A Phonetic Category Learning Road Trip Chris Heffner Blair Linguistics Club 2016-04-06
Categories: Example How do categories work? Introduce NIH vs. NSF – the difference between these categories is behaviorally relevant.
Rule-Based Theories of Category Learning California Surfer hair Up for anything Fusion everything … Minnesota Blond hair, blue eyes Exceedingly gregarious Anti-flavor … Florida Bronzed or wrinkled skin Retired and/or tourism Cuban food … Maryland Constantly wearing Maryland flags Probably a fed Crab forever! … Rule-based account – abstract generalizations are compared via a series of rules. A researcher might satisfy most (thought not all) criteria for one category or the other, and would be judged to be a member of that category
Exemplar-Based Theories of Category Learning Exemplar-based account – categories are made up of the specific memories one has for the individuals in each category. No abstraction is performed. A researcher might be classified as an NIH researcher because they remind someone more of the NIH researchers they know than the NSF researchers.
Dual-System Theories of Category Learning Rules Exemplars
Category Learning in Language Is a /ta/ or a /da/? Rule-Based Exemplar-Based /t/ = VOT > 35 ms /d/ = VOT < 35 ms da ta da ta da ta
Methods: Materials [ç] – palatal [x] – velar 1 2 3 4 5 6 7 8 9 10 Continuum between [ç] and [x] with 10 steps. Created using linear combination. 1 2 3 4 5 6 7 8 9 10
Methods: Procedure x450 (maximum) Brief outline of experiment. Total of about 450 trials maximum (~20 minutes, not for principled reasons), but participants could end early if they have at least 90% of continuum items correct. Participants first heard a sound, and were asked to push keyboard buttons corresponding to the red, yellow, or red categories. They then got feedback. x450 (maximum)
Methods: Conditions [ç] [x] What counts as correct? 3 continuous conditions, 3 discontinuous conditions.
Results: Time to Learn 27 2 Compare learning rates across conditions. Note that continuous (top three) and discontinuous (bottom three) conditions seem to differ in success. 7 8 15
Dual-System Model: Implications Rule-based Exemplar-based
Destination One: Germany
Germany: Rationale
Germany: Background
Germany: Results
Destination Two: Michigan
Michigan: Rationale
Michigan: Background Stuttering is associated with differences in dopamine circuitry in the brain (Giraud et al., 2008; Lu et al., 2010; Wu et al., 1997), which has been computationally modelled as a part of the underlying etiology of stuttering (Civier et al., 2013) Stroke and other injuries to the basal ganglia can result in stuttering (Ciabarra et al., 2000; Tani & Sakai, 2011) “The core dysfunction in stuttering is suggested to be impaired ability of the basal ganglia to produce timing cues for the initiation of the next motor segment in speech.” (Alm, 2004, p. 325) Stuttering may be associated with some differences in language processing and production other than its diagnostic symptoms, such as past tense marking (Bauman et al., 2012)
Michigan: Results
Destination Three: Ohio
Ohio: Rationale
Ohio: Previous Methods (Gabay et al., 2015) Ohio: Previous Methods SMART + what is a good result in SMART?
Ohio: My Methods
Ohio: Background (Maddox et al., 2013) Category learning over the lifespan
Conclusions Germany: surprising similarities between English and German speakers Michigan: kids who stutter seem to be impaired on rule-based learning Ohio: what will happen with varied age groups?
Thank You! heffner@umd.edu