| Congenital amusia,commonly known as tone deafness,is an umbrella term for lifelong musical disabilities which cannot be attributable to brain damage,hearing loss,lack of adequate exposure,or intellectual disability after birth.Recent studies have identified that amusia can affect other domains in addition to music,particularly speech perception.Being a domain-general but pitch-specific disorder,congenital amusia is present in an estimated1.5%–4% of the general population.Individuals afflicted with congenital amusia,called congenital amusics,are frequently reported having difficulties in processing pitch-related materials across music and language,such as musical melodies,intonations,and lexical tones.Although a plethora of studies endeavor to account for amusia,few explore pathological origins of this disorder from the angle of afflicted individuals’ learnability.This dissertation therefore examined amusics’ statistical learning of pitch borrowed from either music or language domain to define whether their deficiencies result from a possibly impaired statistical learning mechanism.This dissertation mainly addresses three research questions: 1)Can amusics learn musical pitch information in novel musical chords through distributional statistical learning? 2)Can amusics learn linguistic pitch information in a second language of Thai through distributional statistical learning? 3)Can amusics track the statistical regularities of both segmental phonemes and suprasegmental lexical tones embedded in the linguistic input of Mandarin despite their overall degraded speech perception? Three experiments conducted in this dissertation corresponded to each research question,respectively.Fifty-eight amusics and 61 musically intact individuals,all of whom spoke Mandarin Chinese as their native language,participated in three statistical learning experiments.Experiments 1 and 2 investigated amusics’ distributional learning,a certain type of statistical learning,of respective musical pitch and linguistic pitch dependent on their subliminal computation of stimulus distributional structure.All experimental materials in the former two experiments were developed based on novelty,such that the first experiment exploited musical chords built upon a new microtonal musical scale,called Prime Number Scale,differing from the general musical scale,and Experiment 2 employed the language of Thai with none of participants having prior knowledge for that.This benefited avoiding contamination in data interpretation of learning outcome from listeners’ pre-existing knowledge.In the third experiment,listeners in the two groups were required to track statistical regularities embedded in the linguistic input of Mandarin in a gated fashion,which involves both frequency information of lexical tones and probability information of co-occurrences of syllable-tone combinations.In each experiment,generalized or linear mixed-effects models were constructed to analyze the collected data,with both fixed and by-subject and by-item random effects included.Working memory was always regarded as the controlled covariate when fitting the models.The first experiment exposed amusics and controls to novel musical chords in the training phase,with two sub-groups of each divided based on the certain distributions they encountered.In detail,following a pretest-training-posttest design,random assignment to either bimodal or unimodal distribution was implemented to both amusics and typical listeners,with bimodally rather than unimodally trained learners expected to gain improvement in posttest relative to pretest.Results of the first experiment demonstrated that in distributional learning of musical chords,amusics showed reduced accuracy than the matched controls in pretest and posttest regardless of distribution conditions.Amusics’ inferior performance in the full spectrum suggested that their musical pitch processing was impaired as compared with typical listeners,pointing to an anomalous recurrent processing network of musical pitch in the amusic brain though musical tokens were based on a scale novel to both groups.However,as revealed by the statistical models,amusics in the bimodal distribution showed significantly higher accuracy in posttest than pretest while those in the unimodal condition did not.These patterns were comparable to normal individuals.Moreover,the perceptual gain was observed in amusics’ processing of all types of musical chords independent of whether or not the chords were trained.This line of finding was indicative of a built category at an abstract level in amusics instead of any practice or repetition effects.Therefore,Experiment 1 manifested that in spite of impaired musical pitch processing,amusics were able to develop the mental musical lexicon via their spared distributional learning of music which allowed them to categorize different musical chords they heard.In a similar vein,the second experiment employed novel Thai lexical tones as linguistic materials for both groups to learn via distributional statistical learning.Results demonstrated that in pretest,amusics exhibited inferior performance in perceiving Thai words;moreover,in posttest,amusics still obtained lower accuracy than controls regardless of exposure either in a bimodal or unimodal manner.This indicated amusics’ degraded linguistic pitch processing as broadly reported in lexical tone perception in the amusia literature.Nevertheless,comparable patterns were unmasked between amusics and typical listeners as both of them were scored higher in posttest than pretest when trained within bimodal rather than unimodal distribution.Importantly,the statistical models reveled that amusics’ improvement was evident in test words across two test dimensions of speaker gender and syllable,irrespective of whether they were trained or non-trained novel tokens during exposure.This implicated that amusics’ distributional learning mechanism of language was spared even the speech materials involved novel lexical tones that they struggled to perceive.The third experiment manifested that amusics showed inferior performance in speech perception to controls across the board.In comparison to typical listeners,amusics exhibited overall lower accuracy in perceiving the gated tokens in terms of correct syllable-only responses,lexical tone-only responses,and syllable-tone word responses,suggesting amusics’ degraded speech perception in line with previous studies in the literature.Nevertheless,analogous to musically intact individuals,amusics showed higher accuracy for listening items with high token frequency than low token frequency across these syllable-only,lexical tone-only,and syllable-tone word measures;besides,the log values of tone error,i.e.,correct-syllable-incorrect-tone responses,calculated based on a formula did not significantly differ between the two groups in the successive gates.The patterns of amusics’ performance remained consistently matched with non-amusic listeners,which clearly certified that amusics were able to compute the statistics of both phonemes and lexical tones embedded in their native language of Mandarin despite their degraded perception at both segmental and suprasegmental levels of speech.This dissertation systematically documented Mandarin-speaking amusics’ statistical learning,or specifically,distributional learning,of both musical and linguistic pitch information embedded in musical chords and spoken words,with the purpose to investigate whether congenital amusia could result from a potential learning factor,equivalent to another issue as to whether deficient processing in one cue or dimension would be connected with impaired statistical learning for that cue or dimension.The observed results answered the research questions by demonstrating that amusics’ pitch processing was poorer than controls in both music and language domains,albeit pitch playing a phonological role in speech;nonetheless,despite their impoverished performance in processing either musical or linguistic pitch,amusics exhibited the reliable patterns of distributional learning in both domains.In sum,this dissertation provided novel and compelling evidence that amusics’ cross-domain pitch processing was impaired but their distributional statistical learning of both musical and linguistic pitch was spared.This advanced the understanding of pathological origins of amusia and other neurodevelopmental disorders as well as encouraged translational research to be implemented to rehabilitate amusia in the future. |
PDF Full Text Request