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

Fairbanks, Grant. Experimental Phonetics – T09

A Physiological Correlative of Vowel Intensity * 1

Grant Fairbanks
University of Illinois

The intensity of a vowel is usually
A taken to involve three anatomical
systems: the breathing apparatus, the
larynx, and the resonance cavities. The
relative roles of these systems and the
interactions between them are not well
understood, but two specific variables,
the vocal fold tone and the modification
of the vocal fold tone by the resonance
system, may be discerned as contributing
to the acoustic end product. It is
with the second of these variables that
the present study is concerned. The
approach may be introduced by two
hypothetical vocal situations.

Assume that a speaker is instructed
to phonate a given vowel and vary its
intensity, keeping the vowel constant.
If it is granted that he kept the vowel
constant by maintaining the anatomical
resonators in a relatively fixed state, the
variation in intensity may be considered
to have resulted from alteration of the
vocal fold tone. The latter has been
produced, presumably, through the mutually
interdependent and balanced interplay
between the firmness of the
glottal closure and the quantity of infra-glottal
air pressure, with related
variations in physiological activity on
the part of the thoracic and laryngeal

Assume, on the other hand, that the
speaker is instructed to keep this physiological
activity constant
(granting for
the sake of illustration that this is possible)
and phonate several different
vowels. Variations in the vocal spectrum,
most experimental phoneticians
would agree, are produced by differential
adjustments of the anatomical resonators,
especially the pharyngeal oral
cavities, so that characteristic differences
in the selective modification of the vocal
fold tone result. The systematic
nature of these adjustments has been
demonstrated repeatedly. While the
problem is complex and may easily be
oversimplified, certain variables may be
identified with considerable confidence.
Manipulation of the cavities by the mandible,
tongue and lips, for example, are
known to all students of speech and
need no summary here. Cavity adjustments
are pertinent to the present discussion,
however, because they not only
affect the vocal spectrum, but some of
them should also affect the vocal intensity.
If the speaker, then, varied vowel
quality by adjustment of resonators,
physiological activity in the glottal region
constant, variations in vocal intensity
should take place. 12

In a recent study by Fairbanks,
House and Stevens 2 3most of the common
American vowels were found to be significantly
different from each other in
mean relative intensity. Measurements
were made in an experimental situation
which appears to permit the conclusion
that these differences resulted primarily
from changes in the supra-glottal cavities.
Briefly, the procedure was as follows:
Eleven vowels were studied, each
being represented by 10 monosyllabic
75words which included voiceless consonant
elements only; 10 adult males
served as subjects, each subject furnishing
acceptable pronunciations of all no
words; the words were printed on individual
cards, intermixed without reference
to vowels, and presented to the subjects
in random order, a different order
being used for each subject; instructions
were to read the words to the experimenter,
who faced the subject across the
microphone and exposed the cards one
at a time. Under these conditions of randomized
multiplication, by words and by
subjects, it seems reasonable to assume
that the physiological activity of the
thoracic and laryngeal structures varied
essentially at random, and was comparable
from vowel to vowel. In other
words, the assumption made in the second
hypothetical case proposed above,
namely, constant physiological activity
at the larynx, was approximately satisfied
on a statistical basis. 3 4If this premise is
accepted provisionally, an explanation of
the obtained intensity differences may be
sought by studying the characteristics of
the vocal cavities during vowel production.

Physiological measurements were not
made in the above investigation, although
such an experiment is in progress.
Meanwhile it is of interest to examine
data from other studies. The
best evidence appears to come from
X-ray photography. Most of the X-ray
investigations, however, are not .directly
comparable to the intensity experiment,
and all of the work on English vowels
known to this writer has been based on
lateral exposures which, since they reveal
the median plane only, have definite
limitations. The findings of Treviño
and Parmenter, 4 5Parmenter, Treviño
and Bevans, 5 6and Parmenter and
Bevans 6 7seem most applicable. The data
are shown in the form of tracings of the
original negatives by methods which are
described in detail and evidence unusual
care. The report by Parmenter and
Bevans presents a group of such tracings
for the same 11 vowels that were used
in the study of Fairbanks, House and
Stevens. The subject, a young, adult,
male college student, native to the Middle
West, also was comparable. For the
present study measurements of these
plates were made and correlated with
the intensity data.


It was considered likely that the intensity
variations might be related to
the cross-sectional area of the vocal channel
at such a point of constriction, for
example, as the mouth opening. It was
not possible to measure area directly,
because the published tracings 7 8are restricted
to the median plane. It was
thought, however, that measurement of
channel diameter in that plane might
provide an approximate index of area
which, although admittedly imperfect
because of varying transverse diameters,
might be useful for purposes of correlation.
Accordingly, measurements of minimum
channel diameter were made at
the following locations: (1) between the
upper and lower lips; (2) between the
cutting edges of the upper and lower incisors;
(3) between the highest point of76

Table I
Mean Relative Intensity Levels of Vowels
in Decibels above Lowest Mean

tableau vowel intensity | mean relative (db)

the tongue and the palate. 8 9The measurements
were made with dividers and a
millimeter scale, and rounded to 0.5mm.
Since the tracings are reduced by a
large and unspecified, but constant, factor,
the data were converted into relative
values. For a given channel point the
ratio of each of the obtained measurements
to the smallest measurement provided
the expression of relative diameter. 9 10
As it happened, the smallest
measurements were equal at all three
points, so that the relative diameters may
be inter-compared.

The mean relative intensity levels of
the 11 vowels, from the study of Fairbanks,
House and Stevens, 10 11are shown
in Table I, where they are expressed in
decibels above the smallest mean, that
for [ɪ]. 11 12For purposes of the present
analysis, in which linear units seem appropriate,
these means were converted

image relative power | relative conduit diameter between incisors | æ | ɔ | ɑ | o | e | ɛ | u | ʌ | i | U | ɪ

Figure 1. — Relative power of vowels vs. relative conduit diameter between upper and lower
incisors. Data from Table II.77

Table II
Relative Power and Relative Channel Diameters of Vowels

image vowel | relative diameter of channel | relative power | lips | incisor teeth | tongue palate | [æ] | [ɔ] | [ɑ] | [o] | [e] | [ɛ] | [u] | [ʌ] | [i] | [U] | [ɪ]

into relative power, considering the
mean for [ɪ] as 1.0. 1213


The relative values for power and for
the three measurements of channel diameter
are presented in Table II, where
the arrangement is in descending rank
order of power. Means for the diameter
measurements are shown as a matter of
interest, and are of expected relative
magnitude. Positive correlation between
power and diameter is apparent
by inspection. Product-moment correlation
coefficients are given in Table
III, where it is seen that the correlations
between power and both lip and incisor
channel diameters are significantly
positive. A word of caution is appropriate
here in view of the limitations of the
procedure of relating data from two separate

Table III
Product-Moment Correlation Coefficients
between Relative Power and Relative
Channel Diameters.

tableau lips | incisors | tongue-palate

experiments, one of which involved
a single subject. By the most conservative
standards, however, it is difficult to
regard the correlation of .95 between
power and channel diameter between
the incisors as unimpressive. The closeness
of the obtained relationship may be
observed in Figure 1, where relative
power is plotted against relative diameter
of the incisor channel.

In consideration of the various limitations,
it does not seem profitable to
carry on additional quantitative treatment
with the present data. The most
important aspect of the study would appear
to be its definite indication that
further experimentation along these
lines should be fruitful. The findings
also call attention to a conception of
the system of anatomical resonators
which is not, in the experience of this
writer, commonly expressed by writers
and teachers in the field of speech. The
connected cavities of the larynx, pharynx,
and mouth may be regarded as constituting,
in the case of vowels, a curved
acoustic conduit, tube, or pipe, irregular
in cross-sectional area and shape
along its length, and changing in these
respects from vowel to vowel. Laryngeal
activity constant, a relationship between
vocal power and the area of the
78mouth of such a conduit, as suggested by
the present study, is plausible, on acoustical
grounds and has implications for
vocal theory. Phoneticians have long
been interested in the spectral functions
of the anatomical resonance system. It
seems fair to suggest that the concentration
on these phenomena, the common
use of the term “cavities,” and the reproduction
of diagrams of the positions
of structures in the median plane have
led to an over-emphasis of the irregularities
of the channel and distracted
attention from a view of it as a conduit.

An attempt to make this point clear
has been made in Figure 2. The midline
tracings of three vowels, [æ], [e]
and [ɪ], have been adapted from the
plates of Parmenter and Bevans. 13 14
These particular vowels were selected
because they are representative of the
ranges of power and channel diameter,
as may be seen in Table II, and because
they are all “front” vowels with physiological
progression in the same region.
In the adaptation the channel outlines
have been changed in two respects.
First, the insignificant, pendent uvula is
not shown, Second, the upper portion
of the epiglottis, posterior to the tongue,
has been omitted, because the cartilage
at that level is normally a curved panel
of relatively small transverse diameter
upthrust in the midline from the anterior
rim of the laryngeal cavity. These
two arbitrary revisions of the channel
wall can, of course, be defended only
for illustrative purposes. The essential
point, however, still stands: The anatomical
prominence of these structures
in the median plane is out of proportion
to their acoustical significance, and gives
an erroneous impression of irregularities
along the length of the conduit. The
diagrams of Figure 2, clearly only approximations,
serve their purpose by

image æ

image e

image ɪ

Figure 2. — Typical changes in the vocal conduit.
Adapted from Parmenter and Bevans.

providing a somewhat more accurate
picture of cross-sectional area. They require
no special comment beyond noting
the already demonstrated changes near
the mouth of the conduit which accompany
variations in the power of its output.
It may be added that the “back”
vowels show greater irregularity in conduit
diameter, the bulk of the tongue
79being displaced posteriorly, but that the
changes at the mouth of the conduit
obey the same general law.


In an attempt to identify some physiological
basis for the significant differences
in the relative intensities of vowels
found by Fairbanks, House and Stevens,
data from an X-ray study by Parmenter
and Bevans were examined, and
measurements of the diameter of the vocal
channel in the median plane were
made at three points of constriction.
The correlation between relative power
of vowels and relative channel diameter
between the upper and lower incisor
teeth was found to be significantly positive.
On the basis of the assumption
that laryngeal activity was essentially
random in the intensity study, this correlation
was tentatively interpreted as reflecting
the effect of variations in the
area of the mouth of the vocal conduit
upon the power of output.80

1* Reprinted from Speech Monographs, Vol. 17, 1950, pp. 390-95.

21 This relationship has been noted by previous
writers, but in general terms. See, for
example. Fletcher, H., Speech and Hearing
(New York, 1929), p. 74. “As would be expected,
the open vowels… have the largest phonetic

32 Fairbanks, G., House, A. S., and Stevens,
E. L, “An Experimental Study of Vowel Intensities,”
J. Acoust. Soc. Amer., XXII (1950).

43 The reader will find small interest in this
study if he is unprepared to give tentative acceptance
to this premise. While the writer
would defend it as reasonable, there is no obvious
way of evaluating it at present.

54 Treviño, S. N. and Parmenter, C E., “Vowel
Positions as Shown by X-Ray,” QJS, XVIII
(1932). 351-369.

65 Parmenter, C. W., Treviño, S .N., and Bevans,
C. A., “A Technique for Radiographing
the Organs of Speech during Articulation,”
Zeits. Exper. Phon., I (1931). 1-22.

76 Parmenter, C E. and Bevans, C. A., “Analysis
of Speech Radiographs,” Amer. Sp., VIII
(1933). 44-56.

87 Ibid., pp. 50-51.

98 Measurements also were made of the minimum
diameter of the mouth opening at any
anterior location. This point was between the
incisors in nine of the 11 vowels, so the measure
was rejected as inferior and duplicative.

109 E.g., between the incisors the smallest value
was 1.5 mm for [ɪ]; [æ], with an obtained
diameter of 6.0 mm, thus had a relative diameter
of 4.0.

1110 Op. cit.

1211 The basic data were recorded on a Sound
Apparatus Co. HPL-E graphic level recorder,
which yields a curve of relative r.m.s. voltage
in db. For each word the measurement was
made at the maximum level reached.

1312 For a given vowel in Table I relative
power is equal to antilog 10 N/10, where N is
the intensity level in question.

1413 Op. cit., p. 50.