Speech errors and morphological structure
A key issue in the discussion of the production of morphological structure is whether morphologically complex words are constructed as and when they are needed, rather than being fully listed in the mental lexicon. If we find that at least some complex words are created on the fly, then a subsidiary question is whether this is true only when certain types of morphology are involved. Most frequently, this involves the distinction between inflectional and derivational morphology, and so we shall look at each of these in turn.
Inflectional morphology
(4.1) and (4.2)– (4.7) illustrate stranding errors. The stranded element is underlined in each case. These errors show that word stems and word endings are treated separately during at least some stages of the process of speaking. In Chapter 2 our description of 4.1 as example 2.6 was that an abstract version of the phrase, something like singular full of plural’, needs to have word stems assigned to the roles indexed by and, and endings produced to indicate singular and plural. The stranding error suggests that the grammatical marking of singular and plural is part of the sentence frame, with the error involving the misplacement of the stems into this frame during lexical insertion. That is, hole in 4.1 is inserted for X instead of for Y, and is marked for singular, i.e. has no overt affix, while floor is inserted for Y instead of X and is given a plural affix, producing floors.
This interpretation of stranding errors suggests that the affixed form is constructed during speech production. If this were not the case, i.e. if in fact the complete affixed form is retrieved from the mental lexicon, then we would predict that the affix remains with the relevant stem, giving a holes full of floor. But such errors are extremely rare.
Stranding errors typically involve grammatical endings, i.e. endings traditionally known as inflections. In 4.1 and 4.2 the inflection is a plural marker. This is one of the most frequent elements involved in stranding errors in English, along with tense markers, as in 4.3 and 4.4. Both plural and tense markers are stranded in 4.5. The example in 4.6 involves the third person singular s ending. It is an error noticed by the author in New Zealand. Given that the - ers and - es endings as in (villagers and villages) are homophonous for many speakers of New Zealand English, it is possible that the entire base bouncer word was moved, and that the error word bouncers was reinterpreted by the author as bounces. In 4.7 the stranded affix is the progressive marker – ing.
Not all English plural and past tense forms involve the simple affixation of endings onto stems. English has plenty of irregular plurals, like ee or l and irregular past tense forms like feet or children and irregular past tense forms like swam or went. Such forms cannot be predicted by a rule such as add -s for plural’, and so it would be reasonable to expect these words to be stored as complete forms in the lexicon and accessed as such rather than being constructed as and when needed. Nevertheless, irregular past tense forms are also involved in English speech errors. Consider the stranding error in 4.8. The exchange is clearly of the underlying morphemes for the stems know and hear, and not of the full forms, which would have resulted in the error I’d heard one if I know it. The location that know has been moved to in the error is a location that has been specified at the functional processing stage as past’, and the subsequent process of specifying word forms results in the insertion of the correct irregular form, rather than a regularised knowed.
Examples like this also show how morphological markers such as past are not the same as the phonological realisation of these markers. That is, what is merged with o in the error in 4.8 is the abstract entity past, rather than a specific form such as /d/.
Words and rules
Because they use the regular, predictable and most widespread way of forming past tenses in English, it seems sensible to suppose that past tense forms like started and sounded (example 4.3) or plurals like kitchens and companies example 4.2 are constructed when needed. Irregular forms like knew and children, however, have to be looked up. But does it make sense to have two routes for finding past-tense forms or for plurals – a rule-based route and a lexical look-up route – rather than just assuming that every past tense is looked up in the dictionary This issue has been a hot topic in psycholinguistics, with many arguments presented for each view see e.g. Pinker, 1999. Let us consider some of the arguments against the idea that all inflected forms are looked up in the mental lexicon.
First, it is clear that the regular forms are used, in a rule-like way, to form past tenses or plurals of new words. We can quite safely predict that if we asked participants to give a plural form for a nonword that we said was a noun, such as bafflack, then they would respond with bafflacks.
A second and related argument is that children learn at quite an early stage that forming plurals – an inflectional process – is highly regular and therefore predictable, as is making past tenses. The so-called wug test’ was used in a study with young children who were shown a cartoon picture of a creature see Figure 4.1, were told it was a and were then asked what they would call two of them Berko, 1958. As predicted, the children said they were wugs.
A third argument involves morpheme shift errors, such as 4.9–4.11. Consider the verb in example 4.10. There is no reason to expect that looking up a stored past tense form of point out in the mental dictionary would produce anything other than pointed out. So where does pointed out come from?
Morphology or phonology?
Since both morphology and phonology involve small bits of words, there will be some errors apparently involving morphemes where an argument in terms of the sound structure of the word might be a plausible alternative to a morphological account. So in 4.10 what might have happened is not a morpheme shift, but a misplacement of the /əd/ sequence of phonemes corresponding to the ed spelling. However, this description is less convincing than one in terms of a coherent unit of word structure, the morpheme. In addition, there are parallel cases, like 4.11, where what looks like the shift of a sound is in fact not. This is because although in the written version of this speech error an s has been moved, in the intended utterance what would have been a /z/ pronunciation in goes, becomes a /s/ pronunciation in backs. This is known as accommodation, i.e. the allomorph of the morpheme third person singular present that occurs in the error is appropriate to the word on which it occurs, and not to the word on which it should have occurred. What this suggests is that such errors do indeed involve the misplacement of a morpheme, rather than of a speech sound, and that subsequent processes specify the appropriate allomorph for the new context.
Derivational morphology
Much of the inflectional morphology involved in the errors above is predictable by rule, and so it is not surprising to find evidence for productive word building for these forms. A different situation arises with English derivational morphology, which is much less predictable. Derivational morphology involves the construction of new words from base forms. (By contrast, a plural form cats is not a new word, in the sense of having a separate lexical entry in a dictionary, but a different grammatical form of the word cat. For example, derivational is from derivation which is in turn from derive. In English, derivational morphology can involve suffixes, such as the -tion and -al endings in the derive / derivation / derivational pattern, the -hood in Nationhood or the -ness in goodness. It can also involve prefixes, such as the -un in unkind or the -in in inadequate. Both of these prefixes -un and -in, indicate a negative meaning, but note that there is no rule that tells speakers that they should use -un in one case and -in in the other. The use of one or the other is something we have to learn. We can test that this is the case by asking people to make a negative version of a nonword. So, we might get them to give a negative form of the invented word armotic. It is likely that some of our informants will say unarmotic, while others might say inarmotic, or even non-armotic. This outcome contrasts with the unanimous response we would expect to receive for a plural of the non-word bafflack.
Errors involving derivational prefixes have also been taken to indicate that morphological structure is represented in the production lexicon. Examples 4.12 to 4.14 all involve self-correction, and so we know what the speaker intended. From the evidence, it seems that the wrong prefix has been added to a base. However, it is unlikely that nave native speakers would for instance know that the target word assert in 4.12 is made up of as+sert, though this may be historically accurate the form is from Latin as-serere, in turn from ad ‘to’ + serere to join, put’ . An alternative explanation is that these errors are a particular type of malapropism see Chapter 3, i.e. that the morphological similarity is coincidental, and that the overall sound similarity is what is important. In 4.13 there could also be phonological interference from the beginning sounds of the preceding and following words, director and development.
Despite these misgivings, the pattern in such putative prefix errors does seem to be that one prefixed form is replaced by another. That is, prefixes are not replaced by non-prefixes (so we do not find review/maview). Also, we do not find substitutions involving two non-prefixed words which overlap after the first syllable (so we do not find kitten/gluten). So, the errors do not simply involve words with sound overlap after the first syllable. This suggests that prefixes may be marked as such in the lexicon, even if they may not be involved in productive rules during word building.
A set of prefix errors that has drawn particular attention involves a shift of a negative prefix, as in 4.15 and 4.16. These errors provide further support for the argument that negative prefixes are stored in an abstract form NEG in the mental lexicon, so that imprecise, disregard, under, nothing are stored as NEG + stem. At the functional level, the abstract form of the target sentence in 4.15 would include a NEG element linked to the object clause i.e. the clause that forms the object of the verb regard. In fleshing out the utterance during positional processing, the NEG gets misplaced, and as a result modifies the verb in the main clause. So instead of modifying precise (giving imprecise), it ends up modifying regard (giving disregard). In the error it is clearly the abstract NEG form that is mis placed, rather than the exact phonological form that would have been attached to the target stem, which would have produced the nonword imregard.
Notice that there is a model for this shift of negative elements between main clauses and object clauses in English, since there is an equivalence between the two sentences in 4.17 and 4.18.
Moving on to suffixes, it has been noted that errors involving the final parts of words – either as the moved elements or as stranded elements – rarely involve derivational affixes. So examples like 4.19–4.24 are far less common than errors such as 4.25, which involves sound sequences which are not morphemes, i.e. –(u)nction and –(u)cture. But note that some of the errors that appear to involve suffixes have plausible alternative or additional explanations. In 4.19 the error could be an anticipation of rent as a syllable rather than as an affix, and similarly the exchanges in 4.20 and 4.21 might be phonological a syllable rather than morpho logical. Indeed, 4.21 is best explained this way, since the -le in s not correspond to a morpheme.
Productivity
Some affix errors involve both inflectional and derivational endings, as in 4.26 and 4.27. The stranded elements in example 4.26 are the plural inflection s which accommodates to /z/ when attached to the stem freeze and the instrumental derivational affix -er. In 4.27 the stems nose and model are exchanged, stranding both the inflectional suffix -ed and the derivational prefix re-.
Errors like these suggest that the distinction between inflectional and derivational endings in English is not as clear as it has often been made out to be in linguistics. In fact, many linguists see the distinction as a false one, and that morphological processes sit on a continuum that might include purely predictable inflectional endings at one extreme and totally unpredictable derivational endings at the other.
A notion that is important in this context is productivity Bauer, 2001 . Productive affixes are the affixes that are most likely to be used on novel words, such as in the wug test described above, or when a new word is coined in the language and inflected or derived forms are based on this. The more productive an affix is, the more predictable it tends to be. The contention here is that it is also more easily separated from its stem, and
that such separability does not discriminate between inflection and derivation. For instance, the instrumental -er ending in 4.26 would count amongst the more productive of the derivational suffixes in English – if we were to invent a new verb, to spling, and asked someone to name an object that you might use to perform the action conveyed by that verb, there is a high probability that they will say it is a splinger. It is therefore small wonder that this affix – which can be so easily added – is also easily separated from its stem in speech errors.
Morphology and lexical stress errors
Lexical stress errors are errors where the correct word has been produced, but with the wrong stress pattern. Since English has many cases of morphologically related words that have differences in their stress pattern, it is interesting to note whether the incorrect stress pattern is in fact the correct stress pattern for a related word. If this is the case, then it could imply that there are connections between morphologically related words in the mental lexicon, and that during the process of selecting the target word there has been some interference from a linked word. Examples 4.28 to 4.32 illustrate stress errors. The morphologically related word with stress in the error position is shown in parentheses after each example, except for 4.32, which was self-corrected by the speaker, a professor of linguistics.
In these errors, the syntactic category specification for the intended word leads to the appropriate ending, but in each case the stress is placed on a syllable which is stressed in a morphologically related word, suggesting confusion between close lexical entries.
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
