28.5 Compressed Speech

Because people were found to learn as effectively from tape recordings as from live speech, a variation on research on auditory comprehension emerged. This research was based on a simple observation: People can comprehend speech faster than narrators can speak. If a technology for speeding up the delivery rate of a tape recording could be developed, an apparent gain in instructional efficiency might be realized. This simple idea has generated a large body of research.

28.5.1 Processing Capacity

One way of measuring cognitive processing capacity is the comprehension of words per minute (wpm). Conversation typically takes place at a rate of 12 to 150 wpm (Benz, 1971; Nichols & Stevens, 1957). Since in conversation, one is simultaneously listening and composing speech, it was assumed that perhaps another 125 to 150 wpm of unused capacity might be available in simple listening. Because the rate for speed reading (Taylor, 1965) is 250 to 300 wpm, it was assumed that that much capacity might be available for listening. In fact Fairbanks, Guttman, and Miron (1957) found that rates of up to 300 wpm were possible with compressed speech.

28.5.2 Compression Technology

In early studies, compressed speech was produced by playing back a recording at a speed faster than that used to make the original recording. While this method is simple to produce, the vocal pitch and intelligibility were affected. The limitations of the "speed changing" technique generally rendered research findings questionable.

Miller and Lichlinder (1950) first demonstrated the tape-sampling method accomplished by deleting segments of the speech signal. A switching device was used that turned off the signal periodically. Garvey (1953) performed further experimentation in compressed speech by editing out segments of the audiotape and splicing the ends of the retained tape together. While Garvey's technique was successful, it was deemed too tedious except for research purposes. Fairbanks, Everitt, and Jaeger (1954) produced the first electromechanical apparatus that allowed both the expansion and compression of recorded tape.

Technological developments in the 1960s improved the Fairbanks technique. Scott (1965) utilized a computer to dispose of empty time intervals between words and to Sample the time interval occupied by words differentially. While the computer proved to be the best means of producing compressed speech, the cost of computer time negated the application of this approach for other than experimental purposes. Electronic developments of the last 2 decades have now allowed the mass marketing of completely electronic compressor/expander tape recorders.

28.5.3 Comprehension

Numerous researchers have varied the rate of compression and measured the resulting effect on comprehension. Fairbanks, Guttman, and Miron (1957) found little difference in comprehension of selections compressed to 141, 201, and 282 wpm. Diehl, White, and Burke (1959) determined that listening comprehension was unaffected by changes between 126 and 272 wpm. Foulke (1962) used both literary and technical presentations and found that listening comprehension was slightly higher in the 175- to 272-wpm range than the 272- to 375-wpm range, at which point an accelerated loss in listening comprehension occurred. Foulke and Sticht (1967) measured a 6% loss in comprehension between 225 and 325 wpm, and a loss of 14% between 325 and 425 wpm. These and other subsequent studies (Boyle, 1969; Carver, 1973; Foulke, 1968; Foulke & Sticht, 1969; Rossiter, 1970; Sticht, 1968; Wasserman & Tedford, 1973; Williams, Moore & Sewell, 1983-84) indicated that as word rate is increased beyond about 250 to 300 wpm, there is a decline in comprehension. Recall that potter's verbal short-term memory buffer has a capacity of about 1.5 seconds. At 300 wpm that is about 7-5 words. Since these words must be processed as concepts to be understood, and our conceptual short-term memory has a capacity of only five to seven items, a limit of about 300 WPM Seems reasonable. Indeed, Carver (1982) found that the Optimal rate (which sacrifices some comprehension) for both reading and listening is about 300 wpm, indicating that there is an innate bottleneck in human information processing beyond which improved technology cannot take us.

However, numerous intervening variables must be considered before a determination of the optimum degree of compression can be made (Duker, 1974). Researchers believe that the ability of subjects to comprehend compressed speech may be dependent on the difficulty of the material. Readability has sometimes been found to influence normal audio comprehension (May & Lumsdaine, 1958; Chall & Dial, 1948), but others (e.g., Molstad, 1955) were unable to replicate this finding. Foulke (1962) determined that the comprehension of a scientific selection was less than the comprehension of a literary work at normal speed. However, at various levels of compression, the comprehension scores of the scientific selection declined less than those for the literary selection. This phenomenon may be because the comprehension scores for the scientific selection were lower at the normal rate; therefore the range in which they could vary was relatively small (Duker, 1974). Fairbanks, Guttman, and Miron (1957) investigated the effect of time compression on messages of various difficulty levels. The study seemed to indicate that, within the range explored, listening comprehension did not depend on the difficulty of the listening material. Goldhaber (1967) and Reid (1968) found that comprehension decreased as wpm increased, and that simplified material was better comprehended than more difficult material. Carver (1982) presented data that relate reading difficulty with efficiency. His data showed that for college students, listening to compressed materials written at the eighth-grade level produced the greatest efficiency. Raising the difficulty level of the materials caused comprehension to drop off abruptly.

Comprehension in ay also be measured in terms of delayed recall. George (1970) studied various rates of compression and two levels of material difficulty. He determined that more was forgotten at the lowest levels of compression in a delayed measurement, I day and I week following treatment. George indicated that although simplified materials were accompanied by some initial forgetting, the amount of forgetting with the passage of time was less than with more difficult materials. Friedman, Freedle, Norris, and Orr (1966) conducted retention studies among college freshmen and sophomores at speeds ranging from 175 to 475 wpm. After a 30-day delay, the posttest was administered again. The retention for the two highest rates, 425 and 475 wpm, were 117% and 90%, respectively, of their first session scores. These authors reported that information presented in compressed format is retained, as well as information that has not been compressed. Foulke (1966) pointed out that forgetting is not limited to the recall of information presented at accelerated word rates. In other words, there is no indication that compression, within broad limits, has a unique effect on the retention of information.

Length of presentation may also be a factor in comprehension and memory. Adelson (1975) examined comprehension by a group of college students listening to a 1-hour lecture at 175 wpm, as compared to the same group of college students listening to an equated 1-hour lecture compressed at 275 wpm for 40 minutes. Compressed materials produced less comprehension than did the normal rate materials. The author concluded that with compressed materials, the length of presentation appeared to be a critical factor, perhaps because of attentional fatigue or other factors.

Narrators of both sexes differ in vocal pitch, average word rate, variation in word rate, pitch, and loudness (Foulke & Sticht, 1969). Foulke (1968) examined the extent to which these factors interacted with word rate in determining listening comprehension. Three versions of a selection were presented to college students at normal and compressed rates by narrators of both sexes. Significant differences in scores on a listening comprehension test were associated with the word rate variable, but narrator's style did not interact with rate. Rossiter (1972) had college undergraduates listen to short informative messages by both male and female presenters. Students were tested on the content of the messages. Data analysis showed an interaction of the sex of the speaker with the sex of the listener, but the author dismissed its importance, concluding that the sex of the speaker was not of much consequence in determining listening scores of subjects participating in the study.

Learner characteristics may influence the comprehension of compressed speech. Some of these variables include the subjects' sex, age, intelligence, and reading ability. Duker (1974) determined that the comprehension scores of male and female subjects revealed no sex-related differences for word rates varying from 174 to 475 wpm. This conclusion is supported by other research studies conducted by Foulke and Sticht (1967), Orr and Friedman (1964), Ross (1964), Bell (1969), Ludrick (1974), and Klavon (1975).

Fergen (1954) and Wood (1966) found that ability to comprehend compressed speech increases with age and grade level of school children. However, beyond age 12, little difference is noted until age 60 or so. Goldhaber (1970) found significantly better comprehension for junior high school students over college freshmen and sophomores. Goldhaber attributed the difference to the interest level and the level of motivation for each population. Duker (1974) agreed that the effects of age and education on the comprehension of compressed speech cannot be generalized. Lysaght (1969) tested elderly, middle-aged, and young adult subjects. He determined that elderly subjects performed lower on a posttest measuring comprehension. Duker (1974) noted that the decline in the ability to comprehend compressed speech may result from "... changes in the central nervous system" as opposed to the variable of age, per se (p. 494).

Aptitude or intelligence may also interact with comprehension. Eckhardt (1970) used a I -hour multimedia presentation at various rates of compression with Air Force recruits of varying aptitudes. Eckhardt concluded that test differences between the groups were due to aptitude and an aptitude-rate interaction. There was a comprehension loss for lower aptitude subjects at the higher compression levels. Sticht and Glasnap (1972) determined that low-aptitude men learned easier material better than more difficult material as a function of decreased wpm. High-aptitude men tended to learn material best at 175 wpm, independent of difficulty level. However, other researchers (Sticht, 1968; Watts, 197 1; Williams, Moore & Sewell, 1983-84) found that subjects with lower aptitudes or lower reading ability performed as well at higher rates of compression as at normal rates. Fergen (1954) found no relationship between the IQ of grade school children and their ability to comprehend compressed speech selections. On the other hand, Goldstein (1940) and Nelson (1948) found a positive relationship between intelligence and comprehension of compressed speech. Cicardo (1974) examined the retention of a compressed speech message presented to junior high school students of various intelligence levels. Cicardo determined that IQ level affects factual retention of material presented, but there appeared to be no interaction with rate. Foulke and Sticht (1969) pointed out that the relationship between lower intelligence and the decline in comprehension may be attributed to the lower scores of less intelligent subjects, which have a lower variance. For this reason, Foulke and Sticht (1969) and Duker (1974) argued that the -difference in comprehension cannot be attributed directly to intelligence level.

Reading ability may influence comprehension of compressed speech. Breed (1977) tested adult vocational technical school students to determine the differences in listening comprehension when subjects were categorized according to reading ability. The subjects in Breed's study listened to tapes that were time expanded and time compressed, varying in rate from 60 to 240 wpm. Breed indicated that listening comprehension and reading ability appear to be related to verbal skills. The poorest readers exhibited the poorest listening comprehension, and better readers were better listeners as measured by scores on tests of listening comprehension. Goldstein (1940) and Orr, Friedman, and Williams (1965) found a positive correlation between reading rate and ability to comprehend compressed speech. Conversely, both studies further determined that practice in listening to compressed speech resulted in an improved reading rate. Robertson (1977) determined that the comprehension of subjects is not affected when they are presented recorded materials within two reading levels below or three reading levels above their particular grade level. In general, it appears that a relationship between better reading ability and the comprehension of compressed speech can be established, although this may reflect an underlying verbal ability.

28.5.4 Training

It has often been speculated that practice might influence comprehension of compressed speech. This research has attempted to provide appropriate training prior to treatment in order to improve comprehension. Voor and Miller (1965) exposed subjects to five listening sessions at 380 wpm. Test scores indicated that comprehension increased as a function of exposure up to 7 minutes, and remained constant thereafter. Orr, Friedman, and Williams (1965) exposed blind subjects to listening material presented initially at 325 wpm and increased in 25 wpm intervals to a rate of 475 wpm. Subjects were tested for comprehension at 475 wpm and compared to equivalent pretraining test scores. An improvement of 29.3% was noted. Friedman, Orr, Freedle, and Norris (1966) compared the comprehension scores of subjects given 35 hours of massed practice with test scores of subjects given 14 to 21 hours of distributed practice in listening to compressed speech. The authors concluded that the comprehension of the distributed-practice group was as good or better than the comprehension demonstrated by the mass-practiced group. Duker (1974) suggested that gradually increasing the wpm rate might have some benefit on comprehension of compressed speech. Klavon (1975) tested this idea, without effect, in at attempt to provide a controlled transition period. In general, studies (Foulke, Amster, Bixler & Nolan, 1962; Friedman, Orr & Norris, 1966) have found that although no particular method of training or practice appears to be any more effective than another, even small exposure to compressed speech can improve comprehension.

28.5.5 Affective Factors

There are some interesting affective factors to consider when using compressed speech (see also 34.7). Listener attitudes toward the speaker are improved significantly (Maclachlan, 1982). Maclachlan notes that people associate fast, fluent speech with confidence, knowledge, and enthusiasm. Because attitude learning is influenced strongly by feelings toward the speaker, compressed speech may have an unexpected application in such situations. Additionally, college students prefer listening to compressed speech over normal tapes (Short, 1977), apparently because of the time savings. Also, for students under about 14 years old, there is a preference for listening over reading, presumably because of their slow reading rates (Boyle, 1969).

28.5.6 Efficiency

Recall that the original impetus, for speech compression was potential efficiency. The instructional implications of using compressed speech for efficiency are limited. When the time saved in compression was used (Fairbanks, Guttman & Miron, 1957) to elaborate certain parts of the text, comprehension for that part of the text increased. But as Sticht (197 1) pointed out~ the time saved in compression was lost in elaboration, and overall comprehension was not improved. Similarly, hearing the same text twice at double speed resulted in no more learning than hearing the same text once at normal speed (Schramm, 1972). Thus it appears that the instructional use of compression cannot be based on efficiency arguments.

28.5.7 Summary

In conclusion, a great body of research has been done on speech compression. It has confirmed the hypothesis that listeners can process at a much higher rate than normal conversational speech, with some loss of comprehension. Usually, any short exposure to compressed speech will result in improved comprehension. In general, no differences were detected for sex, or age between about 12 and 60. However, some differences in the ability to comprehend compressed speech may be due to aptitude or verbal ability. Compressed speech may be preferred to normal speech and may cause positive attributions to the speaker. However, the early hopes that it could lead to more efficient instruction appear to have been unjustified.