Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Pronunciation but not orthography enables generalization of sensorimotor learning during Mandarin Chinese production
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Xin-ran Xie1, Robin Karlin2, Ding-lan Tang1; 1The University of Hong Kong, 2University of Missouri
Introduction: Language involves multiple subsystems that encode distinct types of linguistic information, such as phonology, morphology, semantics, syntax, and orthography. In models of language production, these systems are jointly activated and can mutually influence each other before the motor plans are selected for execution. For example, both orthography and phonology can facilitate word production, though the effects differ at different times of presentation, indicating that the two subsystems are accessed at different times during language planning. However, it is unclear the degree to which non-phonological factors, such as orthography, are referenced by the motor planning and execution systems during speech production. Research in the generalization of motor learning suggests that learning can spread across items that share some degree of similarity. To date, studies have only investigated items that share phonetic (motor) similarity. Here, we use a sensorimotor learning generalization paradigm to investigate the extent to which orthography informs the motor planning stage of language production. We focus on Mandarin, a language with a particularly rich set of items where pronunciation and orthography are dissociated. Methods: Forty-eight native Mandarin speakers participated in the study (16 in each experiment). In all experiments, participants repeatedly produced the training word 答复 /da2 fu4/ “answer” and heard their speech auditory feedback over headphones. During the perturbation phase, the first formant (F1) of the first character (/da2/) was shifted upward by 125 mels before being played over the headphones. To assess the transfer of learning, participants also produced an untrained transfer word before and after learning while their auditory feedback was masked by noise. Three transfer words were used: 1) 达到 /da2 dao4/ “reach” in Experiment 1, which shares the pronunciation but differs from the training word in orthography at the first character; 2) 答应 /da1 ying4/ “agree” in Experiment 2, which shares the orthography but differing in pronunciation at the first character; and 3) 破坏 /po4 huai4/ “disrupt” in Experiment 3, which differs from the training word in both orthography and pronunciation (control). Results: Speakers significantly decreased their F1 in the trained word in response to the upward F1 perturbation in all experiments (Mean changes: Experiment 1: −29 mels; Experiment 2: −37.1 mels; Experiment 3: −23.1 mels; all p ≤ 0.026). Critically, significant transfer was observed when pronunciation was shared: adaptation in “答”复 /da2 fu4/ significantly transferred to “达”到 /da2 dao4/ (−22.1 mels) in Experiment 1. Moreover, transfer magnitudes did not differ significantly from adaptation magnitudes (p = 0.207), suggesting complete transfer. In contrast, no significant transfer was observed when only orthography was shared in Experiment 2 (“答”复 /da2 fu4/ v.s “答”应 /da1 ying4/; −2.5 mels, p = 0.879). As expected, no significant transfer was found when neither pronunciation nor orthography was shared in Experiment 3 (“答”复 /da2 fu4/ v.s “破”坏 /po4 huai4/; −7.8 mels, p = 0.237). Conclusion: We show that only phonetic overlap, and not orthography, provides a pathway for generalization, indicating that the motor planner itself is agnostic to orthography.
Topic Areas: Speech Motor Control, Language Production