Representation and units in phonetics

From both written transcriptions and acoustic representations, it should be clear that all forms of representation of speech are partial, and none gives a complete picture of speech. In the same way, an architect or an estate agent will describe a house in different ways, reflecting different purposes and different levels of detail.

Both transcriptions and (perhaps more surprisingly) acoustic representations have an element of subjectivity in them. As phoneticians, we train ourselves to listen more objectively. Ways to make transcription more of a ‘science’ and less of an ‘art’ include regular practice, collaborating with others whose judgements are trustworthy, or combining the activity of transcription with acoustic observation, which allows for a slower, more piecemeal approach to work and can make it possible to check impressions against acoustics. If our records open the way to other work, then they have served a useful purpose.

Acoustic representations seem more objective: after all, any two people can put the same acoustic signal in and get the same representation out. However, such representations are less objective than they appear. For instance, it is possible to manipulate the way the acoustic signal is processed and the way that spectrograms are drawn so that they appear sharper; or to emphasize the temporal organization of the signal over the frequency aspect (or vice versa); or to draw spectrograms in color rather than black and white; and the Hertz scale does not by any means represent the way the ear and brain analyze the signal. So there are also many unknowns with this kind of representation.

For all these reasons, it is wise be wary of ascribing to any one form of representation some kind of primacy. Made and used carefully, they are all informative in some way.