Psychological Bulletin

1969, Vol. 72, No. 1, 50-62

REVIEW OF RESEARCH ON THE INTELLIGIBILITY AND

COMPREHENSION OF ACCELERATED SPEECH1
EMERSON FOULKE 2 AND THOMAS G. STICHT
University of Louisville Human Resources Research Office
George Washington University

Time-compressed or accelerated speech is speech which has been reproduced in

less than the original production time. Such speech may prove to be useful in a
variety of situations in which people must rely upon listening to obtain the
information specified by language. It may also prove to be a useful tool in
studying the temporal requirements of the listener as he processes spoken lan-
guage. Methods for the generation of time-compressed speech are reviewed.
Methods for the assessment of the effect of compression on word intelligibility
and listening comprehension are discussed. Experiments dealing with the effect
of time compression upon word intelligibility and upon the comprehensibility
of connected discourse, and experiments concerned with the influence of stimulus
variables, such as signal distortion, and organismic variables, such as intelligence,
are reviewed. The general finding that compression in time has a different effect
upon the comprehensibility of connected discourse than upon word intelligibility
is discussed, and a tentative explanation of this difference is offered.

Accelerated speech is speech in which the niques for the acceleration of speech are de-
word rate has been increased. Increasing the scribed, methods for its evaluation are re-
word rate reduces communication time for a viewed, and characteristics of the listener that
given message. Hence, accelerated speech is may affect his perception of time-compressed
often referred to as time-compressed or speech are discussed.
simply compressed speech.
Since the announcement by Fairbanks METHODS FOR THE ACCELERATION OF SPEECH
(Fairbanks, Everitt, & Jaeger, 1954) of a
practical means for the time compression of Speaking Rapidly
recorded speech, there has been an interest in Within limits, word rate is under the con-
its use to enable blind people to read by trol of the speaker, and this method has been
listening at a rate that compares favorably used by several investigators (Calearo &
with the silent visual reading rate (Foulke, Lazzaroni, 1957; deQuiros, 1964; Enc &
Amster, Nolan, & Bixler, 1962; Iverson, Stolurow, 1960; Fergen, 1955; Goldstein,
1956). More recently, time-compressed speech 1940; Harwood, 1955; Nelson, 1948). This
has been considered for use as an audio aid method has the virtue of simplicity and re-
in general education (Friedman, Orr, Freedle, quires no special equipment. However, it is
& Norris, 1966; Orr & Friedman, 1964) and limited by the fact that only a moderate
as a research tool for studying the auditory increase in the rate of articulation of speech
perception of language (Foulke & Sticht, sounds is possible. When the speaker increases
1967). his word rate by talking faster, there are
The present paper is concerned with the changes in vocal inflection and intensity, and
communication problems produced by the
in the relative duration of consonants, vowels,
time compression of speech. Various tech-
1
and pauses (Kozhevnikov & Chistovich,
The work described in the present paper was 1965). When word rate is increased by meth-
performed at the University of Louisville as part of
Project 7-12S4, Cooperative Research Branch, Office ods that alter the rate of reproduction of
of Education, Department of Health, Education, and recorded speech, these changes do not take
Welfare. place. The significance of this fact, with
2
Requests for reprints should be addressed to Dr.
Emerson Foulke, Department of Psychology, Uni- respect to word intelligibility or listening
versity of Louisville, Louisville, Kentucky 40208. comprehension, has not yet been determined.
SO
 INTELLIGIBILITY AND COMPREHENSION OF ACCELERATED SPEECH 51

The Speed-Changing Method poses. However, once the success of the gen-
eral approach was shown, an efficient tech-
The word rate of a recorded message may
nique for accomplishing it was not long to
be changed, simply by reproducing it at a
follow.
different tape or record speed than the one
used during recording. If the playback speed In 1954, Fairbanks et al. published a de-
scription of an electromechanical apparatus
is slower than the recording speed, word rate
for the time compression or expansion of
is decreased, and the speech is expanded in
time. If playback speed is increased, word recorded speech, which embodies a principle
rate is increased, and the speech is compressed adumbrated by Gabor (1946, 1947). The
in time. When word rate is compressed in this Fairbanks apparatus reproduces periodic
manner, there is a shift in the frequencies that samples of a recorded tape. The unreproduced
samples are brief enough so that a discarded
constitute the voice signal, which is propor-
tional to the change in tape or record speed. sample cannot contain an entire speech sound,
If the speed is doubled, the component fre- and the retained samples are abutted in time.
Under these condtions, every speech sound in
quencies will be doubled, and vocal pitch will
be raised one octave. Speech compressed by the original recording is sampled, and the
the speed-changing method has been examined result is a time-compressed reproduction
without alteration in vocal pitch. With this
in several experiments (Fletcher, 1929, pp.
292-294; Foulke, 1966a; Garvey, 19S3b; apparatus, speech can be expanded in time
Klumpp & Webster, 1961; Kurtzrock, 1957; by periodically repeating samples of a re-
McLain, 1962). corded tape. A computer may also be used
for the time compression or expansion of
speech by the sampling method (Scott, 1965).
The Sampling Method
Whereas speech compressors of the Fairbanks
In 1950, Miller and Licklider demonstrated type sample periodically and unselectively, a
the signal redundancy in spoken words, by computer permits a variety of sampling rules.
deleting brief segments of the speech signal. For instance, a computer might be pro-
This was accomplished by a switching ar- grammed to dispose of empty time intervals
rangement that permitted a recorded speech between words, and to sample the time inter-
signal to be turned off periodically during its vals occupied by words differentially, dis-
reproduction. They found that as long as these carding larger fractions of those speech sounds
interruptions occurred at a frequency of 10 with higher signal redundancy. Though, be-
times per second, or more, the interrupted cause of its flexibility, the computer may pro-
speech was easily understood. The intelligibil- vide the most satisfactory method for the
ity of monosyllabic words did not drop below time compression or expansion of speech, at
90% until 50% of the speech signal had been present, computer time is too expensive to
discarded. Thus, it appeared that a large por- justify the employment of a computer in this
tion of the speech signal could be discarded capacity for any but research purposes.
without a serious disruption of communica- The time compression of speech may be
tion. Garvey (1953b), taking cognizance of accomplished by the shortening or the elimi-
these results, reasoned that if the samples of nation of the natural pauses occurring in
a speech signal remaining after periodic inter- speech (Diehl, White, & Burk, 1959; Miron
ruption could be abutted in time, the re- & Brown, 1968). This may be done manually,
sult should be time-compressed, intelligible by removal of blank segments of a recorded
speech, without distortion in vocal pitch. To tape, or by means of a computer, and the
test this notion, he prepared a tape on which remaining speech may be compressed or
speech had been recorded by periodically uncompressed.
cutting out short segments of tape, and by The technique of speech synthesis offers
splicing the ends of the retained tape together another possibility for the compression of
again. Reproduction of this tape achieved the speech in time (Campanella, 1967). The har-
desired effect. Garvey's method was, of course, monic compressor, a device for the time com-
too cumbersome for any but research pur- pression of speech based on research per-
52 EMERSON FOULKE AND THOMAS G. ST1CHT

formed at Bell Laboratories, is now under 174 wpm. The oral reading rate is the rate
construction at the American Foundation for that is relevant to the process under discus-
the Blind. sion since, in most cases, the speech that is
compressed is recorded oral reading. However,
METHODS FOR THE EVALUATION OF the usefulness of average oral reading rates
ACCELERATED SPEECH is limited. The rate of oral reading depends
upon the nature of the material being read,
Some Procedural Problems and this kind of variability can be reduced
There is no common practice in specifying by reporting syllable rate, rather than word
the amount of compression to which a listen- rate (Carroll, 1967). The oral reading rate
ing selection has been subjected. This lack also depends upon the style of the individual
of uniformity can result in confusion, espe- reader. It varies considerably from reader to
cially when the results of different studies reader and from sample to sample of the
are compared (Bellamy, 1966). The amount production of a given reader (Foulke, 1967).
of compression may be specified by the per- There are reasons for believing that speech
centage of the original recording time that is rate is the dimension of which listeners are
saved by reproducing the message at a faster aware. Johnson et al. (1963, pp. 202-203)
word rate. Thirty percent compression means summarized research supporting the conclu-
that 30% of the production time has been sion that perception of the rate of speaking
saved. Conversely, the fraction of original corresponds to the oral reading rate. Hutton
production time remaining after compression (1954) found a logarithmic growth in per-
may be specified. ceived word rate as measured word rate was
Alternatively, specification may be in terms increased linearly.
of the acceleration of the original word rate, A variety of initial or uncompressed word
tape speed, or record speed. An acceleration rates have been used in studies of the effect of
of 1.5 means that the word rate after com- time compression on listening comprehension
pression is l.S times the word rate before (Fairbanks, Guttman, & Miron, 1957c;
compression. In comparing these indices, it Foulke et al., 1962; Goldstein, 1940). These
must be remembered that the relationship be- studies indicated that a rapid decline in com-
tween them is not linear. For instance, prehension commences beyond a word rate
whereas an increase in acceleration from 1.1 of approximately 275 wpm regardless of the
to 1.2 corresponds to an increase in compres- compression which may have been required
sion from 9% to 17%, an increase in accelera- to achieve that word rate. Thus, it seems
tion from 1.9 to 2.0 corresponds to a change advisable to describe compressed speech not
in compression from 47% to 50%. only in terms of the amount of compression
A problem common to both indices is that but also in terms of word rate.
they do not indicate directly the word rate For certain purposes, such as the measure-
of compressed speech. The final word rates ment of intelligibility, single words are com-
of two listening selections, compressed or ac- pressed, and it is, of course, meaningless to
celerated by the same amount, depend upon speak of the word rate of a single word. In
the rates of speaking before compression. these cases, specification must be made in
There is considerable variability in the pub- terms of compression or acceleration ratio.
lished estimates of word rate. Part of this
variability is undoubtedly due to the dif- The Measurement of Intelligibility
ference between spontaneous, conversational The ability to repeat a word, phrase, or
word rate and the word rate of oral reading. short sentence accurately is often taken as an
Nichols and Stevens (1957) found a con- index of the intelligibility of time-compressed
versational speaking rate of 12 S words per speech. A procedure typical of this approach
minute (wpm), while Johnson, Barley, and is one in which words are compressed in time
Spriestersbach (1963, p. 220) found a median by some amount and presented, one at a
oral reading rate of 176.S wpm, and Foulke time, to a listener. The listener's task is to
(1967) found a mean oral reading rate of reproduce them orally, or in writing, and his
 INTELLIGIBILITY AND COMPREHENSION OF ACCELERATED SPEECH 53

score is the correctly identified fraction of Wood (196S) dealt with the problems in-
those words. This procedure is sometimes herent in assessing the listening comprehen-
referred to as an articulation test (Miller, sion of young children by determining their
1954, p. 60). ability to follow brief, verbal instructions,
Disjunctive reaction time (RT) may also compressed in time. Instructions consisted of
be taken as an index of intelligibility (Foulke, imperative statements, such as "buzz like
1965). The underlying rationale, in this case, a bee."
is that reduced discriminability means reduced Some tests of listening comprehension may
intelligibility. It has been shown that if stim- detect differences not detected by others, but
uli are made more similar, and hence less this increased sensitivity may have been pur-
discriminable, choice RT is increased (Wood- chased at the cost of a loss in reliability, or
worth & Schlosberg, 19S4, p. 33). The pro- in ease of test administration and scoring.
cedure for testing intelligibility, under this Bellamy (1966) used both a multiple-choice
approach, is to acquaint the subject with a test and an interview technique to determine
list of response words. The words are then the listening comprehension of a group of
presented to the subject, one at a time, in blind subjects and a comparable group of
random order, for identification. The subject sighted subjects. She reported that the inter-
indicates his choice with a discriminative re- view technique revealed a difference in favor
sponse, for instance, pressing an appropriate of the blind subjects not detected by the
response key. He can then be scored for speed multiple-choice test. Friedman et al. (1966)
and accuracy of reaction. The experiment is used short answer and essay tests to assess
performed using words that have been com- the comprehension of accelerated speech and
pressed in time by several amounts, and found no discernable trend in performance as
changes in RT and/or accuracy are regarded a function of practice in listening to such
as indicative of changes in intelligibility. The speech. On the other hand, a multiple-choice
RT method may be more sensitive than other test revealed considerable improvement. They
methods, since a change in the amount of also found a lack of correlation between the
compression may produce a change in RT to results of short answer and essay tests.
words which are discriminated without error.
Calearo and Lazzaroni (19S7) reported the THE INTELLIGIBILITY OF TIME-
use of a method for testing intelligibility, COMPRESSED SPEECH
familiar to those in clinical audiology, in order
Characteristics of the Signal
to detect the effects of compression. The
minimum intensity required for words to be 1. The method of compression. The intel-
intelligible is determined for words at several ligibility of time-compressed words depends,
compressions. Threshold intelligibility is de- in part, upon the method used for compres-
fined as that intensity at which some per- sion. When a recording is played back at a
centage (usually 50%) of a list of words is speed that is enough faster than the record-
correctly identified. If the threshold for in- ing speed to result in the compressed repro-
telligibility changes as the amount of com- duction of a list of words in two-thirds of their
pression is changed, it is concluded that original production time, there is a loss in
compression has affected intelligibility. intelligibility of 40% or more (Fletcher,
1929; Garvey, 19S3b; Klumpp & Webster,
Tests of Comprehension 1961; Kurtzrock, 1957). On the other hand,
In this approach to the evaluation of the Garvey (1953b) found only a 10% loss in
effects of compression, the listener first hears the intelligibility of a list of words, each of
a listening selection, compressed in time by which was reproduced in 40% of its original
some amount, and is then tested for compre- production time by means of his manual
hension of that selection. Any kind of test sampling method and a 50% loss in intel-
may be used, but researchers have, in most ligibility for words reproduced in 25% of
cases, preferred objective tests of specifiable their original production times. Kurtzrock
reliability. (1957), using the electromechanical sampling
54 EMERSON FOULKE AND THOMAS G. STICHT

method of Fairbanks, obtained an intelligi- corresponding intelligibility scores of 95, 96,

bility score of 50% for a group of words 95, and 86%. In a two-factor experiment in
reproduced in 15% of their original produc- which five discard intervals and eight com-
tion times. Using the same method and similar pressions were represented, Fairbanks and
materials, Fairbanks and Kodman (1957) Kodman (1957) also found a substantial loss
obtained an intelligibility score of 57% for a in intelligibility when the duration of the dis-
group of words reproduced in only 13% of card interval exceeded 80 msec. This was true
their original production times. at all eight compressions.
Compression by either the sampling or the Cramer (1965) reported that when subjects
speed-changing method increases the rate at use earphones to listen to speech that has been
which the discriminable elements of speech compressed in time by the sampling method,
occur. However, whereas the overall spectrum, delaying the signal to one earphone by 7.5
the location of formants within that spectrum, msec, improves intelligibility. This delay pro-
and vocal pitch are unaffected by the sampling vides what Cramer has called "binaural re-
method, they are altered by the speed- dundancy." If, as Garvey (1953a) suggested,
changing method, and these alterations are it is the briefness of highly compressed speech
probably responsible for the difference in in- sounds that makes them unintelligible, bin-
telligibility between the two methods (Nixon, aural redundancy may restore some intelligi-
Mabson, Trimboli, Endicott, & Welch, 1968; bility by increasing the effective duration of
Nixon & Sommer, 1968). speech sounds.
2. Intelligibility and the sampling rule. The Scott (1965) reported a favorable result
message to be compressed may be conceived when subjects use one earphone to listen to
as consisting of a succession of temporal seg- the normally retained samples of compressed
ments, called sampling periods. When speech speech, and the other earphone to listen, at
is compressed by the sampling method, com- the same time, to the normally discarded
pression is accomplished by discarding a samples of the same compressed speech. He
fraction of each sampling period and by refers to such speech as "dichotic speech."
abutting in time the remainders of sampling 3. The rate of occurrence of speech sounds.
periods. It is the retained fraction of the Garvey (1953b) compared the intelligibility
sampling period that determines the amount of words compressed in time by the sampling
of compression. If 10 milliseconds (msec.) of method with the intelligibility, reported by
a 20-msec. sampling period or 30 msec, of a Miller and Licklider (1950), of words that
60-msec. sampling period are retained, the had been interrupted periodically. Garvey's
result is the same—50% compression. For words and Miller and Licklider's words were
any given sampling period, changing the frac- treated alike in that portions of sampling
tion of the sampling period that is retained periods were discarded. However, the retained
changes the amount of compression. samples of Garvey's words were abutted to
When the sampling method is used, the produce time-compressed speech, while the
effect that a given amount of compression has retained samples of Miller and Licklider's
upon the intelligibility of words depends upon words were not abutted, and the resulting
the duration of the discarded portion of the speech, though interrupted, was not com-
sampling period and hence upon the duration pressed in time. There was no difference be-
of the sampling period itself. The duration of tween the intelligibility of time-compressed
the discarded portion of the sampling period words and interrupted words when 50% of
must be short relative to the duration of the each word was discarded. However, when
speech sounds to be sampled. If it is not, a 62% of each word was discarded, interrupted
speech sound may fall entirely within the words were 40% more intelligible than time-
discarded portion of a sampling period, in compressed words. Since the two groups of
which case, it is not sampled at all. Garvey words were alike with respect to the amount
(1953b) used discard intervals of 40, 60, 80, of speech information that had been dis-
and 100 msec, to compress spondaic words to carded, the poorer intelligibility of the time-
50% of their original durations. He obtained compressed words, when 62% of the speech
 INTELLIGIBILITY AND COMPREHENSION OF ACCELERATED SPEECH 55

information was discarded, was probably due the tests were made, whereas the subjects of
to the increased rate of occurrence of speech Miller and Licklider were relatively naive.
sounds. Garvey used spondaic words, whereas Miller and Licklider (1950), using inter-
Miller and Licklider used monosyllabic words. rupted words, and Garvey (1953a), using
Results obtained by Henry (1966) suggest words compressed in time by the sampling
that if Garvey had used monosyllabic words, method, found that repeated exposure to such
or if Miller and Licklider had used spon- words improves their intelligibility.
daic words, the difference in favor of inter- If a group of listeners agree that a par-
rupted speech would have been even more ticular speech sound in a word that has been
pronounced. compressed in time by the sampling method is
4. Intelligibility and linguistic factors. unrecognizable, it may be concluded fairly
Kurtzrock (1957) found that compression by that the difficulty lies with the signal itself.
the speed-changing method degraded the intel- However, Garvey found that subjects dis-
ligibility of vowel sounds more than conso- agreed about the speech sounds that were
nantal sounds and that compression by the rendered unintelligible by compression of the
sampling method degraded the intelligibil- words in which they occurred. Garvey ex-
ity of consonantal sounds more than vowel plained this finding in terms of the differential
sounds. Garvey's (1953a) subjects rated the exposure of subjects to the words in question.
vowel sounds in words that had been com- In this connection, Henry (1966) found a
pressed in time by the sampling method as positive relationship between word frequency
higher in "goodness" than consonantal sounds. in general language, as revealed in the Thorn-
In a study in which the number of phonemes dike and Lorge (1944) word count, and word
per word was varied from three to nine, intelligibility.
Henry (1966) found that increasing the num- 2. Intelligibility and hearing loss. There
ber of phonemes improved the intelligibility appear to be no differential effects of time
of words that had been compressed in time compression upon the intelligibility scores of
by the sampling method. In a similar vein, normal hearing subjects and patients having
Klumpp and Webster (1961) found short conductive or sensorineural hearing losses
phrases, compressed in time by the speed- (Bocca & Calearo, 1963; Calearo & Laz-
changing method, to be more intelligible than zaroni, 1957; deQuiros, 1964; Luterman,
single words. The findings of Henry, and of Welsh, & Melrose, 1966; Sticht & Gray, in
Klumpp and Webster, are probably explained press). However, aged patients, some with
by the increased number of cues available to diffuse cerebral pathology (Calearo & Laz-
subjects because of the redundancy in poly- zaroni, 1957; Sticht & Gray, in press), and
phonemic words and short phrases, and could patients with temporal lobe lesions (Bocca &
have been predicted from the finding of Calearo, 1963; deQuiros, 1964) required
French and Steinberg (1947) that speech is greater intensity for threshold intelligibility
understandable when composed of syllables and showed a higher error rate with supra-
that are only 67% intelligible. threshold words when compression was in-
creased. The latter was true for aged subjects
Characteristics oj the Listener having normal hearing or sensorineural hear-
ing losses (Sticht & Gray, in press). Appar-
1. Intelligibility and prior experience. Fair- ently, the changes accompanying aging reduce
banks and Kodman (1957) found a group of the rate at which speech information can be
words compressed by several amounts to be processed.
more intelligible than a similar group of words
in which the same amounts of speech infor- FACTORS AFFECTING THE COMPREHENSION OF
mation had been discarded by interrupting TIME-COMPRESSED SPEECH
them in the manner of Miller and Licklider.
However, the subjects of Fairbanks and Stimulus Variables
Kodman had received extensive familiariza- 1. Comprehension and word rate. Within
tion with the words to be identified before the range extending from 126 to 272 wpm,
56 EMERSON FOVLKE AND THOMAS G. STICHT

Diehl et al. (1959) found listening compre- made after several retention intervals. In
hension to be unaffected by changes in word general, these studies support the conclusion
rate. In the range bounded by 125 and 225 that differences in the course of forgetting are
wpm, Nelson (1948) and Harwood (1955) due to differences in original learning. Of
found a slight but insignificant loss in listen- course, as has already been shown, the amount
ing comprehension as word rate was increased. of original learning is, in part, a function of
Fairbanks et al. (1957c) found little differ- the word rate at which a listening selection
ence in the comprehension of listening selec- is presented.
tions presented at 141, 201, and 282 wpm. 2. Comprehension and the method of com-
Thereafter, comprehension, as indicated by pression. McLain (1962) and Foulke (1962),
percentage of test questions correctly an- using subjects who were nai've with respect to
swered, declined from 58% at 282 wpm to compressed speech and unaccustomed to read-
26% at 470 wpm, a level of performance near ing by listening, compared the comprehension
chance. Foulke et al. (1962), using both lit- of a listening selection compressed by the
erary and technical listening selections, found sampling method to a rate of 275 wpm with
listening comprehension to be only slightly the comprehension of the same selection com-
affected by increasing word rate in the range pressed to the same word rate by the speed-
bounded by 175 and 275 wpm. However, in changing method. In both instances, a slight
the range extending from 275 to 375 wpm, but statistically significant advantage was
they found an accelerating loss in listening found for the sampling method. However, in
comprehension as word rate was increased. a similar experiment in which blind children,
Foulke and Sticht (1967) found a 6% loss in who were accustomed to reading by listening,
comprehension between 225 and 325 wpm and served as subjects, Foulke (1966a) found no
a loss of 14% between 325 and 425 wpm. The statistically significant difference in favor of
three studies just cited are in agreement re- either method.
garding the finding that as word rate is in- The finding that the obvious superiority of
creased beyond a normal word rate, there is the sampling method when the comparison is
initially a moderate linear decline in compre- based upon a test of the intelligibility of
hension, followed by an accelerating decline. single words is not observed when the com-
Simple comprehension scores do not take parison is based upon a test of the compre-
into account the learning time that is saved hension of connected discourse is of consider-
when speech is presented at an increased word able interest. It suggests that some other
rate. Such an allowance may be made by factor, such as the rate at which words occur,
dividing the comprehension score by the time is also involved in determining the compre-
required to present the listening selection. hension of accelerated speech. A satisfactory
This index of learning efficiency expresses the explanation of such comprehension must,
amount of learning per unit time. Using such therefore, take into account the perceptual
an index, Fairbanks et al. (1957c), Enc and and cognitive processes of the listener.
Stolurow (1960), and Foulke et al. (1962) 3. Comprehension and the difficulty of the
found that learning efficiency increased as compressed material. The extent to which the
word rate was increased until a word rate comprehension of a listening selection is af-
of approximately 280 wpm was reached. fected by compression in time may depend
In a similar approach, Enc and Stolurow upon its difficulty. However, before this ques-
(1960) computed an index of the efficiency tion can be examined satisfactorily, a method
of retention. must be developed for determining the diffi-
The word rate at which a listening selection culty of a listening selection.
is presented apparently has no special effect Using one normal and four accelerated word
on the rate at which forgetting occurs. Enc rates, Foulke et al. (1962) measured the com-
and Stolurow (1960), Friedman et al. (1966), prehension of a scientific selection and a lit-
and Foulke (1966b) performed studies in erary selection. In each case, performance on
which tests of the comprehension of listening a test containing multiple-choice items cover-
selections presented at several word rates were ing the listening selection constituted the evi-
 INTELLIGIBILITY AND COMPREHENSION OF ACCELERATED SPEECH 57

dence for listening comprehension. Compre- scores suggested that the scientific selection
hension of the scientific selection was poorer was relatively more difficult than the literary
than comprehension of the literary selection selection, they were estimated to be equal in
at a normal word rate, suggesting that it was difficulty by the Dale-Chall formula for
relatively more difficult. As word rate was readability. Similar evidence is presented in
increased, comprehension of the scientific a study reported by Enc and Stolurow
selection did not decline as rapidly as compre- (1960). They found considerable variability
hension of the literary selection. Although this in the mean comprehension test scores of ten
interaction was significant, it was probably listening selections, presented at a normal
due to the fact that since comprehension word rate and a slightly accelerated word
scores for the scientific selection were lower at rate, in spite of the fact that the selections
a normal word rate, the range in which they were rated as equal in difficulty by the Dale-
could vary was relatively smaller. Further- Chall formula. Of course, the formula may
more, the apparent difference in difficulty of have failed to detect differences in listening
the two selections may have been due, at difficulty because of a relatively large vari-
least in part, to differences in the tests of ance in the estimates of reading difficulty.
listening comprehension employed. Certainly, However, if the difficulty of an aurally
the apparent difficulty of a selection can be received selection is not the same as the
manipulated by the choice of items used in difficulty of that selection when visually re-
testing for its comprehension. ceived, the explanation may be that differ-
In an investigation of the effect of time ences between the oral and the print display
compression on message units varying in dif- make it necessary for the reader to process
ficulty, Fairbanks et al. (1957c) distributed them differently. The printed page is primar-
the 60 multiple-choice items of a test of listen- ily a spatial display. It permits the kind of
ing comprehension equally among five cate- scanning that helps in understanding long,
gories of item difficulty. The listening selec- complex sentences. On the other hand, when
tions covered by the test of comprehension information is specified by spoken language,
were administered to several groups of sub- it is displayed in a temporal dimension. The
jects, each group experiencing a different ac- only sensory information available to the
celerated word rate. Each subject received five listener at any given instant is the informa-
scores, determined by his responses to the tion specified by the display at that instant.
items in each of the five test-item categories. Unlike the visual reader, the listener must
The mean score for each test-item category depend upon memory alone for the availabil-
decreased as the amount of compression in ity of speech that has already occurred.
time was increased. They concluded that, as- Furthermore, unlike the visual reader, he can
suming item difficulty to be a reflection of the exert no control over the order in which he
difficulty of the message unit to which it per- encounters the syntactic and semantic com-
tained, the effect of time compression on ponents of sentences. The syntactical differ-
listening comprehension, within the range ex- ence between two selections might be inconse-
plored, did not depend upon the difficulty of quential when they are received visually, yet
the listening material. quite significant when they are received
There are formulas for estimating what aurally. The formulas used for estimating
might be called the "absolute difficulty" of a reading difficulty (Dale & Chall, 1948;
selection. These formulas have generally been Flesch, 1948; Rodgers, 1962) are based on
developed for material that is to be read visu- different considerations, and the estimates of
ally (Dale & Chall, 1948; Flesch, 1948). difficulty yielded by these formulas may be
However, it has often been assumed that the expected to vary. However, there has been no
listening difficulty of a selection is the same comparative study of the extent to which the
as its reading difficulty. The results of the effect of word rate on listening comprehension
experiment by Foulke et al. (1962) suggest depends upon the formula used to estimate
that this assumption may not be tenable. In difficulty. The finding of a systematic inter-
this experiment, although comprehension test action between word rate and listening
58 EMERSON FOULKE AND THOMAS G. ST1CHT

difficulty, as estimated by a particular for- their experiments included Grades 1, 3, 4, 5,

mula, would seem to provide a kind of face and 6.
validity for that formula. 3. The intelligence of the listener. In the
4. Comprehension and the oral reader. Oral case of children, the evidence presently avail-
readers differ considerably with respect to able is not sufficient to permit a conclusion
vocal timbre, and, of course, there are con- regarding the effect of intelligence on the
spicuous sex differences in vocal pitch. Oral comprehension of accelerated speech. Fergen
readers also differ with respect to such factors (1955) found no relationship between the
as average word rate and variability in word IQs of grade school children and their ability
rate, pitch, and loudness. Such factors com- to comprehend accelerated listening selections.
bine to define the personal, oral reading style. However, 230 wpm was the fastest word rate
In a preliminary experiment, Foulke (1964) represented in her experiment. Wood (1965)
explored the extent to which oral reading found no relationship between the IQs of chil-
style interacts with word rate in determining dren in the primary grades and their ability
listening comprehension. Three renditions of to follow the instructions conveyed by short,
a listening selection, each read by a different imperative, time-compressed statements. How-
reader (two males and one female), were ever, his procedures resemble more closely
presented to three groups of college students those used in testing for intelligibility. A more
at a normal word rate, and to three compa- definite conclusion is possible in the case of
rable groups at a word rate that was increased adults. Fairbanks et al. (1957b, 1957c),
to 275 wpm by the sampling method. After Goldstein (1940), and Nelson (1948) have all
exposure to the listening selection, all sub- found a positive relationship between intel-
jects took a test of listening comprehension. ligence and the ability to comprehend acceler-
Significant differences in listening comprehen- ated speech. The data of Fairbanks et al.
sion were associated with the reader vari- (1957c) and Goldstein (1940) concur in
able and with the word-rate variable, but the showing a positive relationship between the
reader's effect on listening comprehension did intelligence of the listener and the magnitude
not depend upon the word rate at which the of the decline in listening comprehension as
selection was presented. word rate is increased. This relationship may
be due, at least in part, to the fact that
Listener Variables That Affect Listening intelligent subjects earn higher scores than
Compreh ension less intelligent subjects on comprehension
tests of listening selections presented at nor-
Foulke (1964) called attention to the con- mal word rates. Therefore, the scores they
siderable variation in the ability of listeners earn on tests of the comprehension of mater-
to comprehend accelerated speech. Several ials presented at accelerated word rates have
experiments have been reported in which there a larger range within which to vary.
has been an effort to determine those charac- 4. The visual status of the listener. There
teristics of the listener that may contribute are a priori grounds for expecting blind listen-
to the ability to comprehend accelerated ers to show better comprehension than sighted
speech. listeners. However, the research related to this
1. The sex of the listener. Comparisons of question is meager and inconclusive. In an
male and female listeners have revealed no experiment performed by Hartlage (1963),
sex-related differences in listening comprehen- blind and sighted subjects did not differ with
sion for word rates ranging from 174 to respect to their comprehension of listening
475 wpm (Foulke & Sticht, 1967; Orr & selections presented at a normal word rate.
Friedman, 1964). Foulke (1964) presented evidence that blind
2. The listener's age and educational ex- listeners comprehend time-compressed listen-
perience. Fergen (1955) and Wood (1965) ing selections better than sighted listeners.
found a positive relationship between the age- 5. Reading rate and listening rate. Those
grade level of school children and their ability perceptual and cognitive processes that are
to comprehend accelerated speech. Together, responsible for individual differences in read-
 INTELLIGIBILITY AND COMPREHENSION OF ACCELERATED SPEECH 59

ing rate may also contribute to individual equal to the playback time of the uncom-
differences in the ability to comprehend accel- pressed and unaugmented selection. The ob-
erated speech. If this is true, fast readers jective was to determine whether or not com-
should be able to comprehend speech at a prehension could be improved by trading the
faster word rate than slow readers. This hy- temporal redundancy in the uncompressed
pothesis has been tested by Goldstein (1940) version for the verbal redundancy in the aug-
and by Orr, Friedman, and Williams (1965). mented version. Analysis of the results re-
In both experiments, a significant positive cor- vealed better comprehension only for the
relation was found between reading rate and augmented sections of the listening selection.
the ability to comprehend accelerated speech. There was a decline in comprehension of the
Of course, in all likelihood, a significant posi- unaugmented sections. The explanation of
tive correlation would also have been found this finding may be that subjects associated
between reading rate and reading comprehen- verbal redundancy with importance and dis-
sion. In both experiments, it was also found tributed their attention accordingly.
that practice in listening to accelerated speech Several investigators have explored the pos-
resulted in an improvement in reading rate. sibility of improving the comprehension of
Goldstein (1940) and Jester and Travers accelerated speech by training. The simplest
(1965) compared the comprehension result- and least sophisticated training experience
ing from listening to selections presented at that has been evaluated is mere exposure.
several word rates with the comprehension Voor and Miller (1965) exposed a group of
resulting from reading the same selections at subjects to five listening selections presented
the same word rates. In both cases, compre- at 380 wpm. Total listening time was 17.5
hension declined as word rate was increased. minutes. At the end of each selection, sub-
Listening comprehension was superior to jects were tested for listening comprehension.
reading comprehension up to approximately Mean comprehension scores increased from
200 wpm but inferior to reading comprehen- the first to the third selection but did not
sion thereafter. Simultaneous reading and change significantly thereafter. These re-
listening at 350 wpm resulted in better com- sults probably reflect a simple adjustment to
prehension than could be demonstrated with the initially unfamiliar task of listening to
either mode of presentation alone. accelerated speech.
6. Improving the comprehension of time- Orr et al. (1965) found a 29.3% increase in
compressed speech. In an experiment per- the comprehension of materials presented at
formed by Fairbanks et al (1957b), a mean 475 wpm, following several weeks of training
comprehension score of 63.8% was obtained in which subjects listened to selections, the
by subjects who listened to a selection pre- word rates of which were increased in steps
sented at an uncompressed word rate at 141 of 25 wpm from 325 to 475 wpm. However,
wpm. Subjects who listened to the same selec- since there was no control group that received
tion, compressed by 50% to a word rate of training in listening for comprehension at a
282 wpm, earned a mean comprehension score normal word rate, it is not possible to attribute
of 58%. A third group of subjects, who lis- their results unequivocally to practice in lis-
tened to two consecutive reproductions of the tening to accelerated speech. The improve-
listening selection at 282 wpm, earned a mean ment may have been due simply to practice in
comprehension score of 65.4% which was listening for comprehension.
slightly, but probably not significantly, higher In this regard, Foulke (1964), using blind
than the mean comprehension score resulting subjects who can safely be presumed to have
from a single exposure to the uncompressed had years of experience in listening for com-
selection. In a second study by the same in- prehension, measured comprehension of speech
vestigators (1957a), augmentations were writ- presented at 350 wpm, before and after train-
ten for selected facts in a listening selection. ing. Training consisted of approximately 25
The recorded version of the augmented selec- hours of exposure to (a) speech at a constant
tion was then compressed enough by the rate of 350 wpm, (b) speech that was gradu-
sampling method to produce a playback time ally increased from a normal word rate to a
60 EMERSON FOULKE AND THOMAS G. STICHT

final word rate of 350 wpm, (c) the same as a degraded until a relatively large amount of
but with frequent pauses for questioning signal information has been discarded. The
about the material just heard, and (d) the finding that increasing the amount of com-
same as b but with frequent pauses for ques- pression has a different effect upon listening
tioning about the material just heard. There comprehension than upon word intelligibility
were no significant differences between pre- suggests that decreased intelligibility is not,
and posttraining test scores for any of the in itself, an adequate explanation for the loss
treatment groups. in comprehension that is observed at faster
Friedman et al. (1966) compared the com- word rates. One might expect decreased intel-
prehension test scores of subjects given 35 ligibility to interfere with comprehension to
hours of massed practice in listening to ac- some extent. However, the listener's uncer-
celerated speech with the comprehension test tainty regarding imminent speech is reduced
scores of subjects who received from 12 to 14 because of his ability to estimate the sequen-
hours of distributed practice in listening to tial dependencies in meaningfully connected
accelerated speech. They concluded that the words and syllables, and there is a further
comprehension demonstrated by the distrib- substantial reduction in uncertainty when he
uted-practice group was as good as, or better has heard enough of a message to form a
than, the comprehension demonstrated by the valid hypothesis about its contents. The re-
massed-practice group. duction in message uncertainty should signifi-
From the research reviewed in the present cantly counteract losses in word intelligibility,
paper, it is clear that an adequate training and the finding by French and Steinberg
experience for improving the comprehension (1947), that listeners can understand mes-
of accelerated speech has yet to be found. sages composed with words whose syllables
Simple exposure, at least in the amounts so are only 67% intelligible, suggests that this
far tested, is not adequate. is the case.
The increase in the rate at which compre-
CONCLUSION hension declines beyond 275 wpm suggests
It is possible to provide a fairly accurate that when a certain critical word rate is
description of the relationship between word reached, a factor in addition to signal degra-
rate and listening comprehension on the basis dation begins to determine the loss in com-
of the experimental results that have been prehension. The understanding of spoken
reviewed. There are two general classes of language implies the continuous registration,
results which, when taken together, suggest encoding, and storage of speech information,
that the relationship between word rate and and these operations require time. When the
listening comprehension is structured by more word rate is too high, words cannot be proc-
than one underlying process. First, there are essed as fast as they are received, with the
those studies in which listening comprehen- result that some speech information is lost.
sion has been measured at various word rates To put it another way, when channel capacity
(see Stimulus Variables). When these studies is exceeded, some of the input cannot be
are considered collectively, the relationship recovered at the ouput (Miller, 1953, 1956).
that emerges is one in which listening com- The explanation just suggested is, of course,
prehension declines at a slow rate as word tentative. A good deal of research on sentence,
rate is increased, until a rate of approximately word, and syllable rate, and upon the amount
275 wpm is reached, and at a faster rate of distribution of processing time in connected
thereafter. discourse, will be required in order to provide
In the second class of studies, intelligibility a more substantial basis for the hypothesis.
has been determined for words compressed by
various amounts (see Characteristics of the REFERENCES
Signal). These studies are in general agree-
BELLAMY, M. J. An experimental study to compare
ment regarding the finding that, when com- the comprehension of speeded speech by blind
pression is accomplished by the sampling and sighted children. Unpublished master's thesis,
method, word intelligibility is not seriously University of Texas, 1966.
 INTELLIGIBILITY AND COMPREHENSION OF ACCELERATED SPEECH 61

BOCCA, E., & CALEARO, C. Central hearing processes. compression. Symposium presented at the South-
In J. Jerger (Ed.), Modern developments in audi- eastern Psychological Association, Louisville, March
ology. New York: Academic Press, 1963. 1962.
CALEARO, C., & LAZZARONI, A. Speech intelligibility FOULKE, E. The comprehension of rapid speech by
in relation to the speed of the message. Laryngo- the blind—Part II. (Cooperative Research Project
scope, 1957, 67, 410-419. No. 1370) Washington, D. C.: United States De-
CAMPANELLA, S. J. Signal analysis of speech time- partment of Health, Education, and Welfare, Of-
compression techniques. In, Proceedings of the fice of Education, 1964.
Louisville conference on time compressed speech. FOULKE, E. The comprehension of rapid speech by
Louisville: University of Louisville, 1967. the blind—Part III. (Semi-annual Progress Report,
CARROLL, J. B. Problems of measuring speech rate. Cooperative Research Project No. 2430) Washing-
In, Proceedings of the Louisville conference on ton, D. C.: United States Department of Health,
time compressed speech. Louisville: University of Education, and Welfare, Office of Education, 1965.
Louisville, 1967. FOULKE, E. Comparison of comprehension of two
CRAMER, H. L. Intelligibility of compressed speech as forms of compressed speech. Exceptional Children,
a function of degree and direction of delay of 1966, 33, 169-173. (a)
presentation from one ear to the other. Paper pre- FOULKE, E. The retention of information presented
sented at the meeting of the American Psychologi- at an accelerated word rate. International Journal
cal Association, Chicago, September 196S. for the Education of the Blind, 1966, 14, 11-15.
DALE, E., & CHALL, J. S. A formula for predicting (b)
readability. Educational Research Bulletin, 1948, FOULKE, E. The comprehension of rapid speech by
27, 11-20, 28. the blind—Part III. (Interim Progress Report,
DEQuiROS, J. B. Accelerated speech audiometry, an Cooperative Research Project No. 2430) Washing-
examination of test results. Translations of the ton, D. C.: United States Department of Health,
Beltane Institute for Hearing Research, 1964, No. Education, and Welfare, Office of Education, 1967.
17. FOULKE, E., AMSTER, C. H., NOLAN, C. Y., & BIXLER,
DIEHL, C. F., WHITE, R. C., & BURK, K. Rate and R. H. The comprehension of rapid speech by the
communication. Speech Monographs, 1959, 26, 229- blind. Exceptional Children, 1962, 29, 134-141.
232. FOULKE, E., & STICHT, T. The intelligibility and com-
ENC, M. E., & STOLUROW, L. M. A comparison of prehension of accelerated speech. In, Proceedings of
the effects of two recording speeds on learning and the Louisville conference on time compressed
retention. The New Outlook for the Blind, I960, speech, Louisville: University of Louisville, 1967.
54, 39-48. FRENCH, N. R., & STEINBERG, J. C. Factors governing
FAIRBANKS, G., EVERITT, W. L., & JAEGER, R. P. the intelligibility of speech sounds. Journal of the
Method for time or frequency compression-expan- Acoustical Society of America, 1947, 19, 90-119.
sion of speech. Transactions of the Institute of FRIEDMAN, H. L., ORR, D. B., FREEDLE, R. 0., &
Radio Engineers Professional Group on Audio, NORRIS, C. M. Further research on speeded speech
1954, AU 2, 7-12. as an educational medium. (Progress Report No. 2,
FAIRBANKS, G., GUTTMAN, N., & MIRON, M. S. Audi- Grant No. 7-48-7670-267) Washington, D. C.:
tory comprehension in relation to listening rate United States Department of Health, Education,
and selective verbal redundancy. Journal of Speech and Welfare, Office of Education, 1966.
and Hearing Disorders, 1957, 22, 23-32. (a) GABOR, D. Theory of communication. Journal of the
FAIRBANKS, G., GUTTMAN, N., & MIRON, M. S. Audi- Institution of Electrical Engineers, 1946, 93, Part
tory comprehension of repeated high speed mes- III, 429-457.
sages. Journal of Speech and Hearing Disorders, GABOR, D. New possibilities in speech transmission.
1957, 22, 20-22. (b) Journal of the Institution of Electrical Engineers,
FAIRBANKS, G., GUTTMAN, N., & MIRON, M. S. Ef- 1947, 94, Part III, 369-387.
fects of time compression upon the comprehension
of connected speech. Journal of Speech and Hear- GARVEY, W. D. The intelligibility of abbreviated
ing Disorders, 1957, 22, 10-19. (c) speech patterns. Quarterly Journal of Speech, 1953,
FAIRBANKS, G., & KODMAN, F., JR. Word intelligibility 39, 296-306. (a)
as a function of time compression. Journal of the GARVEY, W. D. The intelligibility of speeded speech.
Acoustical Society of America, 1957, 29, 636-641. Journal of Experimental Psychology, 1953, 45,
FERGEN, G. K. Listening comprehension at controlled 102-108. (b)
rates for children in grades IV, V, VI. (Unpub- GOLDSTEIN, H. Reading and listening comprehension
lished doctoral dissertation, University of Missouri, at various controlled rates. Teachers College Con-
1954) Dissertation Abstracts, 1955, 15, 89. tributions to Education, 1940, No. 821.
FLESCH, R. A new readability yardstick. Journal of HARTLAGE, L. Differences in listening comprehension
Applied Psychology, 1948, 32, 231-233. between blind and sighted subjects. International
FLETCHER, H. Speech and hearing. New York: Van Journal for the Education of the Blind, 1963, 13,
Nostrand, 1929. 1-6.
FOULKE, E. A comparison of two methods of com- HARWOOD, K. A. Listenability and rate of presenta-
pressing speech. In R. H. Bixler (Chm.), Speech tion. Speech Monographs, 1955, 22, 57-59.
62 EMERSON FOULKE AND THOMAS G. ST1CHT

HENRY, W. G., JR. Recognition of time compressed NELSON, H. E. The effect of variations of rates on
speech as a function of word length and frequency the recall by radio listeners of straight newscasts.
of usage. Unpublished doctoral dissertation, Indiana Speech Monographs, 1948, 15, 173-180.
University, 1966. NICHOLS, P. G., & STEVENS, L. A. Are you listening?
HDTTON, C. L., JR. A psychophysical study of speech New York: McGraw Hill, 1957.
rate. (Doctoral dissertation, University of Illinois) NIXON, C. W., MABSON, W. E., TRIMBOLL, F.,
Ann Arbor, Mich.: University Microfilms, 1954, ENDICOTT, J. E., & WELCH, B. E. Observations on
No. 10-494. man in an oxygen-helium environment at 380 mm.
IVERSON, L. Time compression. International Journal Hg total pressure: IV. Communications. Aerospace
for the Education of the Blind, 1956, 5, 78-79. Medicine, 1968, 39, 1-9.
JESTER, R., & TRAVERS, R. M. Comprehension as a NIXON, C. W., & SOMMER, H. C. Subjective analysis
function of rate and modality of presentation. of speech in helium environments. Aerospace Medi-
Paper presented at the meeting of the American cine, 1968, 39, 139-144.
Psychological Association, Chicago, September 1965.
JOHNSON, W., DARLEY, F., & SPRIESTERBACH, D. C. ORR, D. B., & FRIEDMAN, H. L. Research on speeded
Diagnostic methods in speech pathology. New speech as an educational medium. (Progress Re-
port, Grant No. 7-48-7670-203) Washington, D. C.:
York: Harper & Row, 1963.
KLUMPP, R. G., & WEBSTER, J. C. Intelligibility of United States Department of Health, Education,
time-compressed speech. Journal of the Acoustical and Welfare, Office of Education, 1964.
Society of America, 1961, 31, 265-267. ORR, D. B., FRIEDMAN, H. L., & WILLIAMS, J. C. C.
KOZHEVNIKOV, V. A., & CHISTOVICH, L. A. Speech: Trainability of listening comprehension of speeded
Articulation and perception. (Trans. No. JPRS 30, discourse. Journal of Educational Psychology, 1965,
543, 1965) Washington, D. C.: United States De- 56, 148-156.
partment of Commerce, Joint Publications Re- RODGERS, J. R. A formula for predicting the com-
search Service, 1966. prehension level of material to be presented orally.
KURTZROCK, G. H. The effects of time and frequency Journal of Educational Research, 1962, 56, 218-
distortion upon word intelligibility. Speech Mono- 220.
graphs, 1957, 24, 94. SCOTT, R. J. Temporal effects in speech analysis and
LUTERMAN, D. M., WELSH, 0. L., & MELROSE, J. synthesis. Unpublished doctoral dissertation, Uni-
Responses of aged males to time-altered speech versity of Michigan, 1965.
stimuli. Journal of Speech and Hearing Research, STICHT, T. G., & GRAY, B. B. The intelligibility of
1966, 9, 226-230. time compressed words as a function of age and
McLAiN, J. A comparison of two methods of pro- hearing loss. Journal of Speech and Hearing Re-
ducing rapid speech. International Journal for the search, in press.
Education of the Blind, 1962, 12, 40-43. THORNDIKE, E. L., & LORGE, I. The teacher's word
MIIXER, E. C. Effects of learning of variations in
book of 30,000 words. New York: Columbia Uni-
oral presentation. Unpublished doctoral disserta-
versity Press, 1944.
tion, University of Denver, 1954.
MILLER, G. A. What is information measurement? VOOR, J. B., & MILLER, J. M. The effect of practice
American Psychologist, 1953, 8, 3-11. on the comprehension of worded speech. Speech
MILLER, G. A. The magical number seven, plus or Monographs, 1965, 32, 452-455.
minus two: Some limits on our capacity for proc- WOOD, C. D. Comprehension of compressed speech
essing information. Psychological Review, 1956, 63, by elementary school children. (Final Progress Re-
81-97. port, Grant No. 7-24-0210-263) Washington, D. C.:
MILLER, G. A., & LICKLIDER, J. C. R. The intelligi- United States Department of Health, Education,
bility of interrupted speech. Journal of the and Welfare, Office of Education, 1965.
Acoustical Society of America, 1950, 22, 167-173. WOODWORTH, R. S., & SCHLOSBERG, H. Experimental
MIRON, M. S., & BROWN, E. R. Stimulus parameters psychology. New York: Holt, 1954.
in speech compression. Journal of Communication,
1968, 18, 219-235. (Received August 8, 1968)