THE JOURNAL OF COMMUNICATION Vol. 18, September 1968, p. 198-206 Listening Comprehension as a Function of Word Rate EMIERSON FOULKE Abstract Twelve comparable groups of Ss heard a listening selection that differed, from group to group, with respect to word rate. Word rate was varied, in increments of 25 wpm (words per minute), from 125 to 400 wpm, by means of the sampling method for compressing or expanding recorded speech. After listening to the selection, Ss were tested for comprehension by a multiple choice test. Comprehension was not seriously affected by increasing word rate from 125 to 250 wpm, but it declined rapidly thereafter. The suggested explanation of these results is that time is required for the perception of words, and that as word rate is increased beyond a certain point, the perception time available to the listener becomes inadequate, and a rapid deterioration of listening comprehension commences. If word rate is determined for a large number of samples of the oral reading of professional readers, such as radio newscasters or those who read Talking Books, considerable variability will be observed. This variability is the consequence of differences in the nature of the material that is read, and to differences in personal reading style. However, the mean word rate will be approximately 175 wpm. [40] Recent technological developments [15, 23, 631 have made it possible to vary word rate of recorded oral reading over a wide range, either slower or faster than normal, without distortion in vocal pitch. This capability raises the possibility of presenting speech at other rates than the one at which it happens to be produced by an oral reader. On the practical side, recorded speech at a faster than normal rate can provide a needed increase in reading speed for blind people, Dr. Foulke is an associate professor of Psychology, at the University of Louisville. He received his doctoral degree at Washington University in St. Louis. Before joining the psychology faculty at the university of Louisville he worked as a staff psychologist at the Veterans Administration Neuropsychiatric Hospital in Knoxville, Iowa, and a research associate in the Educational Research Department at the American Printing House for the Blind in Louisville. His research interests center around non-visual perceptual systems and non-visual communication.
Foulke: Listening and Word Rate 199 and other people who read by listening. Recorded speech at slower than normal rates may prove to be a useful tool in promoting certain kinds of instruction, such as the learning of a foreign language. In a more theoretical vein, the ability to vary speech rate through a wide range, suggests new avenues for investigating the cognitive processes that underlie the perception of speech. There are several studies in which Comprehension has been measured as a function of word rate; but, in each of these studies, word rate has been varied through a relatively limited range. Therefore, in order to gain an impression of the influence of this variable, it has been necessary to combine the results of several studies. Within the range extending from 126 to 172 wpm, Diehl, White, and Burk [6] found listening comprehension to be unaffected by changes in word rate. In the range extending from 125 to 225 wpm, Nelson [52] and Hanvood [36] found a slight, but insignificant loss in comprehension as word rate was increased. Fairbanks, Guttman, and Miron [ 181 found little difference in the comprehension of listening selections presented at 141, 201, and 282 wpm. Thereafter, comprehension, as indicated by percent of test questions correctly answered, declined from 58% correct at 282 wpm to 26% at 470 wpm. Foulke, Amster, Nolan, and Bixler, [28] using both technical and literary listening selections, found comprehension to be only slightly affected by increasing word rate up to wpm. However, in the range extending from to 375 wpm, they found an accelerated decrease in comprehension as word rate was increased. Fodke and Sticht, [29] using the STEP Listening Test1 found a decrease in comprehension of 6% between 225 and 325 wpm, and a decrease of 14% between 325 and 425 wpm. The last three studies cited are in agreement regarding the finding that there is a change in the rate at which comprehension declines as word rate is increased. A similar relationship has also been found in many other studies in which the determination of the influence of word rate on listening comprehension was not the primary objective. [ 2-51 The purpose of the study reported in this paper is to display Sequential Tests of Educational Progress, Cooperative Test Division, Educational Testing Service, Princeton, New Jersey, 1957.
200 The Journul of Communication, Vol. 18, September 1968 the way in which listening comprehension varies as word rate is varied over a wide range. It is felt that a more certain knowledge of the relationship between these variables will be useful in making decisions about the rate at which to present recorded speech, in both practical and theoretical applications. Subjects METHOD Three-hundred-sixty sighted college students of both sexes, drawn from psychology and education classes at the University of Louisville, served as subjects. In a majority of instances, their service fulfilled a course requirement. Subjects were divided into twelve experimental groups, with 30 subjects per group. Experimental Materials and Apparatus A 2,925 word listening selection, appropriate in interest and difficulty for a college population, was chosen for use in the experiment. [9] A 50 item, four alternative, multiple choice test, with a split half reliability of 0.68, was written for this selection. The selection was read orally by a professional reader and recorded on magnetic tape by an Ampex Tape Recorder, Model, in the Talking Book Studios of the American Printing House for the Blind. This master tape was reproduced on a modified Tempo Regulator, [23] an electromechanical device for the compression or expansion of speech. [Fairbanks, et al., 151 The Tempo Regulator was adjusted for one of the word rates to be used in the experiment, and its output was recorded on magnetic tape by a Crown Tape Recorder, Model 800. Instructions for participating in the experiment were also recorded on this tape. Twelve tape-recorded versions of the listening selection were prepared in this manner, covering the range from 125 through 400 wpm in steps of 25 wpm. The tapes were reproduced on a Wdlensak Tape Recorder, Model T1500. The output of the tape recorder was distributed to the Western Electric Headsets, Type ANB-H-1, fitted with ear cushions, and each headset was provided with a volume control so that the signal level could be adjusted by the subject for comfortable listening.
Procedure Foulke: Listening and Word Rate 201 It was not possible to obtain the assistance of enough subjects at any one time SO that a complete experimental group could be tested at one setting. Therefore, subjects were tested in groups that ranged from 10 to 20 in number, and tests were conducted at a given word rate until the 31) subjects required for that condition of the experiment had been tested. Starting with the slowest word rate used in the experiment, the listening selection was presented to succeeding experimental groups in ascending order of word rate. The experiment was conducted in a large university classroom, with the poor acoustical properties typical of such rooms. However, since all subjects listened by means of headsets fitted with the kind of circumaural ear cushions that completely surround and enclose the external ear, the listening environment was felt to be satisfactory and similar for all subjects. First, test booklets and answer sheets were distributed. Next, subjects heard the recorded instructions for participating in the experiment. Then, the listening selection was presented. Upon its conclusion, subjects proceeded immediately to the test of listening comprehension, and upon its completion, each subject turned in his test materials and quietly left the room. Each Table 1 Means and Standard Deviations of Cosmprehension Test Scores as a Function of Word Rate WPM M SD 125 150 175 200 2.25 250 325 350 375 400 49.33 48.71 44.79 42.39 47.28 45.05 37.96 39.11 30.58 29.87 23.73 20.27 12.86 12.97 14.73 12.79 15.97 15.52 14.17 12.74 17.90 16.18 14.16 11.20
202 The Journul of Communication, Vol. 18, September 1968 W 50 V Lnm +- cz Lnw w040 +cy w ncl W +z u-30 w cin ELL OLn VLn LLLL OL x 20... z 10% z mg i 325 350 375 400 WORDS PER MINUTE Figure 1 Listening comprehension as a function of word rate. experimental session was concluded within the 50 minute class period. RESULTS A corrected test score was determined for each subject by applying to his raw score the formula CS = R - [W + (n - l)] where CS = corrected score, R = right answers, W = wrong answers, and n = the number of alternatives in the test item. [5] The means and standard deviations of corrected test scores, for each of the 12 experimental groups, are shown in Table 1. The relationship between word rate and mean test score, expressed as a percent of the maximum possible score, is displayed graphically in Figure 1. Word rate is scaled on the abscissa, and test score, in percentage units, on the ordinate. Though the curve in Figure 1 is somewhat irregular, the relationship suggested by it is one in which comprehension is relatively unaffected by changes in word rate in the range bounded by 125 and 250 wpm. Beyond this range, however, comprehension declines rapidly as word rate is increased.
Foulke: Listening and Word Rate 203 The test scores used in plotting Figure 1 were examined by an analysis of variance, the results of which are shown in Table 2. The variance in test scores associated with changes in word rate is significant beyond the 0.01 level as shown in row 1 of this table. Table 2 Analysis of Variance of Comprehension Test Scores Source of Variation df ss MS F Between 11 32,524.1534 2,956.7412 14.79% Linear (1).9232,9232 Within 348 69,559.5307 199.8837 Total 359 102,083.6841 * p < 0.01. The significance of the difference between ordered pairs of individual means was examined by means of the Newman-Keuls Test for Ordered Pairs of Means. [76, p. 801 The results of this analysis are shown in Table 3. This table is cast in matrix form, with the word rates at which tests were conducted arranged down the left hand margin and across the top of the matrix in order of increasing magnitude. Entered in each row, under the appropriate column headings, are the word rates for which comprehension scores were not significantly different from the comprehension score associated with the word rate in the left hand margin that identifies the row. The results presented in Table 3 are in general agreement with the impression conveyed by Figure 1. The pattern folrmed by the entries in this table also depict the nature of the relation between word rate and listening comprehension. However, although inspection of Figure 1 suggests that listening comprehension begins to decline rapidly beyond a rate of 250 words per minute, the results displayed in Table 3 indicate that losses in listening comprehension do not reach statistical significance until a word rate of wpm is passed. In evaluating the results of significance testing, one must keep in mind the fact that in view of the considerable variance of test scores as indicated by the standard deviations recorded in Table 1, relatively large differences among mean test scores would be required for statistical significance. The mean comprehension score of 20.27, obtained at 400 wpm, though quite low, was
204 The Journal of Communication, Vol. 18, September 1968 Table 3 Newman-Keuls Analysis of the Significance of Differences Among Group Means WPM 325 350 375 400 125 150 175 200 225 250 325 350 375 400 150 175 200 325 350 325 350 325 350 325 350 375 325 350 375 400 375 400 significantly different from zero, suggesting that there was some comprehension at this word rate. However, in order to be confident that this mean comprehension score had been determined primarily by the listening experience provided the subjects, it would have been necessary to administer the test of comprehension to another group that had not listened to the selection, and this was not done. The relationship between word rate and listening comprehension, suggested by Figure 1 and Table 3, is apparently not linear. The hypothesis of linearity was rejected by the test for linearity shown in Row 2 of Table 2. DISCUSSION The results of the present experiment are in close agreement with those of other experiments in which the relationship between word rate and listening comprehension has been studied. In previous investigations, [18, 281 increasing word rate had little effect on listening comprehension below approximately wpm. Increasing word rate beyond wpm resulted in a rapid decline in comprehension. In the present study, the rapid decline in comprehension set in beyond 250 wpm. From a practical point of view, this study, because of the large number of subjects employed, and because of the large number of word rates at
Foulke: Listening and Word Rate 2Q5 which comprehension was determined, provides a firmer basis for making recommendations regarding the accelerated word rates that might safely be considered in those situations in which speech, compressed in time by the sampling method, is to be used to promote faster aural communication. Of course, relevant experience might be expected to bring about some improvement in the ability to comprehend accelerated speech, and the subjects in this experiment had no such experience prior to the experiment. Voor and Miller, 1741 for instance, found a slight improvement in comprehension during initial practice trials. The results of other training experiences have been equivocal. Foulke [23] found no improvement due to training under any of four conditions of practice. Orr and his co-workers [57, 58, 601 have demonstrated significant improvement in the comprehension of speech presented at approximately 425 wpm. However, training experiences have not yet been devised that will result in good enough comprehension of very rapid speech (400 wpm) to permit its practical application in educational settings, and other situations in which people rely on listening. Until successful training methods are developed, the present findings should constitute a fairly accurate picture of the relationship between word rate and listening comprehension. The present findings also support a hypothesis suggested by Foulke and Sticht [29] regarding the perceptual problems that accelerated word rates create for the listener. According to this hypothesis, the loss in comprehension that attends an increase in the word rate of speech which has been accelerated by the sampling method, is due not only to a degradation in word intelligibility, but also to a reduction in the, perception time needed by the listener to process incoming speech information. Two kinds of evidence can be cited in support of this hypothesis. First, it has been shown [14, 32, 431 that word intelligibility remains at a high level well beyond the compression in time at which the comprehension of connected discourse has begun to decline rapidly. Secondly, the experiments cited earlier in this article in which listening comprehension was determined as a function of word rate, including the present experiment, suggest that listening comprehension is little affected by increasing word rate until a word rate in the neighborhood of 250 or wpm is
T reached, but substantially affected thereafter. It appears that word rate can be increased, to some extent, without depriving the listener of the perception time required to process speech input. However, beyond a certain point, the available perception time is no longer adequate, and comprehension begins to decline rapidly. ACKNOWLEDGMENT The author s research, reviewed in this paper, was performed at the University of Louisville, with financial support from the Office of Education under projects 1005, 1370, and 2430. * * * Mrs. Carol E. Oukrop, Administrative Assodate of the School of Journalism at the University of Iowa, Iowa City, Iowa has prepared an annotated bibliography of 80 items on Semantic Barrim to lntemtiorzal/cross Cultural Communication. Interested readers may obtain a copy by writing her.