An Auditory Cue-Depreciation Effect Author(s): Janet M. Gibson and Michael J. Watkins Source: The American Journal of Psychology, Vol. 104, No. 3 (Autumn, 1991), pp. 439-446 Published by: University of Illinois Press Stable URL: http://www.jstor.org/stable/1423249 Accessed: 15-06-2015 10:50 UTC REFERENCES Linked references are available on JSTOR for this article: http://www.jstor.org/stable/1423249?seq=1&cid=pdf-reference#references_tab_contents You may need to log in to JSTOR to access the linked references.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact
[email protected].
University of Illinois Press is collaborating with JSTOR to digitize, preserve and extend access to The American Journal of Psychology.
http://www.jstor.org
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
An auditorycue-depreciation effect JANET M. GIBSON Grinnell College MICHAEL J. WATKINS Rice University
An experiment is reported in which subjects first heard a list of words and then tried to identify these same wordsfrom degradedutterances.Paralleling previous findings in the visualmodality,the probabilityof identifyinga given utterance was reduced when the utterance was immediately preceded by other, more degraded, utterances of the same word. A second experiment replicated this "cue-depreciationeffect" and in addition found the effect to be weakened, if not eliminated, when the target word was not included in the initial list or when the test was delayed by two days.
This study concerns the identification of degraded stimuli. More particularly, it concerns the finding that a degraded version of a previously exposed word is less likely to be identified if it is preceded by even more degraded versions (Peynircioglu & Watkins, 1986). This "cuedepreciation effect" was observed when subjects were exposed to a list of words (such as RAINDROP) and subsequently asked to complete fragments of the words (e.g., R -I -- ROP):Performance was poorer if the fragments were gradually unfolded (e.g., R______P, R__ RP,R I __ R_ P, R_I __ROP) __ once (R _ I __ ROP).
than if they were shown all at
The cue-depreciation effect is not predicted by any theory of which we are aware, and a meaningful interpretation will, we think, require extensive exploration of its boundary conditions. Peynircioglu (1987) has made a start in this direction with her finding of a cue-depreciation effect when subjects know the semantic category of the to-be-identified words even though the words were not exposed beforehand. She has also found that the effect is influenced by various motivational factors (Peynircioglu, 1989). The present research continues this exploration by asking whether the effect extends to the auditory modality. The procedure parallels closely Peynircioglu and Watkins's (1986) procedure for visual materials. AMERICAN JOURNAL OF PSYCHOLOGY Fall 1991, Vol. 104, No. 3, pp. 439-426 ? 1991 by the Board of Trustees of the University of Illinois
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
440
GIBSON AND WATKINS
EXPERIMENT
1
In the first experiment, subjects heard a list of words and then tried to identify each of these words from a degraded utterance. In some cases, the degraded utterance was preceded by even more degraded utterances of the same word. At issue was whether the addition of these more degraded utterances would reduce identification of the word.
METHOD Subjects Subjects were 30 Rice University undergraduates.
Materials and design There were 35 stimulus words, 3 to illustrate the procedure and the other 32 for the experiment itself. They were selected at random from the pool of words listed in Peynircioglu and Watkins (1986, Appendix); some examples are aircraft, downfall,jubilant, mystique,obstacle,and yearbook.The words were spoken in a female voice and digitized with MacNifty Soundcap on an Apple Macintosh microcomputer. The digital representations were then degraded by dividing each representation into 100-ms segments and replacing all except a slice of each of these segments with silence. Exactly how much of each segment was left intact varied from word to word, according to the results of extensive pilot research. The aim was to create utterances degraded such that, in the absence of prior exposure, they would probably be identified by no more than 50% of the subjects. This calibration was made with the aid of a scale by which the duration in milliseconds of the sounded slice from each successive 100-ms segment (or, equivalently, the percentage of the word that was presented) was 1, 3, 6, 10, 15, 21, 28, 36, 45, 55, 66, or 78; note that each successive point on the scale differs from the immediately preceding point by a steadily increasing amount. The point on the scale that in the pilot research came closest to yielding a 50% identification rate without exceeding that rate was used to define the "critical" version of the test word. This version was used in the standard condition, in which no other version of the word was presented. It was also used as the last and most complete of the sequence of versions in the incremental condition. For each critical version, four additional versions corresponding to the four immediately preceding points on the scale were also prepared. For example, if the critical version of a given word included 36% of its utterance, the four noncritical versions included 10%, 15%, 21%, and 28%, respectively. Beyond the most fragmentary version, the particular slice of sound included in a given version subsumed the slice presented in each lesser version. The four noncritical versions were used for presentation immediately before the critical version in the incremental condition.
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
AUDITORY
CUE DEPRECIATION
441
The stimulus materials were put together on audiotapes. The study, or "exposure," list comprised all 32 words in their intact form, recorded at a rate of one every 3 s. The identification test involved the presentation of degraded versions of these same 32 words in a new random order. For 16 of the words, only the critical degraded version was presented; for the other 16 words, the critical version was preceded by the four noncritical versions. These test formats defined the standard and incremental conditions, respectively. The order of the 32 test words was the same for all subjects. For any given subject, the two conditions were represented in random order in the test sequence, although the use of two complementary versions of the test ensured that each word and each position was assigned equally often to the incremental and standard conditions. Procedure Some subjects were tested individually, others in small groups. They were told that the experiment had to do with the perceptibility of computerspoken words. The procedure was divided into two parts. In the first part, the subjects heard a list of words from a tape recorder. They were not told to remember the words. Rather, they were told to write the first letter of each word as it was being presented. Because the words were presented in their intact form, this task was easy. The second task, which followed immediately, was more difficult. It required identification of the same set of words, but this time the words were presented in degraded form. The subjects were not told that the test words were the same as those presented in the first part of the procedure, although they may well have noticed. For some of the words, there was only one degraded presentation (the standard condition), whereas for others there were five, each being a version that was a little less degraded than the preceding one (the incremental condition). For each word, testing began with a 500-ms warning beep. In the standard condition, the critical version of the test word was presented 4 s after the onset of this beep. In the incremental condition, the first of the five versions of the test word was presented 4 s after the onset of the beep, with the others following at 4-s onset-to-onset intervals. Subjects wrote their identifications on response sheets that had five dashes on each of 32 numbered rows. Each row corresponded to a word, and each space corresponded to one of the five degraded versions of the word. Whenever the subjects thought they had identified the test word, they wrote it in the first space in the case of the standard condition, and in the space corresponding to the version just presented in the case of the incremental condition. They were encouraged to guess when unsure. In the incremental condition, they were also told that they were free to revise their response as they worked across the page, although it was made clear that whenever they did so only their last response would be scored. At each beep, they were to move on to the next row. The subjects practiced the procedure with three words that were not used in the experiment proper. Performance was monitored throughout the experiment to ensure compliance with the instructions.
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
442
GIBSON AND WATKINS
RESULTS AND DISCUSSION For each subject, the words correctly identified were tallied for each condition. Where, in the incremental condition, more than one word had been written in a given row, only the last (i.e., the rightmost) was counted. (In only 3 of a possible 480 cases had a correct response been changed to an incorrect response.) The mean percentage of words correctly identified was 50.4 in the incremental condition and 56.9 in the standard condition. The extent to which performance in the incremental condition fell short of performance in the standard condition was reliable, t(29) = 2.55, p < .01. We may conclude, therefore, that the cue-depreciation effect extends to the auditory modality. EXPERIMENT 2 Peynircioglu and Watkins (1986) found that, in the visual modality, the cue-depreciation effect did not occur for words that had not been shown beforehand. One purpose of Experiment 2 was to see whether this finding would generalize to the auditory modality. We addressed this question by including only some of the test words in the preceding exposure list. A second purpose of Experiment 2 was to explore the effect of delaying the identification test. If, as in the visual modality, the cuedepreciation effect were confined to (or more pronounced for) previously exposed words, then it would be important to inquire how the effect of prior exposure holds up over time. The question has not, to our knowledge, been addressed be'fore, for in previous cuedepreciation studies (Peynircioglu, 1987, 1989; Peynircioglu & Watkins, 1986), the identification test has been presented more or less immediately after the initial exposure. In the present experiment, testing occurred almost immediately for some of the words and after a 48-hr delay for others. The two conditions of Experiment 1 were therefore expanded to eight: the two test conditions (standard and incremental) multiplied by the two exposure conditions (test words previously exposed and previously unexposed) multiplied by two test-delay conditions (immediate test and delayed test). All subjects were tested under all eight combinations of conditions. METHOD Subjects Subjects were 72 Rice University undergraduates.
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
AUDITORY CUE DEPRECIATION
443
Materials From the source used in Experiment 1, 67 words were selected at random. Their preparationfor the identificationtest was as described in Experiment 1. Three of the words were used to illustrate the procedure, and the other 64 were used in the experiment proper. Design The words were randomly assigned to four sets of 16: Sets W, X, Y, and Z. Sets W and X provided the test words of the immediate test, and Sets Y and Z provided those of the delayed test. The words of Sets W and Y were assembled at random into a 32-word list, as were the words of Sets X and Z. For any given subject, one of these lists was presented in the first or "exposure" phase of the experiment. In the identification tests, half of the
words in each of the four sets appeared in the standardcondition, and the other half in the incremental condition. Counterbalancingof the sets and subsets across subjects ensured that, overall, each word was assigned to the two exposure conditions (exposed and unexposed) and to the two test conditions (standardand incremental)equally often. Procedure The procedure followed closely that of Experiment 1, with two exceptions. First, the identification test included words that had not been exposed beforehand as well as wordsthat had. Second, the test was given in two sessions; one took place immediately (or, strictly speaking, after a 5-10-min break for instructionsand practice of the test procedure)and the other 48 hr later. RESULTS AND DISCUSSION The mean percentages of correct identifications for the various conditions are shown in Table 1. The first column shows the immediate test results for the two conditions that were also included in Experiment 1, namely those for the previously exposed words tested in the standard and incremental conditions. The cue-depreciation effect observed in Experiment 1 was replicated, in that performance was reliably lower in the incremental condition than in the standard condition, t(71) = 4.25, p < .01. The second column shows the results Table 1. Mean proportions of test words identified in Experiment 2 Immediate testing 48-hr delayed testing Test condition Exposed Unexposed Exposed Unexposed Standard .63 .37 .59 .46 Incremental .50 .32 .60 .47 .13 .05 -.01 -.01 Cue-depreciationeffect
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
444
GIBSON AND WATKINS
for words of the immediate test that were not previously exposed. The rate of identification of these words, collapsed across test condition, was substantially and reliably lower than that for words that had been exposed, t(71) = 12.19, p < .01, indicating that identification was primed by prior exposure. Of more relevance to the present concern was the effect of test condition on the identification of the unexposed words. In keeping with Peynircioglu and Watkins's (1986) findings for the visual modality, this turned out not to be reliable, t(71) = 1.56, p > .05, although the difference that occurred was in the direction of the cue-depreciation effect. The interaction of the prior exposure and test condition factors was reliable, t(71) = 2.23, p = .03, which was also in keeping with Peynircioglu and Watkins's findings. For the delayed test, as shown in Table 1, an appreciable beneficial effect of prior exposure survived the 48-hr delay. This priming effect was reliable both for the data as a whole, t(71) = 7.65, p < .01, and for the standard and incremental conditions considered separately, ts(71) = 4.34 and 4.65, respectively, p < .01 in both cases. Of more relevance to present purposes, the data from the delayed test provided no evidence of a cue-depreciation effect. Thus, for neither the exposed nor the unexposed words was the level of identification any lower in the incremental condition than in the standard condition. In fact, it was higher, albeit only trivially and not reliably so, t(71) = .10 in both cases. It seems, therefore, that at least for the conditions and materials of the present experiment, the cue-depreciation effect requires not only that the to-be-identified words be exposed beforehand, but also that the identification test not be delayed too long after their exposure. We should, perhaps, note that the overall proportion of words identified was somewhat higher in the delayed test than in the immediate test: 52.9% versus 45.7%. This trend held for both the previously exposed words (59.4% versus 56.7%) and the unexposed words (46.4% versus 34.6%); it was reliable for the unexposed words, t(71) = 5.83, p < .01, but not for the exposed words, t(71) = 1.57, p > .10. The interaction between prior exposure and delay was reliable, t(71) = 3.38, p < .01, reflecting a decline over the delay in the priming effect of prior exposure. The improvement in the identification of the unexposed words from the immediate to the delayed test may have been the result of practice. SUMMARY AND CONCLUSION The results of the two experiments reported here show that the likelihood of identifying a degraded version of a spoken word is
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
AUDITORY
CUE DEPRECIATION
445
reduced when it is immediately preceded by even more degraded versions of the word. We have therefore demonstrated that the cuedepreciation effect extends from the visual modality to the auditory modality. In keeping with previous findings in the visual modality, the auditory cue-depreciation effect was enhanced by, if not dependent upon, pretest exposure of the test words. In addition, no cue-depreciation effect was found when the test was delayed 48 hr. The cue-depreciation effect resembles earlier observations of an impairment in the identification of an out-of-focus picture that arises when its presentation is immediately preceded by even more blurred versions of the picture (Bruner & Potter, 1964; Galloway, 1948; Wyatt & Campbell, 1951). One apparent difference between this phenomenon and the cue-depreciation effect is that, whereas the impairment in picture identification occurs in the absence of a clear exposure of the picture prior to the identification test, the cue-depreciation effect is reduced if not entirely eliminated in the absence of a clear exposure of the test item prior to the test. Whether this difference is fundamental is a question that must await the results of additional research. Whatever the results of such future research may be, the explanation that Bruner and Potter (1964), Galloway (1948), and Wyatt and Campbell (1951) gave for their findings does not appear to generalize to the cue-depreciation effect. These researchers all subscribed to the theory that erroneous hypotheses were formed about the very blurred versions of the pictures, and that these hypotheses inhibited correct identification when the blur was reduced. This theory suffices for the results of Experiment 1, but it leaves unexplained the absence of a cue-depreciation effect under some of the conditions in Experiment 2. An ad hoc argument could be made along the lines that a strategy of forming hypotheses is contingent upon the recent presentation of the target items, in which case no effect would be expected in the delayed test. But such an argument does not explain why, in the immediate test, the effect occurred only for the test words that had been exposed beforehand. These test words were randomly intermixed with words that had not been previously exposed. Thus, if we assume that the subjects could not tell whether an unidentified test item had been exposed beforehand, then prior exposure could not have affected whatever tendency they may have had to form hypotheses. For the same reason, the design of Experiment 2 would seem to preclude any explanation of the cue-depreciation effect based on test strategy or response criterion. The same is true of Peynircioglu and Watkins's (1986) Experiments 4 and 5. We believe that a meaningful understanding of the cue-depreciation effect must await a fuller exploration of the conditions under which
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
446
GIBSON AND WATKINS
it arises. An obvious next step is a more detailed investigation of the effect of delaying the test after exposure of the test items. Does the effect dissipate within a few hours or even a few minutes? And which, if any, features of the initial exposure procedure determine the rate at which the effect dissipates? Does the rate of dissipation depend on the number of other words in the exposure list, or on the number of test words? Does the effect depend on the initial exposure being in the same sensory modality as the test items? Does the effect generalize to nonverbal materials, and how does it relate to the blurred picture findings of Bruner and Potter (1964) and others? Not before we have answers to questions such as these will we have a clear effect. understanding of the cue-depreciation Notes This research was supported by National Institute of Mental Health Grant MH35873 awarded to Michael J. Watkins. Correspondence concerning this article should be addressed to Janet M. Gibson, Department of Psychology, Grinnell College, Grinnell, IA 50112, or via Bitnet at GIBSONJ@GRIN1. Received for publication March 22, 1990; revision received January 23, 1991. References Bruner, J. S., & Potter, M. C. (1964). Interference in visual recognition. Science, 144, 424-425. Galloway, D. W. (1948). An experimental investigation of structural lag in perception. American Psychologist,1, 450 (Abstract). Peynircioglu, Z. F. (1987). Inhibition through incremental fragment cuing with primed items. Journal of Experimental Psychology:Learning, Memory, and Cognition, 13, 569-572. Peynircioglu, Z. F. (1989, November). The cue-depreciationeffect:Trying harder may disrupt cue utilization. Paper presented at the 30th annual meeting of the Psychonomics Society, Atlanta, GA. Peynircioglu, Z . F., & Watkins, M. J. (1986). Cue depreciation: When word fragment completion is undermined by prior exposure to lesser fragments. Journal of ExperimentalPsychology:Learning, Memory,and Cognition, 12, 426-431. Wyatt, D. F., & Campbell, D. T. (1951). On the liability of stereotype or hypothesis. Journal of Abnormal and Social Psychology,46, 496-500.
This content downloaded from 132.174.255.126 on Mon, 15 Jun 2015 10:50:59 UTC All use subject to JSTOR Terms and Conditions
