Types and levels of transcription

Perhaps surprisingly, for any utterance there is more one appropriate phonetic transcription. Different situations make different demands of a transcription, so we need to understand how transcriptions can vary.

For example, if we encounter a new language or a new variety for the first time, there is no way of knowing initially what might turn out to be important, and what might not. In this case it is common to transcribe as many details as possible so that we have rich working notes to refer to. These transcriptions might be personal memoranda to remind ourselves of what we heard. (Most phoneticians have a good auditory memory: reading detailed transcriptions is one way to recall what was heard.)

We might be working on data for a specific linguistic reason, for instance to work out something about the details of place of articulation for [t] sounds within a given variety. In doing this it is best to concentrate on things that are relevant to the problem in hand, so some parts of the transcription might be detailed, while others will be sketchier.

One important dimension is the amount of detail that a transcription contains. At one end of the spectrum, transcriptions can contain representations of as many details as we can observe. This kind of transcription is often called narrow. At the other end of the spectrum are transcriptions that use a restricted set of symbols, and which therefore gloss over many phonetic details on the grounds that they are predictable from the context, and not important in distinguishing word meanings. Such transcriptions are often called broad. Transcriptions in dictionaries are typically broad.

A simple transcription is one which uses familiar Roman letter shapes in preference to non-Roman letters shapes. E.g. the [r] sound in English is often pronounced as [ɹ]; but it can be represented with [r] in a simple transcription unambiguously because although [r] stands for a voiced alveolar trill on the IPA chart, alveolar trills do not usually occur in English.

Transcriptions are sometimes used to compare sounds. For instance, we might want to compare the pronunciation ofin Scottish English and Irish English, so we could use symbols such as [ɾ] (tap), [r] (trill), [ɹ] (approximant), etc., so as to make comparison easier. Transcribing different varieties of a single sound when we hear them produces a comparative (also narrower) transcription.

Systematic transcriptions limit the number of symbols used to a given set. In some circumstances, there are choices about how to represent sounds. Phonemic transcriptions are by definition systematic. For example, the word ‘hue’ starts with palatal approximation, voicelessness and friction. In a systematic transcription, the set of available symbols is restricted. Since [h] and [j] are needed independently (for e.g. ‘who’ and ‘you’), the combination [hj] represents the sound at the start of ‘hue’ unambiguously, without introducing a new symbol, although the symbol [ç] represents a voiceless palatal fricative and is equally accurate in this case.

Phonemic transcriptions embrace the concept that one linguistically meaningful sound should map on to one symbol. (‘Linguistically meaningful’ in this context usually means ‘capable of distinguishing words’.) So the velar plosives in the words ‘kick, cat, cool, skim, school, look, sick’ (which are all slightly different) are all transcribed as [k]. Phonemic transcriptions are necessarily broad. Allophonic transcriptions capture such details, even though they are predictable. Allophonic transcriptions are narrower than phonemic ones. Phonemic and allophonic transcriptions constitute the basis for a phonemic analysis of speech.

A transcription which uses the full potential of the IPA to record much observable detail is called impressionistic. Impressionistic transcriptions (or ‘impressionistic records’) are necessarily narrow.