Lecture 22: Language Learning II
Lecture 22: Language Learning II[Note: as usual, these notes will probably not be much use unless you were also in class - they are meant as an aid to organizing what you have learned in class, rather than as a substitute for attendance.]
A couple of interesting websites related to this class...
1. Innate ability to discriminate speech sounds
e.g. Japanese babies discriminate r/l
2. "Home Sign" languages of deaf children born to hearing parents
3. Non-imitation of parents
4. Creole languages; regularization and extension of input
5. Syntactic Sophistication
Knowledge of wanna-contraction possibilities in English
Generalization: wanna-contraction is not possible when questioning the subject of the subordinate clause.
An Experiment to test wanna-contraction in children aged 3-5 years. This experimental procedure was developed by Prof. Stephen Crain (Univ. of Maryland) and his colleagues, and has been used to test many different aspects of young children's linguistic knowledge.
The first sample experimental protocol is designed to elicit questions in which the object of the subordinate clause is questioned (i.e. what is the action performed upon).
The next protocol is used to elicit questions about the subject of the subordinate clause (i.e. who or what performs the action).
0-8 months: 'Universal listener'
Testing methods:
10-12 monthsNative language listener: only phoneme contrasts in the native language are discriminated reliably, e.g. Japanese babies lose the ability to discriminate [r] and [l] (Tsushima et al. 1994), American babies lose the ability to discriminate the allophones [p] and [ph].
(Gary Marcus et al. 1998 (to appear in the journal Science in December 1998))
'Training': 2 minutes of syllables in an "ABB" pattern
Testing(a) same pattern, new words
(b) different pattern (e.g. "ABA" pattern)
Infants listen longer to the 'novel' pattern than to the pattern that they have heard before, indicating that after just 2 minutes of exposure they have made an appropriate generalization about the pattern they're being exposed to.
(Roberta Golinkoff & Kathy Hirsh-Pasek 1996)
To understand a verb-phrase, a child needs to know that the phrase involves the verb affecting the object, not just the verb and the object existing in the same scene.
14-month-olds succeed at looking longer at video screen that matches the verb phrase on the audio.
[Task involves listening to spoken sentences (played over speaker) while facing pair of video screens. One screen corresponds to the meaning of the verb phrase (e.g. a banana is being eaten), and the other screen corresponds to the component pieces of the verb phrase, but without the meaning of the verb phrase (e.g. a eating an apple, and a banana is also in the scene). Hidden observers measure the time that the infants look at each of the two screens.]
(Lynn Santelmann & Peter Jusczyk 1997)
At the bakery, everybody is often baking bread. One person is neatly mixing the flour and water together. Someone else is quickly adding salt and yeast. In the next room, a big machine is firmly kneading the dough. Another is deftly shaping the loaves for the oven. The whole place is really starting to smell great!
At the bakery, everybody can often baking bread. One person can neatly mixing the flour and water together. Someone else can quickly adding salt and yeast. In the next room, a big machine can firmly kneading the dough. Another can deftly shaping the loaves for the oven. The whole place can really starting to smell great!
Although the speech of young children (e.g. age 2 years) is typically missing many of the words which would be required to form fully grammatical sentences in the adult language, violations of rules of word order is quite rare. For example, subjects overwhelmingly precede verbs in child English, as in adult English, and objects overwhelmingly follow verbs in child English, just as in adult English. Therefore, although young children often omit subjects or objects in their speech, they nevertheless respect the basic SVO word order pattern of English.
We might expect that the children's errors of past tense marking (i.e. overregularizations) are most frequent at the earliest stages, and then become gradually less frequent as time goes on. However, this turns out not to be the case. An initial period of near perfect use of past tense forms is followed by a period in which overregularizations are more common.
- stage 1: perfection (almost)
- stage 2: overregularization (not all the time, but a good deal more frequent than before)
- stage 3: perfection (gradual transition from stage 2 to stage 3)
reflects development of rules!
One way of understanding this course of development is as follows: initially, children treat all verbs as if they are irregular verbs, whose past tense forms must be memorized. They memorize the forms of the small number of verbs that they use well, and so they make few errors. Later, they learn that there is a regular past tense form, which is supplied by a rule ... at this point, there is a tendency for this rule to be overapplied, and therefore overregularizations are observed in the children's speech. Subsequently, however, the correct division of labor between regular and irregular past tense forms is gradually established.
'Wug test': how are 'nonce' words inflected? (Jean Berko, 1957)
Another way of testing children's knowledge of morphological rules is to ask children to give the morphological forms of word-stems that they have never heard before. If they have never heard the stems before, then we can be sure that they could not have memorized the morphological forms of the words.
A famous example of this is an experiment using this technique run by Jean Berko-Gleason in the 1950s. She presented children with pictures of unfamiliar objects, such as an animal that she called a 'wug', and then showed them pictures of pairs of these objects, and asked them to say what the two objects were (i.e. 'wugs'). The children were of course quite successful at producing the plurals of nouns that they had never heard before. This procedure has become known as the 'wug-test'.
This study was conducted by Nader Al Jallad and Chandra Flint (graduate students in Linguistics) and Mary Catherine Richardson (BALS major in Speech Pathology).
Task: using a paradigm very much like Berko's 'wug-test', fill in the blanks in sentences like the following. In each case, the subject has to come up with a past tense form for an unfamiliar verb.
Subjects in the experiment
Overall Rate of Irregular forms supplied
Individual Differences in Irregularization: bars show percentages of subjects making 0-20% irregulars, 20-40% irregulars etc. Although there were some adults who produced many irregular forms, in general higher percentages of children were producing irregular forms at higher rates.
Children
Adults