Segmenting the Signal: A Cross-Age Comparison of Speech and Tone Stimuli

The present study sought to extend findings of an “embeddedness constraint” on statistical learning in the visual domain (Fiser & Aslin, 2005) to the auditory domain. In this set of experiments, we explored the possible manifestations of an embeddedness constraint on speech and tone sequences in studies with adult and infant participants. An embeddedness constraint operant in the linguistic domain could benefit learners by helping them to circumvent the “combinatorial explosion” that arises from large amounts of complex data. This constraint could also prove useful in language acquisition tasks, such as word segmentation, allowing learners to advance more quickly to higher-order tasks, such as semantic and syntactic categorization. Furthermore, findings of an embeddedness constraint on both speech and tone stimuli might suggest that this constraint is broadly domain-general. Present results do not indicate an embeddedness constraint on speech or tone input in adults. However, pending studies with infants may suggest that the existence of the embeddedness constraint is itself constrained by maturation or by experience with language.