Talker Variability as a Desirable Difficulty for Language Learning

Restricted (Penn State Only)
Bulgarelli, Federica
Graduate Program:
Doctor of Philosophy
Document Type:
Date of Defense:
September 05, 2018
Committee Members:
  • Daniel J Weiss, Dissertation Advisor
  • Daniel J Weiss, Committee Chair
  • Janet van Hell, Committee Member
  • Koraly Elisa Perez-Edgar, Committee Member
  • Matthew Thomas Carlson, Outside Member
  • Judith Kroll, Committee Member
  • talker variability
  • desirable difficulty
  • artificial language learning
A well-known challenge for language learners is that the input is typically produced by a variety of speakers, each with distinct vocal characteristics (Liberman, Harris, Hoffman, & Griffith, 1957). Accordingly, many studies have indicated that talker variability leads to processing costs for learners across the lifespan (Jusczyk & Pisoni, 1992; Mullennix, Pisoni, & Martin, 1989; Ryalls & Pisoni, 1997). However, increased talker variability can also be helpful for language learning (Rost & Mcmurray, 2009) by focusing learners on invariant properties of the signal. These discrepant findings may indicate that talker variability is akin to a desirable difficulty (Bjork, 1994) in learning. That is, initial costs in processing may lead to long term benefits for retention and generalization. I investigated the hypothesis that talker variability acts as a desirable difficulty in both adults and preschool-aged-children. I used an artificial grammar task modeled on dependencies found in languages with grammatical gender, which is acquired by 3yos in their native language (Mariscal, 2009), but is notoriously difficult to acquire for adults (Dewaele & Véronique, 2001). Critically, this task afforded learning of novel object labels as well as grammatical dependencies, allowing for tests of retention and generalization. The results suggest that the processing cost associated with contending with multiple speakers for adults may be small and transient, but that high variability may benefit learning a novel grammatical dependency. Results from preschoolers underscore the difficulty of the chosen paradigm, but highlight a differential impact of low talker variability (i.e. 2 speakers) for children and adults, as it facilitated learning for preschoolers but impeded learning for adults. In sum, the results from this dissertation provide limited support that talker variability may act as a desirable difficulty, but that optimal talker variability may vary by age and the content to be learned.