LEXICAL SEGMENTATION

The division of connected speech into words, assisted by cues which indicate where word boundaries lie.

Pausing in connected speech is very variable, but occurs on a rough average about once every 12 syllables. A chunk of speech thus contains several words, with no gaps between them. In principle, some word boundaries are marked by allophonic cues (lengthening, aspiration etc.); but such cues are often not present in informal speech. Furthermore, accommodation (assimilation, elision etc.) may vary the form of a word at precisely the point where the listener needs information most– namely, at the boundary between it and the word that follows.

A number of solutions to the segmentation problem have been proposed:

A linear solution. The listener identifies a word and the end-point of that word marks the beginning of the next. Unfortunately for this view, many words are not unique by their end (few might be the beginning of future), while many others have words embedded in them (catalogue) and others still can be segmented two ways (a sister/assist her).

Aphonological solution. There is evidence that listeners make use of phonotactic cues. For example, the sequence /mgl/ does not occur in any English word, so there must be a boundary between /m/ and /gl/. However, cues of this type are relatively infrequent.

 A ‘unit of segmentation’ solution. Evidence suggests that French listeners segment the speech signal syllable by syllable. However, the syllable is not such a reliable unit in English. There is a marked difference in length and salience between weak and strong syllables (compare the two syllables in PROCEED) and some syllables have ambiguous boundaries (the /m/ in LEMON seems to belong both to lem- and to-mon).

A time-based solution. Computer models of speech recognition (see TRACE) process the message in time-slices. However, this involves a very complex procedure for matching the phonemes within each time-slice against all possible words, and for relating them to the phonemes in previous slices.

A rhythmic solution. The most convincing solution for the segmentation of English exploits the fact that the majority of content words in running English speech begin with a strong syllable. Cutler and Norris propose a Metrical Segmentation Strategy whereby a listener works on the assumption that each strong syllable initiates a new word. An initial weak syllable is assumed to be a weakly stressed function word; other weak syllables are attached to preceding strong ones as a first segmentation hypothesis. This account of English segmentation is supported by evidence from language acquisition. Infants from an early age exercise a preference for the strong–weak rhythmic alternation which characterises English speech.

Computer modelling and numerous on-line experiments have demonstrated the validity of this theory. Note that it defines the strong syllable as one with full vowel quality rather than one that bears stress; the argument is that this enables the listener to make a simple binary judgement (full vowel vs weak vowel) whereas judgement of stress is relative.

There has been considerable cross-linguistic research into lexical segmentation. The working assumption was that languages which are loosely termed ‘stress-based’ (German, Dutch) might lend themselves to a metrical strategy like that of English, while those characterised as ‘syllable-based’ (Spanish, Italian) gave rise to a syllabic strategy like the French one. However, the research findings with European languages have been contradictory. Where strong evidence has emerged is of a third strategy specific to Japanese listeners and based upon the mora, a sub-syllabic unit characteristic of Japanese. Segmentation research has also been extended to language features not specifically metrical: it has been suggested that Finnish listeners employ a combination of stress and vowel harmony, reflecting the unique characteristics of their language.

The question has also been raised of whether a listener can acquire a segmentation strategy specific to a second language and distinct from their native language routine. There is some evidence to suggest that, in these circumstances, they attempt to apply a native strategy or revert to a more generalised one.

The point should be made that lexical segmentation is not an issue for all listeners. It has been estimated that about half of the world’s languages feature fixed lexical stress which occurs exclusively on the first or last syllable of a word. This enables stress to be used demarcatively as a signal of word boundary location.

See also: Bootstrapping, Phonotactic rules, SW (strong-weak) pattern, Unit of perception

Further reading: Cutler (1990, 1996); Cutler and Mehler (1993); Grosjean and Gee (1987)

اخر الاخبار

اخبار العتبة العباسية المقدسة

المجمع العلمي للقرآن الكريم يصدر الكتاب الثاني من سلسلة يخشى

قسم الشعائر يكرّم أصحاب المواكب المشاركة في إحياء ذكرى شهادة السيدة الزهراء (عليها السلام)

جامعة العميد تستقبل طلبة المراحل الأولى المقبولين ضمن كلياتها للعام الدراسي الجديد

سماحة السيد الصافي يكرّم طلبة حفظ القرآن الكريم المتفوّقين ويؤكّد دور التدبّر في الهداية والتكامل

المزيد

الآخبار الصحية

طبيب يكشف تأثير برودة الشتاء على النوم

دون أن تعلم.. 10 عادات يومية تسرع شيخوخة جسمك

تقرير: النيكوتين خطر على القلب مهما كانت الطريقة

أيهما أفضل لصحتك.. زبدة الفول السوداني أم النوتيلا؟

المزيد

الآخبار التكنلوجية

السعودية تدشن أكبر نظام لتخزين الكهرباء بالبطاريات في العالم !

بأمر تنفيذي.. ترامب يؤكد هدف عودة الأميركيين إلى القمر في عام 2028

بعيدا عن البشر.. "قتال جوي في الفضاء" بين أميركا والصين

كشف مذهل في قاع "برمودا" !

المزيد

مواضيع ذات صلة

Problems with the dictionary view

The dictionary view

Dictionaries versus encyclopaedias

The encyclopaedic view of meaning

Embodiment and conceptual structure summary

Schematic systems

Semantic structure

Conceptual structure

Image schemas and abstract thought

Properties of image schemas

What is an image schema?

Image schemas