Bulgarelli, F. & Weiss, D.
Pennsylvania State University
A fundamental challenge of statistical learning is to determine whether variance observed in the input signals a change in the underlying structure. When learners are asked to segment two consecutive artificial languages, they tend to only learn the first structure unless the change is correlated with a contextual cue or exposure to the second structure is lengthened (Gebhart, Aslin, & Newport, 2009). We explored this primacy effect by presenting learners with two consecutive artificial languages, providing them with tests after each minute of familiarization. In one condition, learners received fixed input, whereas in the other they advanced to the second language immediately after learning the first. In the former condition, we replicated the primacy effect, while in the latter, we found that learners tended to learn and retain both languages. Interestingly, contextual cues did not boost performance using this latter paradigm, suggesting that without becoming entrenched in the first language there is no additive effect for such cues. Overall, our findings suggest that anchoring effects may be due to additional exposure after a single structure has been learned, possibly as a function of learners reducing their sampling rate once mastering a set of statistics.