Sunteți pe pagina 1din 7

Journal of Personality and Social Psychology

1979, Vol. 37, No. 5, 728-734

Smiles Can Be Back Channels


Lawrence J. B runner
Department of Behavioral Sciences
University of Chicago
The placement of auditor smile beginnings in the stream of dyadic interaction
was investigated, using detailed transcriptions of the language, paralanguage,
and body motion of the participants in four two-person conversations recorded
on videotape. Auditor smile beginnings showed a strong tendency to occur at
the same kinds of location as "back channel" responses (such as "yeah,"
"uh-huh," and head nods). This finding indicates that the smile can function
as a type of back channel. It is argued that smiles, like other forms of back
channel, make communication more efficient by providing the speaker with
feedback on a number of levels simultaneously.

Most quantitative investigations of smiling


in adults (see reviews by Berlyne, 1969;
Chapman & Foot, 1976; and Goldstein & McGhee, 1972) have treated it as expressive of
underlying states such as mirth or pleasurable
arousal. Possibly because researchers in social psychology and personality tend to focus
their attention more on hypothetical constructs than on the behavior that those constructs are intended to explain (a thesis that
Fiske, 1978, argues cogently and in detail),
the frequent failure to find significant correlations between ratings of amusement and
actions such as smiling and laughing (Chapman, 1976, pp. 156-157) has been attributed
(by Berlyne, 1969, and Chapman, 1976, for
example) to the greater "sensitivity" of selfreport measures.
The present study is deliberately opposite
in approach. Without denying that people
This study is based on a dissertation submitted in
partial fulfillment of the requirements for a doctoral
degree in psychology at the University of Chicago.
The author would like to acknowledge the support
and guidance of his committeeStarkey Duncan,
Jr. (Chairman), Donald W. Fiske, and Norman
Bradburn.
The research was supported by National Science
Foundation Grant SOC 74-24084 to Starkey Duncan,
Jr., and Donald W. Fiske.
Requests for reprints should be sent to Lawrence
J. Brunner, 160S East SSth Street, Chicago, Illinois
60615.

often smile when they are happy or when


something strikes them as funny, it is concerned with smiles as actions rather than expressions of unobservable inner states. It is
an attempt to partly specify the "meaning" of
smilesspecifically, the smiles of the person
who is not speaking in a two-person conversationin what Goffman (1971, p. 235) calls
the "structural" sense of the term. That is,
it is an attempt to describe their placement
in the stream of behavior.
The study is part of a program of investigation (Duncan, 1972, 1974; Duncan & Fiske,
1977; Duncan & Niederehe, 1974) whose
purpose is to discover the organization of
face-to-face interaction. The term organization reflects the assumption that just as language operates according to rules that give it
a definite structure, so does the process of
face-to-face interaction in which language
ordinarily occurs.
The research of Duncan and his associates
on the structure of interaction is based on
detailed transcription of speech and body
motion in two-person conversations. Exploration of these materials has led to the formulation of a set of hypothesized signals and rules
that people apparently use to exchange speaking turns and otherwise regulate the flow of
interaction in conversations. These rules depend on a distinction between the "speaker"
(the person who holds the speaking turn)

Copyright 1979 by the American Psychological Association, Inc. 0022-3514/79/370S-0728$00.75

728

SMILES CAN BE BACK CHANNELS


and the "auditor" (a person who is in the
conversation without holding the turn). To
hold the speaking turn is not the same as to
emit speech sounds; in the conversations we
have studied, brief pauses seldom if ever constitute relinquishment of the speaking turn,
and it is possible for a person to talk without
claiming the speaking turn (for example, to
say things like "yeah" and "uh-huh").
Speaking Turns
Duncan (1972) hypothesized a turn signal
composed of six separate cues in language,
paralanguage, and body motion. If any one or
any combination of these cues is displayed,
the auditor may take the turn if he or she
wishes to. The auditor is not required to
take the turn when the turn signal is displayed; he or she merely has the option of
doing so. There is a marked tendency for the
turn signal to precede orderly (smooth) exchanges of the speaking turn, but it seldom
precedes instances of "simultaneous turns"
in which both participants attempt to hold
the floor at the same time.
Complementary to the turn signal is the
"gesticulation" signal. While the speaker is
gesticulating, the auditor should not attempt
to take the turn, regardless of whether the
turn signal is displayed. The evidence for
this formulation is that the auditor is much
less likely to attempt the turn when the
speaker is gesturing and that if the auditor
does, the result is usually simultaneous turns,
even when one or more turn cues are also
present.
Back Channels
The turn signal and the gesticulation signal
are both conceived to facilitate the smooth
exchange of turns. In further development of
the turn system, attention has been directed
toward speaker-auditor interchanges that occur during speaking turns. Duncan (1974) reported a study of auditor "back channels"
(head nods and statements such as "yeah"
and "uh-huh") that explored some of the
differences between back channels and turn
attempts and related auditor back channels
to the actions of the speaker.

729

Unlike turn attempts, auditor back channels


are not suppressed by the speaker's gesticulation. In addition, back channels and turn attempts are distributed differently with respect to the boundaries of segments into
which the conversations were divided for purposes of statistical analysis. It is likely that
these segments reflect structural properties
of speech and communicative body motion
(Duncan & Fiske, 1977, discuss them in detail). Turn attempts usually occur exactly
between segments, whereas back channels are
more loosely distributed.
Another important distinction between auditor back channels and turn attempts is that
they seem to be related to different sets of
cues given by the speaker. Duncan (1974)
hypothesized a speaker "within-turn" signal
originally consisting of two cues: grammatical
completion and turning of the head toward
the auditor. Brunner (1977) has reanalyzed
the data and found evidence for changing
the head turn cue to simple head direction
toward the auditor. Responses to the withinturn signal, like responses to the turn signal,
are conceived to be optional; the speaker is
not obliged to give a back channel, but may
do so if a signal is present.
The thesis of the present study is that the
placement of smile beginnings in conversations is closely parallel to that of back channels and that, in fact, smiling can be a type
of back channel. This idea was suggested by
the entirely serendipitous finding that most
auditor smile beginnings were preceded by
the within-turn signal. In order to claim, however, that the auditor's smile beginnings occupy a place that is similar to back channels
in the structure of interaction, three other requirements must be met.
1. There must be an acceptable statistical
relation between occurrence of a smile beginning and previous display of the within-turn
signal.
2. Auditor smile beginnings must be more
loosely distributed with respect to segment
boundaries than auditor turn attempts are.
3. The auditor's smile beginning must not
be suppressed by the speaker's gesticulation.

LAWRENCE J. BRUNNER

730

Method
Conversations and Transcription
The four conversations on which this report is
based are a subset of the ones described by Duncan
and Fiske (1977). Conversations 1, 2, and 3 are 7
minutes long. All three were between female law
students and male social work students at the University of Chicago; the participants were paid volunteers who were previously unacquainted. Conversation 4 is 19 minutes long and, unlike the other three
conversations, would have occurred whether or not
it was videotaped. It was between two 40-year-old
male psychotherapists from the Counseling Center at
the University of Chicago. They were good friends
and had known each other for about 10 years. The
subject of their conversation was a client whom both
of them had been seen recently. Duncan and Fiske
give more detailed background information about
these conversations. They refer to Conversations 1
through 4 as numbers S, 6, 7, and 2, respectively.
In all four conversations, the participants were
seated in adjacent chairs turned slightly toward each
other, facing a video camera. The participants in the
first three conversations were requested to get acquainted and carry on a short conversation, and in
the fourth they-were asked to continue a conversation that they had been unable to complete on a
previous occasion.
The conversations have been transcribed in considerable detail, but for the present study, only the
following actions were important: spoken syllables,
completions of subject-predicate clauses, intonation
(according to Trager & Smith's, 19S7, system), gestures, head movements, and smiles. In this study, the
term smile refers to apparent bilateral or unilateral
contraction of the risorius and buccinator muscles.
Each event was carefully located with respect to the
other events in the conversation so as to produce a
transcript in which the sequence of actions is represented, accurate to one syllable.
In transcribing body motion for this study, the
author and a research assistant worked together with
the goal of making a transcription on which they
completely agreed; in cases of doubt, other people
familiar with the transcription system were consulted. Typically, agreement between observers is
very high without discussion, and previous studies
of the reliability of the transcription system used
in the present investigation have yielded a median
kappa (Cohen, 1960) of .81 (roughly 95% agreement) between separate observers (Duncan & Fiske,
1977, p. 342).

Data Analysis
Unit of analysis. In order to assess behavioral
regularities within an interaction, it must be divided
into units to which scores are assigned. The reason is
that in order to present evidence that some action,
A, precedes B more often than one would expect by

chance, it is necessary to know not only how many


times A preceded B and how many times each action
occurred separately but also how often each action
did not occur. In the analyses to be reported here,
the unit of analysis is a stretch of speech that can
be as short as a single phonemic clause (Trager &
Smith, 1957) and as long as an entire speaking turn.
The boundaries of the units of analysis (which
will be called "segments" for convenience) occur
after the boundaries of phonemic clauses that contain one or more of the following actions: unfilled
pause, audible inhalation or exhalation, turning of
the speaker's head toward the auditor, a drop in paralinguistic pitch or loudness, any pitch level - terminal
juncture combination other than 2 2 | at the end of
a phonemic clause, paralinguistic drawl on the final
or the stressed syllable of a phonemic clause, termination of a gesture or relaxation of a tensed hand
position by the speaker, use of a stereotyped expression such as "but uh" or "you know," or the completion of a grammatical clause consisting of a subject and predicate. Duncan and Fiske (1977) explain
segments in much more detail.
In addition to serving as the basic units of analysis, segments are used to describe the distribution of
events within an interaction. The locations of the
auditor's actions with respect to segment boundaries
can be represented in a fourfold typology. The four
categories of location are the following:
1. Speech overlapfrom roughly the middle to
slightly before the end of the segment.
2. Pausebetween the last syllable of the unit in
question and the first syllable of the next unit.
3. Sociocentric sequencebefore or during a sociocentric sequence such as "but uh" or "you know."
4. Postboundaryfrom the very beginning to
roughly the middle of the segment after the one in
question.
Statistics. Evidence for the relation between the
speaker's within-turn signal and the auditor's smile
beginning will be presented in a fourfold table where
the columns indicate presence versus absence of the
signal, and the rows indicate presence versus absence
of a subsequent smile beginning by the auditor. If
the results are parallel to those obtained for back
channels, and the speaker's within-turn signal permits
an auditor smile beginning but does not require one,
almost every auditor smile beginning should be
preceded by the signal. If the signal operated perfectly, the cell representing auditor smile beginnings
in the absence of the within-turn signal would be
zero. There is no requirement, however, that the
other cell in that diagonal (failures to respond to
the signal) be zero or even have a particularly low
frequency; permissive signals allow the option of
not responding.
Consequently, the most useful index of association
between signal and response for the table just described is one that reaches its maximum when only
one cell is empty. The statistic Q (described by Yule,
1900, and further developed by Goodman & Kruskal,
1954, 1959, 1963, 1972) has this property: It equals

SMILES CAN BE BACK CHANNELS


1.0 or 1.0 when one cell or both cells on one diagonal of a 2 X 2 table equals zero. The Q also has
the advantage of being a true index of association,
directly interpretable as percentage of improvement
in predicting the dependent variable when the independent variable is known. The standard error of Q
can be calculated, and the hypothesis that Q is significantly different from zero may be tested by
dividing Q by its standard error and referring the
result to the normal curve. The Q and its two-tailed
probability value will be reported for all the 2 X 2
tables in this article.
Replication. Since the results to be reported here
are the products of very thorough exploration rather
than hypothesis testing, they must be replicated in
order to ensure that they do not merely reflect random contingencies in the data. I have used Conversations 1 and 2 for exploration and Conversations
3 and 4 for replication. Although pooled data are
reported for both the exploratory and replication
samples, the results for all eight participants were
examined separately to make sure that the same
trends held for all individuals.
The analyses to be reported do not include segments with auditor turn attempts or smiles continued
from the previous segment. Segments with turn
attempts were excluded because when a person smiled
as he or she claimed the speaking turn, it was often
impossible to decide whether to call the action a
speaker smile or an auditor smile. Segments with
auditor smile continuations were excluded because

731

the auditor could not begin to smile when he or she


was already smiling.

Results
The Within-Turn Signal
Previous research has shown that the
speaker's within-turn signal regularly precedes auditor back channels whose location
with respect to segment boundaries is late or
between units, but not those which are early
or occur before or during a sociocentric sequence. This is also the case for auditor smile
beginnings.
Table 1 shows that in both the exploratory
and replication conversations, the speaker's
within-turn signal (composed of grammatical
completion or head direction toward the auditor or both) preceded a high percentage of
the auditor's late and between-units smile beginnings and was significantly related to their
occurrence. In Conversations 1 and 2, 23 of
24, or 95.8%, of the auditor's smile beginnings occurred while the within-turn signal
was present. For the 2 X 2 table relating the

Table 1
Relationship Between Display of Speaker's Within-Turn Signal and
Subsequent Display of A uditor Smile Beginning
Speaker's within-turn signal
Auditor's smile
beginning

Proportion of
segments with
signal absent

Proportion of
segments with
signal present

Row total
of segments

Absent
Present

Pause and postboundary smiles: Exploratory sample"


(Conversations 1 and 2)
.27
.72
.04
.96

153
24

Absent
Present

Pause and postboundary smiles: Replication sampleb


(Conversations 3 and 4)
.26
.74
.00
1.00

487
IS

Absent
Present

Speech overlap and sociocentric sequence smiles0


(Conversations 1,2, and 4)
.23
.77
.33
.67

697
12

a Q = .79, p = .00006.
b
Q = 1.00, / > ^ 0 .
"<2 = -.26, p = .36.

732

LAWRENCE J. BRUNNER

auditor's smile beginning to prior display of


the speaker's within-turn signal, Q = .79, p =
.00006. In Conversations 3 and 4, all IS between-units and late auditor smile beginnings
happened in the presence of the within-turn
signal. Q = 1.0, and the associated probability
cannot be calculated because the standard
error is zero.
The bottom third of Table 1 shows that
there is an insignificant negative relation between presence of the speaker's within-turn
signal and auditor smile beginnings located
at speech overlaps or sociocentric sequences.
Similar results are reported by Duncan and
Fiske (1977) for vocal back channels and
head nods that occur in these locations.
Effect

of Speaker's Gesticulation

Auditor turn attempts are suppressed by


the speaker's gesticulation, but back channels
are not. To test whether smile beginnings are
also unaffected by the speaker's gesticulation,
data were arranged in 2 X 2 tables where the
rows indicated presence versus absence of the
speaker's gesticulation and the columns indicated presence versus absence of the auditor's
Table 2
Locations of A uditor Smile Beginnings and
Turn Attempts With Respect to Segment
Boundaries
Auditor turn
attempts

Fre-

Auditor smile
beginnings

Fre-

Location

Exploratory sample (Conversations


16
15.4
Pause
84
80.8
2
Postboundary
1.9
2
Sociocentric
1.9
sequence
104
Total

1 and
8
16
8
1

Replication sample (Conversations


13
22.0
40
Pause
67.8
Postboundary
2
3.4
Sociocentric
4
sequence
6.8
59
Total

3 and 4)
2
11.8
9
52.9
6
35.3

Speech overlap

Speech overlap

2)
24.2
48.5
24.2
3.1

33

0
17

smile beginning. In the exploratory sample


(Conversations 1 and 2), the auditor's smile
and the speaker's gesticulation were not significantly related (Q = -.36, p = .14).
In the replication sample (Conversations
3 and 4), it seemed at first that the auditor
was somewhat less likely to smile while the
speaker was gesticulating (Q = .45, p =
.04). These are not strong results, but it is
strange that smile beginnings should show a
structural similarity to turn attempts rather
than back channels, particularly when this
did not happen in the exploratory study.
Examination of the data from Conversations 3 and 4 separately indicates that there
is virtually no relationship between the speaker's gesticulation and the auditor's smile in
either conversation. Conversation 4 has a low
rate of auditor smiling and a high rate of
speaker gesticulating; Conversation 3 has a
relatively high rate of auditor smiling and an
extremely low rate of speaker gesticulating.
Consequently, although there is no relation
between the speaker's gesticulation and the
auditor's smile beginning in either conversation, an artificial one appears when data from
the two are pooled. This finding shows how
important it is to examine subsets of data
separately before combining them.
Table 2 compares the distributions of auditor smiles and turn attempts with respect
to segment boundaries. Auditor smile beginnings are less concentrated in the "pause"
location (the same finding that has been obtained for back channels), although this tendency is not as pronounced in the replication
study.
Discussion
The organization of auditor smile beginnings in these conversations is parallel to that
of back channels. They are both consistently
preceded by the speaker's within-turn signal,
they are both unaffected by the 'Speaker's
gesticulation, and they are both more loosely
distributed than turn attempts with respect
to segment boundaries. Taken together, these
results are grounds for placing auditor smile
beginnings in the category "back channel,"
along with nods and statements such as
"yeah" and "uh-huh."

SMILES CAN BE BACK CHANNELS

Nods and vocal back channels seem intuitively to belong together, because they both
seem to say, in effect, "Yes, you are understood. Proceed." Although it is not immediately obvious that smiles belong in this category, I believe that auditor smile beginnings,
like auditor back channels, make communication more efficient by providing the speaker
with feedback.
The idea of communicative efficiency lies
behind much of the turn system, though it
has not been given a great deal of emphasis.
For example, the turn signal was constructed
so that at least one cue would be present before each smooth exchange of turns, but not
before instances of simultaneous turns. The
reasoning was that a system of rules for taking speaking turns ought to result in smooth
exchanges when it is working properly, because human beings have only a limited capacity to send and receive information simultaneously.
The people who have written about the actions that we call back channels generally
agree that the auditor uses them to provide
the speaker with information. There is less
unanimity, however, concerning the exact sort
of information that back channels convey.
Fries (1952, p. 49) suggests that they are
primarily signs of attention and involvement
in the conversation. Yngve (Note 1) agrees,
but proposes that they are also "very important in providing for monitoring of the
quality of the communication" (p. 568). Dittmann and Llewellyn (1967) imply that in
addition, they are sometimes used for directly requesting further information or explanation (this is consistent, of course with
Duncan's treatment of short questions as back
channels).
Kendon (1967) distinguishes between
"point granting" and "assenting" signals, both
of which we would describe as back channels.
This distinction merely reflects the fact that
usually the auditor's feedback indicates that
he or she is following what the speaker says,
and sometimes it also indicates agreement. It
is important to mention in this connection
that head shakes and verbal statements of
disagreement were treated as back channels
in the present study and in Duncan's earlier

733

investigations. This was done mostly on intuitive grounds (they "felt" like back channels), but the implicit rationale was that feedback need not always be positive in tone,
although it usually is.
There is a consistency among the various
sorts of information that back channels are
supposed to provide. I would like to suggest
that back channels (including smiles) give
the speaker feedback on three levels. On the
first level, back channels signal the auditor's
involvement and participation in the interaction. They indicate that the auditor is attending to what the speaker says and that a conversation, not a monologue, is occurring.
On the second level, back channels provide
information about the auditor's level of understanding, allowing the speaker to adjust
his or her communicative endeavor so as to
get the ideas across efficiently. "Mhm's" and
head nods tell the speaker that the message
is being decoded successfully; when the location of these signals is "speech overlap"
(before the end of a segment of speech is
reached), they may indicate that the auditor
is "ahead" of the speaker and a little less
elaboration would be in order; a puzzled or
blank look may indicate that more information is needed (no work has been done on the
actions that signify the auditor's lack of understanding, except for clarifying questions).
On the third level, back channels can signal
the auditor's personal response to what the
speaker has just said. This might mean agreement or disagreement, shock, amusement,
scorn, or any number of other reactions. In
general, if an action functions on a higher level
it also functions on the ones below it; any
indication of a personal reaction implies understanding (or the lack of it), and any indication of a degree of understanding implies
some involvement in the conversation. Smiles
are higher level back channels, providing
feedback on all three levels. They indicate a.
positive personal reaction (often merely polite), as well as understanding of the preceding statement and participation in the conversation.
We do not know how far the results of the
present investigation can be generalized. The
findings of the exploratory study were fully

LAWRENCE J. BRUNNER

734

replicated, which indicates that at least they


are not unique to a particular conversation.
Definitive information about the limits within
which they hold can be provided only by further studies of smiling in which many people
are observed in many different settings.
Reference Note
1. Yngve, V. H. On getting a word in edgewise. In
Papers from the sixth regional meeting of the
Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 1970.

References
Berlyne, D. E. Laughter, humor and play. In G.
Lindzey & E. Aronson (Eds.), The handbook of
social psychology (2nd ed.). New York: Academic
Press, 1969.
Brunner, L. J. Smiling and laughter in four two-person conversations. Unpublished doctoral dissertation, University of Chicago, 1977.
Chapman, A. J. Social aspects of humorous laughter.
In A. J. Chapman & H. C Foot (Eds.), Humor and
laughter: Theory, research and applications. New
York: Wiley, 1976.
Chapman, A. J., & Foot, H. C. (Eds.). Humor and
laughter: Theory, research, and applications. New
York: Wiley, 1976.
Cohen, J. A. A coefficient of agreement for nominal
scales. Educational and Psychological Measurement, 1960, 20, 37-46.
Dittmann, A. T., & Llewellyn, L. G. The phonemic
clause as a unit of speech decoding. Journal of Personality and Social Psychology, 1967, 6, 341-349.
Duncan, S., Jr. Some signals and rules for taking
speaking turns in conversations. Journal of Personality and Social Psychology, 1972, 23, 283-292.

Duncan, S., Jr. On the structure of speaker-auditor


interaction during speaking turns. Language in
Society, 1974, 2, 161-180.
Duncan, S., Jr., & Fiske, D. Face-to-face interaction.
Hillsdale, N.J.: Erlbaum, 1977.
Duncan, S., Jr., & Niederehe, G. On signaling that
it's your turn to speak. Journal of Experimental
Social Psychology, 1974,10, 324-247.
Fiske, D. W. Strategies for personality research.
San Francisco: Jossey-Bass, 1978.
Fries, C. C. The structure of English. New York:
Harcourt, Brace, 19S2.
Goffman, E. Relations in public. New York: Harper
& Row, 1971.
Goldstein, J. H., & McGhee, P. E. (Eds.). The psychology of humor: Theoretical perspectives and
empirical issues. New York: Academic Press, 1972.
Goodman, L. A., & Kruskal, W. H. Measures of association for cross-classifications. Journal of the
American Statistical Association, 19S4, 49, 732764.
Goodman, L. A., & Kruskal, W. H. Measures of association for cross-classifications, II: Further discussion and references. Journal of the American
Statistical Association, 1959, 54, 123-163.
Goodman, L. A., & Kruskal, W. H. Measures of association for cross-classifications, III: Approximate
sampling theory. Journal of the American Statistical Association, 1963, 58, 310-364.
Goodman, L. A., & Kruskal, W. H. Measures of association for cross-classifications, IV: Simplification
and asymptotic variances. Journal of the American Statistical Association, 1972, 67, 415-421.
Kendon, A. Some functions of gaze direction in social interaction. Acta Psychologica, 1967, 26, 22-63.
Trager, G. L., & Smith, H. L., Jr. An outline of English structure. Washington, D.C.: American Council of Learned Societies, 1957.
Yule, G. U. On the association of attributes in statistics. Philosophical Transactions of the Royal
Society of London, 1900, A 194, 257-319.
Received February 14, 1978

S-ar putea să vă placă și