JOURNAL

OF COMMUNICATION

DISORDERS

8 (1975) 4>50

PHONETIC INTERFERENCE UNDER VARYING CONDITIONS OF PHONETIC FACILITATION HARRIS Professor

WINITZ*

of Speech Science, 5220 Rockhill Road, University of Missouri-Kansas

Missouri

City, Kansas City,

64110

EVE HERRIMAN Research

Associate,

Department

of Speech Science, City, Missouri

Universiry of Missouri-Kansas 64110

City, Kansas

Introduction

When a child is taught to correct an error sound essentially two processes are involved. A new sound (the standard sound) is acquired either as an isolated element or in a syllable. After the sound is well learned, children are instructed to use their new sound in words. This phase is a difficult point in, the training program, since the incorrect sound competes with the standard sound causing difficulty in “carry-over.” Recently, it has been observed that recall of speech sounds to pictorial stimuli is increased when the substituted sound and the original sound are phonetically similar (Winitz and Bellerose, 1972a), suggesting that the original sound (the error sound) may act as a facilitator rather than as a competitor. Apparently, facilitation takes place because the error sound reinstates all but a few of the distinctive features of the standard sound. We know that newly acquired sounds are difficult to maintain over time (e.g., Shelton, Johnson and Amdt, 1972). It is very possible, then, that facilitation produces interference. Interference is defined as competition between the error sound and the standard sound (Winitz and Bellerose, 1972a,b). The resulting competition is, however, nonmotor, as previous research indicates that over long intervals of time retrieval is difficult, although there is little loss in imitative ability (Winitz and Bellerose, 1972a,b). An example will be helpful here. A young child, who initially substituted 101 for is/, has been taught to produce a correct Is/, imitatively, in response to pictures, and when reading and speaking. The latter three test situations, reading, describing pictures and talking, will be referred to as retrieval functions. This child has no difficulty imitating the /s/ sound correctly at the beginning of each clinical session, but generally has difficulty with /s/ in tests of retrieval in the clinic and outside of the clinic. Our interpretation (Winitz and Bellerose, 1972a) is that /O/ facilitates the recall of Is/, but, since /O/ and /s/ are phonetically similar, the higher habit strength of * Address reprint requests to Harris Winiti. D American

Elsevier Publishing

Company,

Inc.,

1975

43

44

HARRIS

WINITZ

and EVE HERRIMAN

/13/ relative to /s/ will cause competition at the time of retrieval, resulting in production errors. This investigation is an attempt to provide circumstances that can demonstrate the conflicting forces of facilitation and interference (Underwood and Postman, 1960; Turnage and McCullough, 1968). The phonetic similarity of the sounds that are to be replaced will be varied in contrast to the approach used by Winitz and Bellerose (1972a) in which the phonetic similarity of the newly acquired sounds was varied. It has been found that young children frequently substitute I$/ when asked to learn /c/ (Winitz and Bellerose, 1962; Locke, 1969); /s/ is, therefore, a strong competitor of /c/ and it is for this reason that these two sounds were selected for study. The sounds I$/ OR /cl were taught as substitutes for /s/ in soup or /ml in moon. The paradigm may be depicted as follows: IJ up/ Group

I~unl Group

learn llupl for /sup/

learn IJunl for lmunl

lq.~pl Group learn lqupl

Icunl Group

for lsupl

learn /Fun/ for Imunl

Three predictions regarding recall were made: (1) I$/in soup > /$/ in moon. Recall of ljlin the word soup should exceed recall of I$/in moon because the phonetic similarity between I$/and Is/is greater than between ISIand /ml. This prediction follows from the Winitz and Bellerose (1972a) investigation. (2) ISIin soup > /cl in soup. It is assumed that Is/ facilitates equally the recall of IFI and ISI. Recall of I$/ insoup should exceed recall of /cl insoup because/J/ and lsl are strong competitors of IFI, whereas the competition between /IIand Is/ is minimal. The task at the time of retrieval is similar to a matching exercise where answers are provided. Although competition is minimal it still persists when responses are available. It is entirely possible, of course, that Is/does not act as an equal facilitator for Is/ and /E/. Previous research seems to discount this possibility, since recall was unrelated to phonological permissibility for instances in which competition was minimal (Winitz and Bellerose, 1972a). No prediction was made for 1~1in soup and moon. It is difficult to know in advance the relative contributions of facilitation and interference. (3) IJI in moon > /E/ in moon. Recall of I$/m . moon should exceed recall of /c/ in moon. Facilitation in both cases would be reduced relative to soup. However, /cl should evoke greater competition than Is/.

PHONETIC

INTERFERENCE

45

Procedure Reliability Two reliability sessions were conducted. In the first session five children were involved. One examiner reinforced the subjects out of view of the second examiner. Tape recorded stimuli were presented to the subjects. Using a reliability index of agreements/agreements + disagreements, the average percentage of agreements was 89.5 (63-100). After a thorough discussion of the errors a second reliability session was conducted in which an average of 84% (70-90%) was obtained for five children. The error analysis suggested that for /c/ the following were frequent errors: /!I, hjl, /j/ and /h_$, of which lh$l caused the most difficulty in the reliability judgments. It was decided, therefore, that responses that contained an unvoiced /j/ with minimal frication, as would occur in conversational German, would be scored as correct. Although errors in perception on the part of an examiner undoubtedly occurred during the course of the experiment, it should be pointed out that the examiner was unaware of the purpose or intent of this experiment. Learning

and Retention

Tasks

Each subject was to learn one new syllable, which was a substitute response for the standard pronunciation of soup or moon. Children were independently assigned to four groups (N = 18 male and 18 female in each group). The groups are designated by their substitution responses as follows: /cup/, IJupl, Icunl and l~unl. The /cl sound is a fairly difficult pronunciation for young children. Therefore, an elaborate shaping procedure was used. All correct responses were reinforced by placing a peg in a pegboard. The interval between the sound stimuli, as described below, was 6 sec. The first two phases of training were the same for all groups. As will be observed below all children initially learned [cl. From this group of children subjects were assigned to experimental groups. This procedure assured that the groups were potentially equivalent in their abilities to acquire the [c] sound. The first two training phases were as follows: (a) five imitation trials of prerecorded /ju/ and (b) 10 imitation trials of prerecorded /cu/. Only those subjects who correctly articulated /c/ 6 out of 10 times were included. A total of 72 students was disqualified at this point. The subjects were randomly assigned to groups at this time. The subjects in the /cup/ and /cun/ groups continued their training as follows: (c) 10 imi-

46

HARRIS

WINITZ

and EVE HERRIMAN

tation trials of prerecorded /cl, (d) 10 imitations of prerecorded /cup/ or /Gun/, (e) 3 imitations of /cup/ or /Gun/ following the live utterance of the examiner and (f) 6 productions of /cup/ or /Gun/ from memory. In the latter two phases the appropriate picture (either soup or moon) was placed before the child. He was told to “say the word in a different way.” Transfer was relatively easy under these conditions. For the ISI groups (/$ up/ and /$ un/) the first two phases [(a) and (b) above] of training were the same as for the 1~1 groups. In phase (c) the subjects were automatically shifted from /c/ to /$/. This change appeared to give the subjects no particular difficulty. Pretraining on /c/ assured that all subjects would be equivalent in their ability to learn and recall the /c/ sound. Phases (c), (d) and (e) were exactly the same as indicated above with the exception that I$/ was substituted for /cl. Following the final acquisition trial of phase (f) each subject was tested 30 set later and again 7 days later. During this 30 set retention interval the subject marked crosses in circles. For each retention interval each subject was shown either a picture of moon or soup and was asked “to say the word the way you had learned it.” Each subject was given six opportunities to pronounce the word; the experimenter prompted the subject every 6 sec. Responses tested after the 30 set or the 7 day interval were not reinforced. Subjects Children were selected from three elementary schools that represent districts of middle and upper class families. The children were white, monolingual, normal hearing (about 15 dB ASA) and were essentially free of articulatory errors. They were from the third and fourth grades; the average CA of each of the four groups was either 108 or 109 months. The children were assigned by rotation to one of the four groups keeping age and sex equivalent. A number of subjects was initially eliminated because they were regarded as slow learners by their teachers (N = S), because of bilingualism (N = 4) and because of failure to be present during the 1 week test (N = 6).

Results Retention

Scores

Summarized in Table 1 are the number of correct responses for the terminal picture test and the 30 set and 1 week retention intervals. Two procedures were used to score responses: stringent and lenient. Stringent refers to responses in which the test sound, /cl or I$/, and the syllable were correct. Lenient

PHONETIC

INTERFERENCE

41

TABLE 1 Number of Correct Responses for the Terminal Picture Test, 30 set and 1 Week Retention Intervals. Lenient Scores for the Retention Intervals Are Given in Parentheses. Total Possible Correct for Each Cell is 216 Group Condition Terminal 30 set 1 week

IYPI picture

196 191 (195) 128 (133)

YJupl 216 216 (216) 186 (190)

lgmi

IJunl

187 191 (194) 123 (142)

214 215 (216) 144

(168)

refers to responses in which the test sound was correct, whether or not the syllable was correct. The number of correct responses for each of the four groups for each of the three tests (terminal picture, 30 set and 7 day interval) is given in Table 1. Responses scored according to the lenient criterion are indicated in parentheses. All groups showed a decline in the number of responses recalled correctly over the 7 day interval at or beyond the 0.05 level, as measured by the Wilcoxanmatched pairs signed-ranks test. The rate of memory decay appears to be greatest for the /cup/ and Iqml groups. Differences between groups were assessed by the Mann-Whitney U test, corrected for ties. All probabilities reported are one tailed. Hypothesis (1) (the /~up~un/ comparison) was confirmed for stringent scoring (: = 1.74, P < 0.05). However, the significance level for lenient scoring falls short of sighificance (z = 1.17, P = 0.12). In order to test Hypothesis (2) (the /~up-qtp/ comparison) a correction in the initial level of original learning was made, since the scores for the 30 set recall intervals were not the same for all groups. Each subject’s score for the 7 day interval was subtracted from his score for the 30 set interval. These differences were then evaluated by the Mann-Whitney I/ test. Recall of /s up/ was significantly greater than /cup/; z = 1.60, P = 0.054 (stringent scoring) and z = 2.20, P = 0.01 (lenient scoring). The above findings support Hypothesis (2), in which competition was examined while facilitation was kept constant. No statistical evaluations were made between groups /cup/ and Iqml, since the number of correct responses differed for /Gun/ and /Sun/ on The direction of the differences for Hypothesis (3) (the /J‘un%un/ comparison) is not in accord with our prediction. Difference scores were computed, since the number of correct responses differed for /cun/ and /$un/ on the 30 set recall test. For both stringent (z < 1) and lenient (z < 1) scoring

48

HARRIS

WINITZ

and EVE HERRIMAN

the difference in ranks was nonsignificant. Since the contribution of facilitation is minimized for these two groups relative to soup, the effect of interference may have been considerably reduced. An interpretation of these findings is in order. Hypothesis (1) examined recall as a function of phonetic similarity. Correct recall was higher for /sup/ than for /Sun/, suggesting that phonetic similarity facilitates recall. Hypothesis (2) examined recall as a function of interference, keeping phonetic similarity constant. Correct recall was higher for /sup/ than for /cup/. We attribute the decay for /cup/ to associative interference at time of recall. We infer that phonetic similarity is equivalent, but that /$I, a strong competitor of /c/, is aroused at the time of recall. There was no difference between /cup/ and /Gun/. The relative effects of facilitation and interference cannot be partialed out from this comparison. The fact that the scores for /cup/ and Gun/ are nearly equivalent might suggest that differential processes are not operative. Hypothesis (3) examined recall for phonetically distant sounds. Groups /cunl and /$unl recalled equally well, suggesting that interference was not operative. We interpret these findings as consonant with our supposition that interference occurs subsequent to facilitation. In the case of /$un/ and lcun/ phonetic facilitation is reduced, therefore, the associations between /ml and. I$/ and /ml and /cl were direct, that is, /ml does not arouse /s/, a competitor of /cl. Errors

The proportion of error categories is given in Table 2. In comparing /sup/ Examining stimulus /F/, we note and /sup/ some differences emerge. first that there are many errors of similarity (/s/, /$/, /j/, /h/ and /hj/), which for the 7 day interval accounts for 0.75 of all errors. The phonetically similar errors for /$/ involve /s/ and /z/. These sounds account for 0.50 errors on the 7 day retention test. The fact that /cup/ evokes 0.25 more phonetically similar errors than /sup/ suggests that facilitation led to interference. That /m/ substitutions occurred was surprising, a result for which we cannot easily account. Phonetically similar errors for /cun/ and /Sun/ occur, but are reduced relative to /cup/ and /$up/. F or the syllable /cup/, /ml and in/ errors account for only 0.06 of all errors, suggesting that /m/ facilitates little the recall of /c/. Errors of phonetic similarity (/s/, /s/, /h/) occur with a frequency of 0.56 for /cun/ in contrast to 0.75 for /cup/. Also, 126 of the errors for /curt/ are errors of omission (/un/). These findings appear to confirm our initial hypothesis that facilitation and interference work at cross purposes. The errors for $un/ may reflect some interference since 0.21 of the er-

PHONETIC

Proportion

INTERFERENCE

49

TABLE 2 of Error Types for 30 set and 1Week Retention Intervals. The Number of Total Errors Is Given for Each Group (Lenient Scoring) /YPl 30 set 7 days

Number of total errors

(N=21)

(N=83)

/Jup/

/cun/

I unl

30 set

I days

30 set

7 days

30 set

7 days

(N=O)

(N=26)

(N=22)

(N=74)

(N=O)

(N=48)

EiTOi-

/sup/ lzupl /sun/ Other Isl substitutions

0.10 0.19

\:tl/i ! Other $1 substitutions

ljupl hpl lupl ihun/ /hjupl lhjunl Other ihl substitutions /mlsubstitutions lunl /nun/ Other In/ substitutions Other

0.27 0.23

0.05

0.16 0.18

0.14

0.16 0.16 0.08 0.08

0.10

0.01

0.19

0.05

0.03

0.02

0.21

0.03

0.19

0.23

0.15

0.05

0.16 0.02

0.14

0.18

0.02

0.01 0.05 0.26 0.01

0.10 0.02 0.08 0.08

0.09 0.01

0.02 0.46

0.26

0.27

0.19

0.16

0.23

0.05

rots involve /s/ substitutions. However, there was a large number of unclassifiable errors for / un/ (0.46), indicating that facilitation for /Sun/ was reduced relative to I/ up/. In summary, the results of this experiment suggest that facilitation and interference are jointly involved in the recall of speech sounds. Facilitation is governed by phonetic similarity. Interference appears to be governed by the relative response strengths of the error sound and the correct sound. Unfortunately, in this investigation imitation of /cl was not tested subsequent to the 7 day recall test. Previous research (Winitz and Bellerose, 1972ab, in press) has demonstrated that /cl, as well as many other sounds, are easily imitated even though correct productions are not elicited by pictorial stimuli. We feel, therefore, that the results of this experiment are not explainable in terms of a loss in phonetic production. A haunting question remains with us: How can facilitation be maintained while decreasing the effect of interference? The answer to this question, often

50

HARRIS

WINITZ

and EVE HERRIMAN

called the problem of carry-over, will involve intensive examination of the conditions under which facilitative factors continue to function while interference factors no longer operate. The authors wish to thank school oficials of Johnson County, Kansas school district 512 for giving permission to test subjects. The investigation was supported by the following U. S. Public Health grants: 5 KO 4 HD 38907, ROI HD 3571 and NS 10468. References Locke, L. J. Experimentally elicited articulatory behavior. Language Speech, 1969, 12, 187-191. Shelton, R. L., Johnson, A. F., Amdt, W. B. Monitoring and reinforcement by parents as a means of automating articulatory responses. Percept. Mot. Skills, 1972, 35, 759-767. Tumage, T. W., McCullough, T. A. Letter-sequence and unit sequence effects during learning and retention. .I. Exp. Psych., 1968, 76, 141-146. Underwood, B. J., Postman, L. Extraexperimental sources of interference in forgetting. Psychol. Rev., 1960, 67, 73-95. Win&, H., Bellerose, B. Sound discrimination as a function of pretraining conditions. J. Speech Hearing Res., 1962, 5, 34C348. Winitz, H., Bellerose, B. Phonetic interference and motor recall. J. Speech Hearing Res., 1972, 15, 5 18-528. (a) Winitz, H., Bellerose, B. Effect of similarity of sound substitution on retention. J. Speech Hearing Res., 1972, 15, 677-689. (b) Winitz, H., Bellerose, B. Self-retrieval and articulatory retenti0n.J. Speech Hearing Res. (in press).

Phonetic interference under varying conditions of phonetic facilitation.

JOURNAL OF COMMUNICATION DISORDERS 8 (1975) 4>50 PHONETIC INTERFERENCE UNDER VARYING CONDITIONS OF PHONETIC FACILITATION HARRIS Professor WINITZ*...
481KB Sizes 0 Downloads 0 Views