NEUROPSYCHOLOGICAL ASSESSMENT LAB

PAUL JONES, ED.D.
UNIVERSITY OF NEVADA, LAS VEGAS

____________________________________________________________________

Internet/WWW Assessment of Simultaneous Processing

Assessment Lab Report 4-2
June/2000

A variety of techniques are available (Arons, 1992) which result in time-compressed speech, each with outcome of reducing the time required for a user to listen to a message. In simplest form, just the rate of speed is increased; in more complex applications, the sound is modified before presentation by techniques including rate change, eliminating silence, and adjusting the sampling ratio. The essence of most techniques is to produce a sound in which voice pitch is held constant while rate is increased. Prior studies (Heiman, Leo, Leighbody, & Bowler, 1986) have noted that, contingent on the extent of reduction, accurate perception of compressed-speech is possible for most users.

The accuracy of perception of compressed-speech is influenced by whether the task involves identifying isolated words or comprehending a longer passage. Early studies (Arons, 1992) found that connected speech remained comprehensible up to a 50 percent compression while isolated words could remain intelligible with a 10 to 15 percent compression (10 times normal speed).

Of most interest in this context is that from the early studies (Foulke & Sticht, 1969) in which compression was often obtained by cutting and splicing audio tapes, to more current research (Wingfield, Tun, Koh, & Rosen, 1999; Pallier, Sebastian-Galles, Dupoux, Christophe, & Mehler, 1998), individual differences in comprehension of compressed-speech have been evident. This investigator hypothesized that such differences could be associated with the neuropsychological function typically identified as the ability to perform simultaneous processing of information.

Gestalt closure (Kaufman & Kaufman, 1994) is a typical marker test for the simultaneous processing function. This task usually involves the presentation of an incomplete figure with instructions to identify it. The subject, in effect, is asked to perceptually "fill in" the missing parts. Both the gestalt closure and spatial reasoning scales most often used to assess simultaneous processing are problematic in applications with persons with visual disability because of the dependence on visual stimuli. This lab study was designed to provide prelimary investigation of an hypothesis that the underlying task in "filling in" the missing parts in a compressed-speech stimulus would also tap the underlying simultaneous processing function but with stimuli equally applicable for persons with and without visual disability.

Method

Participants

Details about participants and general procedure in the Series Four studies are available in a related report. As described in that report, suspect datasets were eliminated prior to analysis. This is reflected in the results below with some variance in the number of subjects associated with the various analyses. A total of 65 participants provided partial or total datasets.

Instrumentation

The instrument developed for this preliminary study was comprised of 24 test items from eight sound stimuli. Each stimulus was a simple declarative sentence which included an actor, an action, a setting, and a target. The person performing the action was a man, woman, boy, or girl. The action was a constant "to get". The setting was kitchen, garage, office, or storeroom. The target object was a package, a paper, a pencil, or a plate.

The subject listened to the stimulus sentence (e.g. The man was in the kitchen to get a plate), then after pressing a key on the keyboard, responded to three questions for each stimulus sentence, identifying the actor, the setting, and the target. The questions were in multiple-choice format. All responses were with the keyboard.

Test questions were preceded by sample questions to familiarize the subject with the response task and an example of a compressed-speech sound. Although exposure (Arons, 1992) does increase the accuracy of perception of synthetic speech, a lengthy practice period does not appear to be required (Voor & Miller, 1969). The subject heard an example of a standard sentence and an example of a compressed sentence before beginning the test. As will be described below, the extent of compression was then gradually increased with each new test stimulus.

Sentences were prepared using a speech synthesizer based on Microsoft® Agent. The time-warp similarity function in the Goldwave© sound editing program was used to compress the speech sounds. The similarity function in this sound editor uses correlation to add or overlap small, similar sections of the sound. This technique preserves the pitch while reducing (or increasing) the presentation time.

Table 1 below displays the compression rates and mean scores on the three-item sets for each of the eight sentences. With this participant sample, significant levels of compression were required before there was evident decrement in perceptual accuracy.

TABLE 1
COMPRESSION RATES AND MEAN SCORES ON THE THREE-ITEM SETS FOR EACH SENTENCE
n = 65

	compression rate	mean score (3 items)
sentence 1	1.25	2.9
sentence 2	1.50	2.9
sentence 3	1.75	2.9
sentence 4	2.00	2.7
sentence 5	2.50	2.8
sentence 6	3.00	2.4
sentence 7	3.50	1.9
sentence 8	4.00	1.6

____________________________________________________________________

Results

For the 65 participants in this study, the overall mean score on the 24-item CogListening scale was 20.1 with standard deviation of 2.75. The coefficient alpha reliability estimate was .71.

Table 2 below provides product moment correlation coefficients between the CogListening Scale and the other cognitive performance measures used in the Series Four studies. As hypothesized, the only statistically significant relationship was with the gestalt closure anchor test.

TABLE 2
CORRELATION COEFFICIENTS BETWEEN COGLISTENING AND OTHER COGNITIVE PERFORMANCE MEASURES

	CogAtt (n=64)	CogMem (n=63)	CogPlan (n=63)	Closure (n=60)	ProbSolv (n=60)
CogListen	.24	-.05	.19	.26*	-.01

* significant at .05 level

CogAtt: CogAttention Raw Score (maximum is 10)
CogMem: CogMemory Raw Score (maximum is 55)
CogPlan: CogPlanning Condition Contrast Efficiency Score
Closure: Gestalt Closure Anchor Test Raw Score
ProbSolv: Problem Solving Anchor Test Raw Score

____________________________________________________________________

Table 3 provides exploratory factor analysis data regarding the underlying neuropsychological structure of the CogListening scale using scales with known features. As anticipated from prior studies, CogMemory appears to load on a distinct factor. The highest factor loadings for CogListening and Gestalt Closure were on the same factor. Caution is warranted in intepreting these results because of the overall small n in this study, exacerbated by the loss of some datasets. With this caution, these results provide provide tentative support for the hypothesized identification of CogListening as a measure of simultaneous processing.

TABLE 3
FACTOR LOADINGS:
Varimax Rotation
(n = 56)

	Factor 1	Factor 2
CogAtt	.55	.46
CogMem	-.11	.84*
CogListen	.70*	-.26
Closure	.74*	.01
ProbSolv	.40	.44
percent of variance	30%	24%

* factor loading >= .70

____________________________________________________________________

Summary and Discussion

Although additional study is needed and consideration of some instrument modification is warranted, these data appear consistent with an hypothesis that compressed-speech could provide a viable alternative to visual stimuli for assessment of the simultaneous processing function. Correlation coefficients and exploratory factor analysis data were in the anticipated direction. There were no apparent significant difficulties in delivering this content using the Internet/WWW as the test administration modality.

With this participant sample, the floor of the test instrument was unnecessarily low. Increasing the ceiling does not appear viable, but initiating the higher levels of compression at an earlier point in the test could enhance the discrimination effectiveness of the instrument. In future studies there will also be consideration of additional anchor tests as well as more diverse participant samples.

References

Arons, B. (1992). Techniques, Perception, and Applications of Time-Compressed Speech. Proceedings of the American Voice I/O Society Conference, Sep. 1992, pp. 169-177.

Foulke, W., and Sticht, T.G. (1969). Review of research on the intelligibility and comprehension of accelerated speech. Psychological Bulletin, 72, 50-62.

Heiman, G. W., Leo, R. J., Leighbody, G. and  Bowler, K. (1986). Word intelligibility decrements and the comprehension of time-compressed speech. Perception and Psychophysics, 40, 407-411.

Kaufman, A.S., & Kaufman, N.L. (1994). Manual: Kaufman Short Neuropsychological Assessment Procedure . Circle Pines, MN: American Guidance Service, Inc.

Pallier, C., Sebastian-Galles, N., Dupoux, E., Christophe, A., and Mehler, J. (1998). Perceptual adjustment to time-compressed speech: A cross-linguistic study. Memory & Cognition, 26, 844-852.

Voor, J.B., and Miller, J.M. (1965). The effect of practice upon the comprehension of time-compressed speech. Speech Monographs, 32, 452-455.

Wingfield, A., Tun, P.A., Koh, C.K., and Rosen, M.J. (1999). Regaining lost time: Adult aging and the effect of time restoration on recall of time-compressed speech. Psychology and Aging, 14, 380.

____________________________________________________________________

Guide to Reports

____________________________________________________________________