Speech errors and phonological encoding
After the morphological components of a word have been selected, they need to be given form. In speech, this will be the sounds that make up the word. This process of giving sound to the abstract form of the word is widely known as phonological encoding. The lexemes or word-forms are made available via links with the lemmas that have been accessed during the grammatical encoding. So (cat) is linked in the mental dictionary both to the written form (cat) and to the spoken form /k æ t/.
In Chapter 2, we started to show the development of the sentence the cat hunted the mouse, and had followed this as far as the lemma-based structure in 4.33. The process of phonological encoding now inserts the appropriate word-forms into the structure, as in 4.34.
Subsequently, the function words are filled in, along with grammatical endings such as the marker of past tense, resulting in 4.35. A phonetic plan is then generated for this string, which will drive the articulators the speech organs. The detail of the phonetic plan will depend not only on the chosen word-forms but also on the utterance context, because this will affect aspects of pronunciation such as which words will be stressed for emphasis or contrast.
As with morphological aspects of word building, much of the evidence for phonological encoding comes from studying speech errors. Examples of different types of sound errors are given in Table 4.1. As with the word errors discussed in Chapter 3, there are errors that indicate that the wrong element or sequence of elements has been selected, there are errors that show the wrong ordering of selected elements, and there are errors in which elements have been omitted or added. Mis-orderings are the most frequent type of sound error. In error corpora about half of all sound errors are either anticipations e.g. 4.37 and 4.42 or perseverations as in 4.38 and 4.43, with the next largest group being exchanges 4.39 and 4.44.
As we have seen, the majority of sound errors occur within local phrases rather than across larger sentence structures. Unsurprisingly, since there tend not to be many words of the same word class within a phrase, most sound errors also involve two words of different word classes.
Two-thirds of sound errors involve single segments, including initial consonants as in the examples 4.42 to 4.44, as well as final consonants 4.45, and vowels 4.46–4.48. Other errors involve consonant clusters 4.49, but whole syllable errors 4.50 are extremely rare and note that in this case only the element is an entire syllable, i.e. the other element in the exchange – ca – is part of and not the whole syllable, which would be cat.
Investigations of sound errors across corpora of error data have pointed out a number of patterns. The non-randomness of these errors leads to some important conclusions concerning language production at this level of phonological encoding. One observation is that there are important constraints – some positional – on sound errors. For example, the elements involved in mis-orderings are more likely than not to come from equivalent positions in two words. Another is that sound errors are more likely to result in real words, rather than in nonwords, although the examples already given show that this is clearly not always the case.
Metrical structure constraints
The sounds involved in sound errors tend to come from syllables that are either both stressed or both unstressed. So in 4.46, 4.47 and 4.49 the exchanged sounds are in the stressed syllables of each word, and in 4.48 they are in unstressed syllables. The interpretation is that the stored specifications of words include information about their stress patterns, in a sort of metrical frame. The good recall of stress position when people are in the tip-of-the-tongue state is a further piece of evidence for the metrical frame. As we will see later, evidence from tongue-twisters also supports this.
Syllable structure
constraints the most widely accepted account of English syllable structure is a hierarchical one. This is illustrated in 4.51, together with some sample English words. The syllable must minimally have a peak, which is usually a vowel – the example word with this minimum is eye /ai /. The peak is the only obligatory part of the rhyme. The rhyme can also have a final
consonant or sequence of consonants, forming the coda. In 4.51 the word ice /ais/ has a peak and a coda. A coda-less peak can also combine with an onset consonant as in lie. Finally, both onset and coda positions can be occupied: lice /lais/. Syllables in English can be more complex than this, since onsets and codas can contain multiple consonants, as in the word strengths /st ʴe ŋk θ s /, which has three onset consonants and four coda consonants.
Looking at sound errors in terms of syllable structure, we find that peaks exchange with other peaks 4.46, coda consonants swap with other coda consonants 4.45, and onset consonants exchange with other onset consonants as in 4.39 and 4.49. This last class, where onsets swap with other onsets, is the type of speech error known as a spoonerism. Spoonerisms are named after the Reverend William Spooner 1844–1930, who was Warden of New College, Oxford. He is reputed to have frequently commit ted this kind of error, though it is unclear how many of the examples attributed to him were actually his errors – it has been claimed that some were put on by Spooner for effect, and that some were not uttered by him at all, but constructed by his students. Two classic spoonerism utterances attributed to Spooner are reproduced here as 4.52 and 4.53. The target utterances should be obvious.
Spoonerisms seem in fact to be quite noticeable errors, which may reflect a higher frequency of occurrence. But their distinctiveness could equally derive from the fact that the onsets of words are very important for accessing words from the mental dictionary, so that any disruption to onsets will have noticeable effects for the listener. Chapter 8 considers the importance of word onsets in speech comprehension, in particular from the viewpoint of models of spoken word recognition.
Phonetic similarity
The third type of constraint on sound errors is that there is a strong tendency for the sounds involved to be phonetically similar and to come from phonetically similar contexts. Clearly, if there is a preference for onsets to swap with onsets, peaks with peaks, etc., then the elements involved will be similar since for instance the syllable peaks will in most cases both be vowels. But the similarities go beyond this. Although the examples above include some exceptions to this pattern, it has been observed that the onset consonants in spoonerisms are likely to be phonetically similar. The frequently reported error of par carck for car park is a case in point, since the consonants /k/ and /p/ are both voiceless stops. This example also illustrates the finding that the phonetic contexts in which the sounds are found are frequently highly similar – in this case, both the /k/ and the /p/ are followed by the /a/ vowel.
Slots and fillers
The patterns sketched here, that sound errors tend to involve similar sounding elements from similar metrical, syllable-structure and phonetic contexts, have been interpreted in terms of a slots-and-fillers approach to phonological encoding Shattuck-Hufnagel, 1986. This is similar to descriptions given in Chapter 3 and earlier in the current chapter of how sentence structure and morphological structure might be put together. In essence, the approach as applied to phonological encoding suggests that when a lexeme is converted into a sound pattern, phonetic segments are mapped onto a template for the lexeme, in which metrical and syllabic structures are specified. Since the speaker has a number of words lined up for production as we know from word exchanges, there is potential for the sounds of these words to be misallocated. The fact that the patterns mentioned above are characteristic of sound errors implies that the misallocation is not random. Rather, errors reflect the need to match the properties of the sounds to those of their places in the template. If there are competing similarly-defined positions in the template e.g. initial voiceless stop consonants in a stressed syllable, as in a stressed syllable, as in car park), then these are likely to attract the sounds, resulting in error.
Real word bias
We have seen that an important additional aspect to the non-random nature of sound errors is the strong tendency for such errors to result in real words rather than in nonwords. Why should this be the case One possible answer is that the pattern is illusory. That is, it is possible that many errors do in fact result in nonwords, but because we expect speakers to use existing real words rather than nonwords, we misperceive or reinterpret the nonsense resulting from sound errors as real words. This may be one source of under-reporting of sound errors. In Chapter 5 we will review evidence from experimentally induced errors that relates to this question, amongst others.
Ambiguous errors
The discussion above has assumed that there is no controversy over the interpretation of the errors as being of a certain type, i.e. as either sound errors or word errors. In some cases, though, this is not clear. The self-corrected error in 4.54 could result from the anticipation of the vowel in way, or it could be a word substitution of say for speak. In the example in 4.55 there could be a perseveration of the /v/ in you’ve or a word substitution of prevented for presented.
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
