BOOTSTRAPPING
المؤلف:
John Field
المصدر:
Psycholinguistics
الجزء والصفحة:
P38
2025-08-02
444
BOOTSTRAPPING
A prelinguistic infant has no lexicon against which to match the sound sequences encountered in the speech signal. Furthermore, connected speech provides few cues to where word boundaries lie. It is therefore difficult to explain how the language-acquiring infant comes to identify word forms and to map them on to meanings relating to the real world. It has been suggested that the infant can only achieve this task by relying on some kind of technique which gives it a head start just as straps can help one to pull on a pair of boots (the metaphor comes via computer science). This technique might be specific to the process of language acquisition or it might be the product of general cognition, reflecting, for example, a predisposition to impose patterns upon diverse information.
main types of bootstrapping have been proposed:
In prosodic bootstrapping (Cutler and Mehler, 1993), the infant exploits rhythmic regularities in the language it is acquiring. At the phoneme level, it can distinguish a difference between steady-state sequences representing full vowels and transitional sequences representing consonants. It is thus sensitive to syllable structure. From this and from its innate sense of rhythm, the infant acquiring English is able to recognise the difference between longer stressed syllables featuring full vowels and shorter unstressed syllables featuring weak quality vowels. It may be that the infant develops a metrical template (Gerken, 1994) which reflects the tendency of English towards an SW (strong-weak) rhythmic unit. The template encourages the child to seek words which follow an SW pattern, and provides it with the working hypothesis that a stressed syllable in the signal is likely to mark a word onset. This accounts for the following versions of adult words:
It also accounts for evidence of children joining words to form an SW pattern as in: I like-it the elephant.
The concept of prosodic bootstrapping has been applied to larger constituents than the word. It is suggested that infants learn to recognise intonation patterns (especially the placing of the tonic accent) and the regular occurrence of pauses. These features, which are often heightened in Child directed speech, provide infants with cues to phrase boundaries and to the structure of typical phrases.
Syntactic bootstrapping (Gleitman, 1990) assumes that an infant uses surface form to establish syntactic categories. The early mapping process draws upon an assumption (innate or learnt) that there is a word-class which relates to objects in the real world, one which relates to actions and one which relates to attributes. Once this is established, the infant can add less prototypical items to each class (abstract nouns, state verbs) by noticing that they share grammatical properties with words that have already been acquired: in particular, their morphology and their distribution.
It learns to associate count nouns with the frame It’s a... and mass nouns with the frame It’s.... Experiments with non-words (It’s a sib, It’s sib) have demonstrated that infants are capable of making this association as early as 17 months. Infants are also capable of using formal evidence to recognise that non-words like nissing refer to a potential action and non-words like a niss refer to a potential object.
Later on, infants may use syntactic structure to establish distinctions of meaning. Thus, they can distinguish the senses of the words eat and feed by their distribution: eat occurring in the structure Verb þ Noun [edible] while feed occurs in the structure Verb þ Noun [animate]. Among evidence cited in support of syntactic bootstrapping is the fact that blind infants manage to acquire the words see and look without difficulty. The suggestion is that they are able to do so by relating the words to the contexts in which they occur, even though they lack a concept to which to attach them.
Semantic bootstrapping (Pinker, 1994a) hypothesises the reverse process: that infants use their world knowledge in order to recognise syntactic relationships within sentences. Assume an infant has acquired, in isolation, the nouns rabbit and duck. Presented with a sentence such as The rabbit is chasing the duck and evidence from a cartoon film, the infant comes to recognise that the position of the word rabbit in the sentence is reserved for the Agent or syntactic subject and the position of the word duck is reserved for the Patient or syntactic direct object. The assumptions would be confirmed if the cartoon film later showed the reverse situation and the associated sentence was The duck is chasing the rabbit.
As formulated by Pinker, semantic bootstrapping also incorporates the assumption that certain linguistic concepts are innate in the infant: these include the notions of noun and verb as word classes and the notions of agent and patient as roles.
Other bootstrapping theories are:
Perceptual bootstrapping (Nusbaum and Goodman, 1994): where the infant focuses its attention on the most salient parts of the input; this might explain why early utterances do not usually contain weakly stressed function words.
‘Logical bootstrapping’ (Bates and Goodman, 1999): a process whereby an infant systematically directs its attention first to physical objects (nouns), then to events and relationships between the objects (verbs and adjectives) and then to word order and syntax. This step-by-step building of meaning reflects the general pattern of vocabulary acquisition.
See also: Lexical segmentation, Operating Principles, Phonological development, SW (strong-weak) pattern, Vocabulary acquisition
Further reading: Cutler and Mehler (1993); Gleitman (1990); Peters (1983); Pinker (1994a)
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
