CTLF Corpus de textes linguistiques fondamentaux • IMPRIMER • RETOUR ÉCRAN
CTLF - Menu général - Textes

Fairbanks, Grant. Experimental Phonetics – T10

An Experimental Study of Vowel Intensities *1

Grant Fairbanks and Arthur S. House
Speech Research Laboratory, University of Illinois, Urbana, Illinois
and
Eugene L. Stevens
Milwaukee State Teachers College, Milwaukee, Wisconsin
(Received February 6, 1950)

One hundred and ten monosyllabic words, 10 for each of the 11 common American vowels, were spoken in
isolation by each of 10 subjects. Most of the differences between the mean relative intensities of the vowels
were found to be statistically significant. Among the words for a given vowel the intensity of that vowel was
found to differ significantly in most instances, a variation tentatively attributed to consonantal environment.

The general purpose of the investigation was measurement
of the relative intensities of spoken
vowels. More specifically, the experiment was concerned
with two questions: Are there significant differences
between the intensities of vowels that may be attributed
to the vowels as such? Does a given vowel vary significantly
in intensity in relation to its consonantal
environment?

Vowel intensity has been the subject of previous
investigations, of which two seem most pertinent to
the present problem. In 1926 Sacia and Beck 12 published
the results of a study in which 16 subjects each pronounced
the following 11 monosyllabic words: team,
tip, tape, ten, tap, ton, top, talk, tone, took, tool. It will
be noted that the initial consonant was [t] in all words,
and that five different final consonants were used without
apparent phonetic design. Recently, Black 23 reported
a similar experiment in which measurement of
vowel intensity was a partial objective. Sixteen subjects
and 11 monosyllables, one per vowel, were used.
Consonantal environment was held constant; all syllables
began with [p] and ended with [t], providing
four words and seven nonsense syllables. In both studies
each subject contributed one sample of each vowel
sound.

Procedure

The subjects of the present study were 10 young adult
males, enrolled at the University of Illinois, untrained
in speech, and residing in the Middle West. All habitually
spoke forms of the General American dialect. No
subject was aware of the purposes of the investigation.
The materials consisted of the 110 words, 10 for each
of 11 vowels, listed in Table I. 34 It is seen that all words
are meaningful; all are consonant-vowel-consonant
monosyllables; all consonants are voiceless; eight different
consonants are represented.

Each subject spoke all 110 words, which were typed
on separate 3 x 5 cards. At the time of recording he was81

Table I. Stimulus words.

tableau [i] | [ɪ] | [e] | [ɛ] | [æ] | [ʌ] | [ɑ] | [ɔ] | [o] | [U] | [u] | cheek | chick | cape | check | cap | chuck | chock | caught | choke | cook | coop | feet | fit | fake | chess | cat | cup | cot | caulk | coat | foot | coot | keep | kit | fate | Chet | chap | cut | chalk | coke | hoof | hoot | peak | Kate | fetch | fat | fuss | fop | fought | cope | hook | peat | pick | pate | heck | pack | puck | pock | hawk | poke | push | pooch | seat | pit | sate | ketch | pat | putt | pot | sauce | pope | puss | shoot | sheep | ship | shape | peck | sat | shut | shop | sought | shoat | put | soup | sheet | sit | shake | pet | shack | sup | sot | talk | soak | shook | suit | teak | tick | take | set | tack | tuck | tock | taut | soap | soot | toot | teeth | tit | tape | Shep | tat | tut | tot | thought | tote | took | tooth

seated with his mouth positioned 10 in. from the microphone
by means of a wire ring. One experimenter sat
facing him and exposed the stimulus cards in random
order one at a time, with an interval of approximately
four seconds between words. Instructions were simple:
to read the words to the experimenter. No attempt
was made to control articulation; when, as occasionally
happened, words were misread they were reintroduced
into the random order and presented again later. Once
the actual procedure had begun the stimulus words
were presented without interruption. Prior to recording
the subject was given a brief practice period. The above
procedure resulted in a series of responses from each
subject that varied over a moderate but not extreme
intensity range. Later examination of the records
showed that gross, erratic, progressive, or rhythmical
intensity changes were absent.

Recording apparatus was arranged conventionally in
two rooms and consisted of the following: a Shure 9898
crystal microphone; a General Radio Type 759-B
Sound-Level Meter (of which the microphone was a
component) used as an amplifier; a Sound Apparatus
Model HPL-E graphic level recorder, with 0-50 db
input potentiometer, which yields a graphic record of
relative r.m.s. voltage. Recording was direct; a simultaneous
magnetic recording was made for purposes of
identification and reference.

The typical graphic record of a given word was a
relatively smooth, peak-like curve, symmetrically displayed
with respect to duration, and with a clearly
identifiable maximum. The few departures from this
general form seemed to be more closely related to

Table II. Group means and standard deviations. Ten subjects.
Relative intensity in db above arbitrary reference.

tableau AM | SD | [i] | [ɪ] | [e] | [ɛ] | [æ] | [ʌ] | [ɑ] | [ɔ] | [o] | [U] | [u]

subjects than to either words or vowels. For each word
a single value, the level of the maximum in db above a
common arbitrary reference, was obtained. In view
of the phonetic structure of the materials and the
characteristics of the records as described, the assumptions
that the maxima had been reached during the
vowels and that measurement of them furnished valid
basic data concerning the vowels are considered to have
been warranted.

Results

Relative Vowel Intensities

Table II presents the mean relative intensities of the
11 vowels, together with the associated standard deviations.
For each vowel the basic data were the 10 individual
subject means (not shown) of the pronunciations
of the 10 words representing that vowel. The tabled
means are distributed over a range of 4.5 db, a somewhat
larger value than those of Sacia and Beck 45 (3.2 db)
and Black 56 (3.7 db). The rank-order of the means,
however, is generally similar. Rank-difference correlations
of the present data with those of Sacia and Beck
and Black yielded coefficients of 0.61 and 0.65, respectively.

The results of general analysis of variance are shown
in Table m. The obtained F of 15.49 far exceeds the
tabled value at the 1 percent level, and the working
hypothesis of zero difference may be rejected with
confidence. The difference between any two of the
specific vowel means in Table II may be compared to a
requirement of 1.41 db at the 1 percent level and of
1.07 db at the 5 percent level, as calculated by means
of the t statistic Of the total of 55 such differences, 34
exceed the above minimum for the 5 percent level;
of the 34,32 exceed that for the 1 percent level. It may
be concluded that a large proportion of the differences
between vowel intensities, as measured under the conditions
of this experiment, are not to be attributed to
chance. Inspection of the means of Table II indicates
that systematic relationships between these intensity
variations and certain physiological aspects such as lip,
jaw, and tongue position may be closer than previous
results have revealed, suggesting that the intensity82

Table III. Analysis of variance in relative vowel intensity.
Eleven vowels; 10 subjects.

tableau df | sum of squares | variance | F | vowels | subjects | remainder

* F: Vowels/Remainder.
F. 10 & 90:1 percent. 2.53; 5 percent. 1.94.

variations in question are characteristic of the vowels as
such. Study of such relationships is in progress and will
be reported later.

Influence of Consonantal Environment

In the attempt to explore the effects of consonantal
environment upon vowel intensity each vowel was
considered as a separate experimental problem. Study
of Table I will show that, for a given vowel, the stimulus
words were phonetically alike in that all were simple
consonant-vowel-consonant monosyllables in which the
same vowel phoneme was preceded and followed by
voiceless consonant elements, and that they were
phonetically different in the “manner of articulation”
(stop-plosive, fricative, affricative) and/or in the “place
of articulation” (bi-labial, labio-dental, post-dental,
velar, glottal) of the initial and final elements. Eight
initial and seven final consonants were used in all.
These consonant differences, although far from exhaustive,
were regarded as appropriate for general exploration
of the issue in question. The statistical procedure,
results of which are given in Table IV, was to compute
the mean intensity of the vowel of each word and
analyze the variance of these word means for each
vowel separately. Of the 11 F ratios in Table IV, seven
exceed the tabled 5 percent value; of these seven, four
exceed the tabled 1 percent value. This finding indicates,
at least for a majority of the vowels, that significant
variations in intensity occur from word to word, vowel
constant. While it seems reasonable to interpret these
differences as effects of varying consonantal environment,

Table IV. Analysis of variance in relative vowel intensity in
different words. Separate analysis for each vowel: 10 words,
10 subjects.

tableau variance | words | subjects | remainder | F | [i] | [ɪ] | [e] | [ɛ] | [æ] | [ʌ] | [ɑ] | [ɔ] | [o] | [U] | [u]

* df: Words. Subjects. 9; Remainder, 81.
** F: Words/Remainder.
F. 9 & 80:1 percent. 2.64; 5 percent. 1.99.

that conclusion is reached with reservations. The
fundamental requirement that all stimulus syllables be
meaningful words introduces, by definition, a condition
of semantic variability. Examination of Table I shows
that the words ranged widely in meaning, grammatical
function, familiarity, frequency of occurrence, etc. At
any rate, the evidence indicates that further experimentation
along both phonetic and semantic lines
should be productive.

Conclusions

1. When spoken in isolated words most of the
common American vowels are significantly different in
mean intensity; the relationship of these differences to
the physiological differences between vowels requires
additional investigation before conclusions may be
drawn. 2. When the same vowel is spoken in different
isolated words its intensity sometimes varies significantly
from word to word, and it seems probable that
such variations are, in part at least, effects of differing
consonantal environment; the role of semantic differences
between words, however, is not to be overlooked,
and the influence of both factors warrants further study.83

1* Reprinted from The Journal of the Acoustical Society of America, Vol. 22, 1950, pp. 457-59.

21 C. F. Sada and C. J. Beck, “The power of fundamental speech
sounds,” Bell Sys. Tech. J. 5, 391-403 (1926).

32 J. W. Black, “Natural frequency, duration, and intensity of
vowels in reading,” J. Speech and Hearing Dis. 14, 216-221 (1949).

43 Two of the stimulus words, kit and hoot, were duplicated
through a clerical error that was not detected until Table I was
prepared for publication. Examination of the measurements and
computations, however, showed that the duplications had had
negligible effect upon the results.

54 See reference 1, p. 400.

65 See reference 2, p. 219.