I
ON THE DEVELOPMENT OF THE CHILD'S ABILITY TO IDENTIFY AFFECTIVE STATE FROM THE PROSODIC FEATURES OF ADULT SPEECH
Piroska Klintbjer Åsa Nilsonne Greta Ågren
Linköping University Departrnent of Conununication Studies SIC 9, 1984 LiU-Terna K-RB-84-9 ISSN 0280- 5634 ISBN 91 - 7372- 86 1- 6
Address : Departrnen t of Conununication Studies Linköping University S-581 83 LINKÖPING Sweden
Tel: 013-28 10 00
Vid filosofiska fakulteten vid Linköpings universitet bedrivs forskning och ges forskarutbildning med utaångspunkt från breda problemområden - teman - i stället för traditionella akademiska discipliner. Det finns fyra teman: Hälso- och sjukvården i samhället, Kommunikation, Teknik och social förändring samt Vatten i natur och samhälle. Forskningen inom varje tema sker i samverkan mellan forskare med olika vetenskaplig balegrund. Verksamheten startade 1980 och efter en uppbyggnadspcriod om 6-7 år kommer varje tema att ha 20-25 fasta forskartjänster - professorer, docenter etc - och cirka 40 forskarstuderande.
Tidigar e utgivna rapporter:
SIC 1 Linell , P.er. Modeller och metafor e·r för kommunika tion. 1982 .
SIC 2
SIC 3
SIC 4
SIC 5
SIC 6
SIC 7
SIC 8
SIC 9
Linell , Per .The ~ritten Zanguage Bias ~n Linguistics. 1982.
Svensson , Cai Om_ Kqmmun ikation , 1 . Ett urva l för e drag från tema - K: s sympos ium den 25 - 26 maj 198 2 . · 1982 . ·
Gustavsson, L & Hult , H {utg}
Te xt och bild i Z~romedel ;· Sju ana lyse r av svenska som . andrr:is pråk . 1983 .
Severinson Eklundh, ~erstin The Not~on ·of Language Game -A Natural Unit ~f Dialogu~ and Discours e . 1 983.
Granström, Kjell (red ) Om Kommunikati on , 2 .. Ett urval för e drag från tema-K : s sympos ium den 26 - 27 maj 1983 . 1983 .
Hansson , Gunnar (red)
Nordenfe l t, Lennart
Litteratur som korrnnmunikation : Förståelser , tolkningar, värderingar . 1983 .
Five studies in action theory . '1984 .
Klintbjer , Piroska On the development of the child ' s ab1'. l7'.ty to identify affec tive s tat e from the prosodic f eat
,~~-rrr / li,y ur es of adult speech ~r.v4 1..1 'f-0~ ~ ~ ~~~C)
~1; ·~-.; f Oj, «.,4
'11\/c ut-i' ..J
ISSN 0280 - 563 4
III
CONTENTS
Introduction
2 Material and methods 3
3 Results 5
3. 1 Sex diff erences 5
3.2 Ontogeny 5
3. 2.1 "Anger" 8
3.2.2 "Sur prise'' 8
3.2.3 "Sadness" 8
3.2.4 "Happiness" 9
3.3 Judgements of the male and f emale voice 9
3.4 Judgements of the diff erent emotions 13
3.5 Misjudgements 13
4 Discussion 1 5
5 Ref erences 20
IV
Addresses :
Pir oska Klintbjer
Department of Communication Studies University of Linköping
Åsa Nilsonne
Karolinska Institute
Department of Psychiatry
S:t Görans Hospital
Stockholm
Greta Ågren
Department of Zoology
University of Stockholm
Abstract
The ability to judge emotional states (anger, surprise, sadness and
happiness) from the non- verbal components of speech was studied in
169 children aged from 4 to 10 years (20- 43 children per group, one
gr oup for each year of age except 9) in Uppsala, Sweden . 19 adults
were used as a control group. The stimulus material consisted of
speech samples, in which a male and a female actor had been in
structed to portray the indicated emotional states .
In general, the ability to make correct judgements increased with
increased age. This was especially the case with the recognition of
"surprise", "sadness" and "happiness", for which emotions increasing
improvements were seen up to t he age of 8, where the ability was
close to that of the adult controls. The abi l ity to recognise an
angry voice appeared to be well developed already at the age of 4. The most difficult item was the male voi ce portraying "happiness",
where the children did not show and increase in their scores until
the age of 8 . Gi rls were found to have a more developed abilit y to
identify the expressed emotions in all age groups .
In most cases where erroneous judgements were made, the male voice
was judged to be angry and the female to be sad, irrespective of
intended emotion and of the sex of the child.
Introduction
Human communication can be divided into verbal and non- verbal com
ponents. The verbal component is expressed in words and sentences.
Non- verbal vocal communication consists of the prosodic features of
speech and vocal expressions as laughter, giggles, screams, sobbing
etc. Non- verbal communication also includes body language , touch and
chemical signals.
In the study of prosody the stress is on how something is said rath
er than on what is being said, i.e. the information that can be
transmitted through the intonation, pitch, amplitude and pauses of a
speech sample.
The laryngeal muscles are under voluntary control which enables a
speaker to modify his or her voice as for instance ac t ors do. On the
other hand, under conditions of emotional arousal, there is also a
non-voluntary influence on the voice mediated by changes in the
breathing pattern and in the action of the laryngeal muscl es. It has
been suggested that the limbic system of the brain is involved i n
these processes (Scherer 1979).
The fundamental frequency of the voice (pitch) is determined by the
interaction between the laryngeal and breathing muscles . Several
reports have shown that the fundamental frequency contour varies
with the emotional state of the speaker. For instance, speech tempo
is high in an angry or frightened voice, but slow in a sad voice.
Anger is loud, sorrow is quiet (Williams & Stevens 1972 , Kotlyar & Mo rozov 1976, Scherer & Oshinsky 1977).
The prosodic qualities of speech can thus mediate information about
emotional state. When studying this type of human communication, the
prosody must be isolated from the verbal content of the utterance .
Many different methods have been developed to this end.
Test stimuli without linguistic content as for instance t he alphabet
have been used (Dusenbury & Knower 1939, Davitz & Davitz 1959) ;
2
alternatively test sentences without emotional content - neutral
sentences - have been presented in different emotional modes
(Dimitrovsky 1964, Fenster & Goldstein 1971) . The ver bal content can
be masked by presenting only the fundamental frequency (Göranson et
al. 1981) or by using randomized splicing (Scherer 1971, Scherer et
al. 1972) . Synthetic stimuli - electronically generated tones - can
be varied in a controlled way and have also been used in this kind
of investigations (Scherer et al. 1977).
Prosody is the dimension of language which is first discovered and
understood by children (Wood 1976).
Wood (1976) proposes five developmental stages in the child's under
standing of prosody.
Stage 1 (4 months)
Stage 2 (18 months)
Stage 3 (2 1/2 yrs)
Stage 4 (5- 12 yrs)
Rough recognition of the prosody in a happy
vs. unhappy voice .
Children exaggerate the intonation when they
begi n to speak in order to make themselves
understood.
The prosody becomes complementary to the
linguistic context of the child's speech.
The child lear ns to identify a wider range
emotions with increasing reliability. Verbal
and non-verbal information can be understood
separately.
Stage 5 (after 12 yrs) The child learns to understand conflicting
messages, i . e. when the prosody does not
carry the same informati on as the verbal
content of the utterance.
The aim of this study was to study the ontogeny (Stage 4 according
to Wood) of the child's ability to identi fy different emotional
states in the speaker from prosodic information alone.
3
The specific Questions investigated were:
Are there sex differences in this ability?
How developed is this ability at different ages?
Is thc rnale and the female voice perceived in the same way?
Does identification of the different emotions present
different degrees of difficulty?
2 Material and methods
A test tape (kindly provided by Anne- Marie Öster, Royal School of
Technology, Stockholm) containing recordings of a male and female
actor reading six neutral sentences in four different emotional
modes was used. The emotions were: anger, surprise, sadness and hap
piness. The sentences were: A. "Fröken kom för sent till skolan"
(The teacher was late for school), B. "Bollen studsade in genom
fönstret", (The ball bounced in through the window), C. "Det var
Olle som vann tävlingen", (It was Olle who won the competition), D. "Sommarlovet börjar sent i år", (Summer vacation begins late this
year), E. "De kommer på torsdag", ( They will arri ve on Thursday),
and F . "Det finns en råtta i skafferiet", (There is a rat in the
larder). The tape was played to children and adult controls who were
asked to try to identify the emotional state of the speaker. Fach
emotion was presented four times, the test thus consisted of 16
stimuli .
The children and the adult controls were provided with answer forms. For each sentence there were four possible answers of which only one
was correct. The sentences were played with intervals of 10 seconds
and the test lasted for approximately 10 minutes. The children were
aged between 4 and 10 years (see table 1), they were recruited from
11 different groups of children at day care centers, pre- school and
primary school .
Table 1.
Sex
Female
Male
4
Age distribution of the children
4 5
15 9
15 15
30 24
Age (years)
6
10
10
20
7
11
11
22
8
20
23
43
1 O* Adul t
?
?
20
8
11
19
73
86
178
* Data of sex distribution not available (data kindly provided by A- M Öster . )
The 4- 7 year old children were tested individually and they gave
their answers orally . The 8 and 10- year olds and the adult controls
were tested in groups and gave their answers on the forms provided
for them. The test session started with an oral instruction by the test leader, the instructions were the same for every individual or
group. The children were shown schematic pictures of faces portray
ing the emotions in question in order to make sure that they could
recognize and name the investigated emotions. After this there was a short discussion about different means of expressing emotions. Four practice sentences were played and evaluated together with the test
leader . Finall y the test subjects were asked whether they had any
questions regarding the test procedure . When eventual questions had been answered the test was administered as described .
In the statistical analysis the x2- test, Fisher's exact probability test and the sign test were used .
5
3 Results
69% of the total amount of ratings were correct . The distribution of correct and incorrect answers in relationship to sex and age is
shown in figs . 1-4 (see last page).
3.1 Sex differences
A comparison between the results obtained by the sexes within each
age group shows that the girls have an overall better score than the
boys (P<0 . 01) (table 2).
Table 2.
Sex
Female
Male
Percentage correct answer s within each age group in relationship to sex.
4
51
47
5
55
59
Age (years)
6
76
69
7
73
73
8
92
82
71
67
Adults
99
93
The distribution of correct answers for gir ls and boys respectively regar ding the different emotional states as presented by the male
and the female voice is shown in fig 5 . After the age of 7 the girls
have better scores than the boys as regards most emot i onal states . This difference persists in the adult contr ol group. The material is
not large enough to permit the difference to be proved statisti
cally.
3.2 Ontogeny
The children achieve better scores the older they are (table 2) . The
di fference bet ween 4- and 5- year old girls is significant (P<0 . 01), the same applies to the differences between 5- and 6-year olds
(P<0.01) , no difference is found, however, between the results of
6
MALE VOICE O/o ANGER O/o HAPPINESS 100
--f:f- -~ 100
o_ _ ... - • .-....... ;;;> • 80 0 80
0
60 60 • / /
/ /
40 40 0 / /
e/'.'.'. • 20 20
/
/ / 0
0 / /
Il > ,, / dl • I >
' 4 5 6 7 8 years /4 5 6 7 8 years
Fema l es; r =0 . 95 , df=4' p <0 .01 Fema les; r =O .80, df=4. NS Ma les • r=0.89, df=4, p <0 .05 Males • r=0.30, df=4, NS
FEMALE VOICE O/o ANGER O/o HAPPINESS 100 0 • 100 ---- o ....... - ./
.......... .......... ......
80 80 ........ .» .......... o...,..... ........
0 .......... ........ 60 • 60 -6
..........
40 40
20 20
ll I> I I I > I '4 4 5 6 7 8 years 5 6 7 8 years Females; r =0 .55, df=4, NS Females; r=0.94, df=4 ' p <0 . 01 Males r=0.43, df=4, NS Males
' r=0.95, df=4' p <0 . 01
Fig. 5. Percentage correct judgements relat ive to the age of the mal e ( 1 ---and the fema l e ( o ) subjects for the male and the female vo ice, and the four emotions .
0/o SURPRISE 100
80
60
40
20
5
/
/ /
0 ~ Q 0
•
6 7
Femal es; r=0 .95, df=4 , P <0.01 Males ; r=0 . 89 , df=4 , P <0. 05
O/o SURPRISE 100
80 /
/ ./ 0
60 ./ ./
/ . / 0 ./
./ 40 ./
/ /
0
20
I 4 5 6 7 "
Fema les; r=0 .83, df=4, P <0.05 Males ; r=0.95 , df=4, P <0. 01
7
MALE VOICE
0
' /
'> 8 years
0/o SAONESS 100
80
60
40 °
20 0
5 6 7 Fema l es; r=0 . 50, df =4, NS Ma les ; r=0 . 98 , df=4 , P <0.0 1
FEMALE VOICE
0
'> 8 years
O/o SADNESS 100
• <It :?"
-?' 80
/ /
/ /
60 / /
9-/ 0
40 •
20
I
4 s 6 7
Femal es; r=0.96 , df=4, P <0. 01 Ma l es ; r=0 . 88 , df=4, P <0.05
0
'> 8 years
'> 8 years
8
the 6- and 7- year old girls. The 8 -year old girls again have sig
nifi cantly bett er scores than the 7- year olds girls (P<0.01 ). A
final improvement of the results can be seen between the 8 - year old
girls and adult women (P <0.025) .
Among the boys there is no difference between the results of the 4-
and the 5-year ol ds . Between the ages of 5 and 6, however , ther e is
an increase i n the number of correct ratings (P<0.05) . Between 6-and 7- year old boys there is no measureable improvement which cor
responds to the findings among the girls. The 8- year old boys again
improve as compared to the 7- years ol ds (P<0. 05). Finally the adult
men had a greater amount of correct ratings than the 8- year old boys
( P< O. 01 ) .
As the differences between the sexes were smal l , the results from
boys and girls have been pooled within each age group to allow
further statistical evaluation.
3 . 2 . 1 "Anger"
The "angry" voice of the male, as well as the female voice was
recognizable al ready t o the 4-year olds (P<0 . 01) . An i mprovement of
the ratings was seen at t he age of 7 (P<0.05) . No further development after that age was seen (fig 6) .
3.2 . 2 11 Surprise"
The 4- year olds could also recognize both t he male and t he female
rendition of 11sur prise" (P<0.05) . This ability did not improve until
the age of 6 (P<0 . 001). There was no significant difference between
the 6 and 7-year olds, but there was an improvement between the ages of 7 and 8 (P<0.025) . No fur ther improvement was noted (fig 6).
"Sadness"
In the case of the male voice , it was not until the age of 8 that
the children were able to identify sadness correctly
(P<0 . 01) (fig 6) .
9
The development then parallels that of "surprise" . An improvement
between the ages of 4 and 6 (P<0 . 05) but no difference between the 6- and 7- year old boys was found. Between the ages of 7 and 8 there was again an amelioration of the results (P<0.05)(fig 6) .
The f emale voice was correctly assessed already by the 4- year olds (P<0. 001). There was a significant increase in correct answers at the age of 5 (P<0 . 001 ). Another increase was seen between the ages
of 7 and 8 (P<0.01)(fig 6) .
"Happiness"
The ability to identify the mal e voice portraying happiness was not demonstrated until the age of 8 (P<0.001) . The female voi ce was correctly assessed already by the 4-year olds (P<0.001). The develop
ment of the ability to correctly assess the female voice again
showed the same pattern as the ability to judge the surprised voice ,
that is there was a difference between the results of t he 4- and 6-year olds (P< 0.001 ) . There was not reliable difference between the 6- and 7- year olds but between the 7- and 8- year olds there was
another improvement (P<0 . 01) (fig 6).
3. 3 Judgement of the male and femal e voice
66% of the judgements of the male voice were correct as cornpared to
71% of the judgements of the female voice. This constitutes a significant difference (P<0. 05) . Table 3 shows the results for the var ious age groups as regards t heir ability t o correctly recognize
the emotional content of the male and female voice respectively .
10
O/o ANGER 100
_____......----. ' 0 0 0 0 0 0 0 ' 0 ','•Mo•.'.'~ .. .. .. ·
80
60
40
20
~'-'~--~----~--~~~--~~~--/~,~~----~~~....,._~ r 4 5 6 7 8 10 years ' / ~ Adults
O/o SURPRISE 100
80
60
40
20
I > I I I I I ,~....____.~_._~_._~_._~~--~~~~-r1
4 5 6 7 8 10 years Adults
Fi g. 6 Percentage correct judgements of t he ma l e (• -------- ) and the fema l e (6 -------- ) vo i ce per age group and emotion.
O/o
100
80
60
40
20
SADNESS
I I
I ~
/
f /{
1 1
--6----- ··· ·· ···· ·· ···6 .L:s- --_,, .... ... .... p" .... .. .. ~
/
I I /1 > '1-4.____.5.____.6 _ _.7_ ..... 8 __ ___..10--y-ea_r_s~' />---A-d ...... ul-ts-~
0/o HAPPINESS 100
80
60
40
20
~ å-"
~ I
/ ~
IS /
/
A----······ ··· ··· ·····6
~--........... ··· ····· ·
I I /1 > ',_4"'---5'--__.6.____7.___8.__ _ _.....10--ye-a-rs _ __,, ,r---A-d-'-u-l ts-~
Table 3 .
Sex
Age (years)
Male voice
Fem ale voice
p
Sex
Age (years)
Mal e voice
Female voice
p
12
Percentage of corr ect answers regarding the male respectively female voice per sex and age group.
Female
4 5 6 7 8 Adults
34 52 78 62 90 98
53 57 74 82 93 100
< o. 001 < 0 . 05 < 0 . 01
Male
4 5 6 7 8 Adults
35 39 66 62 83 94
45 62 72 81 82 95
< 0.001
From table 3 it can be seen that the female voice is more often correctly identified than the male voice throughout the age groups . In
the cases where the differ ences are significant the P- values are entered into the table .
In a comparison between the ability of the boys and that of the
girls to judge the male and the female voice , no difference could be found except in the case of the 8- year olds judging the female voice where the girls had 93% right answers which is significantly better than that of the boys who had 82% correct answers (P<0 . 001 ).
1 3
3.4 Judgements of the different emotions
The "angry" voice was the easiest to identify (table 4), and "anger"
in the male voice was more readily identified than in the f emale
voice (P<0.01). The female voice portraying "sadness" and "happi
ness" were the next two most easily identified stimuli.The differen
ces in judgements of the male and female "surprised" voice were not
significant.
Table 4 Percentage correct answers r egarding the male and female voice in relation to the expressed emotions .
Expressed emotion "Anger"
Male voice 88
Female voice 80
p < o. 01
"Surprise" "Sadness"
61 43
62 77
< 0 . 01
"Happiness"
33
68
< 0.01
The most difficult voices to identify were the male voice portraying
'1 sadness" and "happiness". The "happy" male voice was correctly
identified less frequently than any of the other st imuli (P <0.001) .
3.5 Misjudgements
On the whole the girls and the boys chose the same alternatives when
making incorrect judgements (figs 1-4). The boys aged 4-7 chose 11 anger 11 as the main incorrect alternative to all the expressed emo
tions of the male voice (P<0.05). The girls have made similar
assessments, except in the case of "sadness" as expressed by the
male voice where "surprise" was the main wrong choice (P<0 .05) .
The female voice was in most cases judged to be expressing "sadness"
when a misjudgement was made. This applies to children of both sexes
up to the age of 7.
Fo 1 Hzl
200
100
0 1 2 time Is)
A. SURPRISE. HALE VOICE .
Fo !Hz)
200 ~
100
0 1 2 time ( s)
B. ANGER . FEHALE VOICE.
Fo !Hzl
400
200
0 1 2 3 t i me ( s)
C. SURPRISE. FEHALE VOICE.
F0 <Hzl
200
100 \ t :;::...
0 1 2 t ime ( s)
0 . SURPRISE . MALE VOICE .
14 Fo (Hz)
200
~ 100
0 1 2 time (s)
A. ANGER. MALE VOICE.
Fo !Hz)
200 V\!'!\ 100
't~_._~~~--~~~ ............. ~---> 0 1 2 time (s)
B. HAPPINESS. MALE VOICE.
F0 !Hzl
200
100
r. 0 1 2 3time(s)
C. SADNESS. FEMALE VOICE.
Fo (Hz)
200
J~ vJ 100 \
c :;::... 0 1 2 3 4 t ime (s)
D. SADNE SS. MALE VOICE.
Fig . 7. Fundamenta l frequencies (F0) of some of the test sentences (A to D, see methods) .
15
After the age of 7 there was no clear pattern in the choice of in
correct alternatives.
The fundamental frequency (fig 7) of the various test sentences has
been extracted in order to see if there were similarities between
the stimuli that were confused. The male voice, unlike the female one, did not show a lesser variation of fundamental frequency in the "sad" recordings . This could have made identification difficult, es
pecially since the fundamental frequency contour was not much different from "surprise" and "anger" resp., although t he mean fundamental frequency was lower.
4 Discussi on
The results of the present study were mostly in agreement with pre
vious work. Firstly, the results of this investigation suggest that
girls are more successful in perceiving the non- verbal emotional content of a spaken utterance than boys are . In studies where a
larger number of persons were tested (Dimitrovsky 1964: N=225,
Rosenthal et al. 1978: N=2615) this sex difference was more pronounced. However, in a pilot study of 163 persons Scherer & Oshinsky
(1977) did not find any difference between the sexes.
Secondly, the children improved in their ability to correctly assess the emotional state of the speaker from year t o year up to the age
of 8 at which age their scores did not differ from those of the adult control group. These results are also in agreement with previ
ous findings (Dimitrovsky 1964, Fenster et al 1971, Scherer & Oshinsky 1977 , Rosenthal et al . 1978).
Thirdly, it was found that "anger" seemed to be more easily recog
nized than "happiness", findings that coincide with those of Dimitrovsky (1964), Kotlyar et al. (1976) and Rosenthal et al.
(1978) . Scherer & Oshinsky (1977), on the other hand, finds "happiness" to be more easily recognizable than "anger" in a pilot study .
1 6
Fourthly, when the children made misjudgements in this study,
"anger" and "sadness" were the most frequently chosen alternatives.
This is in agreement with Dimistrovsky's (1964) finding that, when
mistakes were made, positive emotions were judged to be negative .
Finally, the results of this study suggest that the female adult
voice was in general easier to interpret than the male voice, in
particular when expressing "sadness" and "happiness". The male voice
communicated "anger" more successfully. This feature has not pre
viously been reported. However, in this study the test sentences
were read by only one person of each sex. Thus, it cannot be estab
lished whether the above-mentioned difference was due to sex diff
erences in general or ind ividual differences between the actors.This
aspect needs further investigation. A further methodological diffi
cul ty was that the emotional content of the sentences was simu
lated. Possibly there is a difference in the acted and spontaneous
expression of emotion. However, the present test sentences were
interpreted by professional actors, trained to simulate emotions.
Furthermore, the females in the adult control group gave correct
answers to 99% of the stimuli, and adult men and girls of 8 years of
age gave correct answers to more than 90%. Thus, the actors ex
pressed the sentences in a way that was easily interpreted by
adults. A related difficulty stems from the verbal content of the
sentences whi ch the children may not have been able to disregard.
Thus, some sentences may not have appeared neutral to them. (E . g.
there's a rat in the larder) -
Investigators of prosody rarely discuss their results f rom a bio
logical point of view. However, the physiological background of the
gradually improving ability to assess the prosodic qualities of an
utterance has received attention in studies of neural development
(Wood 1976). The finding of several investigators, that girls begin
to talk earlier than boys do, and that they achieve better results
in tests of verbal skills such as spelling, language acquisition and
vocabulary has also been discussed in this context (e.g . Freedman
1979). Thus, Freedman (op. cit.) relates these sex differences to
differences in dominance between the two brain hemispheres . The left
17
hemisphere reportedly controls verbal tasks, while the right one
controls spatial tasks (Bogen 1969, Bower 1970, Dennis & Whitaker 1976) . Kimura et al. (1967) showed that the left hemisphere becomes dominant as regards speech 2 years earlier in girls than in boys.
The present findings can also be discussed in terms of a more general biological perspective: especially in social species, that is
in species where individuals live in groups, the communication of
emotional state enables the different individuals to interact with one another in an adequate way . The ability to encode and decode
emotional information can thus be discussed from the point of view of function against the background of the evolutionary past of human
communication. There is a disagreement about many aspects of early
man and his predecessors, but hardly about his social nature. Thus
it seems likely that man had a well- functioning system for communication of i mportant emotional states bef ore the development of sym
bolic language. Language in this highly developed form is considered
to have become possible with the development of the neocortex; the
older system for vocal, non- verbal communication underlies and colours speech production - especially in emotional situations
(Chevalier-Skolnikoff 1973).
Prosodic features of speech can thus be seen as expressions of this
primitive system - this can be illustrated by the communicative pro
perties of the infants' cries . The crying becomes differentiated at
an early age allowing mothers to identify hunger, fear and discomfort. Tankova- Yampolskaya (1968) analyzed the intonation curves of the newborn infant's cry and adult vocalizations in different emo
tional states. She found that the intonation patterns when expres
sing discomfort were identical. This is probably an effect of the "buil t in" vocal communicati ve system as proposed by Darwin already in 1872. However, the proficiency of the individual develops with
age through interactions with other individuals, which allows lear
ning of the more complex and culturally determined forms of signalling emotion, and through neuronal interaction (cf. above).
18
For example, in social mammals, including primates such as man, so
cial signals are involved in the establishment of dominant rank
orders which are of intrinsic importance to many aspects of social life, e .g . in the distribution of food.
Relevant to the present investigation are the results of several studies of children born blind in different cultures. These studies support the concept of an ability to express and interpret aggres
siveness which is in evidence early in life and which does not seem
to reQuire learning. Our finding of a well developed ability to perceive anger even in the youngest group of children (4 years old) is
in accord with this concept. Eibl-Eibesfeldt (1976) also describes
how small children compete for the mother's breast, for food and toys, and how they try to establish a rank order. Similarly, Palluk
and Esser (1971) report that children defend their rank already at one to one and a half years of age.
The function of a rank order as a means of reducing the amount of overt aggression depends on the ability to recognize the threat signals . In this context the observed well- developed ability of the
small child to recognize anger makes sense. Accordingly, misinterpretation of "happiness" should involve lower risks than any misinterpretation of "anger". Hence , in case of uncertainty, an interpretation of "anger" would be the safest alternative, which also
corresponds with the experimental results from this and other i nvestigations .
However, the "angry" and the "happy" voice show many similarities as
to freQuency, pitch, volume and tempo (Scherer & Oshinsky, 1977). It
may therefore be difficult to judge whether a voice is "happy" or "angry" from intonation only. In this context Van Hooff's (1972) hypothesis concerning the phylogenetic development of laughter from
threat through signals of submission is interesting, If this is true, the ability to show threat is the more primitive system.
Considering the importance of being able to encode and decode threat
signals correctly, it is likely that the different non- verbal chan-
1 9
nels for signalling emotion - that is prosody, facial expression,
post ure etc . should, each by i tself, be suffi cient to allow recogni
tion of the threat. If the "happiness" signal did develop from a threat the differences must be large enough to avoid confusion. Pos
si bly, for this reason and the relative recency of this type of sig
nal (van Hooff, op.cit.), the communication of "happiness" may require reinforcement via addit i onal communicative channels. This explain why the children in this and other studies found it difficult
to identify the emotion "happiness" on the basis of the voice alone,
in particular in the male voice. However, not only "happiness", but also "sadness" in the male voice was more difficult to recognize
than in the female voice. These sex differences could possibly be
attributed to the cultural bias against expressions of sadness in males which could influence both the children and the male actor .
Thus some sex differences were found in communicative skills. These
differences can be discussed in terms of the social organization of
man . A social organization similar to that of e.g. hamadryas or gelada baboons and chi mpanzees has been considered to be relevant for early hominids. This is characterized by male aggressive domi
nance and longterm maternal care (Jolly 1972, Wilson 1975).
The longterm close proximity between the juvenile and its mother to ensure its protection and nourishment also means, that the mother
will be of crucial importance in the socialization process . This
also applies to modern men . In this situation a superior communicative ability in females should be of value . The males may achieve their goals by efficiently signalling and perceiving anger which is
in accord with our finding that the male voice was more efficient than the female voice in encoding "anger".
20
5 Ref erences
Bogen, J.E., (1969). The other side of the brain. II: An appositional mind. Bulletin of the Los Angeles Neurological Societies, 34:135- 162.
Bower, G. H. , (1970) . Anal ysis of a mnemonic device. American Scientist, 58:496- 510.
Chevalier- Skolnikoff, S. (1973) . Facial expression of emotion in nonhuman primates. In: P. Ekman (ed) . Darwin and facial expression: Acentury of research in review. Academic Press . New York . pp 11 - 89 .
Darwin , C. (1872). The expression of the emoti ons in man and ani mals. John Murray. London. (Reprinted 1965, Univ. of Chicago Press).
Davitz, J . R. & Davitz, L. J., (1959). The communication of feelings by content- f r ee speech. Jour nal of Communication, 9, 6- 13.
Dennis, M. & Whitaker, H. A., (1976) . Language acquisi t ion fol l owing hemidecortication: linguistic superiority of the l eft over the right hemisphere . Brain and Language, 3:404- 483.
Dimitrovsky, L. (1964). The ability to identi fy the emotional meaning of vocal expression at successive age levels . In Davitz, J . R. The communication of emotional meaning. Greenwood. New York.
Dusenbury, D. & Knower, F. H. , (1939) . Experimental studies on the symbolism of action and voice, Quarterly Journal of Speech, 25, 67- 75 .
Eibl- Ei besfeldt , I. (1976). Den programmer ade människan . Natur & Kultur . Stockholm .
Fenster , A. & Goldstein, A.M. (1971). The emotional world of children ' Vi s A Vis' the emot ional world of adults: An examination of vocal communication, The Journal of communicati on, XXI, 353- 362 .
Freedman, D.G. , (1979). Human sociobiology, The free press, Collier Macmillan publ . London.
Göransson, A., Jansson, I . , Johansson, B. & Perfekt, R. , (1981 ). Vad hör man på tonfallet, Bruce, G. , red., tema prosodi, seminarieuppsatser i fonetik, Praktisk lingvistik, nr 6, Inst. för l i ngvistik, Lunds univ .
Hooff, J.A.R . A. M. van, (1972). A comparative approach to the phylogeny of laughter and smiling. In: Hinde, R.A . (1974) . Biological bases of human social behavior, 131. McGraw-Hill Book Co . N.Y.
2 1
Jolly, J. (1972) . The evolution of primate behavior. Macmillan Publ Co. N. Y. p . 1 31 .
Kimura, D., (1967) . Functional asymmetry of the brain in dichotic listening. Cortex 3:163- 178.
Kotlyar, G.M & Morozov, V.P., (1976). Acoustical correlates of the emotional content of vocalized speech, Soviet Physics Acoustics 22, 208- 211. In : Sundberg, J. (1980), Röstlära .
Palluk, R.J. & Esser, A.H. (1971) . Controlled experimental modification of aggressive behavior in territories of severely retarded boys, Amer ican Journal of Mental Deficiency, 76, 23-29 .
Rosenthal, R., Hall, J .A. , Archer, D. , Di Matteo, R.M. & Rogers, P.L., (1978) . The pons test. Measuring sensitivity to non-verbal cues. In. Weitz , S., ed. (1979). Non- verbal communication , Readings with commentary, 2nd edition . Oxford Univ. press, N.Y.
Scherer, K. R. , (1971). Randomized splicing. A note on a simple techni~ue for masking speech content, Journal of experimental research in personality, 5, 155- 159·
Scherer, K.R . , Koivumaki, J., och Rosenthal, R. (1972) . Minimal cues in the vocal communication of aff'ect : Judging emotions from contentmarket speech; Journal of Psycholinguistic Research 1, 269- 285 .
Scher er, K.R. & Oshinsky, J.S., ( 1977) . Cue utilization in emoti on attribution from auditory stimuli, Motivation and emot ion , 1 (4), 331 - 346.
Scherer, K.R., (1979). Non- linguist ic vocal indicators of emotion and psychopathology, In: Izard, C.E. (ed) Emotions and Pshychopathology, Plenum Pr ess .N. Y. pp. 493-529.
Tonkova- Yampolskaya, R.V. (1968). Development of speech intonation in infants during t he first two years of life. Translation in: Soviet Psychology ( 1969), 7, 48- 54.
Williams, C.E. & Stevens, K.N., (1972). Emotions and speech : some acoustical correlates, JASA 52, 1238- 1250,
Wilson, E.O. (1976). Sociobiology, The new synthesis . Howard Univ. Pr ess, Cambridge.
Wood, B.S. , (1976). Children and communication: Verbal and nonverbal language development . Prentice Hall Inc . Englewood Cliffs, New Jersey .
100%
100 °lo
100 °lo 100 °/ o
100 °lo
100 °lo
100 °lo
100 °lo
Adults
6 years
5 years
CORRECT ANSWERS
• female subjects
111 male subjects
INCORRECT ANSWERS
N N
An Su Sa Ha An
MALE VOICE
Su Sa Ha
FEMALE VOICE
4 years ~
D female subjects
male subjects
Fig 4. Happiness. Distribution of the answers among the alternative emotions anger (An), surpi se (Su), sadness (Sa) and happiness (Ha), when happiness was expressed . In each pa i r of bars the left bar represents the answers of the female subjects and the r i ght one the answers of the male subjects .
100%
An Su Sa Ha An Su Sa Ha
MALE VOICE FEMALE VOICE
4 years
Adu lts
7 years
CORRECT ANSWERS
• female subjects
R male subjects
INCORRECT ANSWERS
• D fem ale su bjects
male subjects
Fig. 1. Anger. Di stributi on of answers among the alternative emotions anger (An), surprise (Su), sadness (Sa) and ha ppiness (Ha), when angerwas expressed . In each pair of bars the left bar represents the answers of the female subjects and the right one the answers of the male subjects.
100%
100 °lo
100% 100%
100%
100%
100%
An Su Sa Ha An Su Sa
MALE VOICE FEMALE VOICE
100 °lo
100%
years
4 years
Ha
Adu lts
CORRECT ANSWERS
• fernale subjects
• male subjects
INCORRECT ANSWERS
~
D fernale subjects
male subjects
Fig. 2. Surpri se . Distribution fo answers among t he alternative emot i ons anger (An) , surprise (Su), sadness (Sa) and happines s (Ha), when surpri se was expressed. In each pair of bars the left bar represents the answers of the female subjects and t he r ight one the answer s of the ma le subjects.
100%
100 °lo
An
100 °lo
100 °lo
Su Sa
MALE VOICE
100 °lo
Ha
100 °lo
100 °lo
An Su Sa
FEM AL E VO I C E
100 °lo
100%
5 years
4 years Ha
Adults
CORRECT ANSWERS
• female subjects
• male sub jects
INCORRECT ANSWERS
• D female sub jects
male subjects
Fi g. 3. Sadness. Distribut ion of answers among the al ternative emotions anger(An), surprise (Su), sadness (Sa) and happiness (Ha), when sadness was expressed . In each pair of bars t he left bar represents the answers of the female subjects and the right ones the answers of the male subjects .