Adele E Goldberg How argument structure constructions are
Adele E. Goldberg. How argument structure constructions are learned.
l Children have the ability to extract word forms from continuous speech. This observation is based on transitional probabilities between syllables. l 1. Bananas with milk. l This phrase contains four transitional probabilities across syllables: ba to na; na to nas; nas to with; with to milk.
l The chance that ba will be followed by na, and that (ba)na will be followed by nas is obviously higher than the probability that nas will be followed by with. l Generally, transitional probabilities are higher within words than across words.
l These basic learning abilities are neither unique to humans nor language specific. Hauser, Newport, and Aslin (2001) discovered that cotton-top tamarin monkeys also track transitional probabilities, and are able to notice word boundaries.
l Children have the ability to discover syntactic regularities between categories of words as well as the regularities in sound patterns. l The presence of an article like the or a predicts that a noun is appearing soon. Here learners can use this cue to find out syntactic phrase boundaries.
l Children are able to combine word level and syntactic level information. l Twelve months old children can use the discovered word boundaries to find out regularities in the syntactic distributions.
l It can be argued now that the association of meanings with particular forms is biologically determined. l According to this view, the input for a child is not ‘rich’ enough in order to be able to learn the relevant generalizations.
l Thus, proponents of the nativist view believe that language is genetically encoded in the child. l There exists a special language device which is inherent in all humans.
l Goldberg presents evidence that at least certain patterns in language are learnable.
l How can children learn form and meaning correspondences on the basis of their input?
l Table 4. 1, page 73.
l Some researchers argue that children already have specific linking rules at a young age. This view implies that the linking rules are innate and not learned from the input. l Pinker (1989: 248) defines linking rules as ‘near-universal in their essential aspects and therefore (they) may not be learned at all’. l Universal linking rules include the mapping of agent to subject, patient to object, and goal to oblique.
l Proponents of the other side emphasized the conservative nature of the early learning of children. They demonstrated that children fail to generalize beyond their input until they have been exposed to an enormous amount of data at the age of 3. 5 or older. l Here the implication is that constructions must be learned, because they are acquired rather late and in a gradual way.
l Kaschak and Glenberg (2004) have investigated the online processing of the construction This shirt needs washed of adults. l They found that the speakers were able to read instances of this construction more fluently after hearing or reading other instances of the construction.
l Facilitation was also found for the same pattern involving the verb wants instead of needs, but after training on the original construction. l Thus the facilitation effect transferred to a related verb. The increased fluency, measured by shorter reading times, indicates that speakers learned to understand the construction.
l The fact that the participants demonstrated increased reading times for semantically inconsistent follow-up sentences, even in the initial testing trials, suggests that they were able right from the beginning to comprehend the construction. l The study demonstrates that experience with a construction can be acquired quickly.
l Hudson & Newport (1999) showed in their study the overgeneralizations adult second language learners perform. l The result of their study indicates that when aspects of distribution are not predictable, the learners tend to generalize the more regular aspects.
l Goldberg et alli. (2004) examined the early speech of children in order to investigate the input children receive in more detail.
l Analyzing the speech of mothers, they found a strong tendency of one verb occurring with very high frequency compared to other verbs used in the constructions. l The verb go for example accounts for thirty-nine per cent of the uses of the intransitive motion construction in the speech of mothers talking to their twenty-eight months old children.
l The high percentage is significant because this construction is used with thirty-nine different verbs in the speech of mothers in the corpus. l Table 4. 2, page 76.
l The same patterns which are noted in the speech of mothers to children are mirrored in their early speech.
l Why do verbs like go, put, and give appear so frequently in the input that children receive?
l After comparing the verbs go and amble, or put and shelve, it is obvious that go and put are used more frequently. l The reason is that they apply to a wider range of arguments and therefore they are relevant in a larger variety of contexts.
l Additionally, each of these verbs designates a basic pattern of experience. l Someone causing someone to receive something (give) Something causing something to move (put) Something moving (go) l l l These meanings involve concrete actions, so they are readily accessible to the child.
l Obviously, the verb meanings must be accessible as well as occur highly frequent in the input children receive in order to be frequently produced in their early language.
l Table 4. 3, page 78.
l The generality of the meanings of these verbs as well as their highly frequent and early appearance in the speech of children suggests that they could help children generalizing patterns from the input.
l Ninio (1999) suggested that also syntactic patterns can emerge from the generalization of certain verbs. l She observed that children often start to use a specific verb with a direct object before a direct object appears with other verbs. l Also, these verbs are mostly part of the set of general purpose verbs.
l Ninio observed that SVO and VO patterns are at first produced with only one or a few verbs for a longer period. l After that more verbs are used in an exponentially increasing way.
l This increase could be explained by the fact that children gradually abstract a more general syntactic pattern based on the early verbs. l Therefore the growing generalization allows them to use new verbs in this syntactic pattern more easily.
l An increase in vocabulary is connected with an increase in the ‘power’ of the generalization. l It becomes progressively easier to assimilate new verbs into the patterns.
l The current assumption is that the high frequency of certain verbs in certain constructions facilitates children to unconsciously establish a correlation between the meaning of a particular verb in a construction and the construction itself. l This will result in an association between form and meaning.
- Slides: 32