Examining variability and generalization in dimension-based statistical learning for speech: The case of place of articulation.

In this study, the question of generalization for dimension-based statistical learning in speech perception is revisited. Learning for F0 and voice-onset-time as cues to stop voicing has been suggested to be fairly specific to a particular contrast, which was previously shown not to generalize between two places of articulation. The present study seeks to replicate generalization for the same place of articulation, using more varied stimuli and a different design. Then, it is tested if increased evidence for a distributional pattern, i.e., two places of articulation showing that pattern, leads to generalization to a third place of articulation. Same place of articulation learning is replicated, and no generalization across place of articulation is found, reaffirming that learning appears to be quite specific.