Interaural alternation, information load, and speech intelligibility Arthur Wingfield and John L. Wheale Department of Psychology,Brandeis University, Waltham, Massachusetts02154 (Received 7 November 1974; revised 10 February 1975) Speaksand Trooien [J. Acoust. Soc. Am. 56, 640-644 (1974)] report a failure to replicate selective intelligibility loss typically associatedwith rapid alternation of speech between the two ears. Results from several ,studies show the effect to be real, but that listeners must be under some load for the

loss to appear. This can be accomplishedby requiring "shadowing" as a responsemeasure, using linguistically complex messages,or increasinginput rate of the speechsignal. Subject Classification: 70.30, 70.35.

In a recentpaper, SpeaksandTrooienx report a failure to replicate previous findings of intelligibility loss associated with rapid alternation of speech between the two ears. This is typically reported to occur at alternation rates between four to five complete switching cycles per

second. •'-4 Becauseextendedpassagesof 100 or more words have typically been used, "shadowing" has served as the usual response measure. The subject must both listen and repeat what is heard, simultaneously. When subjective estimates of intelligibility or stenographic transcription were substituted for verbal shadowing,

ing is also important in this regard, but not necessary. Also shown in Fig. I are data taken from a larger study where shadowing was eliminated by inserting 4-sec silent periods between every ten words of 100-word passages. This gave the listener time to repeat what was heard before the onset of more speech. This function (open circles), using the same scientific prose, but without shadowing, shows only a minimal intelligibility loss in the region of the critical alternation rates. The final curve (open squares) shows the effects of

SpeaksandTrooienx report intelligibilityto remain high across all rates of alternation, with no suggestion of the usual V-shaped intelligibility function. Shadowing lowered overall intelligibility scores, but did not produce a differential loss at previously reported "critical" alternation rates. The speech materials used were from a popular magazine, recorded at 135 and 194 words per

ioo

9o

minute (wpm). Speaks and Trooien thus reject all three past explanations for the phenomenon: recognition time in rapid at-

8o

tentionswitching, •' contralateralinhibition,3 andsyllabic segmentation.4 The effect observedin earlier studies is attributed not to alternation, but to the shadowing itself, "interfering with the intelligibility of alternated

7o

speech..." (p. 644).x Althoughshadowing,andshadowing strategies,• undoubtedly influence intelligibility scores, there is a more likely explanation for Speaks and Trooien's failure to replicate. Alternation at the previously reported critical rates does affect intelligibility, but the fragmentation of speech produced by alternation can be overcome by the

listener's use of contextual redundancy to reconstruct the passage with good accuracy. For the effect to appear, the listener must be placed under sufficient load to minimize this activity. Figure I shows shadowing

•,

6o

z

50

4o

scores for 100-word passages of highly redundant light

fiction, recorded at 180 wpm (closed circles).

This

function is similar to that of Speaks and Trooien, both in its high intelligibility and in the absence of a selective dip at the critical alternation rates. A comparable group of subjects shadowing more complex scientific prose, however, shows the expected V-shaped function clearly (closed squares). The passages were taken from recent issues of the American Scientist, recorded at 185 wpm. Increasing complexity of speech materials is one way to induce overload on the listener. Shadow1219

J. Acoust. Soc. Am., Vol. 57, No. 5, May 1975

=

= SHADOWING LIGHT FICTION

•- ....

•> SCIENTIFIC PROSE, NO SHADOWING

=

-' SHADOWING SCIENTIFIC

o- ....



PROSE

COMPRESSED SPEECH, NO SHADOWING

3o

2

ALTERNATION

4

CYCLES

8

PER

16

SECOND

FIG. 1. Intelligibility of alternated speech for shadowing light fiction and scientific prose, and for scientific prose at normal

and fast speaking rates (time compression) without shadowing. Copyright¸ 1975 by the AcousticalSociety of America

1219

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 141.209.100.60 On: Tue, 23 Dec 2014 02:24:47

1220

Le{tersto the Editor

1220

increasing the task demands by time compression of the

cues to overcome the perceptual fragmentation

speechmaterials to 70%of normal playingtime, or

surely caused by alternation. It is of no special importance whether this load is imposed by shadowing of extended messages of low redundancy (as opposedto mono-

approximately 264 wpm. Even though silent periods were again employed to eliminate the need for shadowing, the effect is clear.

which is

syllabiesS),linguisticcomplexityof the speechmaterials,

In all cases, alternation was accomplished by passing recorded speech through a Grason-Stadler 829•E electronic switch pulsed externally by two Hunter 131C digital timers. The alternated speech was monitored over Sharp HA-10 earphones at average intensities of

or high input rate. be fully explained.

68-70 dB SPL re 0.0002 dynes/cm•'. Between14 and18

from

subjects took part in each of the experiments described. Each study was fully balanced for both order and passage effects, with all subjects hearing an equal number of passages at all rates. Order of presentation of rates and the passages heard at each rate within the studies were varied and balanced across subjects. Despite some variability in overall level of performance between subjects, the four curves in Fig. I are an accurate reflection of individual subject trends.

The time-compressed speech was prepared with the assistance of Dr. Emerson Foulke, Center for Rate Controlled Recordings, University of Louisville.

The detrimental effect of alternation on speech intelligibility does occur, and without shadowing as a response measure. Any explanation of the effect relating directly to shadowing itself, such as the competitive

feedback of Speaks and Trooien (presumably inducing special interference at critical rates), seems unacceptable.

The effect, however, is clearly a weaker one than

may usually be appreciated. listeners

As in the examples shown,

must be under sufficient load to deprive them

of time and/or processingcapacityto utilize contextual

The effect is real, and remains to

ACKNOWLEDGMENTS

This work was supported by PHS Grant NS-09767-01 the National

Institutes

of Health

to the first

author.

1C. SpeaksandT. T. Trooien, "InterauralAlternationand SpeechIntelligibility,"

J. Acoust. Soc. Am. 56, 640-644

(1974).

2E. C. CherryandW. K. Taylor, "SomeFurtherExperiments Upon the Recognition of Speech, With One and Two Ears," J. Acoust. Soc. Am. 26, 554-559

(1.954).

3E. D. SchubertandC. D. Parker, "Additionto Cherry and Taylor's Findings on Switching SpeechBetween the Two Ears, J. Acoust. Soc. Am. 27, 792-794 (1.955).

4A. W. F. Huggins,"Distortionof the TemporalPatternof Speech, Interruption and Alternation," 36, 1.055-1064 (1964).

J. Acoust. Soc. Am.

5N. T. Hopkinson,"Combined Effectsof InterruptionandInteraural Alternation on SpeechIntelligibility," Lang. Speech10, 234-243

(1967).

J. Acoust. Soc. Am., Vol. 57, No. 5, May 1975

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 141.209.100.60 On: Tue, 23 Dec 2014 02:24:47

Interaural alternation, information load, and speech intelligibility.

Interaural alternation, information load, and speech intelligibility Arthur Wingfield and John L. Wheale Department of Psychology,Brandeis University,...
216KB Sizes 0 Downloads 0 Views