Fairbanks, Grant. Experimental Phonetics – T23

An Acoustical Comparison of Vocal Pitch in
Seven- and Eight-year-old Children *1

Grant Fairbanks
University of Illinois

The present paper brings together for inter-comparison the results of
two previously reported acoustical studies of vocal pitch in the speech of
children. In one of these investigations Fairbanks, Wiley, and Lassman (3)
employed two groups of boys at ages seven and eight chronological years.
In the other investigation Fairbanks, Herbert, and Hammond (2) were
concerned with comparable female subjects. While both studies focused
primarily on age comparisons, their design permitted combination of the
data for four-group statistical analysis bearing on sex as well as age differences.
The following summary of the basic investigations is presented as
background.

I. Procedures and Results of Basic Investigations

Experimental procedures were alike in all instances. Each subject was
brought into the laboratory individually, and read aloud a specially constructed
192primary-level test passage of 52 words, which was set up as the
central portion of a longer passage (3). A high-quality disc phonograph
recording was made of each performance. Oscillograms of these samples
were made and subjected to fundamental frequency measurement. Acoustical
data thus secured were tabulated and further measurements were
made following routines previously established (1). For each subject the
data yielded a number of final measurements of various aspects of vocal
pitch which were used in the group computations.

The differences between the means are impressively small. All apart
from statistical tests of the significance of differences, persons familiar
with the usual experimental values obtained in studies of this type will
immediately recognize the perceptual smallness of the maximum differences.
Thus, for example, the means for pitch level, descriptive of the
central tendencies of the distributions of pitches used, are confined within
a musical extent of 0.5 tone, or one semitone. In the basic studies the
differences between the age groups were not statistically significant 1 2In
both instances the design was a two-group comparison between ages seven
and eight, sex constant, the test of significance being the t. Comparisons
between the sex groups were not made.

II. Results

Analysis of Variance

Analysis of Variance. The present study tested the hypothesis that the
four basic experimental groups, representing both sexes at two age levels,
were random samples from the same population with respect to the pitch
characteristics in question. Table III presents the results of simple analysis
of variance, carrying the four groups in parallel. With F required to equal
4.16 or 2.78 at the 1 per cent or 5 per cent levels, respectively, 2 3it will be
seen in the third column of the table that all obtained values fall far
short of significance.

As a further test, “factorial” analysis was performed, employing an age-by-sex
arrangement, with the results shown in Table IV. Age, sex, and
interaction variances are given, and the within groups variances repeated
from Table HI. The F ratios are superfluous, as may be seen by inspection,
but were calculated for record.

In general, neither Table III nor Table IV presents any evidence for rejection
of the null hypothesis. This finding probably is particularly important
with respect to pitch level and number of voice breaks, because of the relationship
of these variables to development of the larynx as a function of
age, and because of the obvious differences between the sexes at later age
levels. It provides, furthermore, a suitable base of departure for the larger
research program which has been planned, and which will consist of
similar studies at successive age levels throughout childhood. Of special
interest in these studies will be the determinations of the age levels at
which significant differences between the sexes first appear, and at which
differences between age levels first appear, sex constant.

Such studies will also increase the confidence with which the data of
the present study may be interpreted. It has been shown here that the
variances between groups are both absolutely and relatively small, and that
all groups resemble mid-adolescent males in many respects. So radically
do such findings depart from tradition that they must be accepted with
great caution. Differences between fact and tradition, however, are hardly
rare phenomena, and it should be pointed out that, within the knowledge
of the author, the present study constitutes the first extensive, objective
investigation of this problem. Nevertheless, two observations are in order:
first, it is possible, although seemingly unlikely, that the four groups were
unrepresentative of the sex and age populations from which they were
randomly drawn; second, it is possible, although seemingly unlikely, that
these seven- and eight-year-old boys and girls were not preadolescents, as
ordinarily conceived.

Comparison of Variances. Sex Groups

On the basis of these indications the five significant measures of Table
V were further examined by reverting to the basic groups of 15 subjects
each, and comparing males to females at the two age levels separately. This
procedure yielded the 10 sets of values given in Table VI. Study of this
table shows that the variance was larger in the male group in nine of the
ten comparisons, that the difference in variance was significant beyond the
10 per cent level in six of these nine, and that five of these six significant
differences were at age eight. Again it is observed that three of the six
significant differences occurred in related measures of extent of pitch shifts.
In none of the comparisons of Tables V and VI was the variance significantly
larger in the female. It would appear justifiable, therefore, to record
a slight tendency for variance to be larger in the male, especially at age
eight.197

Comparison of Variances. Age Groups

Comparison of Variances. Age Groups. As in the preceding section,
the four basic groups were then combined into two age groups of 30 subjects
each, and the same procedures followed. The results are shown in
Table VII. In 10 out of the 12 measures the variance is larger for the
eight-year-old children, and in two of these ten the difference is significant
at the 10 per cent level. In number of downward voice breaks and in
number of total voice breaks the variance is larger at age seven, and the
difference is significant at the 10 per cent level in the former measure. The
three measures in which significant differences were found were singled
out as above for similar comparisons employing the basic groups, with
the results shown in Table VIII. The variance in number of downward
voice breaks, represented by the first two sets of values, is seen to be larger
at age seven in both males and females, the difference between the two
female groups being significant at the 10 per cent level. In the remaining
four comparisons the difference is in the opposite direction in three
instances, and reaches significance at the 10 per cent level once. Except for
the inexplicable reversal in the case of downward voice breaks, the data of
Tables VII and VIII seem to indicate, then, a mild and inconclusive trend
toward larger variance at age eight.

III. Summary

References

1. Fairbanks, G. Recent experimental investigations of vocal pitch in
speech. J. Acoust. Soc. Amer., 1940, 11, 457-466.

2. Fairbanks, G., Herbert, E. L. and Hammond, J. M. An acoustical
study of vocal pitch in seven- and eight-year-old girls. Child Develpm.,
1949, 20, 71-78.

3. Fairbanks, G., Wiley, J. H. and Lassman, F. L. An acoustical study
of vocal pitch in seven- and eight-year-old boys. Child Develpm.,
1949, 20, 63-69.

4. Snedecor, G. W. Statistical Methods. (4th Ed.) Ames: Iowa State
College Press, 1946.

Manuscript received November 3, 1949.200