Thesis (Ph. D.)--University of Rochester. Dept. of Brain & Cognitive Sciences, 2010.
When linguistic input provides inconsistent evidence for grammatical
structures, children tend to regularize. For example, children learning languages from
parents who are imperfect users of the language regularize their parents’ inconsistent
usages (Singleton & Newport, 2004). Previous studies (Hudson Kam & Newport,
2005, 2009) have examined this phenomenon using artificial languages and have
found that children, but not adults, regularized this type of variable input. However,
little is known about precisely how and when such regularization occurs.
In a series of 5 experiments, we investigate how various aspects of the
distribution of the input probabilities affect such behavior, as well as whether this
behavior varies by age. Adults and children were exposed for 5 days to one of five
artificial languages. Languages differed in the probability with which determiners
were used. After exposure, subjects produced novel sentences; determiner use was
scored.
The first two experiments investigated the role of the type of variability
present in the input. When taught a language with unpredictable, inconsistent
variability, five- and six- year old children regularized the variability in their own
language productions, always using the most frequent determiner more often than
occurred during exposure. When taught a language with predictable, lexically consistent
variability, children did not regularize, but instead learned the pattern responsible for the variability, and used this pattern in their own productions. In both
cases, adult’s productions matched those present in their input.
The second set of experiments investigated which feature of the distribution
children are sensitive to, as well as whether this effect varies by age. Adults and 5-8
year old children were exposed to one of three artificial languages, differing in the
distribution of determiners. Each language had a dominant determiner that was used
more frequently than any of the other determiners, and a minority determiner that was
used less often. The probability with which these determiners were used, was
systematically varied such that a given probability of usage (40%) was the dominant
determiner in one language, but the minority determiner in the other.
In line with previous results, the 5-6 year old children in these experiments
regularized the dominant determiner in their productions; adult productions reflected
the probabilities present in their input. Children ages 7-8 regularized less than
younger children, but more than adults. This suggests that the tendency to regularize
begins to drop off at an early age, but does so gradually.
Moreover, though a determiner is used equally frequently across two
languages (40%), children regularize it only when it is the dominant determiner. It is
thus the relationships between alternating forms – and not simple frequency – that
affects the youngest learners.