Producing Speech after It Is Planned
The abstract phonetic representation of the speaker’s sentence is sent to the central motor areas of the brain, where it is converted into instructions to the vocal tract to produce the required sounds. Speaking is an incredibly complex motor activity, involving over 100 muscles moving in precise synchrony to produce speech at a rate of 10 to 15 phonetic units per second (Liberman et al. 1967). During silence, the amount of time needed for inhaling is about the same as for exhaling. Respirationduring speech is very different: the time for inhaling is drastically reduced, sometimes to less than half a second, and much more time is spent exhaling, sometimes up to several seconds. During speech, air from the lungs must be released with exactly the correct pressure. The respiratory system works with the muscles of the larynx to control the rate of vibration of the vocal folds, providing the necessary variations in pitch, loudness, and duration for the segmental (consonants and vowels) and suprasegmental (prosody) content of the utterance.
SEMINARII
Tasks:
1. Read the text “The Hearer: Speech Perception and Lexical Access” and give definitions of the terms in bold.
2. Describe the process of the lexical decision task.
3. How does the frequency and ambiguity of lexical items affect subjects’ performance on a lexical decision task? Do these variables have the same effect when words are processed in sentences?
4. Give your own examples of interlingual homophones and homographs.
5. Prepare 5 more questionson the text to check the understanding of it by your groupmates.
THE HEARER: SPEECH PERCEPTION AND LEXICAL ACCESS*
Accessing the Lexicon
The speaker enters the lexicon using information about meaning so she can retrieve the phonological structure of the appropriate words to convey the meaning she is constructing for a sentence. The hearer’s (or reader’s) task is the opposite. He uses a phonological representation (decoded using information from the acoustic signal) to retrieve information about meaning. The hearer looks for a lexical entry whose phonological representation matches the one he has heard. When there is a match, a word is retrieved, and information about the word’s meaning and structural requirements is then available. The speed of lexical retrieval is remarkable – it takes a mere fraction of a second to find a word in a lexicon consisting of some 80,000 items. The lexicon is searched by meanings in production and by phonological forms in perception. Evidence about both the process of retrieval and the way the lexicon is organized is provided by studies that examine how lexical access is affected by meaning and form relations among words, as well by variables such as phonotactics, word frequency, and lexical ambiguity.
|
|
|
A technique widely used to investigate lexical access is the lexical decision task. Participants are briefly shown a string of letters and asked to push one button if the letters constitute a word in their language, and a different button if they do not. Responses in a lexical decision task tend to be very rapid, ranging between 400 and 600 milliseconds. In a lexical decision experiment, participants will see equal amounts of words and non-words, and within the many words they will see throughout the experiment, a subset of those is of interest to the investigator: those words contain a contrast being investigated in the experiment.
To simulate how a lexical decision task works, consider the 16 letter strings in Table 1, and write Y or N next to each one, to indicate for each whether it is a word of English. Try to write your responses as quickly as possible.
Table 1. Word list for simulated lexical decision task. For each string, write Y if it is a word of English, N if it is not.
| CLOCK | DOCTOR | ZNER | FLOOP |
| SKERN | NURSE | ABLE | FABLE |
| BANK | TLAT | URN | MROCK |
| MOTHER | PLIM | HUT | BAT |
You probably wrote N next to six of the letter strings, and might have even noticed that you responded to three of them very quickly – TLAT, ZNER, and MROCK – and to the other three somewhat more slowly – SKERN, PLIM, and FLOOP. All six strings are non-words in English, but the first three violate the phonotactic constraints of the language. Impossible non-words, like TLAT, ZNER, and MROCK, are rejected very rapidly in a lexical decision task. It is as if the lexical retrieval system were carrying out a phonological screening of sorts, not bothering to look in the lexicon when the string is not a possible word in the language. In contrast, possible non-words, like SKERN, PLIM, and FLOOP, take longer to reject, as if the retrieval system conducted an exhaustive, ultimately unsuccessful, search for their entries in the lexicon.
|
|
|
Of the real words in Table 6.1, you probably responded faster to the more frequent ones (like CLOCK and BANK) than to the less frequent ones (like HUT and URN). The lexical frequency of a word can be measured by counting how many times a particular word occurs in a large corpus for that language. Lexical frequency is correlated with lexical decision times and with responses to other types of lexical access tasks: more frequent words are responded to faster (Forster and Chambers 1973; Forster 1981). Words that are used often are evidently more available to the lexical retrieval system.
Another property of words that has been used to study lexical retrieval is lexical ambiguity. Lexically ambiguous words are words that have more than one meaning. Some research has examined whether such words have more than one lexical entry, and whether having more than one lexical entry can lead to retrieval advantages. Consider the word bank, which as a noun can be a money bank, a river bank, or a snow bank; bank can also be a verb. Some lexically ambiguous words have multiple meanings that are completely unrelated (e.g., the noun punch can refer to a type of drink, or to a blow with the fist, or to a piercing instrument); such ambiguous words are called homonyms. Other ambiguous words have meanings that appear to have a systematic relationship to each other (e.g., the noun eye refers to an organ used for vision, or to the opening in a needle, or the aperture of a camera); these words are polysemous. Rodd, Gaskell, and Marslen-Wilson (2002) compared these two types of ambiguity in a series of lexical decision experiments, and found that ambiguous words with related senses (polysemous words like eye) are retrieved faster than ambiguous words with unrelated senses (homonyms like punch). Homonyms have multiple meanings that compete against each other, resulting in delayed recognition. In contrast, the semantic relationships between the multiple senses of polysemous words facilitate their retrieval.
|
|
|
One final variable we will discuss affecting lexical access routines is priming (Meyer and Schvaneveldt 1971). Priming is actually a very general property of human cognition: a stimulus you just experienced will affect how you respond to a later stimulus – and this associative response is true not just with linguistic stimuli, but with stimuli of any type (pictures, smells, non-linguistic sounds, etc.). In the list in Table 1, the words DOCTOR and NURSE are related semantically, and the words TABLE and FABLE related phonologically. Reading the words in each pair consecutively might have influenced how quickly you responded to the second member of the pair.
How does priming work? When you encounter a stimulus of a given type, you activate its mental representation, but as you search for the unique mental representation for the stimulus, you activate associates for that stimulus, as well. Priming, then, is residual activation from previously experienced stimuli.
In a lexical decision experiment concerned with measuring priming effects, a prime word is presented for a brief amount of time; it then disappears and a target word takes its place; Figure 6.6 illustrates this graphically. (In many priming experiments, primes are presented in small letters, while targets are presented in capitals, and participants are asked to make lexical decisions only on words presented in capital letters.) The experiment includes primes that are related to the target (e.g., for a target like DOCTOR, a related prime would be “nurse”), as well as primes that are unrelated to the target. Responses to the target will be faster when it is preceded by a related than by an unrelated prime.
Many studies have used semantic priming techniques to study to what extent semantic representations are shared between translation equivalent words in bilingual lexicons. This research has confirmed that when a prime and a target are in different languages, a semantic relation between them facilitates retrieval of the target word; for a French-English bilingual, access to cat is facilitated by both dog and chien (Kroll and Sunderman 2003). The strength of the priming can be asymmetric: priming is typically stronger from the bilingual’s dominant language (which is usually, though not always, the bilingual’s first language) than from the non-dominant language. The idea is that the dominant (or first language) lexicon is bigger, since it was learned first, making the links to (non-linguistic) conceptual representations stronger for words in the dominant language than in the non-dominant language (Kroll and Dijkstra 2002).
|
|
|
Sometimes words that have the same (or very similar) form between two languages are not related semantically at all. A pair of languages can have interlingual homographs (words that are written the same way between the two languages), like coin in English and coin (‘corner’) in French, as well as interlingual homophones (words that sound the same in the two languages), like aid in English and eed (‘oath’) in Dutch. Notice that these two examples are pairs of words that are not translation equivalent, but rather interlingual “false friends.” False friends have become useful in research that examines to what extent bilinguals are able to inhibit one language while retrieving lexical entries in a unilingual mode. In research like this, bilingual participants perform lexical decisions in one language only, and the experiment compares reaction times to interlingual homographs and frequency-matched controls, for example, on the assumption that false friends will result in processing cost if the other language is not inhibited. Generally, studies such as this have found that interlingual homographs take as long to process as control words, suggesting that the bilinguals’ other language is inhibited during unilingual processing; however, other experiments using priming techniques (discussed in more detail in the following section) have demonstrated that words with orthographic and semantic overlap in the two languages can affect processing time (Dijkstra 2005). One such study (Beauvillain and Grainger 1987) asked participants to make a lexical decision on word pairs consisting of one word in French (like coin) followed by a word in English (like money). Beauvillain and Grainger found that when the prime and target were semantically related, in English, reaction times on the target word (money) were faster, suggesting that even though the prime had been accessed in French, the corresponding English lexical representation had been activated as well.
SEMINAR III
Tasks:
1. Read the text “First Language Acquisition”and give definitions of the terms in bold.
2. Ask your parents about the stages of language acquisition you went through. Share your experience.
3. Prepare 5 more questionson the text to check the understanding of it by your groupmates.
FIRST LANGUAGE ACQUISITION*
Дата добавления: 2018-06-01; просмотров: 784; Мы поможем в написании вашей работы! |
Мы поможем в написании ваших работ!