Teaching and assessing phonetic transcription: a roundtable discussion

© Dr Barry Heselwood, University of Leeds

Table of contents

• In attendance
• Teaching
• Assessment
• Bibliography

In attendance

Eric Atwell (University of Leeds), Sue Barry (Manchester Metropolitan University), Claire Brierley (University of Leeds), Paul Carter (University of Wales, Bangor), Ian Crookston (Leeds Metropolitan University), Anthea Fraser-Gupta (University of Leeds), Frank Herrmann (University of Sheffield), Barry Heselwood (University of Leeds), Sara Howard (University of Sheffield), Ghada Khattab (University of Newcastle upon Tyne), Pippa Noble (De Montfort University), Sue Peppe (Queen Margaret University College, Edinburgh), Leendert Plug (University of Sheffield), Joe Reynolds (Leeds Speech & Language Therapy Service), Fiona Skilling (University of Leeds), Vesna Stojanovik (University of Reading), Fay Windsor (Manchester Metropolitan University). Between us we represented a number of different perspectives and specialisms, including general phonetics, clinical phonetics and phonology, corpus linguistics, sociophonetics, English language.

Teaching

The first part of the afternoon was devoted to issues around the teaching of phonetic transcription, specifically issues of context (different degree programmes and types of students), content (what kinds of transcription should be taught), and delivery (how should it be taught).

Context

There was complete agreement that phonetic transcription needs to be taught in the context of phonetic theory, otherwise transcription conventions are not interpretable and students will not be able to understand what the symbols etc. represent. An early link between phonetic theory and transcription comes in what Abercrombie (1967, p. 127) refers to as the notational use of symbols, where they stand for general types of sounds as on the IPA chart.

It is important to recognise that students learning phonetic transcription are by no means a homogeneous group. Thought needs to be given to the needs of students on different degree programmes, many of whom may however come together on phonetics modules. The observation was made that transcription is perhaps most strongly present now on speech and language therapy degrees where it is seen as an essential tool in understanding and assessing clients’ speech problems as a basis for remediation. It was also noted that there is a tendency in general phonetics teaching for instrumental analysis, now so widely available and easy to use, to take precedence over auditory-perceptual analysis, and thus for instrumental records such as spectrograms to be seen as more important and more reliable than transcriptional records. Whatever the validity of this latter point, good transcription skills should be highly valued in phoneticians as a useful and convenient means of expressing a phonetic understanding of speech as it is heard.

The different kinds of degree programmes identified as being ones where phonetic transcription might be taught at least to some level are:

• General linguistics and phonetics
• Speech and language therapy (SLT)
• English language
• Foreign languages
• Training of language teachers

Requirements regarding transcription skills will not be the same for all these students. Students on general linguistics and phonetics programmes will normally have more scope to look at a wider range of speech data from different languages as well as English accent data and atypical speech, and perhaps also forensic methods involving transcription.  SLT students will focus mostly on atypical English speech, although familiarity with the phonetics of ethnic community languages will also be beneficial. Students of English language will focus on accent variation, while students of foreign languages might look at accent variation in those languages. Language teachers may not often go beyond the transcriptions found in dictionaries.

Within these student constituencies, there are obviously students at different levels of study from first year undergraduates through to students on taught postgraduate programmes. Students also come to their programmes with different personal linguistic backgrounds – some are monolingual, some bilingual, and for some English will not be their native language.

Content

Bearing in mind differences of context, the content of transcription teaching should be tailored to student needs and the value of transcription for those needs should be made clear.

The main issues regarding content are what types of transcription should be taught, and what kinds of data should be used for developing transcription skills. Running through these issues is the question of the aims of a transcription, specifically whether a particular transcription is speaker- or listener-oriented, i.e. aiming to represent how the speaker produced the speech, or the impression it made on the listener. This division becomes quite sharp when dealing with markedly atypical speech, but is present in all transcription activities. The fact that transcription conventions tend to be defined in articulatory terms biases the interpretation of a transcription towards the speaker, but it is in fact the listener’s phonetically-informed impressions which are actually recorded. Awareness of this potential tension needs to be developed in students, but care needs to be taken not to problematise the transcription process too early on or it might have a demoralising effect (the same problem attends the issue of what counts as a correct or incorrect transcription).

Transcription types can be divided broadly into those where it is the level of detail that differs, and those where it is the system of representation that differs. It is level of detail that differs across phonemic, broad, narrow, systematic and impressionistic transcriptions, and system of representation that differs across segmental, suprasegmental and parametric transcriptions. Ideally, students should be conversant with all these at least to the extent of knowing what they are used for.

Phonemic – it was felt generally that there is little value in a purely phonemic transcription. Knowledge of phonemic contrasts and the assigning of allophones to phonemes, and the concept and form of a phonemic representation, should be dealt with in those parts of the curriculum where phonology is covered. It seems like bad teaching/learning practice to teach a system where one has to mark something wrong which is phonetically correct, e.g. use of a glottal stop symbol when a glottal stop is heard in English.

Broad phonetic – this was felt to be the most appropriate level of transcription to introduce to beginners before moving on to more narrow transcriptions. It gets students familiar with the conventions and gives them the confidence to use them if they are not expected to incorporate much fine detail at this stage. Degree of broadness is to some extent language-specific – transcribing all realisations of English /p/ as [p] might seem overly broad to a native speaker of a language that distinguishes /p/ and /ph/ (e.g. Thai, Korean, Punjabi) but rather narrower to a native speaker of a language that has only one bilabial plosive (e.g. Arabic). The rationale for leaving out phonetic detail is motivated by ease of perception rather than by any phonologically motivated considerations.

Narrow phonetic – there was some discussion as to whether a useful distinction can be drawn between ‘narrow’ and ‘impressionistic’ transcription. ‘Narrow’ was taken to mean a transcription where the transcriber is interested in specific details, e.g. obstruent voicing, vowel nasalization, whereas ‘impressionistic’ was taken to mean trying to capture as much detail as possible. Narrow transcription was identified as a useful development to deepen broad transcription skills and so should be brought into the curriculum later and in a gradual fashion by directing students’ attention to specific phonetic dimensions, as might be appropriate in accent and sociophonetic variationist studies, comparative dialectology, and clinical work.

Impressionistic – the ultimate aim of transcription teaching in general should be to make students competent in impressionistic transcription, by which we mean the ability to transcribe data in an unknown language in as much detail as transcription conventions allow. It is recognised, however, that this is not often going to be the main application of transcription skills. For most practical research purposes, narrow transcription in the sense outlined above is what is practised, i.e. trying to achieve the same level of detail as an impressionistic transcription but directed to specific defined parameters.

Systematic – perhaps the most appropriate for sociophonetic studies where the phonological system is known and there are specific phonological variables with known variants, but where the speaker’s choice of variant is not predictable. The context of the variable might be transcribed phonemically/broadly, but the variant itself more or less narrowly, according to the transcriber’s interest in level of detail.

Segmental – the easiest system to use, and the one in most widespread use, is the alphabetic system of the IPA and its diacritics. This is the system that should be central to the teaching of segmental transcription, supplemented by the Extensions to the IPA (ExtIPA) as necessary. Although the ExtIPA symbols and diacritics have been adopted specifically for the transcription of atypicaleech, some of them are very useful for narrow transcription of normal speech where idiosyncratic behaviours are to be represented.

Suprasegmental – it was commented on that this is often under-emphasized in transcription teaching, partly because there is less certainty about the phonetic categories of suprasegmental dimensions of speech, partly because of the lack of an agreed widely-used system (although the Voice Quality Symbols (VoQS) conventions plug part of this gap), and partly also because of the perceived difficulty of suprasegmental phonetic analysis. There was general agreement, however, that this is not a satisfactory state of affairs, and more attention should be given to it.

Parametric – the main benefit of parametric transcription was agreed to be in explaining how speech is produced and to show students that speech sounds are not primes, but that they emerge from the interaction of time-varying parameters and vocal tract gestures. Getting students to produce a parametric transcription of a polysyllabic word is a good way to prompt them to introspect using proprioception and kinaesthesia guided by phonetic theory.

Types of data can range from single word citation/canonical form utterances produced live by the tutor through to lengthier passages of recorded spontaneous speech; real words, nonsense words, a known language, an unknown language, immature and atypical speech; audio recordings, video+audio recordings. An important consideration is the varying difficulty of different kinds of data for transcription. Broad transcription of single words/non-words delivered live with several repetitions is an appropriate way to begin teaching transcription, leaving ‘real’ speech data until students have begun to master narrower transcription. Many samples of recorded speech have been recorded in quite noisy environments and may not be clear enough to use for transcription practice, so it is important to select good quality samples.

The ethical issue of using recorded speech was raised. Tutors need to adhere to university data protection/ethical policies, and also local clinical regulations when using clinical data. It should not be thought that clinical data is only appropriate for teaching transcription to SLT students – non-clinical students find them fascinating, and the problems posed are very similar to those faced by fieldworkers working on hitherto undescribed languages/dialects. This of course means that tutors using such data have to ensure they are not in breach of codes regulating their use. It was widely felt that it would be a great shame were heightened sensitivity to data protection to seriously impede the use of  certain kinds of data. It is an issue that needs to be addressed at a national level but it is not clear in what forum.

Delivery

Transcription tends to be taught in ‘practical phonetics’ classes where production and perception work is also done. Class size is an important issue and often one that is caught between pedagogical considerations and financial/resource pressures. Eight students per class was generally thought to be an ideal number, with ten as a comfortable maximum. An even number means students can work in pairs, and keeping numbers down means the tutor can give individual attention. It also facilitates consensus transcription as a way of getting students used to the idea of transcription agreement/disagreement and how to compare and discuss transcriptional differences. Number of hours spent teaching transcription will necessarily vary across different degree programmes and is not likely to be great, given the other aspects of phonetics that also need to be taught. Self-study materials can help to make up for this, including use of electronically available language corpora that have phonetic transcriptions with them as well as in-house material.

In live delivery for transcription of IPA-type nonsense words, the issue of performance from transcription arises. This reverses the usual relation between sound and symbol in that the symbol now comes first! Decisions have to be made on such questions as ‘how much voicing should there be in an IPA [b]?, how much nasalization is represented by [ ̃  ], how ‘gulpy’ is a canonical implosive?’

Rather than teach phonetic transcription as an isolated skill, it was felt to be important to relate it to other areas of the curriculum in addition to phonetic theory. Students can be shown, through good selection of data, how it is an important tool in various areas of linguistic investigation, such as sociolinguistics (sociophonetics, sociophonology), speech pathology and therapy, forensic phonetics, for annotating instrumental records, and for providing data for phonological analysis to test phonological and psycholinguistic theories, e.g. whether the kind of observed systematic variation that can be represented in narrow phonetic transcription can be accounted for in Optimality Theory; whether models of bilingualism can account for observed narrow phonetic distinctions in the two languages of an individual.

Assessment

This was the theme for the second part of the afternoon. Discussion revolved around the kinds of materials appropriate for assessing transcription ability, assessment conditions and procedures, and how student transcriptions should be marked.

Materials

To have a range of transcription tasks in the assessment with the weighting of each task determined by the amount of content was felt to be appropriate. For example, short nonsense words in broad transcription would have a much lower weighting than an impressionistic or narrow segmental and suprasegmental transcription of a complete intonation group of several words in an unknown language taken from a sample of spontaneous speech. Tasks clearly need to be suited and graded to what students have been taught up to the time of assessment.

The kinds of tasks might include:

• Broad transcription of short nonsense words
• Broad transcription of a passage of English or other known language
• Narrow transcription of certain aspects of a passage in an unknown language
• Full impressionistic segmental and suprasegmental transcription of immature/atypical speech  sample
• Transcription of conversational data
• Parametric transcription of a polysyllabic word (this could be part of a written phonetic theory exam)

A further consideration in deciding on the weighting of a task is how much time was spent on it during the module/course.

Procedures

Broad transcription tasks might best be delivered live by the tutor with a controlled number of repetitions, whereas data for narrow and impressionistic transcription is probably best done from recordings. It would be good to have at least one task done from a video+audio recording. The question then is under what conditions? Students can either be given recorded data to take away and hand in a transcription say 48 hours later, or they could do the transcription under exam conditions, e.g. in a phonetics/language laboratory. The latter ensures the transcription is the student’s own work, but it doesn’t allow students to take extended breaks and return later to the task with refreshed ears. A mixture of both is of course possible.

Marking

How to mark transcriptions was acknowledged by all as a real problem, not so much for dictated broad transcription where the tutor is confident of what counts as correct, but certainly for narrow/impressionistic transcriptions where the tutor has no privileged knowledge of the data. In these cases, there has to be some latitude in what is taken as correct, but how much?  One way is to have recourse to consensus transcriptions produced by the teaching team, or two or three ‘professional’ transcriptions which are taken as indicating the limits of allowed transcriptional variation. In all cases, however, it is important to recognise degrees of incorrectness and not just to mark a transcription symbol as either right or wrong – if [ð] is the sound the tutor produced, [z] is probably a better answer than [ʒ]. Rather than merely looking at the symbols students put, we should think what it is they are trying to represent, what they seem to have noticed; i.e. we should be as analytic in our marking as we want students to be in their transcriptions.

It was agreed that although there often isn’t a single correct version of a transcription, nevertheless there are versions which can confidently be rejected as incorrect, e.g. [n] for [ð]. There is of course a problem in telling a student who insists they heard [n] that they didn’t!

Once a marking version of a transcription has been created, there is the question of how marks should be awarded. Practice among members of the group varied somewhat between negative and positive marking. Nobody strongly favoured one or the other overall but it was thought that some tasks might lend themselves more to positive marking, some to negative. A mixture of both kinds might be a good balance to have. But however a final mark is arrived at, it will need to be made to fit the university’s marking scales for presentation at exam boards and the calculation of students’ overall marks. It was noted that often the external examiner for a programme will not be a phonetician and may need persuading of the fairness and propriety of whatever marking scheme is used – the more complicated the scheme, the more they might need persuading. Clearly there is an issue of transparency and accountability here, but there was a strong feeling that the marker’s professional judgment should be trusted, and that a mark is a qualitative judgment of a student’s performance; the primary function of a marking scheme is to ensure consistency across scripts. There was some feeling that students who are clearly trying to produce a narrow/impressionistic transcription as required but are making mistakes should be rewarded more than those who do not go beyond a broad transcription.

The issue of students with dyslexia and hearing impairments was raised but unfortunately there was not time to go into this at the meeting. It could perhaps be discussed at a future meeting.

Bibliography

Referencing this article

Below are the possible formats for citing Good Practice Guide articles. If you are writing for a journal, please check the author instructions for full details before submitting your article.

• MLA style:
  Canning, John. "Disability and Residence Abroad". Southampton, 2004. Subject Centre for Languages, Linguistics and Area Studies Guide to Good Practice. 7 October 2008. http://www.llas.ac.uk/resources/gpg/2241.
• Author (Date) style:
  Canning, J. (2004). "Disability and residence abroad." Subject Centre for Languages, Linguistics and Area Studies Good Practice Guide. Retrieved 7 October 2008, from http://www.llas.ac.uk/resources/gpg/2241.