10 Decoding: sounds, syllables and words
10.4 Processing words
THIS is the DOG that CHASED the CAT that KILLED the RAT that LIVED in the HOUSE that JACK BUILT.
BOX of MATCHes WAITed at HOME WENT to the BANK LOOKING at the SKY GUESSes the TIME PLACes in SPAIN WANTed to KNOW PIECes of CAKE RULES for DRIVers The GLASS is BROKen.
determine where in a piece of continuous speech one word ends and the next begins.
The operation, known as lexical segmentation, might seem a straight-forward task if one has the vocabulary knowledge. Surely all a lis-tener has to do is to match a group of sounds to a word. Once the match has been made, she can locate the end of the word and go on to seek a match for the next group of sounds. In fact, the process is not so easy. Suppose an L1 listener hears the word catalogue. The simplest matching routine would suggest that it consists of not one word but three: cat + a + log. Or consider a longer sequence such as The captain’s in cabin eight. Possible matches that a listener might make include (beside the correct ones): cap, capped, tins, zinc, cab, in, innate.
One easy solution to this question suggests that the listener is capable of making use of context and co-text (including grammar) to resolve any ambiguities of the type described. Yes, indeed; but as we saw in Chapter 8, employing contextual and co-textual cues is both slow and demanding upon mental resources. The listener needs some other, more automatic, routine which enables her to make rapid matches with a low likelihood of failure.
Several theories have been put forward about the techniques that L1 listeners employ to locate word boundaries. The most convincing holds that listeners exploit the phonology of the language in ques-tion. For English, it appears that native listeners employ a metrical segmentation strategy (Cutler, 1990). In segmenting the speech stream into possible words, they are guided by the assumption that each stressed4 syllable marks a likely beginning for a word. The strategy makes sense because it has been calculated that somebody listening to English has a 90.2% chance of being right if she operates on the basis that each stressed syllable marks the beginning of a new content word.5
Another way of locating word boundaries lies in detecting frequently occurring syllables in the form of prefixes (marking the beginning of a word) and suffixes (marking the end). Prefixes and suffixes, like func-tion words, are often of very low prominence; but it seems likely that
4 The terms stressed and unstressed are used here, but this involves a deliberate sim-plification. Cutler (1990) refers to a strong syllable, meaning one with a full-quality vowel as against a weak syllable with a weak-quality vowel. Other commentators on segmentation, however, make the distinction on the basis of stress.
5 This figure allows for the fact that many content words consist of one syllable.
It also allows for the relative frequency of words in a typical piece of connected speech.
L1 listeners become adept at recognising them because they are so frequent.
Second language practice
Evidence exists (Field, 2001b) of French intermediate learners adopt-ing an English-like segmentation strategy when transcribadopt-ing short pieces of speech recorded by an English speaker. Nobody had trained these learners in segmenting speech; they just showed themselves sensitive to rhythmic regularities in the target language. They had learnt from expe-rience the value of inserting word boundaries before stressed syllables, without realising that learning had taken place. This strongly suggests that training learners in segmentation would pay dividends. It might ensure that they acquire the requisite techniques more rapidly than if left to their own devices. It might also alleviate some of the early frustration of the novice listener at the apparent impenetrability of connected speech.
Table 10.6 gives examples of the kinds of exercise that can be used to practise stressed-based lexical segmentation like that of English.
It has to be said that, for certain languages, segmentation does not pose a major problem. These are languages where the words carry a regular stress (on the first, penultimate or final syllable), providing the listener with a reliable and consistent guide to where each one begins or ends. Nevertheless, it still makes sense to demonstrate to learners how to use the appropriate cues when locating word boundaries.
With languages that do not have this kind of fixed stress, teachers should bear in mind that a lexical segmentation strategy represents a rule-of-thumb technique: one that succeeds in many instances but that may not always do so. A first attempt may have to be revised if it does not give rise to a group of phonemes that match a word. So, while it is important to ensure that L2 listeners are given practice in the strategy appropriate to the language being acquired, it is also important to ensure that they treat segmentation outcomes as provisional and are prepared to adjust their first hypotheses.
A second type of segmentation practice relevant to many languages draws attention to weak syllables that potentially represent common prefixes or suffixes. It is necessary to train the L2 listener to recognise them as word-boundary markers because of their low prominence in connected speech. The exercise carries an additional benefit when L2 listeners encounter an unknown word in a piece of spoken input. While they may find it difficult to recall the phonological form of the word or to work out its meaning, they will at least be able to identify its word class from the presence of a familiar suffix. This may help in fitting the unknown word into a larger grammatical structure. See Table 10.7.
Standard segmentation practice. Teacher plays a sentence excised from a piece of natural speech. Learners transcribe the words they understand. Teacher replays, learners add more words to their transcriptions. Learners compare answers, teacher replays. Repeat with several sentences. Ideal also for self-study.
Gap filling. ‘Listen and fill in the missing words.’ Teacher gives learners a transcript, in which groups of words (not just single words) have been omitted.
Awareness raising. ‘Write what you hear.’ Teacher dictates ambiguous sequences to the learners; then adds an unexpected ending.
a nice cream. . . dress [learner writes ‘an ice cream’, then has to revise it]
The way to cut it. . . is like this. [‘the waiter cut it’ → ‘the way to cut it’]
Some boxes have. . . arrived. [‘some boxes of’ → ‘some boxes have’]
I want to drive a. . . train. [‘a driver’ → ‘to drive a’]
Ear-training. Practice in recognising the specific feature in L2 that assists lexical segmentation. If the feature is word stress, the teacher dictates short sentences and learners mark on a
transcript the syllables which carry stress. Teacher draws attention to the role stress plays, e.g. in English, the fact that most words are monosyllabic or carry stress on the first syllable.
Segmentation of strings. ‘You will hear one, two or three new words. Try to write them down.’ Teacher dictates sequences containing non-words whose boundaries are marked by the segmentation system used in L2. Teacher reviews answers and explains the L2 segmentation system. Examples for the English strategy might be:
TANdion HEAKal LECKnif KEDGern PIFEtal KIFFnoke SAMPren TINGlort
Sample sentences. ‘You will not understand everything. But try to guess how many words there are.’ Teacher reads aloud a sequence containing unfamiliar words whose boundaries are marked by the segmentation system used in L2. Teacher replays and learners attempt to write them down. Teacher explains the L2 segmentation system. For English, take examples of the rhythmic pattern from poetry:
Dirty British coaster with a salt-caked smokestack.
Table 10.7 Prefix and suffix exercises
Affix spotting. ‘What is a prefix? What is a suffix? Give examples. You will not understand many of the words in this recording. Just write down each prefix or suffix that you hear.’
Teacher plays a short section of authentic speech beyond the learners’ language level. Teacher later replays the text, pointing out the value of affixes as word-boundary markers.
Suffix awareness. ‘Listen to the words I say. Then listen again and write down any syllable which occurs twice.’
government, happiness, disappear, careless, excitement, revisit, blindness, disappointed, rebuild, hopeless
Cued affixes. ‘Read aloud the prefixes and suffixes written on the board. Listen to the recording and count how often each one occurs.’ A useful source of material is Longman Dictionary of Contemporary English, Appendix B7.
Affix segmentation. Teacher plays a short extract (three or four sentences) from an authentic recording beyond the level of the learners. Learners are asked to listen for prefixes and suffixes.
Teacher replays the recording, and learners try to transcribe only the words beginning with prefixes or ending with suffixes.
Suffixes and word class. ‘Listen and write down these words.
Guess their meanings, then write a sentence containing them.’
Teacher dictates unknown words which contain known words as their stems.
washable postage imagination musical strongly ending smoker modernise replacement mountainous loudness sleepy warmth
10.4.2 Activation and automatic processes First language processing
A picture was painted in Section 10.1 above of a listening process that draws upon multiple pieces of information, including evidence from the speech signal (the sounds and syllables that have been detected), vocabulary knowledge of the existence of a range of words, and evidence from co-text and context. In making a word match, the listener draws upon all these sources, weighing one against another. But the weighing
of evidence also receives support from other processes that are so well established that a competent listener is not even fully aware of them. They derive from the wide vocabulary base that a native speaker possesses, and from many years’ experience of hearing the spoken forms of words used in everyday encounters. The most important are:
Frequency. Associated with the vocabulary store is an internalised knowledge of how frequent different words are. This affects the way in which an experienced listener processes the input, lending extra weight to interpretations that favour common words.
Current activation. When a word has been heard quite recently, it remains partly activated in the mind of the listener, assisting her if and when she hears it again.
Spreading activation. The vocabulary of somebody who knows a lan-guage well draws upon elaborate connections between word meanings as well as word forms. When a listener hears a word that is part of one of these networks, it activates a number of other words that are closely associated with it. On hearing DOCTOR, the words patient, hospital, etc. become foregrounded, so that the listener recognises them more easily if and when they occur.
Second language practice
These three sources of support are available to most second language listeners only to a limited degree. An awareness of the relative frequency of words develops only through extended exposure to a language. It represents one of the obvious benefits of the comprehension approach and of the kind of individualised practice recommended in Chapter 4, but it is clearly a gradual process. Current activation depends critically upon the listener’s ability to recognise examples of individual words with a fair degree of confidence. The novice second language listener is acutely aware of the approximate nature of the matches that she makes at phoneme and word level, which does not provide a very strong basis for identifying a word that might recur.
Most limiting of all is the fact that most L2 listeners have a much more rudimentary knowledge of vocabulary than their L1 listener coun-terparts. Their word store is in a transitional state, with gaps for words not yet acquired and with connections between words still in the process of being established (Meara, 1997). In the early stages of acquiring a second language, this situation denies the listener much of the support that an L1 listener draws from spreading activation. Part of the solution is provided, of course, by the vocabulary teacher – and in particular by the tradition of introducing words in lexical sets which emphasise the
conceptual links between them. But the listening teacher also has a tech-nique available in the long-established practice of asking learners to pre-dict the vocabulary content of a passage that they are about to hear and then to check which of the items actually occur. This exercise demands a much more conscious process than spreading activation, but nevertheless one that introduces word association into the way the target language is handled. Other possible exercise types are shown in Table 10.8.
Table 10.8 Exercises to promote spreading activation
Word association. ‘I will say a word and point to one of you. As fast as possible, say a word that is connected to it in meaning.’
Teacher calls out a range of words (nouns, verbs, adjectives) within the learner’s vocabulary.
Anticipating words. ‘You are going to hear a recording about X.
What words do you expect to hear?’ Teacher writes the words on the board. Teacher plays the recording and learners put up their hands every time they hear one of the predicted words.
Online activation. ‘You have just heard part of a recording.
What were the important words? Think of other words that you might expect to hear in the next part of the recording. Now listen to see if you were right.’
10.4.3 A note on meaning
A perceptive reader may have noticed that the account given here has steered round the question of word meaning. One reason is that the emphasis in this chapter has been upon how listeners manage to form auditory matches. Most current accounts of listening assume a very close connection between recognising a spoken word and gaining access to its meaning. In a process known as lexical access, a listener opens up a mental record of a word as soon as the word has been identified as a possible match for what has been heard. But the exact sense of the word cannot be fully ascertained without taking account of the context within which it occurs. We will therefore look in detail at issues of word meaning in Chapter 12.