口吃者加工汉语歧义短语的神经过程

Neural processing of ambiguous Chinese phrases of stutters

韵律边界是口语韵律特征的有机组成部分,在语言理解中发挥着重要作用。口吃作为一种言语节律性障碍,主要表现为音节经常性的重复、拖长或停顿等。本研究采用ERP,考察口吃者完成词汇判断和结构判断两种任务时,加工歧义短语(动宾/偏正歧义结构)内部韵律边界的认知过程。结果发现口吃者和言语流畅者在加工汉语歧义短语过程中,所诱发的反映韵律切分的脑电成分CPS (closure positive shift)不存在显著差异。在0～300 ms,不论中线还是两侧,口吃者和言语流畅者加工两类短语时,动宾短语韵律边界诱发正效应的头皮分布范围小于偏正短语。在300～600 ms,中线上,口吃和言语流畅者在完成两种任务时,两类短语的韵律边界均稳定诱发了正效应;在两侧,结构判断任务中两类短语的韵律边界都诱发了正效应,但词汇判断任务中只有偏正短语稳定诱发该效应。综上,口吃者和言语流畅者一样对口语韵律边界敏感,并且他们加工歧义短语内部韵律边界时,诱发的脑电效应受到实验任务和短语结构类型的影响。

Prosodic boundary is an integrative part of spoken language that segments ongoing utterance into prosodic units. These boundaries are correlated with the perception of a pause, a lengthening of the pre-boundary syllable and tonal movement at the end of the phrase. Stuttering is characterized by involuntary disruptions in the flow and rhythm of speech, which was reflected by repetitions of words, sounds or syllables, prolongations and silent blocks. Behavioral response and neural processing results in the past few years indicated that adults who stutters exhibit processing differences compared with fluent speakers during syntactic, semantic and phonological （rhyme） processing. However, existing studies did not examine whether stutters encounter difficulty during perception of prosodic boundary. The present study aims to explore how stutters and fluent speakers process prosodic boundary of ambiguous Chinese phrases （Verb NP1 Aux NP2） in lexical and structural judgment task using ERPs. We used 168 typical ambiguous Chinese phrases as experimental materials. These phrases were temporarily ambiguous between modifier–noun construction （MNC） and narrative–object structure （NOS）. Eighty-four phrases without ambiguity were used as fillers. Twenty-four （20 males） undergraduates/graduates participated in the experiment. They were told to listen carefully to pairs of phrase in two sessions with the same materials. In session one, they completed a lexical judgment task （to determine whether a visually presented word appeared in the pairs of phrase they heard）, while in session two they were asked to complete a structural judgment task （to judge whether the pairs of phrase they heard belong to one kind of structure or not）. Electrophysiological data were recorded by a set of 64 electrodes from eegmagine （ANT Neuro） according to the extended 10-20 positioning system. EEG data were time-locked to the offset of verb and Aux （de） of the first phrase using a 100-msec pre-stimulus baseline and an averaging time window of 800 msec. We selected two time windows （0~300 ms and 300~600 ms） for statistical analysis in the midline and lateral areas. During 0~300 ms, we found that prosodic boundary （v.s. non-boundary） elicited positivity in the midline, F （1, 22） = 24.28, p 〈 0.001, ηp2 = 0.52, and lateral areas, F （1, 22） = 45.51, p 〈 0.001, ηp2 = 0.67. Besides, the interaction between Structure and Boundary was significant in the midline, F （1, 22） = 5.84, p 〈 0.05, ηp2 = 0.21, and lateral areas, F （1, 22） = 4.18, p = 0.053, ηp2 = 0.16. Simple effect analysis indicated that prosodic boundary elicited positive effect for MNC in the midline, while in the lateral areas prosodic boundary elicited positivity for both of NOS and MNC, F （1, 22） = 10.35, p 〈 0.005, ηp2 = 0.32; F （1, 22） = 29.69, p 〈 0.001, ηp2 = 0.57. During 300~600 ms, we found that prosodic boundary （v.s. non-boundary） elicited positivity in the midline, F （1, 22） = 36.61, p 〈 0.001, ηp2 = 0.61, and lateral areas, F （1, 22） = 36.59, p 〈 0.001, ηp2 = 0.71. Besides, the interaction between Region and Boundary was significant in the midline, F （2, 44） = 10.07, p 〈 0.005, ηp2 = 0.31,and lateral areas, F （2, 44） = 24.16, p 〈 0.001, ηp2 = 0.52. Simple effect analysis indicated that although the positivity elicited by prosodic boundary was broadly distributed in the whole scalp area, it was prominent in frontal-central area. More importantly, the interaction between Task, Boundary and Structure was significant in the lateral area, F （2, 44） = 3.95, p 〈 0.05, ηp2 = 0.15. Simple effect analysis indicated that in lexical judgment task, prosodic boundary of MNP elicited positive shift, F （1, 22） = 23.41, p 〈 0.001, ηp2 = 0.52, but NOS didn’t, F （1, 22） = 2.47, p = 0.131. However, prosodic boundaries of both MNP and NOS elicited positive effect in structure judgment task, F （1, 22） = 17.02, p 〈 0.001, ηp2 = 0.44; F （1, 22） = 11.65, p 〈 0.005, ηp2 = 0.35. Overall, we found that stutters and fluent speakers exhibit similar neural process during prosodic boundary processing. This finding was reflected by the fact that the stable CPS was elicited by prosodic boundaries of both MNP and NOS. The positive effect elicited by MNC in an earlier time window was distributed more broadly in scalp than that elicited by NOS in both kinds of task. In a later time window, prosodic boundaries of both MNC and NOS elicited the stable CPS regardless of the kind of experimental task in the midline. In the lateral areas, the CPS was detected in the prosodic boundary of MNC in both kinds of task, whereas the CPS was stably observed at the boundary of NOS in structure judgment task. In conclusion, we contend that stutters and fluent speakers are both sensitive to prosodic boundary and their processing was influenced by the structure of ambiguous phrases and experimental task.