categorical perception appears to be rather unusual, and it may well be an aspect of the special capacities which humans have for language mentioned in our
main introduction. A further observation supporting this possibility comes from the behaviour of infants. As we mentioned in the last section, it is possible to perform experiments with very young babies, and to use measures such as head turning, sucking or heart rate as indicators that they do or do not perceive a difference between two sounds. When this is done with babies that have been
exposed to some form of language, it is discovered that they too perceive VOT
categorically, many months before they start trying to pronounce adult plosive sounds themselves (exercise 1).
Speech production
At some stage in the production of speech, the speaker has to for-
mulate plans for moving the articulators in such a way as to produce the required sounds in the required order. This is far and away the most complex motor-control problem faced by human beings. The number of different muscles involved is
enormous and the fine-tuning required to get even an approximation to human
speech is extremely delicate. The complexity of the process is seen to be even greater when we realise that we can and do introduce extremely subtle changes into our normal speech, by altering its rhythm and loudness, and especially our tone of voice (intonation), so as to achieve different nuances of meaning. We can
114
sounds
even play with our speech, by imitating other accents or modes of speaking. When we speak to someone with a different accent, we unconsciously accommodate to that accent in a fashion that is only really apparent to a person who is trained in phonetics (see section 4 for sociolinguistic perspectives on this phenomenon).
Given the complexity of the problem, it is all the more remarkable that we speak with relatively few errors. However, errors are made in normal speech and these throw considerable light on the nature of the speech production process. Later in the book, we shall be looking at speech errors made by people which involve
whole words, and how these might be used to investigate the nature of the mental lexicon. Here, we shall focus on errors which indicate the importance of individual sound segments and syllable structure with a view to understanding the process of speech production. In most cases, the errors we cite have been collected by
linguists or psycholinguists listening to conversations, lectures, or TV and radio programmes.
One of the types of speech error that we all make, and which everyone is aware of, is in the context of the tongue-twister. In every language there are certain sequences of sounds or syllables, which, for some reason, are particularly hard to pronounce. Some of these can be remarkably innocent-looking. For instance, you can get friends to try saying the name Peggy Babcock three times very
quickly. Make a note of what they actually say, using phonetic transcription (you will probably find it necessary to record their attempts), and see what types of error are made.
The problem posed by tongue-twisters is one of vocal gymnastics, something
akin to patting your head and rubbing your tummy at the same time. However,
there are different sorts of errors which, in many ways, are more interesting, because they don’t have such obvious correlates in non-speech motor control. One of the most famous types of speech error is illustrated in (74):
(74) a.
You have hissed all my mystery lectures [missed all my history lectures]
b.
You have tasted the whole worm [wasted the whole term]
c.
our queer old dean [our dear old Queen]
These are examples of spoonerisms, allegedly uttered by the Reverend William Spooner, a lecturer at Oxford University in the last century. (‘Allegedly’, because undergraduates were in the habit of making up such things and attributing them to their notorious mentor.) What is happening here is that two sets of sounds are being exchanged, as shown for (74b) in (75): (75)
(t)erm
(w)asted the whole
⇒
(t)asted the whole (w)orm
Example (74a) is similar in that single whole segments are exchanged, but (74c)
is different, as we can see if we refer to phonetic transcription, as in (76):
e
I e
(76)
i
[(d) ould (kw)i n]
⇒
[kw ould di n]
Processing sounds
115
In terms of segments, we are exchanging two sounds for one here; however, our discussion of syllable structure in section 5 has shown us how we can construe this as an exchange of one unit for another. The cluster kw- in queer/queen is the onset of the syllables which make up these words, and it is this onset which is being exchanged with the onset of the syllable [diːn]. Indeed, it turns out that syllable structure is important in analysing speech, since it is only onsets that get exchanged for onsets, or codas for codas. We don’t find constituents of the syllable getting confused with each other in exchanges. In other words, we don’t find the onset of one word being exchanged with the coda of a later word, i.e. we don’t observe errors of the form shown in (77):
(77)
a ca(t)
and
a (d)og
⇒
a tog and a cad (unattested error type)
Simple as this observation is, it provides a very direct indication of the involvement of syllable structure in speech production. If the speech production mechanisms did not have access to this structure, there would be no reason to expect that such errors would not occur – logically, they are just as plausible as those involving the switching of onsets or codas.
Exchanges are not the only kind of speech error involving individual sounds.
In (78–82) we see a number of other reasonably common types (in each case collected by researchers from ordinary conversations):
(78) a.
it’s a meal mystery [real mystery]
b.
fonal phonology [tonal phonology]
(79) a.
give the goy [give the boy]
b.
Michael Malliday [Michael Halliday]
(80)
his retters [letters]
(81)
country presents [peasants]
(82)
the Britch [British]
The examples in (78) are anticipations, in which a sound is anticipated from a following word, whereas those in (79) are perseverations, in which a sound is repeated from an earlier word. Example (80) is a substitution of one phoneme by another, while (81) is the addition of a phoneme (producing incidentally a real word). Finally, in (82) we see a case of omission of a phoneme. (exercise 2).
Exchanges are a relatively commonplace type of error, so it may not be
immediately apparent that they pose an important theoretical problem for the modelling of action. In fact, they indicate very clearly that we formulate a plan of what we are about to say before we actually get round to saying it. As
early as 1951, the psychologist Karl Lashley used this as an argument against Behaviourism, a psychological position that maintains that all our actions
are governed by habitual responses to stimuli. Lashley pointed out that errors of serial order of the kind illustrated by exchanges demonstrate that we must
116
sounds
Figure 38 A simplified version of the scan-copier model of speech production plan ahead, and that we don’t simply respond to whatever stimulus has just
impinged on us.
The idea of forward planning is enshrined in an influential model of speech
production called the scan-copier model. According to this model, we first form an abstract representation of the next phrase we are about to utter. Then we copy that representation into a ‘buffer’. This then gets translated into movements of the articulators. As we saw from example (74c), syllable structure is very important, so it is appropriate to assume that the scan-copier is sensitive to syllable structure.
Читать дальше