The effectiveness of phonemic transcription education in speech & language therapy: a preliminary investigation

Ian Crookston

Speech & Language Sciences,

Leeds Metropolitan University

All Speech & Language Therapists (SLT's) in the UK are trained in phonetic transcription. How this transcription relates to the work of the practising SLT is a barely researched question, yet one which is worth researching. Phonetic transcription education is understandably unpopular with a body of students selected for their people skills rather than their interest in the minutiae of language, and the practising SLT has considerable autonomy in assessment and management decisions. To what extent are phonetic skills deployed in clinical work, and how?

It would be very ambitious to attempt to provide systematic evidence for an answer to this question, and this study does not do so. What is attempted here is to make some limited progress towards an answer to a subsidiary question. How do phonetic transcription skills develop after the completion of training? Do they progress, remain static, or degenerate? An answer to this question would put a piece into place in the jigsaw puzzle involved in answering the broader question just discussed.

One significant universal building-block of SLT phonetic training is either phonemic or broad phonetic transcription of material in a formal speech style in a well-known accent of English. This element lends itself to research for practical reasons. It can be relatively objectively assessed, and a short test in it can be delivered and answered on paper.

55 questionnaires were distributed to graduates of the same Speech & Language Therapy course. Some of the recipients had graduated approximately 3 1/2 years previously, and some approximately 2 1/2. 24 of the questionnaires or 44% were returned. The question under analysis here asked the respondent:

Please transcribe the following words as they would be said by a speaker of RP or any educated South-Eastern England accent.

Six words were then supplied. Each respondent's transcriptions were then compared with transcriptions of the same words by the same SLT on the occasion of the final formal assessment in phonemic transcription. This final assessment was a dictated passage of the familiar and traditional type.

The principal instrument of comparison was the Net Change in Error Count (NCEC). If a respondent had made 2 errors over the six words as a student, and made 1 error in the questionnaire responses, this was registered as an NCEC of "+1", which it is convenient to read as "1 to the good". Similarly, if a respondent had made 3 errors as a student and 4 in the questionnaire responses, this would be registered as an NCEC of "-1" or "1 to the bad".

The independent variable in this comparison is purportedly the date of performing the transcription, and the dependent is the error count. There are, unfortunately, a number of confounds: see table 1.

Narrow transcription education in the event had very little overt effect on the questionnaire responses: it proved easy to ignore the few diacritics which presented themselves, as well as the few choices of symbol which departed from standard phonemic symbol-sets. Likewise there was no overt effect of the transcriber’s own accent, as deduced from a question on the respondent’s region of origin: not a single error in the questionnaire responses could be attributed to this source. (This was possibly in part a side-effect of the selection of items for the questionnaire).

The fact that the questionnaire respondents were self-selected is perhaps potentially the most serious confound, but again results allayed this fear considerably: the average respondent when a student had 93% of the error count of the average for their group. In other words, the sample of respondents was only slightly biased towards the better students.

The fact that the questionnaire left respondents to imagine a pronunciation to transcribe seems to have had some effects on the outcome in the area of central vowels, as will be discussed below.

The other confounds are imponderables and an opportunity for practicable research which excludes them is awaited.

Overall results give cause for a degree of professional and pedagogical satisfaction. The average NCEC per respondent was -0.21. In other words, respondents performed almost exactly as well in their questionnaire as in their student assessment. In analysis of the results, some specific hypotheses were tested.

HYPOTHESIS 1: There will be decline in performance on fricatives

Most beginners in phonetics seem to have exactly as much awareness of sounds as is helpful in learning to spell, and not a pennyweight more. Thus most seem to experience a slow learning curve on all fricative voicing contrasts except /f v/ (which are even accurately transcribed in English nonsense words). Phonemic awareness in general does not appear to be natural for most people: western pre-literate children, western adult illiterates, and Chinese readers who do not read alphabetised Chinese all lack it (Perfetti 1994, Rayner & Pollatsek 1989). It is thus plausible that phonemic awareness often only develops strictly to the extent necessary in learning to read.

However, once the slow learning curve on fricatives is completed by a phonetics student, the resulting knowledge appears to be stable. Only 3 respondents showed an NCEC on fricatives, and in those three it was positive.

HYPOTHESIS 2: There will be two clusters of post-assessment learners

This hypothesis was suggested by two things. One is my own persistent impression that there are two sharply differentiated types of successful learner in phonemic transcription, namely, a small group in which learning appears almost effortless and performance is high, and a large group in which learning appears effortful and performance is not as high. The other is the marking of nonsense word transcription, which, given a certain approach, always seems to yield a small cluster of very high marks alongside a large, and remarkably tight, cluster of adequate marks.

However, if there are two types of successful listener, this investigation failed to differentiate them. Total NCEC was in fact close enough to a normal distribution, and it is hard to argue for even a suggestion of bimodality:

-3 2 respondents

-2 1 respondent

-1 8 respondents

0 5 respondents

+1 6 respondents

+2 1 respondent

+3 1 respondent

HYPOTHESIS 3: There will be a decline in performance on central vowels /I @ U V 3/

It seems possible that when transcribers pays conscious attention to sounds, they internally transform unstressed syllables into some stressed equivalent. This could affect even stressed central vowels by a knock-on effect. If for example the word "water" is transcribed /wOt3/ and this is then corrected to /wOt@/, this could produce uncertainty with both central vowels involved.

Central vowels in fact produced variability in NCEC:

-3 1 respondent

-1 6 respondents

0 10 respondents

+1 6 respondents

+2 1 respondent

This variability was much greater than that noted for fricatives above, and central vowels were the only category which produced an NCEC (other than zero) in anything like the majority of respondents. Thus it does seem that some speech therapists experience continued uncertainty with central vowels after the completion of formal phonetics education.

The methodological confound noted above, that the questionnaire required respondents to imagine a pronunciation while the dictation gave a pronunciation, had further effects on these figures. Where there are two alternative pronunciations in the target accent, it could not of course be known which had been imagined. There were six responses in which could be observed a contribution to NCEC of +1 by some pattern such as the following:

In other words, the aggregate NCEC in this category includes an element of +6 associated with imperfections in the methodology. It is reasonable to assume that at least some of these cases would have produced errors in a tighter questionnaire methodology, thus yielding an overall NCEC that would have been somewhat to the bad as well as variable from respondent to respondent.

Central vowels are thus an area where the effectiveness of phonemic transcription education for Speech and Language Therapists is open to question. The fact that after phonemic transcription education an SLT student generally proceeds to some form of narrow phonetic transcription or analysis suggests that phonemic transcription does not represent a sufficiently detailed level of awareness of speech sounds for a fully-trained SLT. These two observations taken together should spur phonetics teachers to look for better ways of teaching the perception of central vowels.

Two main conclusions emerge. In general, practising Speech and Language Therapists retain the facility for phonemic transcription which they acquired in training, as far as can be tested by a small transcription exercise. More specifically, as a group they seem to be uncertain in the differential perception of central vowels.

REFERENCES

Perfetti C (1994) "Psycholinguistics and reading ability" in M Gernsbacher ed Handbook of Psycholinguistics, Academic Press, San Diego, Ca, 849-894

Rayner K & A Pollatsek (1989) The Psychology of Reading, Lawrence Erlbaum, Hillsdale, NJ

KEY TO PHONETIC SYMBOLS

I=kit vowel

U=hood vowel

@=schwa

3=nurse vowel

V=strut vowel

O=thought vowel