The Negative Effect of Prior Exposure on Remote Dependency Learning

Remote dependency-learning is a mechanism that may aid the acquisition of morpho-syntactic dependencies such as 'Wellington IS always praisING Sharpe' (Gómez, 2002). Infants become sensitive to such dependencies in their language at around 18 months (Santelmann & Jusczyk, 1998), although they recognize the functional morphemes that form the dependencies earlier (8-11 months, Shi & Lepage, 2008). In natural languages, therefore, dependencies are presumably acquired at a stage where the dependent elements are already familiar to learners. We tested the effect of prior familiarity on remote dependency-learning in an artificial grammar paradigm with adults. We employed a simple grammar aXb where subjects were meant to detect dependencies between a and b at familiarization, adding a pre-familiarization phase of varying lengths (1, 2, or 8-minute) where they were exposed to aY/Zb strings to facilitate learning the individual a/b elements. Results showed that participants successfully learned dependencies with a short pre-familiarization, but performance decreased with longer pre-familarization. Increasing acoustic prominence of the a/b tokens, or presenting a/b tokens in isolation at pre-familiarization also had a significantly negative effect on learning, showing that higher salience of a/b elements at pre-familiarization inhibited learning. We concluded that prior familiarity with dependent elements inhibits remote dependency-learning.