THE EFFECT OF ADDED PAUSE TME UPON THE COMPREHENSEBEUTY OF COMPRESSED SPEECH Thesis for the Degree of Ph. D._ MICHIGAN STATE UNIVERSITY FRANK ROCCO ' 1969 gnaw: ‘m-w-rw a... LIBRAP. y g" . Michigan 3: University This is to certify that the thesis entitled THE EFFECT OF ADDED PAUSE TIME UPON THE COLIPREHENSIBILITY OF COMPRESSED SPEECH presented by Frank Rocco has been accepted towards fulfillment of the requirements for M degree in M11011 Maior h‘rnfpecnr Date. z_23._69.____- - 0-169 3“! a!!! r« 3;- . ‘. ~ va‘fi=mp .". 0" ,— ABSTRACT THE EFFECT OF ADDED PAUSE TIME UPON THE COMPREHENSIBILITY OF COMPRESSED SPEECH By Frank. Rocco. At present, the medium.used most by blind children as a means of information gathering is braille. Recently the talking book record and tape recorded materials have come into use as a supplementary system to braille. The blind high school student reads braille at a rate of approx- imately 90 words per minute. Record or tape recorded mat- erials, when recorded by professional readers, are presented at an average rate of approximately 175 me. In contrast, a sighted high school reader has an average silent reading rate of 251 me. In addition, the sighted reader has the capability of visual scanning and overviewing, as well as the ability to quickly and easily reverse course in order to review or retrace the sequence of events. The printed page by virtue of its format permits the sighted reader 'visually based organizational cues by use of varied spacing, ‘bold type and center and side headings. It is obvious that the blind child is at a significant educational disadvantage 'because of the slow rate at which.he must listen or read. One method of more favorably equating this disparity .in.rate of information.acquisition.is that of presenting Frank Rocco recorded material at a word rate in excess of 175 me. It is possible to raise the me.rate of recorded material amp ploying speech compression technology to any level desired without alteration in pitch of the original presentation. However, comprehensibility of this compressed speech begins to decline at approximately 275 me. Although this rate (275 wpm) is slightly higher than the average silent read- ing rate quoted earlier for sighted readers, the oral method of presentation does not compensate for the visual advan- tages listed for sighted readers. It is advantageous there- fore to attempt to raise the me.rate of compressed speech to speeds beyond 275 me without loss in comprehension so that the blind listener might be permitted two "readings" in the time originally required for one, or one "reading" for previewing content and a second or third for mastery. An experiment was conducted to determine if it is the lack of information processing time in compressed pas- sages of connected discourse that causes loss in comprehen- sion at word rates greater than 275 me. It was assumed that S requires time to make the associations which make full comprehension possible and that this time is greatly reduced by the process of compression. A 1,319 word fictional story was tape recorded and compressed, using a sampling method with discard intervals of 20 milliseconds, to word rates of 250, 300, and 350 wpm. .. 9. . .sm (e‘vs'i'T'g—v .3!" 'Jrq‘e—vr” | I I l I Pause 30 fj g; *1. on rates with obta: Were Drevi Frank Rocco Pause times of l, 2, and 3 seconds duration were added at 30 fixed intervals (common to all tapes) within the story. This resulted in four tapes at each of the three compression rates previously listed, one with no pause time added, one with 1 second pauses.added, etc. In all, 12 tapes were obtained in this manner. Three hundred and forty-eight college students were randomly assigned to the 12 treatment conditions and asked to complete a 38 item.mulitple—choice examination in order that comprehension scores might be obtained for each treat- ment. The test results were submitted to an analysis of variance to determine if pause addition was beneficial to comprehension. It was hypothesized that (a) the main effect for pause time will be statistically significant, greater comprehension being associated with the presence and the magnitude of pause time and, (b) the interaction effect of pause time and compression rate will be statistically sig- nificant, greater comprehension being associated with longer pause.times at higher rates of compression. Results of the analysis of variance showed a signif- icant main effect for compression. Comprehension scores of the 250 and 350 wpm rates of compression were found to be significantly different. The 300 and 350 me rate scores were also significantly different. This was in keeping with previous literature on the subject. No significant difference UBGX‘ any I Pas STOP; llll'lll'nl'llal'll Il| ‘I I‘ll Pres: +v flu at .t Wm. u f d l % ma nu m Ma .t II'I lll'llllll'llllvlll'llllli Frank;Rocco was found for either the main effect of pause addition of the interaction term (compression x pause addition). It was concluded therefore, that under the conditions of the present study, there appears to be no beneficial effect for compression attributable to the addition of pause times of the frequency and duration used. An investigation was undertaken to determine if this unexpected negative finding could in part be attributed to any of the following variables: complexity of the stimulus passage, test ceiling, validity of the measure of comprehen— sion, reliability of the test, appropriateness of the comp pression rates selected for testing, S's motivation, place- ment of pauses within the test selection, and sex make-up of the treatment groups. While these variables may have had an effect under different conditions there was no inter- nal evidence indicating that these variables could be held responsible for the negative findings of this research. TEE EFFECT OF ADDED PAUSE TIME UPON IHE COMPREIENSIBILITY OF COMPRESSED SPEECH By Frank Rocco A THESIS Submitted to Michigan State University in partial fulfillment of the requirements. for the degree of DOCTOR OF PHILOSOPHY Department of Elementary and Special Education 1969 DEDICATION This dissertation is affectionately dedicated to my father Antonio Rocco and to the memory of my dear mother Anna Rosa Rocco who together embarked upon the great journey to America in order that their children might have a better life. 11 —-—-— .— —. .—— —__-— .- — 1‘- n w effc side Sell ~38 howe free Free 581 I Servi r989a Feadi for 8' and fc ACKNOWLEDGEMENTS The research reported herein represents a cooperative effort on the part of many individuals who contributed con- siderable time and energy toward its completion. Special appreciation is extended to Dr. James.E. Keller who as Guidance Committee Chairman advised and directed the study from.conception to completion. More than this, however, Dr. Keller provided a model of humanity that re- freshed this observer and committed him anew to his task. Grateful appreciation is also extended to Dr. Donald Freeman who provided subjects for the study as well as counp sel concerning design and statistical analysis. Gratitude is due the many individuals who provided direct assistance in the study. Among those who gave unsel- fishly of their time special mention is given the following: mra. Becky Briggs for scheduling and rescheduling of the language laboratory testing site; Mrs. Linda Nutting for serving as girl Friday throughout the testing phase of the research; mrs. Kathy Dickerman for the many hours of proof- reading the capy; Dr. Andrew Porter and Dr. Rosanne Bulgarella for statistical counsel; Dr. Richard DeMott for design counsel and for serving as test prOctor; Steven Miller for serving 111 ... , .JJXT pj— — as test proctor and general assistant; Carl Oldsen for pro- viding resource information for the review of literature; Dr. Emerson Foulke for technical counsel and materials dev- elopment; Joseph Levine and John Shultz for technical counsel regarding pause addition to tapes; Mrsi Brenda Pecor for typing the test booklet and thesis proposal; George Churchhill for counsel and assistance regarding test booklet duplication; Dennis Allen, Mike Maurer, Harold Van Morgan and Mike Brooks for serving as laboratory technicians. To Sam.Beavers, Matt Klein, Ross Dailey, Miss Barbara Nicholas, Mrs. Lani Ford, John Hoogstra and Mrs. Terry Wood grateful appreciation is expressed for their invaluable assis- tance in "encouraging" subjects to participate in the study. Appreciation is also extended to the more than 400 Education 200 students who volunteered to be subjects for this study. Without their cooperation there would have been no study. Finally, I wish to express my heartfelt thanks to a very wise and beautiful woman, my wife Joyce. She unselfish- 1y served in every aspect of the research as pretest subject, scorer, editor and typist. Her constant assistance and gentle encouragement made completion of this research a reality. Her concern for the quality of life of the people about her has made my tenure as,a doctoral candidate thorough- ly enjoyable. iv ‘r'w E1, TABLE OF CONTENTS Page DEDICATION.....................ii Acmowxmncmms..................111 .LISTOFTABLES...................vi Chapter I. INTRODUCTION AND REVIEW OF THE LITERATURE . . 1 Statement of the Problem Review of the Literature II. METHOIDLOGY o o o o o o o o o o o o o o o o 0 21 Hypotheses, Design Subjects materials Procedure III. FINDINGS AND DISCUSSION 0 o o o o o o o o o 0 39 Findings Discussion IV. SUMMARY AND CONCLUSIONS . . . . . . . . . . . 59 Summary Conclusions LIST OF REFERENCES . . . . . . . . . . . . . . . . . 66 APPENDICES A. Characteristics of Subjects by Treatments 72 B. Directions for Testing, Test Story and Pause Placement Points . . . . . . . . . . 95 0- 0 meat mmnt C C I O O O O O O O O C C C O 105 LIST OF TABLES Table Page 1. Designoi’theStudy............. 23 2. Sex, Mean Age and Grade of Participating Subjects by Treatment . . . . . . . . . . . 26 3. Time Required to Present Audio Testing “tori-315 O O O O O O O O O O O O O O O O O 31 4. Comprehension Test Reliability Data by Treatment Conditions . . . . . . . . . . 33 5. ’ Sample Distribution by Testing Dates . . . . 36 6. Means and Standard Deviations for Comprehen- sion Scores by Treatment Conditions . . . . 40 7. Analysis of Variance of Comprehension Test scores 0 O O O O O O O O O O C O O O O 41 8. Tukey's Method for Multiple Comparisons Pair Wise Contrasts Between Row Means . . . 43 9. Cmnprehension Test Score Frequency Distribution 0 o o o o o o e o o o o o e o 45 Vi "I. h? '1 ['3 " I“ b: 61 fc is CHAPTER I INTRODUCTION AND REVIEW OF THE LITERATURE Statement 93 the Problem The blind child has long been at a disadvantage when compared with his normally sighted counterpart with , respect to the acquisition speed of printed material. Harris (1947) reports the median silent visual reading rate for sighted high school students at 251 words per minute (m). For blind high school students the mean braille reading rate is 90 wpm (Bixler, Foulke, Amster, Nolan, 1961). Recently two methods have come under exper- imentation that show some promise of enabling the blind to cover material at rates comparable to print readers. One has been the introduction of telegraphic braille materials by Martin and Alonso (1967). In this 'method the traditional or non-telegraphic presentation of braille material is reduced to its kernel or significant elements by eliminating extraneous content (unnecessary for comprehension) and restructuring the remaining mater- 181 in the manner commonly found in telegrams. This method 18 yet experimental and requires that each passage be ' l analysed . rental 3‘ speech. and hes 3 other rec eed speec who condu recorded 1118 arr" the speec. found the- Percent 01 Ge time 58p could be 6 Bibility c He °°nstm nail? disc the r9221: 88d 335680} 51' equipment 1954 ) that analysed and rewritten to be effective. The second method, also still only in the experi- mental stage, is that of time compression of recorded speech. This is a listening rather than reading method and has particular applicability to talking books and other recordings for the blind.“ Interest in time compres- sed speech began with a study by Miller and Licklider (1950) who conducted research on the comprehensibility of pre- recorded words (not connected discourse). A manual switch- ing arrangement made the systematic off-on interruption of the speech signal of each word possible. The investigators found that word intelligibility did not decrease until 50 percent of the speech signal was discarded. Garvey (1953) surmised from this research that the time gap caused by the interruption of the speech signal could be eliminated thereby providing the same comprehen- sibility of signal in less time without distortion in pitch. Re constructed such a tape (of connected discourse) by man- ually discarding portions of the speech signal and splicing the remaining tape together. This resulted in time-compres- sed speech. Since Garvey's pioneering research electromechanical equipment has been deve10ped (Fairbanks, Everitt, & Jaeger, 1954) that compresses tapes to any desired number of words per minun a listening by Foulke, after, an a: Williams ucetion 12F tioo prooeH Tr. ”5 18 app (1962) has 81Krillfioeh sed speech “I 'Ould ; men the . nificant d; Va“3898 oi View do D01 bran-1° and Prefient tat: 0r °°mPreag Poaaible th undiag- 0 The per minute. The use of such a process with the blind to shorten listening time for recorded materials was first recognized by Foulke, Amster, Nolan, & Bixler (1962). Shortly there- after, an effort (Orr a Friedman, 1967, 1968; Orr, Friedmsn a: Williams, 1965) was made to extend the application to ed- ucation in general and to the broad area of human infome- tion processing. The normal oral reading rate of professional read- ers is approximately 175 m. The work of Foulke, et a1. , (1962) has established that in the range from 275 to 325 wpm significant loss in comprehension of recorded time-compres- sed speech occurs. Listening rates of approximately 250 wpm would permit comparable rates of information intake be- tween the sighted reader and the blind listener. Yet, sig- nificant disadvantages still present themselves. The ad- vantages of previewing, scanning, overviewing or rapid re- view do not lend themselves well to touch perception in braille andare not possible with recorded materials using present technology. However, if more rapid presentations of cospressed speech without loss of comprehension were possible the blind listener might be permitted a second “reading” or time for note taking, etc., in the same space of time that one "reading" originally required. The problem being explored in this research is whe ther of con-1pm ished 83;: ingt‘he: l" '1 d Stri: the 4 whether the decline in comprehensibility for faster rates. of compressed speech is in part attributable to the dimin- ished amount of time available to the listener for process— ing the information he is receiving. According to Foulke and Stricht (1967b, p. 19) The perception of speech entails the registration, encoding and storage of speech information, and these Operations require time. When the word rate is too high, words cannot be processed as fast as they are received with the result that some of the words and their associated meanings are lost. It may be useful to conceive of at least two fac- tors in the comprehension of connected discourse: the in- telligibility of individual words, and the meaningfulness of groups of words; that is, of phrases, sentences, and larger units in a passage. The serious loss in comprehen- sion at compression rates exceeding approximately 275 wnm .may be due, not so much to the loss in intelligibility of individual words, but to the failure of the listener to .relate the larger semantic units of the material. It is perhaps of relevance in this connection that :in the electromechanical devices used for compressing speech, 'the normal pause times between phrases and sentences are :ilso reduced. In addition, such "semantic processing" time is further reduced by the time eliminated by the speeding up of the words. It may be that by the reintroduction of some of the lost natural "pause" time in the form of “‘ 3 W36.- ~ .1-“— erbitrar; higher 0’ presumab‘ the gain '- portance of res-dig. bLind C21: En integr of the lo Speech is Tfilms resu. Course, 1 (1.6., am the lOSt ¢ ‘6 he requ: the £681; them the 68tab: Pause ‘ \ aQLlI 5 arbitrary pause units, comprehension can be improved at the higher compression rates. The insertion of pause times of presumably effective length would undo only a small part of the gain achieved by time compression. The resolution of this problem is of practical im- portance to work with the blind as the disparity in rate of reading between braille and print readers places the blind child at an educational disadvantage particularly in an integrated classroom setting. If in fact, a major portion of the loss of comprehension reported in studies of rapid speech is attributable to the limiting of natural pause time resulting from compression of recorded connected dis- course, it may be possible to artificially replace pauses (i.e., add time) at strategic intervals thereby reclaiming the lost comprehension without significantly increasing the time required for compressed presentation. Many areas of inquiry based on this assumption sug- gest themselves; however, the most basic of these deals with the establishment of a cause and effect relationship between ;panse addition and the comprehension of rapid speech. feeter (C: Bergen, l,- lees than' in vocal : relativeh p. 3.) All sFacial en; the Playbe. Speed, '50 d‘dced wnic Speed. In 33% Compre: 6 Review of the Literature1 Spgech Compression Methods An obvious way to produce rapid speech is to speak faster (Calearo & Lazzaroni, 1957; Enc & Stoluran, 1960; Fergen, 1955; Harwood, 1955; Nelson, 1948). This method is less than satisfactory since it "introduces undesired changes in vocal inflection and fluctuations in rate, and (has) a relatively low upper limit ..." (Foulke & Stricht, 1967b, P0 30) Another method which requires little in the way of special equipment is the "speed changing" procedure in which the playback speed is greater than the original recording speed. However, a rise in the pitch (frequency) is intro- duced which is directly prOportional to the increase in speed. Intelligibility falls off rapidly; for example, at 33% compressionz obtained by speed changing, there is a 40% 1For an exhaustive review of the literature the reader is directed to Foulke's (1967) invaluable volume on which the writer relied heavily in the formulation of this review. 2Studies of compressed speech usually report either a compression ratio or a word per minute (me) rate. Comp pression ratio refers to the savings in time of the’original recording; thus, a 30% ratio would indicate that 30% of the original recording time was saved. Because the ratio is dependent not only on compression procedures, but also on original wpm rate, Foulke (1967b, p. 6) advocates reporting the final wpm rate rather than a ratio. Miron and Brown (1968) believe that syllables per minute would be an.even Inore accurate indication of rate. loss of 1 the rate the subje E‘oulke, J I advocated from dist sampling 1 moving t between w moved and Within a that impo °°mputer 7 loss of intelligibility (Foulke, 1968, p. 6 ). Changing the rate of speech by the speed changing method has been the subject of several pieces of research (Barabasz, 1968; Foulke, 1966a; Garvey, 1953; McLain, 1962). The sampling technique is probably the most widely advocated method for producing rapid speech as it is free from distortion found in the "speed changing method.” The sampling method reduces the original recording time by re- moving tiny (20-100 milliseconds) bits of words and/or pauses. between words. If a computer is used, both the interval re- moved and the frequency of these removals can be varied within a single selection, over a wide range, thus insuring that important components are not sampled out. However, the computer technique is enormously expensive, involving the use of both a digital and an analog computer (Scott, 1965). Feasibility of the sampling method was determined by Miller and Licklider (1950) when they demonstrated speech redundancy by removing parts of the speech signal. Their experiment illustrated that monosyllabic words did not drop below 90% of intelligibility until 50% of the speech signal was removed. Recognizing the significance of this finding, Garvey (1953) removed the semnts of interrupted speech signal, a.._ 1“— m" spliced compress pensive use of r as the ‘0- heads spliced the tape together again and the result was time- compressed intelligible speech. An electromechanical device, the Tempo-Regulator manufactured by Telefonbau und Normalzeit, Frankfurt-am- Main, Germany, is commonly used in the sampling method. Although the equipment is quite expensive, it is less ex- pensive than the process using computers. The device makes use of recording heads mounted on a cylinder, which rotates as the tape passes over it. Consequently, the recording heads make contact with the tape only at preselected inter- vals and for preselected durations. Once selected, the sampling frequency and intervals remain constant, resulting in random deletions. Because the sampling technique does not change the frequency of the playback speed, there is no distortion in pitch. Detailed explanation of a similar device is found in Fairbanks, Everitt, and Jaeger (1954). One other method, according to Foulke (1968), is that of "harmonic compression." The speech signal is fed through a filter, which reduces it to half its original frequency. The signal can then be played back at twice its original rate, without a change in pitch and without any loss of original signal content. The device is appar- ently under construction, as no studies have been reported 1181113 the technique . —- Intellg' pressed discrete tests be widely L1: in which phrases b O ically or: port two s Which redu tine requi identific: 9 Intelligibility of Time-Compressed Speech Intelligibility, one method of evaluating time-com- pressed speech, may be defined as the recognition of short, discrete passages (usually isolated words). A number of tests have been devised to test intelligibility, the most widely used is the "articulation test" (Foulke, 1967, p. 7) in which the listener is asked to write down the words or phrases he hears as they are presented. Other intelligibility measures are more psycholog- ically oriented. Foulke and Stricht, (1967b, p. 7-8) re- port two such measures: one, disjunctive reaction time in which reduced intelligibility resulted in increased reaction time required on the part of the listener for making correct identification (Foulke, 1965); and two, threshold intensity in which the relationship between compression and the min- imum threshold intensity needed for comprehension is ex- Plored (Calearo d: Iazzaroni, 1957). More specifically, the investigators established a minimum intensity required for Words to be intelligible when recorded at various levels 0f compression. Changes in threshold intensity, " that in- tensity at which some percent of a list of words (e.g., 50%) are correctly identified," (Foulke 8c Stricht, 1967b, p. 7) would indicate an alteration of intelligibility. A number of studies (Garvey, 1953; Kurtzrock, 1957) have demonstrated that there is little intelligibility loss even at 3 been 1188‘ 7 no correi I coaprehe: the obvic level W01 I and comp: eteLy int ation of no signif tion intr Study wni by speedi £0qu no ; material . erial Dre; a°°°mPlisl I‘ecordmg VQueue U m rate. % Th 10 even at high word rates (when the sampling technique has been used). Unfortunately, there appears to be little or no correlation between intelligibility of single words and comprehension of compressed connected discourse other than the obvious statement that some minimum intelligibility level would be required for any comprehension to occur. To explore the relation between word intelligibility and comprehension of larger units, Foulke (1966b) deliber- ately introduced degrees of distortion (by using a combin- ation of speed changing and sampling techniques), and found no significant relationships between the amount of distor- tion introduced and the actual comprehension. In another study which.explored the distortion of word intelligibility by speeding the rate of playback presentation, Foulke (1966a) found no significant differences in comprehension between material presented by sampling technique and the same mat- erial presented by the speed changing technique (which is accomplished by changing the rate of playback speed of a recording). In both studies, it was found that the only variable which exerted influence on comprehension was the wpm.rate. Comprehension'gg Time-Compressed Speech The.measurement of comprehension is usually accomP plished by a short-answer or multiple-choice objective test. Sf tests rm: and lute: revious sion" so: reading ‘ doubtful of the o: of the 5's Many rese (Foulke, 11 test. Selection of listening passages and construction of tests must consider a number of factors such as: reading and interest levels; contamination (where variations in previous knowledge account for differences in "comprehen- sion" scores); validity and reliability. Most standardized reading tests are unquestionably contaminated or are of doubtful reliability when used as tests of listening. One of the often used published tests is the listening sub-test of the Sequential Tests of Educational Progress (STEP). Many researchers, however, have develOped their own tests (Foulke, 1964; Foulke, et al., 1962; McLain, 1962). figfig,-Perhaps the most widely quoted study in regard to rate and comprehension is Foulke's (1966b) rate study using 360 college students. A selection was presented at 12 rates, in increments of 25 wpm from 125 to 400 me. The results indicate that the relationship between rate and comprehension is not a linear one; specifically, comprehen- sion was reasonably good until 250 me, where it began to fall off rapidly. At 275 wpm the comprehension loss was significant. These findings are in accord with other studies (Bixler, et 81., 1961; Fairbanks, Guttman, & Miran, 1957a) which demonstrate good comprehension up to about 275-300 wpm, at which.point comprehensibility begins to decline rapidly. The me.rate at which there is no loss in comle variables of mater: Listener pression Stricht , Wood (196 SOhOOl g: cock and 12 in comprehension varies from S to S aa a function of variables relating to the listener and to the type of material used (i.e., literary or scientific). Listener Variables _S_e_gc_.--No sex differences in comprehension for com- pression rates up to 475 m have been found (Foulke & Stricht, 1967b; Orr, et 81., 1965). $59 and Educational Experience.-Fergen (1954) and Wood (1965) have found a positive relationship between school grade and comprehension of compressed speech. Wood- cock and Clark (1968) also found a relationship between lis- tener age and comprehension in elementary school children. Their data suggest it may not be advantageous to introduce accelerated speech in elementary grades. Foulke and Stricht (1967b, p. 15) note that while high school and college students have been used in compre- hension studies, the relationships between age and other variables and comprehension have not been explored. They further suggest that studies of intelligibility (Calearo 8: Lazzaroni, 1957) with aged Ss lead one to suspect that comprehension is affected by changes. in the central ner- vous system which accompany old age. Intelligence.-Foulke and Stricht (1967b, p. 15-16) state that there is not sufficient evidence in regard to "m. children ' relations et a1. , 1 positive comprehen S ship betw Forces Qua coupresse between :- appear to ”‘1 8Peach 1; far lhare 1 listening 8ishted, aLeo be 5-8 “"6 to 1 (Poms O: dame that Sign. 50' how in 1‘64 (39118 , . 301ml rate I l3 children's 1.0. and listening comprehension to suggest a relationship. However, studies with adults (Fairbanks, et al., 1957a, 1957b; Nelson, 1948) have demonstrated a positive correlation between intelligence and ability to comprehend time-compressed speech. Stricht (1968) found a highly significant relation- ship between mental aptitude (as measured by the Armed Forces Qualification Test) and the ability to comprehend compressed speech although he did not find an interaction between rate and intelligence. That is, aptitude did not appear to play a greater role in comprehension of compres- sed speech than it did in comprehension of normal speech. Visual Status.-Because the blind use their hearing far'more than do the sighted, one might expect that their listening comprehension would be superior to that of the sighted. In addition to this‘“use hypothesis," it might also be assumed that lack of visual distractions should serve to increase listening attention and comprehension (Foulke & Stricht, 1967b, p. 16). Foulke (1964) found evi- dence that blind Se were superior in listening comprehen- sion. However, conflicting evidence has been reported, both in regard to comprehension of compressed speech (Bellamy, 1966) and in comprehension of speech presented at normal rates (Hartlage, 1963). Bellamr, in a study l4 utilizing blind and sighted Ss who were exposed to a passage of compressed material on which they were tested using a multiple-choice examination, found no difference in the comprehension abilities of the blind and sighted. Hartlage's study was similar to Bellamy's except that he used only uncompressed material. "It was concluded that with the age, sex and intelligence variables controlled, no differences in listening comprehension existed between the blind and sighted groups." (Hartlage, 1963 p. 4) Present Reading M.—Orr, et al., (1965) tested the hypothesis that fast readers would be able to compre- hend time-compressed speech better than slow readers and found a significant positive correlation between reading rate and comprehension of compressed speech. Both experi- ments also determined that training in use of compressed speech improved reading rate. On the other hand, Nelson (1948) found no correlation between reading rate and comp prehension of accelerated speech. "Coggitive.§3113"‘gg'Ability Patterns.-—Two studies have attempted to discover individual differences in comp prehension of compressed speech as a function of 88' char- acteristic cognitive approaches or abilities. Friedman and Orr (in Foulke, 1967, p. 73-74) noted that better listeners tended to hear phrases or sentences rather than individual fie. w . ,.. 0." 15 words. A battery of aptitude and information tests showed a surprising inverse relationship between specific memory skills and listening ability, "suggesting that excessive attention to detail is probably antithetical to good lis- tening in these conditions" (p. 74). Friedman and Johnson (1968) ran a factor analysis on a battery of 10 reading and listening tests to discover which (if any) could be used as predictors of success in comprehending compressed speech. The @353 W _I:J_a_m_e_ M, which required Ss to infer a class relationship in a set of objects, showed increasing correlation with comprehension as wpm increased. The authors concluded that this test defined an ability important to comprehension of compressed speech which is the ability to make "rapid comparison of alternative responses to find one which is most similar to a stimulus" (p. 213). This ability could account for dif- ferences between individuals in comprehending compressed speech. Stimulus Variables The first class of stimulus variables has to do with the technical aspect of producing the auditory signal. This class would include the method of compression, frequency and intervals of sampling, and final me.rate-- all of which have been discussed in previous sections. It would also 16 include such considerations as signal/noise ratio. Miron and Brown (1968) write of the high level of background noise in comparison with the signal strength in the Fair- banks device. This is technically referred to as the S/N (signal/noise) ratio. It should be noted that factors such as this are seldom.tctally accounted for in studies on comp pressed speech, and.many studies (especially the early ones) used recordings which were poor in quality. A second set of variables deals with the message itself--its content and readability. In regard to read- ability, most studies at least consider the relationship to comprehension with phrases like "the material was judged suitable for the grade levels tested." Many studies use either the Flesch (1948) or the Dale-Chall (1948) formulas for estimating the difficulty of reading passages. However, Miran and Brown (1968) observed that readability levels can be measured throughout any message and can be seen to vary. Hence, a total readability index may not reflect the actual readability of individual portions of a selection. Another speculation would be that readability levels of test pas- sages might best be kept well below Ss' ceiling readhmg level in order to control for variance attributable to complexity of the material. Bixler, et a1., (1961) clearly demonstrated that the content of a selection affects comprehension, both at ova—v. z-w.flma F. 17 normal and at accelerated rates. The study compared comp prehension of a literary selection and a scientific selec- tion, presented at several wpm rates. Comprehension of literary material was found to be higher throughout, and there was a significant interaction with respect to compre- hension between complexity of material and compression rate. Improvigg Listenigg Comprehension Through Training From the preceding discussion, it will be noted that comprehension declines sharply with only moderate accel- eration in word rate after attaining about 275 me, and that informational or explanatory type reading seems to be more seriously affected than literary material. Several studies have attempted to reduce this comprehension loss through training of listening skills. Foulke (1964) tested several training procedures for increasing comprehension of time- compressed speech. He compared training using a gradual increase in rate with training using a constant high rate; he also tested whether frequent question—and-answer periods could enhance comprehension. No evidence was found to sup- port the superiority of one training method over the other or to support the superiority in comprehension of trained Ss over naive Ss. In fact, no evidence was found for a "practice effect" commonly noticed among naive listeners upon their first exposure to compressed speech. There are, 18 however, serious limitations to Foulke's study. He used very small groups (7-11 members). In addition, he noted that the selection on which Ss were trained may have been too short to accomplish its purpose, and he acknowledged that motivation had not been considered. In a recent study Friedman and Orr (in Foulke, 1967, p. 69-75) report results that appear Optimistic with regard to training. These researchers used college students as paid Se, and further motivated them through bonuses for in- creased comprehension scores. This procedure resulted in significant gains for the experimental group. One group obtained almost 80% comprehension (based on ccmprehens ion at normal rate) of material presented at 425 m. Results of an unpublished study by Nolan and Morris (1968) show that the deleterious effects of compression on comprehens ion may appear at wpm rates as low as 225 wpm when control Ss at 175 wpm are highly motivated. Blind Ss were rewarded with a box of candy for doing well. The findings of the study indicate that high 3 motivation produces better comprehension for 175 and 225 wpm rates, having a particular- 1y beneficial effect upon the comprehension of moonpressed material (175 wpm). One may conclude then, that training for improving the comprehensibility of rapid speech is feasible. The studies just reviewed suggest that simple exposure to accel- erated speech is not sufficient for improvement however, and 19 that.motivaticn:must be considered an important aspect of the training program. Retention Foulke (1966c) apparently believed that retention of the material in a passage of compressed speech.may vary as a function of compression rate. Perhaps comprehension as measured by a test immediately following presentation of the material may be a better indicator of retention over long time intervals for some compression rates than for others. To examine the effect compression had on the reten- tion of rapid speech Foulke (1966c) compared retention among four experimental wpm.rates (175, 225, 275, and 325 wpm) at intervals of 0 days (immediate recall), 7 days and 30 days. Foulke found that lower retention over time was a function of lower immediate recall (comprehension) and concluded that retention of accelerated material is.not a special case. That is, retention is dependent upon original comprehension and therefore only indirectly effected over time by compres- sion rate. Other studies add support to this conclusion. Bars- basz (1968) studied retention among college students and found.nc interaction between word rates and retention. These studies support the conclusion of McGoech and Irion (1952) that retention is generally a function of initial 20 learning and as such can be fostered by improving the learn- ing aspects of S's first encounter with the subject matter. sin-Lari The sampling method appears best for the comprehen- sion of pre-recorded speech as no distortion in pitch of original recording results in playback at faster wpm rates. COanrehension of compressed speech begins to decline sharply at approximately 275 Win. This decline is due in part to both listener and stimulus related variables. Lis- tener related variables which correlate with comprehension are: age and educational experience, reading ability, "cog- nitive listening style“ and intelligence of the listener. Sex and visual status of the listener are not shown to be clearly related to comprehension. Stimulus related variables to be considered in comprehension are: rate of presentation (compression), method of compression, readability level of the passage used in compression and type of material used (i.e., scientific or literary style). Improvement of comprehension through training of Ss' listening skills appears feasible. Retention of com- pressed material would be helped by such training as good retention over time appears to be a function of good initial comprehens ion. ej'V'M .p' _ ‘1- .- ‘o ‘1 CHAPTER II METHODOLOGY W The basic aim of this research was to determine whether the introduction of pauses in compressed connected discourse would reduce the loss in comprehension that usually occurs at high compression rates. It was hypothesized that this would occur. It also seemed reasonable to assume that long pauses, as opposed to short ones, would be more help- ful to Ss exposed to high compression rates.than to those exposed to lower compression rates. Pause times of O, l, 2, and 3 seconds were used at compression rates of 250, 300, and 350 me. Thus, it was assumed that at the slower rate of 250 me a 1 second pause would facilitate comprehension as effectively as a 3 second pause, whereas at 350 wpm a 3 second pause would be more helpful to a lis- tener than a 1 second pause. An analysis of variance (ANOVA) design was utilized. In.terms of this design the following hypotheses were tested. 21 .fi yr” Mr. 1 22 Hypothesis'I.-The main effect for pause time will be statistically significant, greater comprehension being associated with the presence and the magnitude of pause time. Hypothesis.;I.-The interaction effect of pause time and compression rate will be statistically significant, greater comprehension being associated with longer pause times at higher rates of compression. Common sense would suggest, perhaps, that the greater the pause time, within reasonable limits, the bet- ter the comprehension. ‘However, exploratory pre-testing suggested that the length of the pauses would not be a crucial variable; that comprehension would vary only accord- ing to the presence or absence of some pause time. In pre- testing, the 3 second interval seemed to be perceived by SELas.a_long delay, particularly at high compression. rates. It seemed plausible to speculate that relatively long pauses might encourage irrelevant ideaticn which would represent a distraction from.the task at hand. The ANOVA design permitted exploration of whether or not these assump- tions may be tenable. Desigg A twoaway fixed effect ANOVA design utilizing comp pression rates of 250, 300, and 350 wpm and pause time 23 additions of O, 1, 2, and 3 seconds were used to test the hypotheses. Ss were randomly assigned to the 12 treatments represented in Table 1. TABLE 1 DESIGN OF THE STUDY Compression Pause Times Rate 0 sec. 1 sec. 2 sec. 3 sec. 250 wpm N829 29 . 29 29 300 wpm 29 29' 29 29 350 wpm 29 29 29 29 Total N=3¢8 The compression rates were chosen to span those rates where, according to past research, comprehension drape off most rapidly as a function of increased rate. Selection of these rates would serve to maximize the pos- sibility of observing an interaction effect. The selection of pause time intervals of l, 2, and 3 seconds duration was somewhat arbitrary as no evidence could be found in the literature as to optimal pause length for listening behavior. Some attempt at reducing the ar- bitrary nature of this decision was made, however. 24 Preliminary exploratory testing showed that pauses greater than 3 seconds in duration, when added to recorded material compressed at a rate of 350 wpm were perceived as being too long a delay in material presentation. Pause intervals of less than 1 second duration, on the other hand, were diffi- cult to perceive at the slowest rate used (250 wpm). Subjects Three hundred and fortybeight students enrolled in Education 200, The Individual and the School, spring quar- ter, 1969, at Michigan State University were used as Ss. Ss were drawn from.14 discussion sections of the course and were given credit toward their final grade for partic- ipation. Participation was somewhat voluntary in that Ss, could choose from.a number of alternatives (i.e., one page book or article critique, 10 minute class presentation,