ABSTRACT

"VOCAL" COMMUNICATION: THE EFFECTS OF RATE
(SPEED) AND INTENSITY (LOUDNESS) ON RESPONSE
TO SPOKEN MESSAGES

By
Alfred Esimatemi Opubor

People respond not only to the referential content
(or the what) of spoken messages, but also to the non-
lexical or "vocal" content, (the how). This study investi-
gated the effects of two "vocal" speech variables, rate
(speed) and intensity (loudness) on response to messages
spoken in a language unfamiliar to listeners.

Each message had been mechanically manipulated to
combine one of three levels of rate with one of three
levels of intensity. Nine combinations were obtained:

Loud + Fast, Loud + Normal, Loud + Slow; Medium + Fast,
Medium + Normal, Medium + Slow; Soft + Fast, Soft + Normal,
Soft + slow. The basic or "control" message combined
Medium intensity and Normal rate. The recordings were in
Marathi, a language of western India.

Four hundred forty-five English-speaking high school
students were randomly assigned to nine groups, one control
and eight experimental. Each group listened over head-

phones to a recording of the basic message and one

Alfred Esimatemi Opubor

appropriate experimental message. They responded to each
message on rating forms containing semantic-differential-
type scales, representing three dimension of effect, namely,
evaluation, strength and activity. Mean differences between
group responses to the basic and experimental messages were
calculated for each effect dimension.

Analyses of variance and t—tests involving these
means were used to test six main effects hypotheses, four
alternative main effects hypotheses and three interaction
hypotheses.

The results indicate that in general slower and
softer messages are more favorably evaluated and are con-
sidered more "forceful" than any other combinations, but
that slower ang_medium intensity messages are considered
most "dynamic" by listeners.

The implications of these results for a strategy of
message preparation, as well as the possibilities and nec-
essity for cross-cultural investigation of various "vocal"

phenomena were discussed.

"VOCAL" COMMUNICATION: THE EFFECTS OF RATE
(SPEED) AND INTENSITY (LOUDNESS) ON

RESPONSE TO SPOKEN MESSAGES

By

Alfred Esimatemi Opubor

A THESIS

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

DOCTOR OF PHILOSOPHY

Department of Communication

1969

IECopyright by
ALFRED ES IMATEMI OPUBOR

1970

Accepted by the faculty of the Department of Communi-
cation, College of Communication Arts, Michigan State Uni-
versity, in partial fulfillment of the requirements for the

Doctor of Philosophy degree.

Director of Theéis

 

Guidance Committee: My WA

 

airm n V”
mix/5M4

7 V

 

 

 

ACKNOWLEDGMENTS

The present study is but one manifestation of the
convergence of the different intellectual traditions,
influences and interests to which I have been willingly
captive in the last twelve years. During that time, I
have had the good fortune of contact with many people on
three continents. Some of them were fellow students;
some were teachers who became friends and colleagues.

Among the latter, I recall with special delight
and gratitude, John w. Spencer, who first stimulated and
then encouraged my interest in the serious study of
language.

My doctoral study in Communication beneffited greatly
from the academic guidance and humanity of Dr. Erwin P.
Bettinghaus, who also supervised this study and generously
provided funds for analyzing the data.

The other members of my guidance committee--Drs.
Randall P. Harrison, Hideya Kumata, Irvine Richardson and
Everett M. Rogers gave advice and illuminating feedback
at various stages of this study. Dr. James A. Noonan, who
also served on the Committee, helped shape my thoughts

early in the game, before his departure from the

ii

University. I am indebted to all these scholars and to
Dr. Oscar Tosi of the Department of Audiology and Speech
Sciences who patiently answered many questions and helped
record the messages used in this study.

Encouragement and assistance of a non-academic
nature pertinent to the completion of this study were
provided by the following:

Professor James P. Wang, Chairman, Department of
Linguistics, made funds available for data processing.

Mr. William Helder, Coordinator in the Social Science
Teaching Institute, made my entry into the Lansing School
System enjoyable and productive.

Mrs. Vivian Riddle, Chairman, Department of Foreign
Languages, Everett High School, Lansing, cooperated with
the demands of this experiment and gave of her time and
the facilities of the language laboratories at Everett to
ensure its success.

Mr. Steve Lucas assisted with the collection of data,
as did also Miss Signe Nelson who undertook a number of
busy administrative and coding tasks.

My deepest gratitude goes to Amelia, wife and friend,

whose love makes all good things possible.

111

TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS . . . . . . . . . . . ii

LIST OF TABLES O O O O 0 O O O O O I V

LIST OF FIGURES . . . . . . . . . . . vi
Chapter

I. THE PROBLEM AREA . . . . . . . . 1

Introduction. . . . . . . . . 1

"Vocal" vs. Verbal Communication . . l

The Problem . . . . . . . . . 5

Review of Relevant Literature . . . 6

Theoretic Hypotheses . . . . . . 15

II. METHOD AND PROCEDURES. . . . . . . 19

Message Construction . . . . . . 19

Subjects . . . . . . . . . 23

Experimental Design . . . . . . 2“

Data Analysis . . . . . . . . 26

Experimental Procedures . . . . . 27

III. RESULTS . . . . . . . . . 30

Test of the Main Hypotheses. . . . 31

Summary and Discussion . . . . . 35

IV. CONCLUSIONS AND FURTHER RESEARCH . . . 38

BIBLIOGRAPHY . . . . . . . . . A9

APPENDICES. . . . . . . . . . 53

iv

LIST OF TABLES

Mean Evaluation "Effect" Scores for All
Nine Groups . . . . . . . . .

Analysis of Variance Among Evaluation
"Effect" Means . . . . . . . .

Mean Potency "Effect" Scores for All
Nine Groups . . . . . . . . .

Analysis of Variance Using Above Means
(Potency). . . . . . . . . .

Mean Activity "Effect" Scores for All
Nine Groups . . . . . . . . .

Analysis of Variance Using Mean Activity
"Effect" Scores. . . . . . . .

Page

32

32

33

3A

3A

35

Figure

1.

LIST OF FIGURES

The Relationship Between Psycholinguistic,

Extralinguistic and Paralinguistic

Studies (as deduced from the discussion

in Mahl and Schulze, 196A) . . .

Set Up for Intensity Recording. . .

 

Message Treatment Groups. . . . .

Graphic Representation of Hypotheses.

vi

Page

22
25
27

CHAPTER I

THE PROBLEM AREA

Introduction

 

The study of messages is an inevitable aspect of
the study of "the process of communication." Indeed, the
process of communication can be profitably looked upon as
"the process of message manipulation." Sources encode and
transmit messages; receivers decode and are affected by
messages.

Messages come in different forms; they are produced
and apprehended through many different mechanisms. In
communication research, the division of messages into

verbal and nonverbal is fairly widely accepted. So also

 

is the subdivision of nonverbal messages into such modality-
dependent types as kinesics, "the systematic study of the
visually sensible aspects of nonverbal interpersonal communi—

cation" (Birdwhistell, 1952) and proxemics, the study of the

 

use of space in communication (Hall, 1959).

”Vocal" vs. Verbal Communication

 

°In the literature it has been mentioned with increas-
ing frequency that verbal messages--at least in their pri-

mary spoken form—-are a composite of two largely independent

systems. The observations that led to this view are
common-place enough. Speakers of most languages seem to
recognize the phenomenon reflected by the statement:

"it is not E232 he said, but how he said it." Or again,
if we consider the situation of mediated communication
e.g., via the telephone, or a loudspeaker system, we find
that peOple are generally capable of making fairly accur-
ate guesses about certain demographic and even psycho—
logical attributes of their interlocutor regardless of
what he is talking about.

Soskin first formalized these impressions in com—
munication terms by suggesting that speech comprises two
channels of communication—-the voca1,and the verbal or
lexical. According to Starkweather "the verbal component
is the pattern of sound which results in words and phrases,
and other linguistic units making up the content of speech.
The vocal components are all the remaining characteristics
of sound which may be called tonal variation or voice
quality" (1956, p. 39“).

The verbal or linguistic channel carries the poten-
tial referential information, while the vocal or non-
linguistic channel is the repository of potential affective
information. Spoken messages without the verbal component
may be said to be "content free." A typescript of content-

free speech would be almost impossible to achieve.

The term "content-free" as used by Starkweather and
his associates describes specially produced and treated
recordings of speech. Generally, by passing magnetic
tapes of speech through mechanical filters, they remove
the high frequency components of the speech signals.
Since much of the verbal component of speech--the vowels
and consonants--is carried by the higher frequencies,
much of this kind of information is alleged to be lost in
the filtering process. However, the lower frequencies
which contain the vocal components of speech remain
intact.

The major concern of Starkweather and others who
are similarly inclined, is to remove, from the effect of
a given message, the contribution of semantic information.
In this way, they hope to be able to specify more accur-
ately what the vocal channel contributes to speech com-
munication. The vocal channel has different components,

among them the following:

Intensity Rhythm
Pitch Rate
Quality

Together these include what have been variously re-
ferred to at different times, and by different authors, as
paralinguistic, prosodic, non—lexical, extralinguistic,
supra- and sub-segmental phenomena (Trager 1958; Crystal

and Quirk, 196A; Mahl and Schulze, 196A; Pike, 1967).

Electronic filtering devices are only one method

of producing "content—free" messages. Older attempts

have included the use of:

8..

"Meaningless" content such as numerals and

 

letters of the alphabet. In a number of
experiments summarized by Kramer (1962)
speakers were asked to portray certain
emotions by recording only numerals or
letters of the alphabet. Listener recog-
nition of the portrayed emotions were better
than chance (Skinner, 1953; Knower, 19Al;
Pfaff, 195A; Davitz and Davitz, 1959).

Constant content. The rationale for this

 

method is that if the words are held constant,
and if they do not contain emotional content,
their contribution to differences between
utterances is negligible. Through the use

of actors, different emotions are portrayed
and are identified by listeners (Fairbanks,
1940; Fairbanks and Hoaglin, 19Al).

Ignoring content. Verbal content is "ignored"

 

through the measurement of non—verbal properties
of speech such as speech rate, breathing rate,
duration of pauses, the nature and duration of
non-fluencies. These prOperties of speech are

regarded as largely independent of content

thus they can be used to differentiate
different emotions (Goldman-Eisler, 1955, 1956;
Mahl, I959; Dittman and Wynne, 1961).

d. Use of foreign language. The verbal content of

 

a message in a foreign language conveys virtually
no meaning to someone who does not know the
language. Using this rationale, Kramer asked
native Japanese actors to portray different
emotions using Japanese utterances. American
college students who knew no Japanese were

then asked to identify the emotions portrayed

and did so better than chance expectations

(Kramer, 1962).

The Problem

 

The present study focuses on two vocal message vari-

ables—-intensity and rate of speech.

 

We are interested in these two variables for two
reasons. Firstly, previous research has operationalized
them in ways that leave something to be desired. Secondly,
the relationship of the intensity and rate of a message to
its effect has not been well studied; perhaps because, as
Mahl and Schulze suggest, prior work in this area has not
been interested in the response of the listener (196A).

The research question we are asking in this study is:
what is the effect on a listener of differences in the loud-

ness and speed of a spoken message?

Review of Relevant Literature

 

Interest in psychological aspects of language goes
back a long time; long before the term "psycholinguistics"
came to be applied to particular conceptual approaches to
the study of human language. Certainly psychologists have
enaged in speculation, though few have actually conducted
research, on questions related to the interests of this
study.

One of these questions concerns the kinds of in-
ferences that can be made from a person's voice about his
personality. In his 1960 review of voice and personality
studies by psychologists, Diehl traces the history of the
study of the voice—personality relationship through the
preoccupation with rhetoric in 18th century England and
nineteenth century America to the more "scientific" and
technologically—assisted work beginning in the early 1930's
and continuing to the present.

During the last three decades approximately 80

publications have appeared in the area of voice

and personality. Inquiries have taken four major

directions:

(a) Judgments of physical and psychological
characteristics from the recorded or
concealed voice;

(b) Relation between effective public speak—
ing and personality;

(c) Specific vocal characteristics repre—
senting personality traits; and

(d) Vocal characteristics of psychoneuroses
and psychoses (Diehl 1960, pp. 179—180).

The first and third lines of inquiry above are

closest to the one pursued in this study.

Diehl cites research to Show tendencies in the
relationship between certain voice characteristics such
as pitch, force, quality,speech melody, horaseness, etc.,
and various psychological characteristics such as happi—
ness and other emotions, social adjustment, intelligence,
extroversion-introversion, emotional stability, etc. His
summary of existing research is that;

. . . although developments in voice science have

provided more objective ways to study relationships

between voice and personality, a large portion of
the work still remains to be done. There seems to
be sufficient theory based on empirical evidence;
however, the paucity of experimental data at the
present time makes it difficult, if not impossible,
to apply in a systematic way any of the relation—

ships which may be suspected (Diehl, 1960, pp. 19A-
195).

A more extensive and more recent survey of relevant
research from a psychological perspective is by Mahl and
Schulze (196A). Their review covers what they call the
"extralinguistic area," which is broader than Diehl's
voice-personality area. It subsumes Trager's paralanguage

and includes "the functional relationships between the

 

extralinguistic phenomena and the non-linguistic states
and processes in the speaker and listener interacting in
the communication situation" (Mahl and Schulze, 1964,

p. 52). But the extralinguistic area is itself included
within psycholinguistics. The relationships may be

reduced to the following diagram:

 

Psycholinguistics

The extralinguistic
area

 

Paralanguage (Trager)

 

FIGURE l.--The relationship between psycholinguistic,
extralinguistic and paralinguistic studies (as deduced from
the discussion in Mahl and Schulze, 196A).

Psycholinguistics is the broadest and most inclusive
of the three areas depicted. An early working definition
by Osgood and Sebeok states that psycholinguistics "is
concerned in the broadest sense with relations between
messages and the characteristics of human individuals who
select and interpret them" (Osgood and Sebeok, 1965, p. A).
Trager's work is the first major attempt to systematize
the study of nonlinguistic properties of speech, by first
delimiting the area and then providing a taxonomy and
methodology. His pioneering effort has had obvious in—
fluence on subsequent work in the field. He subdivides

paralanguage into two components: voice qualities and

 

vocalizations, each of which is further subdivided thus:

 

Voice Qualities Vocalizations

 

 

Pitch range Vocal characteristics

Vocalic control
Glottis control Vocal qualifiers
Pitch control

Articulation control Vocal segregates
Rhythm control

Resonance

Tempo

Trager then provides verbal descriptions of how
these features are manifested in speech, how they can be
described. But his work does not go beyond describing the
message code to consider the characteristics of either
receiver or sender as Mahl and Schulze do.

Mahl and Schulze divide former systems for classify-
ing extralinguistic phenomena according to two dimensions:
Behavioral-Paralinguistic, depending on the methods of
analysis which were employed in-these systems. This

. . . distinction between Behavioral and Para-

linguistic methods does not imply that . . .

different classes of phenomena or different levels

of discourse are involved. The basis for making a

distinction . . . seems . . . largely a matter of

emphasis. Those investigators who use Paralinguis—
tic methods of analysis usually are more concerned
with individual—-that is, discretely identified-—
instances of certain features of utterances, in-
cluding deviations from characteristic patterns.

Usually an exhaustive analysis is made, in which

scores attempt to identify all instances of all

phenomena included in the entire classification

system of paralanguage (196“, p. 53).

In the second section of their survey, Mahl and
Schulze look at "systematic studies of extralinguistic
phenomena." They conclude from the work examined in

this section that:

10

One of the most striking revelations of the survey
is the absence of systematic study of how the extra-
linguistic phenomena affect the listener's spon-
taneous,_communicative behavior of his underlying
psychological states and processes (p. 78).

This statement is closely related to the reasons
why the present study is being undertaken, and reveals the
tangential nature of previous source-oriented research
findings to our goals here. But such studies can provide
a basis of hypothesizing certain receiver responses to
given values of intensity and rate.

Mahl and Schulze make the following summary of
available research with regard to rate (speed) of speech.

Most of the investigations of Speech Rate support
the hypotheses of Kanfer and Miller, that Speech Rate
is_positively related to anxiety. However, since
Speech Rate has been showngto be largely determined
by duration and frequency of pauses, the many dis-
crepancies found among investigations of Silent
Pauses suggest that the generality of findings obr
tained from Speech Rate studies may also be limited.
So far, very little work has been done with Articu—
lation Rate, even though it has been shown more
sensitive to certain physiological changes than
Speech Rate (1964, p. 98).

 

 

As for Loudness, the one study reported in this
section is in the "Paralinguistic methods of analysis"
subsection:

Eldred and Price . . . studied 15 psychotherapy

interviews of one patient. They did not find any

evidence of a relationship between anxiety ratings

and Pitch, Rate, or Loudness (p. 101).

Waskow, in her study of the effect of drugs on

speech behavior, was interested in the changes in affect

which occur as a result of the administration of drugs

11

such as LSD to subjects. She surmised that vocal aspects
of speech would be sensitive to changes in affect. She
therefore obtained a sample of speech from subjects who
had received doses of drugs known to have different
somatic effects, subjected these speech samples to a
mechanical filtering device, thus rendering them "content
free," and had the resultant vocal samples rated on forty
semantic—differential rating scales.
Since her ultimate purpose was to determine some of
"the dimensions along with listeners respond to the voice,"
. . . the mean ratings (of five raters) on the
forty scales were intercorrelated and a principal
components factor analysis was performed. . . . .
After study of the factors that emerged, fifteen
scales were selected as looking particularly pro-
mising, and these were then used in the first
application of the scales. . . .
She found that the fifteen scales grouped themselves in
"3 clusters of 3 scales each, and 6 additional scales."
The results she presents are for one of the three—scale
Clusters, consisting of the scales "annoyed--pleased,"
"tense-—relaxed" and "complaining--content." She then
correlated the scores on this factor with three of the
six factors on the Clyde Mood Scale. She found that
"scales which, a priori, would seem to be related to
vocal tension were correlated with the vocal scores."
In other words, then, listeners were able to identify

the emotional state of speakers by listening to vocal

messages produced by these speakers (Waskow, 1967).

12

Perhaps the most directly relevant study for our
purposes is that by Davitz (1964). For that reason we
will discuss and quote from this study at some length
to indicate where our practices differ from his and
demonstrate the need for this study. Davitz was pri—
marily interested in

. . . some of the auditory cues associated with
vocal expressions of emotional meaning. To in-
vestigate this problem, two sets of variables were
selected, one set concerned with dimensions of
emotional meaning, and a second concerned with
auditory characteristics of speech. The variables
of emotional meaning were based on Osgood, Suci,
and Tannenbaum's research, which suggested three
aspects of emotional meaning: (1) Valence, (2)
Strength, and (3) Activity. Selection of speech
variables was based on the characteristics of
sound commonly considered in psychophysical studies
of audition: (1) loudness, (2) pitch, (3) timbre,
and (A) rate of speech. The research considered
the relation between each of the three variables
of emotional meaning, valence, strength, and
activity, and each of the four speech variables

of loudness, pitch, timbre and rate (p. 101).

What Davitz refers to here as "Valence" is more commonly
known as "Evaluation" or the "Evaluative" dimension. We
will use these more common labels throughout this study.
Davitz asked seven speakers to read two sentences
in fourteen different contexts so that fourteen different
emotions were expressed. Each speaker also read the
sentences in a "non emotional" context.
These readings were tape recorded, and then the
readings of the two sentences identical in content
for all paragraphs were taken out of context and
spliced so as to provide a recording of different

emotional expressions with standard content (196A,
p. 102).

13

The recordings of all seven speakers were
judged by 20 persons who were given the list of
IA feelings and asked to identify the feeling ex-
pressed by each reading of the two standard
sentences after hearing the non emotional 'normal
conversation' reading by each speaker. . . . The
final tape used for ratings of vocal characteris-
tics consisted of the reading of each feeling
identified most frequently for male Speakers and
for female speakers, plus the non emotional
"normal conversation" reading by each speaker
(1964, p. 103).

The final tape was played to a second set of
20 judges who were asked to rate each expression
on four seven—point scales dealing with (l) loud-
ness (loud to soft); (2) pitch (high to low);
(3) timbre (blaring to resonant); and (A) rate
of speech (fast to slow). A speaker's expression
of a given feeling was played twice, each time
followed by a recording of that apeaker's non
emotional, "normal conversation," which served
as a baseline against which the motional expres-
sion was rated (196A, p. 104).

Davitz's results revealed that "activity was posi-
tively and linearly related to the speech variables, but
both valence and strength appeared to be related curvi-
linearly, if at all, to the speech variables." He there-

fore generalized as follows: "In the vocal communication

 

of emotional meanings, auditory cues of loudness, pitch,
timbre, and rate are a function of the subjectively

rated activity level of the feeling communicated (p. 108).

 

He further states that "Presumably, valence and strength
are communicated by other, perhaps more subtle and complex,
auditory cues" (p. 122).

Many aspects of the Davitz study invite comment.
First, the number of judges on whose ratings his results

are based--twenty--is extremely small. Davitz himself

1A

admits that this limits the extent to which his findings
may be generalized. He also considers as a drawback "the
relatively simple vocal characteristics considered . . ."
and "the specificity of the categories of emotion."
Furthermore, each emotion is a composite of different
values on each vocal variable, and changes in the con-
text of communication could lead to changes in the
particular value of each variable called into play.

Thus anger may be accompanied by excessive loudness in
one instance, and by a general tautness and quiteness in
another (1964, p. 106).

Davitz used the "content standard" method to elimi—
nate bias introduced through the verbal channel, but it
is possible to raise questions about this method. How
suitable are the two embedded sentences to the expression
of each of the emotions elicited? For if the verbal con-
tent of a message is more appropriate to the expression
of one emotion than another, we cannot correctly argue
that such content is "neutral." In fact, we might have
considerable interaction effects between verbal and vocal
variables in such circumstances.

The Davitz study is the most ambitious attempt to
investigate the relationship of vocal variables to
aspects of connotative or emotional meaning. However,

it raises fundamental reservations in terms of the

15

operationalization of vocal variables and the interaction
between verbal and vocal channels. The present study is
an attempt to remedy some of the more obvious limitations
of the Davitz study, and thus provide a sounder basis for
the investigation of the dynamics of communication through

the vocal channel.

Theoretic Hypotheses

 

To recapitulate, the research question we are ask-
ing is: what is the effect on a listener of differences
in the rate (speed) and intensity (loudness) of a spoken
message? The hypotheses are based on the assumption that

we can match message variables with response variables;

 

 

that given certain message variables, we can predict
their effect in terms of receiver response. Specifically,
we hypothesize as follows:
H—l. Rate of speech is positively related to the
perceived potency of a message, i.e., the

faster a message is, the more "potent" it
will be rated by a listener.

 

H—2. Rate of speech is positively related to the
perceived activity of a message, i.e., the
faster a message is, the more "active" it
will be rated by a listener.

H-3. Rate of speech is positively related to the
evaluation of a message, i.e., the faster a
message is, the more favorably it will be
rated by a listener.

 

H-A. Intensity of speech is positively related to
the perceived potency of a message, i.e., the
louder a message, is, the more "potent" it
will be rated by a listener.

 

H-5. Intensity of speech is positively related to
the perceived activity of a message.

 

16

H—6. Intensity of speech is positively related to
the evaluation of a message.

 

 

These hypotheses presuppose linear relationships be-
tween message variables and response variables. That this
may not be the case has been suggested by Davitz who found
non significant correlation ratios between his four message
variables and two of his response variables, evaluation and
strength. He states, however, that "Although the corre-
lation ratios obtained indicate curvilinear relations be—
tween valence and strength and the four speech variables,
the magnitude of the ratios do not permit generalization
beyond the present data" (1964, p. 106). Assuming that
what Davitz found holds true for other data, we will pro—
pose the following alternative hypotheses:

AH-l. Rate of speech is curvilinearly and positively
related to the perceived potency of a message.

AH—2. Rate of speech is curvilinearly and positively
related to the evaluation of a message.

 

AH—3. Intensity of speech is curvilinearly and
positively related to the perceived potency
of a message.

 

AH-A. Intensity of speech is curvilinearly and
positively related to the evaluation of a
message.

 

 

There are no alternative hypotheses regarding message
effects on activity as Davitz's findings agree with the
relationship which we have hypothesized here.

The foregoing hypotheses and alternative hypotheses
refer to main effects. However, as we are dealing with

two message variables, both co—present in all messages,

17

we must take their possible joint effects into account.
There is not much suggestion in the literature about
what the interaction effects of these two variables
would be.

In View of this situation, we will invoke some
intuitive reactions to the nature of the message elements
involved, as well as the reasoning which led to our main
effects hypotheses, and suggest the following interaction
hypotheses:

Interaction Hypotheses

 

H—7. In combinations of gate and intensity, the
perceived activity of the message will be
positively and linearly related to rate of
speech, and positively curvilinearly re-
lated to the intensity of speech.

 

 

H-8. In combinations of rate and intensity, the
perceived strength of the message will be
positively curvilinearly related to the
intensity of the message and negatively
related to the rate of the message.

 

 

H-9. In combinations of rate and intensigy the
evaluation of the message will be curvi-
linearly and positively related to the rate
of speech and curvilinearly and negatively
related to the intensity of speech.

 

 

 

To test these hypotheses, a study was conducted in
which subjects were required to respond on scales to three
different levels of the two vocal variables--rate and in—
tensity--and combinations thereof. Intact groups of
subjects were randomly assigned to one of nine treatment
groups-—eight experimental and one control. Each group

was then exposed to one of the experimental messages and

18

the control message, the exposures being separated by a
time interval. The procedure adOpted is more fully

described in the next chapter.

CHAPTER II

METHOD AND PROCEDURES

This chapter describes the construction of messages,

the operationalization of the independent and dependent

variables, the sample, the experimental design and tech-

niques of analysis.

Message Construction

 

As was stated earlier (p. A), many methods for iso-
lating the vocal channel have been attempted in previous
studies. [The method adopted here is the use of a foreign
language--i.e., a language unfamiliar to the receivers.
The rationale behind this method is that the verbal con-
tent of a message in a foreign language is completely in-
accessible to the receiver; but this is not necessarily
the case for the nonlexical features of the message.
Studies by Miron (1961) and Kramer (1962) among others
suggest that insofar as recognition of portrayed emotion
is concerned, receivers are able to respond with better
than chance accuracy to vocal features. In effect then,
certain vocal characteristics seem to Operate as communi-

cation cues regardless of the language in which they occur.

19

20

A male speaker of Marathi, a language of western
India, translated the following passage from English into
Marathi: "There is no other answer. You've asked me that
before. My reply has always been the same, and it will
always be the same." This passage is the same as that
used in the study by Kramer, and is a modified version
of that used in the work of Fairbanks and his associates.
A version of it was also used by Davitz.

The Marathi speaker recorded this passage in his
language in a sound-proofed studio. He was instructed to
read it with "as little emotion as possible" at a moderate
speed. This version 10 seconds long was retained as the
"basic" message insofar as pgpg'was concerned.

The tape of this basic message was then sent to the
Gotham Audio Corporation in New York City for expansion
and compression, on their "Electro Information Rate
Changer." This is a device which can compress a message
up to nearly half (53%) of its original length, or expand
it up to twice (200%) its original length with very little
distortion.

We asked Gotham to compress the Marathi tape to
75 per cent of the original and expand it 150 per cent.
These two versions operationalized the "fast" and "slow"
levels of message rate, while the original tape was

retained as the "normal" level.

21

We decided to manipulate intensity also in objective
terms. The three "rate" versions described above were put
on one tape. This tape was then dubbed on equipment dia-
grammed in Figure 2, at the Michigan State University
Audiology and Speech Sciences Laboratory.

The dubbing produced three tapes with different
levels of intensity--40 decibels, 60 decibels and 80
decibels. These levels operationally define "soft,"
"medium," and "loud" intensity as used in this study.

At the end of this manipulation, we emerged with
three sets of recordings each with three versions of the
original message. The levels of the two variables are

as follows:

 

Intensity Rapg

Loud (80 db) Fast (150%)
Medium (60 db) Normal (100%)
Soft (no db) Slow (75%)

The combinations which these levels produce are as follows:

Loud + Fast Medium + Fast Soft + Fast
Loud + Normal Medium + Normal Soft + Normal
Loud + Slow Medium + Slow Soft + Slow

The dependent variable in this study is message
Effect, which has three dimensions, viz:

A. Evaluation represented by the scales:

 

Good - Bad
Pleasant - Unpleasant

Beautiful - Ugly

22

 

es H

.wcHUhoOOL smpamcmpcfi: you as pmmll.m mmDUHm

 

 

 

 

hop ooom nozom Ho>oq

 

poopso

 

AHV -

Appooomv me hoohooom mama

 

 

 

G O

 

 

 

Qomnmmae :1 monsoomm mama

23

B. Strength represented by the scales:
Strong — Weak
Large - Small
Thick - Thin
Heavy - Light
C. Activity represented by the scales:
Active - Passive
Fast - Slow
Shrill - Mellow
Sharp — Dull
For each group of subjects, scores for the scales
representing each facet are summed and the mean score is
obtained. The difference between the group mean score
for the "basic" message and the group mean score for the
experimental message is the effect of the experimental

message for the particular dimension under consideration.

Subjects

The subjects for this study were 4A5 students at
Everett High School in Lansing. Since the mechanics of
administering the messages required language laboratory
facilities, most of the subjects were enrolled in modern
language courses studying Spanish, French and German.
About one-sixth of the subjects were, however, not en—
rolled in these language classes, but participated in the
experiment in their English and American literature

classes. All exposure to the messages was nevertheless

2A

in the language laboratory, during periods regularly
scheduled for classes. The breakdown of the pOpulation

according to demographic variables is as follows:

 

 

11% .8335.
I 15.9 Male = 172
s.d. 1.6 Female = 270
No report = 3
7474—57
Year in School Language Studied
SOphomores 223 French 130
Juniors 153 Spanish 67
Seniors 66 German 78
No report 3 English &
American
Literature 169
No report 1
Ira—5' mm

Experimental Design

 

The independent variables were: (1) 2&2! versus
normal versus £a§p_message rate, and (2) £233 versus medium
versus lppd_message intensity. The dependent variable is
message effect in terms of perceived valence, strength and
activity.

Basically, we have a two variable 3 x 3 factorial
design, as demonstrated in Figure 3. The groups were
randomized to the different treatment conditions.

These nine combinations represent all possible
combinations of our two independent variables-~INTENSITY
and RATE.

Each group received two messages--the basic or

"control" message, and one of the experimental messages--

25

 

 

 

RATE
FAST NORMAL SLOW
LOUD Group #1 Group #2 Group #3
Group #5
INTENSITY MEDIUM Group #A Basic or Group #6
"Control"
Message
SOFT Group #7 Group #8 Group #9

 

 

 

 

 

FIGURE 3.-—Message treatment groups.

except the control group (#5) who received the basic message
twice. Because of the logistics, both messages were ad-
ministered about half an hour apart. For example subjects
in Group #1 received the basic or control message first;
then they engaged in activity related to their school
work—-the teacher made some announcements in French and
answered questions. Then the subjects received the “Loud+
Fast” message.

The experiment was conducted on two separate days.
On the first day, all subjects used the language laboratory
equipment, and heard the messages through earphones. It
was discovered, on this occasion, that the "soft" level

on the intensipijariable was extremely difficult for

 

subjects to hear as the ambient noise of the laboratory

26

machines was about 35 db. The groups affected were num—
bers 7, 8 and 9, especially the last two. During the
second session, the messages were therefore played over
loud speakers. In our sample, all the subjects in
Groups number 8 and 9 and most of Group 7 heard the
messages only over loud speakers. We were therefore
interested in the possibility of differences in equip-
ment—-between earphones and loudspeakers——introducing
bias into our results. To check on the effect, if any,
of these differences, we randomly assigned subjects on
the second (i.e., the loudspeaker) day to Groups 1-7.

In our analysis, we will compare the response of those
subjects in Groups 1-7 who heard the message over loud-
speakers with the response of those who heard it through

earphones.

Data Analysis

Responses to both control and experimental messages
were matched on the basis of name, so that we could compare
a subject's response to the experimental message with his
response to the control message.

All the theoretic hypotheses and their alternatives
require that we relate each of the message variables--
rate and intensity--with each of the facets of the effect
variable--evaluation, strength and activity. We have

hypothesized the following relationships:

27

 

 

 

 

 

 

 

 

 

Evalua- Potency Activity
tion
Intensity (H6) Intensity (HA) Intensity (H5)
or or or
Rate (H3) Rate (H1) Rate (H2)
Evalua-
tion Potency
Rate (AH-l) Rate (AH-2)
or or
Intensity (AH-3) Intensity (AH—A)

FIGURE A.--Graphic representation of hypotheses.

Furthermore, H-l through H-6 are tested through
various t-tests and analyses of variance comparing
differences in mean effect scores among the various

treatment groups.

Experimental Procedures

 

At both sessions, the following procedures were

followed in sequence:

1. Subjects entered the language laboratory and
sat in their regular places (in the case of
modern language students) or in any available
booth (in the case of the English and American
literature students). The classroom teacher

then introduced the experimenter who explained

28

that he was involved in a study of "Language
Behavior" under the auspices of the Department
of Communication at Michigan State University,
and sought the cooperation of the subjects.
Subjects were then asked to read the first

page of materials placed in the booths prior

to their arrival. This page explained the
kinds of situations in which the study was
interested (see Appendix I, p.55 ).

Subjects were informed that they would hear
some tapes in a language they did not know,

and would be asked to react to what they had
heard. They were urged to listen carefully

as the taped segments were quite brief.
Subjects then read the second page of materials
which explained the use of semantic differential-
type scales. The experimenter answered all
questions and directed subject's attention to
the rating forms on which their responses would
be recorded.

According to which subgroup they were assigned
to (randomly by seating arrangement, e.g., first
two rows versus the rest or first two columns
versus the other three) subjects then listened
to a recorded message and rated it on the forms

provided.

29

Each message was played twice. If any subjects
so requested, the message was played a third
time. Usually, there were no requests for

a fourth playing.

When all subjects had completed the rating,

all the forms were collected, and the class
teacher conducted part of her lesson or made
announcements about classwork.

The experimenter than administered a second
message to the subjects, who again rated it in

the same way as before.

CHAPTER III

RESULTS

The main findings of the study are reported in this
chapter. As was indicated in the previous chapter (p. 25)
one of the major difficulties of the experimental situation
arose out of the level of ambient noise in the laboratory
machines. It therefore became necessary to use loud—
speakers rather than headphones for administering the
messages, In groups 1, 3, 5, and 7, some subjects re-
ceived the messages over loudspeakers while some received
them over headphones. A test of differences between these
two subsets of subjects within each group, revealed no
significant differences on the criterion variables. We
can assume therefore that the kind of equipment used for
administering the messages did not affect the results ob-
tained.

Similarly, because we wished to check on the effects
due to the order in which scales were presented on the
rating sheets, two versions (Form I and Form II in Appendix
I, pp, 58—59) were used. The scores on each criterion

variable showed a correlation of .80 or above between

30

31

the two versions; leading to the conclusion that the order
of presentation of the scales did not affect the obtained
results.

Another possible source of bias was the language
studied by subjects. Again there were no significant
differences between languages on the criterion variables.
Of course the use of a foreign language unknown to any of
the subjects may have been largely responsible for the
absence of significant differences.

All these checks seem to support the assumption
that any differences obtained on the criterion variables
between or among groups is likely to be due to the differ-

ent message treatments administered to the nine groups.

Test of the Main Hypotheses

 

The instrument on which the data for analysis are
based, consists of semantic—differential-type scales. Each
criterion variable is expressed as a mean score, derived
from the sum of relevant scales. The data analyzed are
the mean difference or "effect" scores between responses
to the basic or "control" message, and those to each
experimental message.

The data analysis involved 3 x 3 analyses of vari-
ance. The results of theseanalysis are provided in the

following tables.

32

TABLE l.-—Mean evaluation "effect" scores for all nine
groups.

 

 

 

 

Rate
Intensity
Fast Normal Slow Overall
Loud 1.13 0.24 1.30 0.923
Medium 0.22 -0.10 0.31 0.143
Soft 0.67 -O.12 0.46 0.337
Overall 0.673 0.006 0.523 0.434

 

TABLE 2.——Analysis of variance among evaluation "effect"
means.

 

 

Source df MS F p

 

Rate 2 16.46 14.49 0.0005
Intensity 2 18.86 16.49 0.0005
Rate x Intensity 4 4.57 4.02 0.003
Error 436 1.14

 

Although both main rate and intensity effects are

 

significant, we also have a significant interaction effect.
This significant interaction effect "suggests that the
differences among the simple effects of the treatments in
one classification may be different for each treatment (in
the other classification) with which they may be_combined"

(Lindquist, 1953, p. 211). In our case this means that

33

the differences in means among the different levels of
rate are not distributed in the same manner as are the

differences in means among the intensity levels. In

 

other words, although differences in either the rate or
intensity of a message lead to differences in the evalu-
ation of the message, we cannot say that an increment in
rate or intensity is linked with an increment in evalu—
ation. An inspection of the cell means show that in
general, the lowest mean occurs in cells in the middle
of each axis of the table, i.e., the extremes of rate
and intensity have higher means than the middle level.
What we have then is a curvilinear relationship. For
overall 3323, the curvilinearity shows a somewhat posi-

tive trend; for overall intensity, it shows a somewhat

 

negative trend.
All this evidence together supports Hypotheses 6,
9 and Alternative Hypothesis 2. Hypothesis 3 is not

supported.

TABLE 3.--Mean potency "effect" scores for all nine groups.

 

 

 

Rate
Intensity
Fast Normal Slow Overall
LOUd -00u9 -00018 -00206 —Oc26
Medium 0.247 0.450 0.166 0.069
Soft 0.229 0.876 0.857 0.615

overall -0019 Oouo 0.2“ 00315

 

34

TABLE 4.—-Analysis of variance using above means (potency).

 

Source of

 

Variance df ms F p
Rate 2 14.79 11.88 0.0005
Intensity 2 27.61 22.19 0.0005
Rate x Intensity 4 0.534 0.43 n.s.
Error 436 1.24

 

Both rate and intensity yield significant F-ratios,

with no significant interaction effects.

The means Show

that for rate, there is a negative, curvilinear trend.

And for intensity a negative linear trend.

 

The results

for the rate effects support Alternate Hypothesis 1, but

fail to support Hypothesis 1, whereas those for the_ipf

tensity effects fail to support either Hypothesis 4 or

Alternate Hypothesis 3.

Hypothesis 8.

The results also fail to support

TABLE 5.-—Mean activity "effect" scores for all nine groups.

 

 

 

Rate
Intensity
Fast Normal Slow Overall
Loud -l.06l 0.494 0.534 -0.09
Medium -0.209 1.13 1.13 0.42
Soft -0.745 0.817 1.54 0.36
Overall -0.720 0.766 1.02

 

35

TABLE 6.——Analysis of variance using mean activity "effect"
scores (above).

 

 

Source of

Variance df ms F p
Rate 2 130.69 135.56 0.0005
Intensity 2 11.65 12.09 0.0005
Rate x Intensity 4 4.47 4.64 0.001
Error 436 0.964

 

While both main effects are significant, so is the
interaction. The simple effects show a negative linear
trend for rate; and the same, though less marked for

intensity. The marginals for intensity show something

 

 

of a negative curvilinear trend, but this is not pro-
nounced at all.
The rate results fail to support Hypothesis 2. The

intensity results do not support Hypothesis 5. Hypothesis 7

 

is also not supported.

Summary and Discussion

 

The hypotheses investigated in this study are con-
cerned with whether or not differences in certain "vocal"
variables make a difference when a listener reacts to
spoken messages. Hypotheses l, 2, and 3 suggest that the
ppggd at which a message is spoken would make a difference.

Specifically they state that the faster a message is, the

36

better, more powerful and more dynamic a listener would
perceive it to be.

The results obtained from the analysis of our data
do not provide support for any of these hypotheses.

Rather they indicate (in support of Alternative Hypothesis
1) that if a message is either fast or slow, it will be
regarded as "stronger" or more powerful than if it is
neither fast nor slow. The results also indicate (in
support of Alternative Hypothesis 2) that a fast or slow
message is considered "better" than a message which is
neither fast nor slow. But when a listener judges the
dynamism ("active-ness") of a message, he tends to favor
Slower messages over faster ones. This finding is sur-
prising, and counter-intuitive. We expect "dynamic"
speakers to speak rather faster than slower.

Hypotheses 4, 5, and 6 were concerned with the effect
of the loudness of a message on a listener's response to it.
They stated that the louder a message the "better," more
forceful more dynamic it would be considered to be. The
results of this study do not altogether support these
statements. Our findings indicate that while the louder
a message, the more favorably it is evaluated, the reverse
is true when the "potency" or "forcefulness" of the
message is being judged, so that the softer a message is
spoken, the more potent it is considered. But a soft or

medium loud message is considered more "active" or

37

"dynamic" than a loud message, with the medium loud
message being the most highly rated.

In Hypotheses 7, 8, and 9, we looked at the com-
bined effects of both the loudness and sp§§d_of a message
on the way it is received. Here again, we stated in
Hypothesis 9 that if a message was either fast or slow

as well as soft, it would be more favorably received

 

than if it was otherwise. This hypothesis was supported
by the results. But Hypothesis 8 which stated that the
louder apd slower a message, the more "forceful" it would
be rated, was not supported; the results in fact lead to
the Opposite conclusion. While these results are con-
sistent with those obtained for the isolated effect of
message loudness by itself, they are slightly different
from those obtained for the isolated effect of message
speed. The fact that we do not have a significant inter-
action effect here is interesting. It is intuitively
reasonable that when we combine extremes of loudness and
speed, the most effective combination for conveying
"strength" or "potency" should be one which contains the
highest level of one variable with the lowest level of
the other; but it is not easy to predict which of the
two possible combinations "fast" + "soft" or "slow" +
loud" would achieve this effect.

These results raise certain questions and suggest
certain practical applications which will be considered

in the next chapter.

CHAPTER IV

CONCLUSIONS AND FURTHER RESEARCH

It is clear from our results that how fast or how
loud one speaks affects a listener's response. What is
not so clear is the pattern of relationships between
loudness or speed or both of them together, and how
peOple react to messages.

This study has attempted to be "clinically pure"
in its isolation of "vocal" variables from the spoken
message. We have assumed, with reasonable justification,
that the verbal or referential content of the message did
not affect the results we obtained. But as soon as we
move into any realistic Situation, we will have to con-
front the verbal channel. The most interesting and most
useful kinds of results will be those which deal with the
interaction between yhat is said and hpw_it is said.
Obviously then, we have to work with languages which are
understood by listeners.

The combination of verbal and "vocal" channels leads
to complexity--but it is a ubiquitous complexity which

human beings deal with constantly. One difficulty for

38

39

the communication researcher in such situations is specify—
ing the whip of the resultant messages. It is a notorious
fact that communication research is weak in the analysis
and categorization of messages. Linguists, whose chief
concern is to describe the verbal code, have yet to

achieve the kind of description and analysis which are
relevant for communication research.

In the specification of the EEEE of a message, to
which we wish to relate a p93, one approach is to cate-
gorize messages according to the situation in which they
are produced. Thus we can distinguish the category
"Inaugural Address," "After-dinner speech," "Sunday ser-
mon," "Bull session." Hymes (1962, p. 24) refers to such
categories as "Speech Events."

One kind of research then would be concerned with
varying the hpw_while keeping the speech event constant.
For example, does it make a difference if a nursery rhyme
is said fast or slow, loud or soft?

The mention of nursery rhymes leads to another
possible method of classifying the verbal content of
messages. Nursery rhymes, like cultural truisms, are
marked by a large amount of redundancy; they are highly
predictable at all levels. Thus if we were to use nursery
rhymes in studies of the combination of the verbal and
vocal channels, we would really be able to convey new

information only through the vocal channel. The research

40

question here would be: How do differences in the re-
dundancy of the verbal Channel affect response to a
spoken message?

Some prior researchers suggest that one of the
factors which will interact with response to verbal
channel redundancy is social class. This is because
children, in their acquisition of language, tend to
learn the value of vocal features much before they have
achieved mastery of the verbal code. Their use of the
vocal as well as the verbal channel in later life is
therefore not arbitrary, but largely dependent on the
environment into which they have been socialized. This
is the central thesis of the work of Basil Bernstein, the
British Sociolinguist. Bernstein approaches language be-
havior from a sociological perspective, and suggests that
social structure and the position of the individual within
it, is related to the possession of certain codes of
communication. For British society, he identifies two

coding procedures: elaborated and restricted. The pure

 

 

or ideal form of restricted code, which has many variants,
is characterized in general by this attribute: "the
verbal component of the message, given the social context,
is highly predictable," from the point of view of "an
observer who knows the code." Bernstein then makes a
distinction between

. . the verbal component of the message and the
extraverbal components. The verbal channel . . .

41

refers only to the transmission of words. The
extraverbal channels include messages transmitted
through the expressive associates of the words
(intonation, etc.) and messages transmitted through
gestures, physical set, and facial modification
(1964, p. 58).
In other words, Bernstein distinguishes verbal from non-
verbal message channels, and in the latter case, between
what we have referred to as the "vocal" and gestural bands.
The three variants of the restricted code are distinguished
one from another in terms of the locus of message predicta-
bility in either the verbal or nonverbal channels. "The
third variant refers to an order of communication where
the verbal component approaches maximal redundancy but
where the extraverbal channels permit messages of a rela-
tively low order of prediction" (1964, p. 59). In other
words, in the third variant, new information can be trans—
mitted only through extraverbal channels.

Bernstein says that although all of the English

children whom he studied possessed a restricted code, the

 

working class ones, were largely limited to this code,
whereas the middle class children also possessed an

elaborated code. We would expect therefore, that working—

 

class children in Bernstein's sample would make greater
use of, and be more sensitive to, extraverbal channels.
A possible study based on Bernstein's work would be to
compare American school children from different ethnic

and socio—economic backgrounds on the extent to which

42

they were able to create and respond to nonverbal messages.
If Bernstein's findings are crossculturally applicable,

one should find that social-structural differentiating
variables correlate highly with ability in identifying
and/or reacting to, emotion portrayed through vocal and
visual channels.

In summary of the foregoing, then: the literature
suggests that just as children master verbal skills with
age, they learn to master the use of vocal and other non-
verbal codes. Furthermore, since these codes have social
importance, they are used differentially by different
social groups.

A tantalizing extrapolation from one of Bernstein's
findings is to attempt to relate the dialectic of physical
violence in American society to the use of different codes.
One may say for example that the incidence of physically
violent and aggressive behavior in certain minority groups
or socio—economic strata is related to the fact that in
these areas of society, the verbal channel is not regarded
as appropriate for communicating new information. The
resort to the nonverbal channel is then seen as logical
for expressing feelings of frustration and new self
awareness--for saying something ESE to the rest of society.

If the use of vocal and verbal codes varies from
subculture to subculture, would it not vary even more
from language to language? Is there a cross-cultural

"vocal" language?

43

Previous research findings are inconclusive about
the extent to which vocal characteristics constitute uni-
versals of human communication. That they are part of

the expression system cannot be denied; the physiolOgy

 

and acoustics of speech production and reception make
some of the vocal features inevitable. What is not clear
is how general are interpretations of differential exploi-
tations of aspects of the vocal expression system.

A related and long standing research tradition,
that involving "phonetic symbolism" has attempted to in-
vestigate the generality of symbolic connotations associ—
ated with speech sounds. Wicker (1968) summarizes the
major results of this research:

Speech sounds have been said to have symbolic
connotations in themselves, beyond those imposed by
the arbitrarily assigned meanings of the words they
comprise. This doctrine of phonetic symbolism was
given experimental support when Sapir (1929) dis-
covered that CVC trigrams containing high vowels
were judged to be more appropriate labels of small
objects than trigrams containing low vowels. This
effect seems easy to replicate (Newman, 1933), and
there has been little disagreement among psycholo-
gists as to its validity.

There has been, however, much disagreement
regarding the generality of the phenomenon. It has
been debated, for example, whether phonetic symbolism
is universal: whether it is independent of the cul-
tural and linguistic background of subjects (Taylor,
1963; Weiss, 1964). There has been disagreement on
whether it is reflected only in subjective judgments
about verbal materials (Newman, 1933) or whether it
fian alsg be seen in real languages (Johnson, 1967)

p- 175 .

Miron's study of cross-linguistic phonetic symbolism
attempted to broaden the cultural base of research in this

area .

44

Nonsense syllables, composed of phonemes which

occur in both Japanese and American—English, were

presented on tape recordings to Japanese and Ameri-

can subjects. The two groups both rated the
syllables on the semantic differential . . . and
marked common tendencies in their ratings were
interpreted as evidence for the existence of
cross-linguistic phonetic symbolism. Since the
vocal expression of the phonetician who recorded

the syllables may have been influenced by his own

reaction to the syllables, it seems possible that

non—verbal vocal cues may have been produced which
were similarly interpreted in both American and

Japanese (Kramer, 1962, p. 11).

Kramer's 1962 study was designed to see whether,
among other things, "American listeners will be able to
categorize the emotions being portrayed in Japanese by
Japanese speakers with an accuracy significantly better
than chance" (p. 11). Although Kramer found that American
actors could in fact identify the emotions intended by
Japanese actors in the recorded messages he noted that
"not all emotions proved equally easy for American
listeners to recognize in Japanese. Furthermore, some
emotions may be portrayed in Japanese in more than one
way, not all of which are recognizable to Americans"

(p. 34).

Such studies demonstrate the universality of emotions
and their expression through the vocal channel. They sug-
gest that human beings can interpret vocal cues in largely
similar ways regardless of their cultural background, but
that the extent to which they do so is yet to be deter-

mined.

45

The present study carries the investigation one
stage further. Instead of asking: What emotion is being
portrayed, we ask: How do you react to what you have
heard? This is a more general, more open—ended question
even though our methodology requires a particular set of
responses. A replication of this study with subjects of
different language backgrounds will move research in
"vocal" communication more firmly towards cross—cultural
generalizability.

It would be interesting to find out for example
whether subjects whose languages consistently and systemati-
cally manipulate some of these vocal variables will differ
in their response from other subjects. Speakers of tone
languages where pitch variation is a feature of the sylla-
ble may conceivably react differently to vocal cues than
speakers of English in whose language pitch variation
(intonation) is spread over a longer linguistic stretch--
the phrase or utterance. It is probably not accidental
that it is speakers of tone languages who have developed
drum, whistle and gong languages--surrogates of speech in
which the vocal features of speech form the basis of com—
munication. In fact, West African drum languages depend
largely on the reproduction of tonal relationships (i.e.,
relative pitch differences) rather than vowels and
consonants, for signalling messages. But the messages

which they send are mostly referential and formulaic

46

(Herzog, 1947; Carrington, 1949; Stern, 1954; Opubor, 1969).
It would be interesting to investigate the connotative
possibilities of drum languages, to see what combinations
of their codes are associated with what kinds of conno-
tative meaning.

Another related question is the extent to which
stereotypes of various languages are related to the
realities of their "vocal" behavior. The French and the
Italians are said to speak very rapidly; Southern Ameri-
cans to speak slowly. Are these borne out by objective
measurements of rate of speaking? Do homosexuals generally
speak at higher frequencies than other males? Is a high
pitched male voice necessarily associated with perjorative
connotations across cultures?

Answers to such questions can lead to a more rational
strategy of verbal communication across national boundaries.
If we can determine which aspects of response to vocal cues
are universal and which are language or culture specific,
we can devise ways of ensuring that international communi—
cators, both in face-to-face and mass media situations,
maximize the effect they wish to achieve through skillful
use of all communication channels, including the "vocal."

The results of this study suggest the following

elements of such a strategy:

47

1. If a speaker wants an audience to react favor-
ably to what he says, he must speak either more
slowly, or speak faster, than his normal speed.
He would be even more successful in achieving
his goal if he also spoke softly. However
should he wish to speak at his normal speed,
he must speak louder than normal.

2. If a speaker wishes an audience to consider
him "powerful" or "forceful," he must speak
more softly apd more slowly than normal.

3. If a speaker's goal is to be considered "dynamic"
by his audience, he must speak more slowly than
normal but keep the volume of his speech at
about normal.

These are strategy conclusions from the results of
this study, and indicate the kinds of practical appli-
cations to which such studies can be put. The accuracy
and effectiveness of such strategies are limited by the
constraints intrinsic to a study such as ours. We cannot
generalize beyond the group of high school students used
in the present study. We cannot say that the strategy out-
lined would hold if the speaker uses a language which his
audience knows, or if the nationality of his audience will
lead to different strategies. Our results indicate that

the sex of his audience or the foreign languages which

48

they know should make no difference. But this is con—
ceivably only an artifact of the particular circumstances
of this study.

These and similar reservations can be obviated
through the kinds of further research which we outlined

earlier.

BIBLIOGRAPHY

BIBLIOGRAPHY

Bernstein, Basil. Elaborated and restricted codes:
Their social origins and some consequences. In
American Anthropologist, 1964, 66, 2, 55—69.

 

Birdwhistell, R. L. Introduction to Kinesics.
Washington, D. C.: Foreign Service Institute,
Department of State, 1952.

Carrington, J. F. Talking Drums of Africa. London:
Carey Kingsgate Press, 1949.

 

Crystal, David and Quirk, Randolph. Systems of Prosodic
and Paralinguistic Features in English. The Hague:
Mouton and Co., 1964.

 

 

Davitz, J. and Davitz, Lois. The communication of feel—
ings by content-free speech. Journal of Communi-
cation, 1959, 9, 6—13.

 

Davitz, Joel R. The Communication of Emotional Meaning.
New York: McGraw-Hill Publishing Company, 1964.

 

Diehl, Charles F. Voice and Personality. In Dominick A.
Barbara (Ed.), Psychological and Psychiatric Aspects

 

of Speech and Hearing. Springfield, Illinois:
Charles C. Thomas, 1960.

 

Dittman, A. T., and Wynne, L. C. Linguistic techniques
and the analysis of emotionality in interviews.
J. Abnorm. Soc. Psychol., 1961, 63, 201—204.

 

Fairbanks, G. Recent experimental investigations of
vocal pitch in voice. J. Acoust. Soc. Amer., 1940,
ii, 457-466.

 

Fairbanks, G. and Hoaglin, L. W. An experimental study
of the durational characteristics of the voice
during the expression of emotion. Speech Monogr.,
1941, 8, 85-90.

 

50

51

Goldman—Eisler, Frieda. Speech-breathing activity--a
measure of tension and affect during interviews.
Brit. J. Psychol., 1955, 46, 53—63.

Goldman-Eisler, Frieda. Speech-breathing activity and
content in psychiatric interviews. Brit. J. Med.

Psychol., 1956, 29, 35—48.

Hall, E. T. The Silent Languag_. Greenwich, Conn.:
Fawcett Publications, 1959.

 

Herzog, George. Drum—signalling in a West African tribe.
Word, 1945, 1, 217-238.

Hymes, Dell H. The ethnography of speaking. In Thomas
Gladwin and William C. Sturtevant (eds.) Anthropology
and Human Behavior. Washington, D. C.: The Anthro-
pological Society of Washington, 1962.

 

Johnson, R. C. Magnitude symbolism of English words.
J. Verbal Learning and Verbal Behavior, 1967, 6,
508-511.

Knower, F. H. Analysis of some experimental variations
of simulated vocal expressions. J. Soc. Psyehol.,
1941, 14, 369-372.

Kramer, E. Judgment of portrayed emotion from normal
English, filtered English and Japanese speech.
In Experimental Analysis of the Control of Speech
Production and Perception, III. Ann Arbor:
University of Michigan, Office of Research, No.
04411-3-F, 1962.

 

Lindquist, E. F. Design and Analysis of Experiments in
Psychology and Education. Boston: Houghton Mifflin

Co., 1953.

 

Mahl, G. F. Measuring the patient's anxiety during inter—
view from "expressive" aspects of his speech.
Trans. New York Acad. Sci., 1959, 21, 249—257.

Mahl, G. F. and G. Schulze. Psychological research in the
extralinguistic area. In T. A. Sebeok, A. S. Hayes,
and M. C. Bateson (eds.) Approaches to Semiotics.
The Hague: Mouton, 1964.

Miron, M. E. A cross—linguistic investigation of phonetic
symbolism. J. Abnorm. Soc. Psychol., 1961, 62,
623-630.

52

Newman, S. Further experiments in phonetic symbolism.
Am. Journal Psychology, 1933, 45, 53-75.

Opubor, Alfred E. Review of "Hausa Music I" by D. W.
Ames. Journal African Languages, 1969, 7, I
(forthcoming):

Osgood, C. E. and Sebeok, T. A. (Eds.). Psycholinguistics.
Bloomington: Indiana University Press, 1965.

Pfaff, P. L. An experimental study of communication of
feeling without contextual material. Speech Monogr.,
1954, 2l, 155-1560

 

Pike, K. L. Language in Relation to a Unified Theory of
Human Behavior. The Hague: Mouton, 1967.

 

Sapir, E. A study in phonetic symbolism. J. Exper.
Psychol., 1929, 12, 225-239-

Skinner, E. R. A calibrated recording and analysis of
the pitch, force, and quality of vocal tones express-
ing happiness and sadness. Speech Monogr., 1935, 2,
81-1370

Starkweather, John A. Content-free speech as a source of
information about the speaker. Journ. Abnorm. Soc.
Psychol., 1956, 52, 394-402.

Stern, Theodore. Drum and whistle languages: An analysis
of speech surrogates. Amer. Anthro., 1957, 59,
487-5060

 

Taylor, I. K. Phonetic symbolism re-examined. Psychol.
Bull., 1963, 60, 200-209.

Trager, C. L. Paralanguage: A first approximation.
Studies in Linguistics, 1958, 13, 1-12.

Waskow, Irene E. Vocal measures and drug effects. In
R. Salzberger (ed.) Research in Verbal Behavior
and Some Neurophysiological Implications.

New York: Academic Press, Inc.,'l967, 393-402.

Weiss, J. H. Phonetic symbolism re-examined. Psychol.
Bull., 1964, 61, 454—458.

Wicker, F. W. Scaling studies of phonetic symbolism.
J. Personal. Soc. Psychol., 1968, 10, 2, 175-182.

 

APPENDICES

APPENDIX I

THE INSTRUMENT

THE INSTRUMENT

STUDIES IN LANGUAGE BEHAVIOR
Department of Communication
Michigan State University

General Instructions:

 

You are about to participate in a study of language
behavior.

We are interested, in this study, in how peOple re-
spond to the human voice and to speech. Imagine a situation
like this:

The telephone rings in your home or apartment or

room. You pick it up and speak to the other per-
son. You do not recognize the voice; the person

at the other end of the line is a stranger.

In spite of the fact that the caller is a stranger,
you would probably have no difficulty in making certain
statements about them. You would usually be able to say
whether they were male or female, young or old, happy or
unhappy, sometimes even whether they were friendly or un-
friendly. Most times your guesses would be correct.

In this study we will ask you to make judgments about
voices on tape. Try to respond as carefully as you can in
the way we have suggested on the questionnaire.

Thank you for your cooperation.

55

56

STUDIES IN LANGUAGE BEHAVIOR
Department of Communication
Michigan State University

 

 

 

NAME SCHOOL STUDENT NUMBER
AGE SEX: MALE FEMALE YEAR IN SCHOOL MAJOR
INSTRUCTIONS:

 

On each page you will find a series of adjectives in
pairs. Between each pair of adjectives there are five
spaces, for example:

++ + 0 - --
friendly unfriendly
You will hear a voice on tape. When the tape is
stOpped, you will be asked to judge what you have heard on
the tape, using the adjective pairs. Make ppe judgment
for each pair of adjectives.

If you feel that the voice on the tape sounds "very
friendly," you would place a check in the first space,
under the double plus (++) thus:

++ + 0 - -—
friendly X, unfriendly

If the voice sounds "friendly" but not extremely so,
place the check in the second space, below the single plus
(+) thus:

++ + 0 - --
friendly X unfriendly

If the voice sounds neither friendly nor unfriendly,
or if you are uncertain, place the check in the middle space
below the zero (0) thus:

++ + 0 - —-
friendly X unfriendly

Also if you do not think that the voice can be de-
scribed in terms of "friendly" and "unfriendly," check the

Iniddle space.

57

If the voice sounds "not friendly," check the fourth
space, under the single minus (—), thus:

++ + 0 - —-
friendly X unfriendly
If the voice sounds "very unfriendly," check the

fifth space, under the double minus (-—), thus:

++ + 0 - -—
friendly X unfriendly

 

 

THERE ARE NO WRONG OR RIGHT ANSWERS. The best response
is what you feel is appropriate. We are interested in
what the voices mean to ypu.

 

58

Rating Scales

 

Voice Sample #

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

high low

good bad

tense relaxed
active passive
shrill mellow

hard soft

calm excited

sick healthy

ugly beautiful
fast slow

happy sad

thick thin
masculine feminine
slow fast

weak strong
normal breathless
dull sharp
pleasant unpleasant
naive sophisticated
large small

 

Form I

 

good
large
relaxed
naive
hard
pleasant
sharp
excited
strong

normal

fast
passive
sharill
feminine
healthy
ugly
high
happy
thin

strong

Form II

59

Rating Scales

 

Voice Sample #

bad

 

small

 

tense

 

sophisticated

 

soft

 

unpleasant

 

dull

 

calm

 

weak

 

breathless

 

Slow

 

active

 

mellow

 

masculine

 

sick

 

beautiful

 

low

 

sad

 

thick

 

weak

 

" _—"_w
I'-