Grama, I. 1, 2 & Wijnen, F. 1, 2
1 Utrecht University
2 Utrecht Institute of Linguistics OTS
The Starting Small hypothesis (Newport, 1990; Elman, 1993) claims that young learners benefit from initially attending to smaller units of language, and subsequently examining the complex structures these units are part of. Conversely, Starting Big (Arnon 2009; Arnon & Ramscar, 2012) proposes that learning is beneficial when it first internalizes more complex structures and only subsequently breaks them down into smaller units. Arnon & Ramscar (2012) found that participants learned adjacent relationships between words better when ?starting big? (being initially presented with word-combinations) rather than ?starting small? (being initially presented with the words in isolation).
I test the effect of Starting Small with non-adjacent dependencies: I pre-familiarize learners with the individual a/b words of an artificial aXb language (Gomez, 2002), and subsequently expose them to the language itself. Initial findings support the results of Arnon & Ramscar (2012): Starting Small appears detrimental to learning. However, I show that these results are due to a primacy effect: learners exposed to two consecutive learning phases show poorer recall for the second phase (Gebhart, Aslin & Newport, 2009). I control for this confound, and show that the effect of Starting Big is in fact modulated by individual differences in working memory and word-learning.