Kathleen Currie Hall

Notes on: Beckman, Mary E., & Edwards, Jan. (2000). Lexical frequency effects on young children's imitative productions. In Michael B. Broe & Janet B. Pierrehumbert (Eds.), Papers in Laboratory Phonology V: Acquisition and the lexicon (pp. 208-218). Cambridge: Cambridge University Press.

Summary:
In this paper, the authors describe how the lexical frequency of particular _combinations_ or sequences of sounds influences children's ability to produce those sequences, providing evidence for a model in which phonological subunits such as phonemes are emergent from the representation of larger sequences.

Details:
1. Previous research has shown that linguistic development hinges on vocabulary growth -- i.e., amount of experience with the ambient language.  The focus of this paper is to demonstrate this point experimentally for small phonological units such as the phoneme.

2. The notion of duality of patterning in human language divides language into meaningful units on the one hand and subunits on the other, which have no inherent meaning but which can usefully be put together to form meaningful units (i.e., there is not a one-to-one correspondence between every unit of sound and every unit of meaning).  The question, however, is how these non-meaningful subunits are acquired; the claim here is that they emerge from larger sequences of words as children become more familiar with overlapping sequences in different words.

3. If these generalizations really are emergent, and if children really do acquire sequences of phonemes before they acquire the individual phonemes, then children should have an easier time imitating nonsense words with familiar sequences than with unfamiliar sequences, even when both types of sequences are made up of phones that the child has mastered.

4. Two experiments were run.  In each, children were asked to repeat back nonsense words with sequences that were either frequent or virtually non-existent in English.  The words were controlled for how well adults could repeat them and how "wordlike" adults thought they were.  The second experiment differed from the first in the exact identity of a few of the nonsense words.  The number of phones produced correctly in each word was counted as a measure of accuracy.

5. In both experiments, it was found that familiar sequences were produced more accurately than unfamiliar sequences.  This effect was true, however, only for CV and CC sequences, not for VC sequences.

6. The second experiment was also run on phonologically disordered children.  Here, the effects were even larger and more consistent across all three types of sequences (CV, VC, and CC) -- familiar sequences of familiar phones are easier than unfamiliar sequences of familiar phones.

7. The authors conclude that children generate sequences of sounds in their phonological representation, based on the frequency of sequences in the language they are learning.  That is, children start by learning whole words; then realize that there are subunits of words that can be put together to form words and start storing these; eventually, the individual phonemes will emerge from these larger sequences. 

Questions:
1. How was "correctness" decided in production?

2. Not sure I understand how "wordlikeness" was actually controlled.  Of the ratings given, some seem misleading -- e.g. familiar sequence [m'oftin] as 3.8 on a scale of 1-5 for wordlikeness, while its unfamiliar sequence counterpart [m'ofken] was rated 2.4.  How close did the ratings have to be?  Doesn't this have an effect?

3. Why did VC sequences not have any effect of frequency?