerned, seeking agreement, encouraging, polite,
approving, deferential, positive, empathic, and professional).
The affiliative composite variable consisted exclusively of posi-
tive politeness strategies. Circumspect consisted of negative po-
liteness strategies (uncertain, avoidant, and apologetic) and one
off-the-record strategy (indirect), The last composite, other ori-
ented, was a mixed bag of positive strategies that acknowledged
the listener's needs (attentive, concerned, seeking agreement,
encouraging, polite, approving, positive, and empathic) and
negative politeness strategies that minimized the imposition
(deferential and professional). Although P. Brown and Levin-
son ( 1987 ) considered all politeness to be oriented toward oth-
ers, our analysis separated other-orientation from affiliation and
circumspection.
The principal-components analysis addressed the explor-
atory question about the kinds of strategies individuals use
when nonlinguistic channels are taken into account. The typol-
ogy of strategies showed a considerable amount of overlap with
the structure of linguistic strategies proposed in politeness the-
ory. Affiliativeness and circumspection consisted of positive and
negative politeness strategies, respectively. Although other-ori-
ented strategies consisted of a mix of both positive and negative
strategies, this result is consistent with past research suggesting
that the delineation of positive and negative strategies may not
be clear cut and that individuals tend to use a mixture of both
types of strategies (Craig et al., 1986; Lim & Bowers, 1991;
Scollon & Scollon, 1983 ).
Rel i abi l i t y of Judges" Rat i ngs
The effective reliabilities (Rosenthal & Rosnow, 1991 ) of the
judges were calculated for each composite and each channel.
The interjudge reliabilities ranged from .35 to .84, with a mean
of .63, for the Korean sample. The interjudge reliabilities
ranged from .48 to .80, with a mean of .65, for the American
sample (see Table 3 ).
St udy 1: Korean Sampl e
Separate analyses of variance were performed for each de-
pendent variable (other oriented, affiliative, and circumspect);
hierarchy (communicating to superior, peer, or subordinate),
news (good news or bad news), and channel (full channel with
video and audio, video only, audio only, or transcript) were in-
Table 4
Hierarchy, News, and Channel Effects Across Composite
Ratings Among Koreans and Americans
Other oriented At f i l i at i ve Circumspect
Source df F r F r F r
Koreans
Channel 3 19.28** .60 28.26** .67 81.46"* .84
Hierarchy 2 16.95"* .81 25.54** .65 12.26"* .51
News 1 1.42 .20 57.72** .79 121.90"* .88
Americans
Channel 3 6.04* .42 9.06** .49 167.30"* .93
Hierarchy 2 0.40 .12 15.25"* .59 6.22* .43
News 1 67.08** .84 66.73** .84 118.30"* .90
* p < .005. **p < .0001.
NONLINGUISTIC POLITENESS 1003
6.5
6.0
5.5
..= 5.0
=o
~. 4.5
4.0
3.5
Other-oriented
Affiliative
Circumspect
I "1
Video+Audio Video Only Audio Only Transcripts
Channels
Figure 1. Ratings of other-oriented, affiliative, and circumspect strategies across different communication
channels: Korean participants.
dependent vari abl es. Tabl e 4 i l l ust rat es t he resul t s of t he anal y-
ses of var i ance for t he hi erarchy, news, and c ha nne l effects for
each composi t e rat i ng.
Channel effects. Anal yses of vari ance revealed significant
channel effects for all t hree composi t e variables (see Table 4 for
F values, significance levels, and effect sizes). These effects were
exami ned furt her i n a pl anned cont rast compar i ng t he purel y non-
linguistic channel (si l ent vi deo) and t he channel s i ncorporat i ng
linguistic i nf or mat i on; weights o f - 1 (ful l ), - 1 ( audi o) , - I
( t r anscr i pt ) , and 3 ( vi deo) were used. Thi s cont rast was significant
for all t hree composites: ot her ori ent at i on, F( l , 70) = 9.93, p =
.0024, r = .472; affiliativeness, F ( l , 70) = 28.65, p < .0001, r =
.67; and ci rcumspect i on, F ( l , 70) = 101.43, p < .0001, r = .86.
The means of these t hree variables across t he f our channel s are
present ed i n Fi gure 1. Kor ean part i ci pant s were rat ed as most cir-
cumspect i n t he nonl i ngui st i c channel . I n contrast, ratings of
affiliation and ot her or i ent at i on showed t he opposite t rend; these
rat i ngs were lowest i n the nonl i ngui st i c channel , suggesting t hat
these strategies are communi cat ed t hrough language. Thi s fi ndi ng
provides suppor t for Hypothesis l : For Kor ean part i ci pant s, the
i ndi rect strategy ( ci r cumspect i on) was conveyed most l y t hr ough
t he nonl i ngui st i c channel .
Hi erarchy effects. Anal yses of var i ance of ot her - or i ent ed,
afliliative, and ci r cums pect r at i ngs all showed si gni fi cant effects
for hi er ar chy (see Tabl e 4 for F val ues, si gni fi cance values, and
effect si zes). To e xa mi ne t he effect of power, we conduct ed con-
t r ast anal yses usi ng wei ght s of - 1 ( s ubor di na t e ) , - I ( peer ) ,
and 2 ( s uper i or ) ; power effects pr edi ct t hat i ndi vi dual s will be-
have di fferent l y t owar d super i or s t ha n t hey do t owar d peers and
s ubor di nat es ) . These cont r ast s were si gni f i cant for ot her or i en-
t at i on, F( 1, 70) = 30. 70, p < . 0001, r = .68, and ci r cums pec-
2 Effect size estimates were calculated by means of the following for-
mula ( Rosenthal & Rosnow, 1991 ):
~ F dfnumerat or
r = Vi F x dfnu--~erat--~r) + d - ~ n o mi n a t o r '
where the dfnumerat or is always 1 and the dfdenomi nat or is always the
number of participants minus 1.
1004 AMBADY, KOO, LEE, AND ROSENTHAL
.
6.5
6.0
5.5 m
5.0
4.5
4.0
3.5 --
Other-oriented
Affiliadve
Circumspect
t I I
Subordinate Peer Boss
Communication Target
Figure 2. Ratings of other-oriented, atfiliative, and circumspect strategies across different hierarchy levels:
Korean participants.
tion, F( 1, 70) = 24.49, p < .0001, r = .64, but did not reach
conventional levels of significance for affiliation.
To test the effect of social distance, we conducted a quadratic
contrast using the weights - 1 (superior), 2 (peer), and - 1
(subordinate); distance effects predict that individuals will be-
have differently toward peers than they do toward superiors or
subordinates. This quadratic contrast was significant for affil-
iation, F( 1, 70) = 48.28, p < .0001, r = .76. The means of the
other-oriented, circumspect, and affiliative ratings across hier-
archies are illustrated in Figure 2.
These results provide partial support for Hypothesis 2. Some
strategies (namely other-orientation and circumspection) were
more sensitive to power effects and were used significantly more
when communicating with superiors than when communicat-
ing with peers and subordinates. Other strategies (namely
affiliation) were more sensitive to distance effects and were used
more when communicating with peers.
News effects. The means of good and bad news for affil-
iative and circumspect ratings are shown in Figure 3. Analy-
ses of variance revealed that Korean participants were rated
as significantly more circumspect and significantly less
affiliative when delivering bad news rather than good news
(see Table 4 for F values, significance values, and effect
sizes). There was no difference on other-orientedness in the
delivering of good and bad news. Overall, support for Hy-
pothesis 3 was mixed. As Hypothesis 3 predicted, more cir-
cumspection was used for bad news than for good news; how-
ever, instead of being more affiliative when the imposition was
larger (bad news), individuals were more affiliative when the
imposition was smaller (good news).
St udy 2: Ameri can Sampl e
The same analyses of variance, with similar contrast
weights, were conducted with the American data.
Channel effects. The channel effect was significant for all
three composites (see Table 4 for F values, significance val-
ues, and effect sizes). The means across the four channels are
shown in Figure 4. The pl anned contrast exami ni ng whether
the video channel was different from the other channels, us-
ing weights of - 1 ( f ul l ) , - 1 ( a udi o) , - 1 (t ranscri pt ), and 3
(video), was significant for ratings of affiliativeness, F( 1, 84 )
NONLINGUISTIC POLITENESS 1005
o
6.5
6.0
5.5
5.0
4.5
4.0 - -
3.5 - -
Other-oriented
y
Affiliative
Circumspect
BadNews GoodNews
Type of News
Figure 3. Ratings of other-oriented, afl~liative, and circumspect strategies across different messages: Ko-
rean participants.
= 11. 60, p = . 001, r = . 54, a nd c i r c ums pe c t i on, F ( 1, 84) =
413. 41, p < . 0001, r = .91. For t he ot her - or i ent ed c ompos i t e
var i abl e, t he p l a n n e d c ont r a s t was n o t si gni f i cant , F ( 1, 84) =
0. 02, p = . 8810, r = - . 0 3 . Agai n, t hese f i ndi ngs pr ovi de s up-
por t for Hypot hes i s 1: Of f - t he- r ecor d, i ndi r e c t st r at egi es
( c i r c u ms p e c t i o n ) were mos t l y c o mmu n i c a t e d vi a t he n o n l i n -
gui s t i c c ha nne l .
Hi erarchy effects. The me a n s of t he ot her - or i ent ed, ci r-
c ums pe c t , a nd affi l i at i ve r at i ngs acr oss hi er ar chi es ar e i l l us-
t r at ed i n Fi gur e 5. The anal ys i s o f va r i a nc e showed a si gni fi -
c a nt hi e r a r c hy effect for af f i l i at i on a n d c i r c u ms p e c t i o n ( see
Tabl e 4 for F val ues, s i gni f i cance val ues, a nd effect si zes) .
The c ont r a s t anal ysi s, us i ng t he wei ght s - 1 ( s u b o r d i n a t e ) ,
- 1 ( pe e r ) , a nd 2 ( s upe r i or ) t o t est t he effect o f power, was
s i gni f i cant for r at i ngs o f c i r c ums pe c t i on, F ( 1, 56) = 10. 81, p
= . 0017, r = .53. The qua dr a t i c cont r as t , us i ng t he wei ght s
- 1 ( s u b o r d i n a t e ) , 2 ( pe e r ) , a nd - 1 ( s upe r i or ) t o t est t he
effect of di s t ance, was s i gni f i cant for af f i l i at i on, F ( 1, 56) =
26. 92, p < . 0001, r = .70. Thus , Ame r i c a n pa r t i c i pa nt s were
mor e c i r c ums pe c t t o s uper i or s a nd mo r e affi l i at i ve t o peers.
Thes e f i ndi ngs pr ovi de par t i al s uppor t for Hypot he s i s 2:
Some st r at egi es ( c i r c u ms p e c t i o n ) ar e mor e sensi t i ve t o power
di fferences, a nd s ome st r at egi es ( af f i l i at i on) ar e mo r e sensi -
t i ve t o di s t ance di fferences.
News effects. Significant news effects were f ound for all of t he
politeness composi t e ratings. Amer i can part i ci pant s were rat ed as
significantly mor e ot her ori ent ed and afffliative when delivering
good news and as significantly mor e ci rcumspect when delivering
bad news (see Table 4 for F values, significance values, and effect
sizes). The means of good and bad news for the atraliative and
ci rcumspect ratings are shown i n Fi gure 6. Suppor t for Hypot hesi s
3 was mi xed: Indi vi dual s were mor e ci rcumspect when delivering
bad news but were mor e ot her ori ent ed and affiliative when deliv-
ering good news.
1006 AMBADY, KOO, LEE, AND R OS E NT HAL
6.0 -
5.5 -
5.0 ~
4,5
4. 0 -
3.5
3.0
pect
Vi deo+Audi o Video Only Audi o Onl y Transcripts
ted
Affi l i at i ve
._=
Channels
Figure 4. Ra t i n g s o f ot he r - or i e nt e d, af f i l i at i ve, a n d c i r c u ms p e c t s t r a t e g i e s a c r o s s di f f e r e nt c o mmu n i c a t i o n
c h a n n e l s : Ame r i c a n p a r t i c i p a n t s .
Met a-Anal ysi s: Compar i s on of the Two Cul t ur es
How did the effects of hierarchy, channel, and news differ
across cultures? We compared the effects from the Korean and
American samples using meta-analytic techniques. The analy-
sis revealed cross-cultural similarities in the usage of affiliation
and circumspection strategies but marked differences in other-
orientation.
Similarities Between Samples
Table 5 lists, for each effect, the significance of the effect size
comparisons between the Korean and American samples. Strik-
ing similarities between the two samples are apparent in ratings
of affiliation and circumspection for the hierarchy, channel, and
news main effects. The effect sizes of the contrasts we examined
not only were in the same direction but were similar in magni-
tude. Examination of the recta-analytic comparisons shows
that, by and large, these effects and significance levels were not
significantly different from one another (p >. 1 ). The only ex-
ception was for the comparison of ratings of circumspection for
the channel effect; the effect size comparison was marginally
significant (Zr = 3.08, p = .001 ). However, the effect sizes were
in the same direction and were both significant (r = .86 in the
Korean sample, r --- .97 in the American sample).
The effect sizes and significance levels of the affiliative and
circumspect ratings in the hierarchy, channel, and news effects
were combined meta-analytically. The combined effect sizes
and significance values are listed in Table 6. All in all, strategy
usage was consistent across both cultures. Furthermore, chan-
nel, news, and hierarchy effects were similar across both cul-
tures for affiliation and circumspection.
Differences Between Samples
Table 5 shows that the contrast effects for other-orientation
were different between the Korean and American samples. Ko-
reans were rated as significantly more other oriented to superi-
ors than to subordinates and peers, but this effect was not ap-
parent in the American sample. Meta-analytic comparisons re-
vealed that the Korean and American effect sizes were
NONLINGUISTIC POLITENESS 1007
._=
6.5
6.0
5.5
5.0
4.5
4.0
3.5
Other-oriented
Affiliative
Circumspect
I I t
Subordinate Peer Boss
Communication Target
Figure 5. Ratings of other-oriented, affiliative, and circumspect strategies across different hierarchy levels:
American participants.
significantly different (Zr = 3.31, p = .0005), as were the sig-
nificance levels (Z = 4.26, p = .00001 ).
Koreans were rated as significantly less other oriented in the
video-only channel as compared with the other channels.
Again, this trend was not apparent for the American sample. A
comparison of the Korean and American effect sizes and sig-
nificance levels shows that these values were significantly
different (Zr = 2.08, p = .02, for the effect size comparison and
Z = 2.82, p = .002 for the significance level comparison).
In the American sample, good news was rated as more other
oriented than bad news. The same effect was evident in the Ko-
rean sample, but the effect was considerably smaller. The effect
sizes for the news variable were significantly different between
the Korean and American samples (Zr = 3.92, p = .00004), as
were the significance levels of the news effect (Zr = 3.60, p =
.0002).
These effects provide support for Hypothesis 5. Ratings of
other-orientation differed on the basis of power and distance for
Koreans but not for Americans, suggesting that Koreans rely
more on relational cues in using other-oriented strategies. In
contrast, ratings of other-orientation differed on the basis of the
type of news for Americans but not for Koreans, suggesting that
Americans rely more on contextual cues in using other-oriented
strategies.
Di scussi on
These findings have important theoretical implications for
politeness theory. First, these results highlight the value of ex-
amining nonlinguistic aspects of politeness theory. Quite apart
from contextualizing linguistic utterances, nonlinguistic chan-
nels communicate politeness independent of linguistic channels
of communication. A second contribution of this study is the
finding that even when nonlinguistic channels are taken into
account, strategy usage is affected by social and contextual cues
in ways similar to the propositions of linguistic politeness put
forth by P. Brown and Levinson (1987). A third contribution
is that some aspects of the usage of politeness strategies seem
to be universal, whereas others appear to be culture specific.
Specifically, Koreans and Americans used the strategies of
affiliation and circumspection similarly, but they differed in the
manner in which they used the strategy of other-orientation.
1008 AMBADY, KOO, LEE, AND ROSENTHAL
6.0
5.5
5.0
4.5
4.0
3.5
3.0
I I
Bad News Good News
Other-oriented
Affiliative
Circumspect
Type of News
Figure 6. Ratings of other-oriented, affiliative, and circumspect strategies across different messages: Amer-
ican participants.
In addi t i on to t he t heoret i cal findings, t hi s art i cl e involves
two key met hodol ogi cal i nnovat i ons for pol i t eness research.
First, nonl i ngui st i c channel s of communi cat i on were consid-
ered in addi t i on t o l i ngui st i c channel s of communi cat i on. We
compar ed t he communi cat i on of pol i t eness vi a a pur el y linguis-
t i c channel ( t r anscr i pt s) , a pur el y nonl i ngui st i c channel ( vi deo
onl y) , and two mi xed channel s ( audi o and ful l -channel vi deo
pl us audi o) of communi cat i on. Thi s met hodol ogy per mi t t ed an
exami nat i on of t he manner i n whi ch different channel s convey
different messages. In addi t i on, t hi s fi ne-grai ned exami nat i on
of different channel s of communi cat i on was i mpor t ant because
it cont r i but es t o t he ecological val i di t y of pol i t eness t heory: The
two mi xed channel s represent some of t he mor e common
modes of i nt erpersonal communi cat i on such as face-to-face and
t el ephone communi cat i on. Previous research has, t o some ex-
t ent , neglected t he ecological val i di t y of pol i t eness t heor y by
exami ni ng only t r anscr i pt s and language. Second, t he present
research used a nat ur al i st i c appr oach t o exami ne politeness.
I nst ead of respondi ng t o experi ment er-concoct ed vignettes,
par t i ci pant s generat ed t hei r own responses t o real i st i c target
persons about realistic face-t hreat eni ng i nf or mat i on. The cod-
ing of t he pol i t eness strategies was also mor e nat ural i st i c. Cod-
ers di d not assign each behavi or t o one of t he superstrategies
pr escr i bed by pol i t eness t heory; rather, t hey r at ed t he behavi or
i nt ui t i vel y on a vari et y of substrategies. These rat i ngs were t hen
combi ned t o creat e t he composi t e strategies. In essence, we
coded t he strategies bot t om- up from t he dat a as opposed t o top-
down from t he theory.
S u mma r y and Int erpret at i on o f Resul t s
Similarities across cultures. Across bot h linguistic and
nonl i ngui st i c channel s of communi cat i on, t hr ee mai n t ypes of
NONLINGUISTIC POLITENESS
Tabl e 5
Effect Si ze Comparisons Between t he Korean and Ameri can Sampl es
Source (contrast weights) Korean r American r
Effect size
comparison
~ )
Hierarchy (subordinate, peer, superior)
Other oriented ( - 1, - I, +2) .68 - . 03 .0005
Atiiliative ( - I , +2, - 1 ) .76 .70 .31
Circumspect ( - 1, - I, +2) .64 .53 .26
Channel (video + audio, audio, transcript, video)
Other oriented ( - 1, - 1, - 1, +3) .47 - . 03 .02
Attiliative ( - 1, - 1, - 1, +3) .67 .54 .22
Circumspect ( - 1, - 1, - 1, +3) .86 .91 .001
News
Other oriented .20 .84 .00004
Afliliative .79 .84 .28
Circumspect .88 .90 .36
a Pearson rs were converted to Fisher's zrs, and the Z value of the difference between the two effect sizes was
I + ~ 1 It/2. The value reported here is that
computed with the formula Z = (Zrl - Zr2) + ~ 3/2 - 3] P
associated with the Z value (Rosenthal & Rosnow, 1991 ).
1009
st r at egi es wer e used i n bot h t he Kor e a n and Ame r i c a n sampl es:
ot her - or i ent at i on, affi l i at i on, and ci r cums pect i on. These st rat -
egy compos i t es r oughl y cor r es pond t o P. Br own and Le vi ns on' s
( 1987) di f f er ent i at i on bet ween posi t i ve st rat egi es t hat acknowl -
edge ot her s ' needs and negat i ve st rat egi es t hat mi n i mi z e t hr eat ;
affi l i at i on compr i s es pur el y posi t i ve st rat egi es, c i r c ums pe c t i on
compr i s es negat i ve and of f - t he- r ecor d st rat egi es, and ot her - or i -
ent at i on compr i s es bot h posi t i ve and negat i ve strategies.
Acr oss bot h cul t ur es, par t i ci pant s used t he st rat egi es o f
affi l i at i on and c i r c ums pe c t i on i n a s i mi l ar manner . Th a t is, bot h
Ame r i c a ns and Kor e a ns wer e mor e affi l i at i ve t owar d peer s and
mor e c i r c ums pe c t t owar d superi ors. Si mi l arl y, bot h were mor e
Tabl e 6
Combi ni ng Effect Si zes and Significance Values
Source Combined r a Combined pb
Hierarchy
Other oriented .42 .02
Affiliative .73 .00001
Circumspect .59 .002
Channel
Other oriented .26 .18
Afflliative .62 .0001
Circumspect .93 .00000001
News
Other oriented .58 .02
Atfiliative .81 .000004
Circumspect .89 .00000004
a The combined r was calculated by converting the Pearson rs to Fisher's
z transformation of r, computing the mean of the Fisher' s zrs, and then
converting the Fisher's zrs back to Pearson rs (Rosenthal & Rosnow,
1991 ). t'The Z values corresponding to the p values from the two cul-
tures were found, and the sum of the two Z values was divided by the
square root of two (Z = ZI + Z2/V2). Finally, t hep value associated with
the new Z value is presented as the combined p value.
affiliative and less c i r c ums pe c t when del i ver i ng good news. Fur -
t her mor e, f or bot h Ame r i c a ns and Kor eans, c i r c ums pe c t i on
was mor e appar ent f r om t he pur el y nonver bal channel o f be-
havi or (i . e. , t he si l ent vi deo c ha nne l ) . These fi ndi ngs s uppor t
t he pr edi ct i ons o f pol i t eness t heory. Fi rst , c i r c ums pe c t i on
( c ompos e d o f t he mor e pol i t e negat i ve pol i t eness and off-t he-
r ecor d st r at egi es) s houl d be exhi bi t ed mor e t owar d superi ors.
Second, t he st r at egy o f bei ng c i r c ums pe c t was associ at ed wi t h
del i ver i ng bad news, wher eas t he st r at egy o f bei ng affiliative was
associ at ed wi t h del i ver i ng good news. The fi ndi ng t hat , i n bot h
cul t ur es, t he nonl i ngui st i c vi deo channel seems t o be used t he
mos t i n t he expr essi on o f c i r c ums pe c t i on suggests t hat negat i ve
and of f - t he- r ecor d st rat egi es may be c o mmu n i c a t e d t hr ough
nonl i ngui st i c behavi or.
Our results also reveal ed some unexpect ed findings regardi ng
affiliation strategies. Al t hough di st ance differences predi ct less po-
liteness wi t h peers ( because of closer social di st ance) and less 190-
liteness when delivering good news ( because o f lower i mposi t i on) ,
we f ound that, in bot h t he Kor ean sampl e and t he Amer i can sam-
ple, i ndi vi dual s were mor e affiliative when communi cat i ng t o
peers and when communi cat i ng good news. Thi s suggests that, i n
addi t i on t o redressing face-t hreat eni ng i nf or mat i on, specific posi-
tive politeness strategies may be used t o mai nt ai n close i nt erper-
sonal rel at i onshi ps and t o accent uat e face-enhanci ng event s such
as good news. Thi s is consi st ent wi t h P. Br own and Levi nson' s
( 1987) assertion t hat positive politeness strategies ar e not onl y
used t o redress face t hreat s but ar e used as a ki nd o f social accel-
erat or in whi ch t he speaker i ndi cat es t hat he or she wants t o " c ome
closer" t o t he target. Affiliation, a distance-sensitive strategy, seems
t o per f or m this role.
Differences across cultures. Th e st r at egy t hat exhi bi t ed t he
mos t cr os s - cul t ur al var i abi l i t y was ot her - or i ent at i on ( a st r at egy
c ompos e d o f s ome posi t i ve and s ome negat i ve pol i t eness
st r at egi es) . Kor e a ns used di f f er ent l evel s o f ot her - or i ent at i on
dependi ng on t he st at us o f t he t arget ( t hey wer e mos t ot her ori -
1010 AMBADY, KOO, LEE, AND ROSENTHAL
ented toward their superiors), whereas Americans were equally
other oriented toward all three targets, suggesting that Koreans
pay more attention to interpersonal and relational cues. In con-
trast, Americans used other-oriented strategies when they deliv-
ered good news rather than bad news, but Koreans did not vary
in other-orientation depending on the type of information being
communicated, suggesting that Americans are affected more by
the content of the message than by their relationship with the
target of the message.
All in all, these findings support assertions that, in "low-
context" cultures (e.g., American culture; Cohen, 1987; Hall,
1976), the content of communication is extremely important:
Directness and accuracy are valued, and relationships are based
on equality rather than hierarchy (Hynson, 1990). In "high-
context" cultures (e.g., Korean culture), however, the relational
and hierarchical context rather than the content of the message
per se conveys meaning (Hofstede, 1980). In such cultures, in-
dividual identity is embedded in socially defined roles. There-
fore, contextual factors such as group membership and the roles
of target and speaker are used in conveying and interpreting
meaning.
Co n c l u s i o n s
Future research should address some of the limitations of this
study. It would be worthwhile to examine the pure vocal chan-
nel using only tone-of-voice cues. As a result of loud back-
ground noise, the audio material from Korea did not permit
content filtering to extract vocal cues. Furthermore, the nature
of the hierarchy and the content of the communication differed
for the two samples in this study. The American sample con-
sisted of graduate students, whereas the Korean sample con-
sisted of brokers. Each sample was given different types of mes-
sages to convey because of the differences in the work environ-
ment, and this may have confounded the results. Future studies
should compare similar samples of participants from each cul-
ture. Furthermore, although it is clear that politeness is con-
veyed nonlinguistically, it would be useful to identify specific
nonverbal behaviors that convey politeness and to establish
whether these behaviors are universal.
Despite these limitations, this article makes some noteworthy
contributions in the social psychological examination of polite-
ness theory. First, methodologically, we used a more naturalistic
situation to examine politeness strategies: Participants role-
played situations in which they generated their own strategies
rather than responding to vignettes. Second, whereas previous
studies have focused on examining transcripts alone, we exam-
ined several different channels of communication such as silent
video, audio, and the full audio and video channel. This allowed
for a detailed, in-depth analysis of politeness behavior not only
in linguistic channels but also in nonlinguistic channels. Third,
instead of coding superstrategies, as outlined by E Brown and
Levinson (1987), we coded substrategies and then combined
them, as suggested by the data, into three composite strategies.
This permitted us to combine positive and negative strategies,
as suggested by recent critiques of Brown and Levinson' s work,
and to define strategies that were more specific, ecologically
valid, and situationally relevant to our study. Fourth, previous
research has examined politeness as a response to face-threat-
ening acts (delivering bad news), but this study suggests that
politeness strategies are used as a response to non-face-threat-
ening acts as well. Participants across both cultures used more
affiliative strategies in delivering good news to peers,
Clearly, future investigations of politeness theory should pay
close attention to nonverbal channels of communication. Po-
liteness is communicated through both language and nonverbal
behavior. Furthermore, whereas some linguistic as well as non-
verbal strategies seem to be universal, others are culture spe-
cific. Perhaps many misunderstandings that occur between cul-
tures are due to the miscommunication of politeness. Consider
the example of Mary at the beginning of the article. If Mary
was a Korean telling her American subordinate about a paper
rejection, the subordinate might feel that Mary did not exhibit
appropriate levels of politeness based on the imposition of the
bad news because Koreans' usage of strategies is more sensitive
to relational cues than to the content of the message. Had Mary
been an American telling her Korean advisor about good news,
the same problem might have occurred. Mary would be more
sensitive to the content of the message and might exhibit a level
of politeness deemed inappropriate when speaking to her advi-
sor. The present research suggests that these miscommunica-
tions can be fruitfully examined by focusing on cross-cultural
similarities and differences in the communication of politeness.
References
Baxter, L. (1984). An investigation of compliance-gaining on polite-
ness. Human Communication Research, 10, 427-456.
Brown, P., & Levinson, S. (1987). Politeness: Some universals in lan-
guage usage. Cambridge, England: Cambridge University Press.
Brown, R., & Gilman, A. (1989). Politeness theory in Shakespeare's
four major tragedies. Language in Society. 18, 159-212.
Cohen, R. ( 1987 ). Problems of intercultural communication in Egyp-
tian-American diplomatic relations. International Journal of Inter-
cultural Relations, 11, 29-47.
Craig, R. T., Tracy, C., & Spisak, F. (1986). The discourse of requests:
Assessment of a politeness approach. Human Communication Re-
search, 12, 437-468.
DeMente, B. ( 1988 ). Korean etiquette and ethics in business. Lincoln-
wood, 1L: NTC Business Books.
Ekman, P., & Friesen, W. V. (1975). Unmasking the face. Englewood
Cliffs, N J: Prentice Hall.
Field, S. O. ( 1991 ). On saying unpleasant things: An experimental in-
vestigation of politeness theory. Unpublished doctoral dissertation,
Harvard University, Cambridge, MA.
Goffman, E. ( 1967 ). Interaction ritual: Essays on jace-to-face behavior.
Garden City, NY: Anchor & Doubleday.
Hall, E. T. (1976). Beyond culture. New York: Doubleday.
Hofstede, G. (1980). Culture's consequences. International differences
in work-related values. Beverly Hills, CA: Sage.
Holtgraves, T., & Yang, J. (1990). Politeness as universal: Cross-cultural
perceptions of request strategies and inferences based on their use.
Journal of Personality and Social Psychology, 59, 719-729.
Holtgraves, T., & Yang, J. (1992). Interpersonal underpinnings of re-
quest strategies: General principles and differences due to culture and
gender. Journal of Personality and Social Psychology, 62, 246-256.
Hynson, L. M. (1990). Doing business with South Korea: A handbook
for executives in the public and private sectors. New York: Quorum.
Kanter, R. ( 1977 ). Men and women of the corporation. New York: Basic
Books.
NONLINGUISTIC POLITENESS 101 1
Lee, E ( 1993 ). Being polite and keeping MUM: How bad news is com-
municated in organizational hierarchies. Journal of Applied Social
Psychology, 23, 1124-1149.
Lim, T,, & Bowers, J. W. ( 1991 ). Facework, solidarity, approbation, and
tact. Human Communication Research, 17, 415-450.
Markus, H. R., & Kitayama, S. ( 1991 ), Culture and the self: Implica-
tions for cognition, emotion, and motivation. Psychological Review,
98, 224-253.
Mintzberg, H. (1973). The nature of managerial work. New York:
Harper & Row.
Park, M. (1987). Communication styles in two different cultures: Ko-
rean and American. Seoul, Korea: Hart Shin.
Rosenthal, R., & Rosnow, R. L. ( 1991 ). Essentials of behavior research:
Methods and data analysis. New York: McGraw-Hill.
Scollon, R., & Scollon, S. B. K. (1983). Face in interethnic communi-
cation. In J. C. Richards & R. W. Schmidt (Eds.), Language and
communication (pp. 156-188 ). London: Longman.
Shweder, R. A., & Bourne, E. J. (1984). Does the concept of person
vary cross-culturally? In R. A. Shweder & R. A. Levine (Eds.), Cul-
ture theory: Essays on mind, self, and emotion (pp. 158-199). Cam-
bridge, England: Cambridge University Press.
Smith-Hefner, N. (1988). Women and politeness: The Javanese exam-
ple. Language in Society, 17, 535-554.
Triandis, H. C. (1989). The self and social behavior in differing psycho-
logical contexts. Psychological Review,, 96, 506-520.
Recei ved J a nua r y 13, 1995
Revi si on r ecei ved No v e mb e r 7, 1995
Accept ed De c e mb e r 6, 1995