关键词: 口语词汇产生, 多重音韵激活, 独立两阶段模型, 交互激活模型

Abstract:

A debatable issue between serial discrete models and interactive models is whether non-target lemmas activate their phonological words in spoken word production. Serial discrete models assume that only target lemma activates its corresponding phonological node to articulation, whereas interactive models assume that the semantic and phonological nodes linked to multiple candidates are co-activated during the retrieval of target word. Multiple phonological activation has been supported by evidences from alphabetic languages, but it remains unknown whether this finding can be generalized to non-alphabetic languages. Therefore, the current study aimed to investigate whether the not-to-be named pictures activate their phonological nodes in Chinese spoken word production.
Using electrophysiological measures, the present study employed a word translation task in native Chinese speakers with a high level of English proficiency. Thirty-two participants (13 males, average 22.94 years) were presented with an English probe word and a context picture (semantically related or unrelated, phonologically related or unrelated to target word) simultaneously. Eighty-six English probe words from CELEX database and forty-three black and white line pictures from a standardized picture database in Chinese were chosen as stimuli. Participants were asked to translate English probe words into Chinese as accurately and quickly as possible while ignoring context pictures presented simultaneously.
Behavioral results showed a typical semantic facilitation effect, with faster translation latencies in the semantically related condition than in the semantically unrelated condition. More importantly, phonological overlap, which generally elicits priming in Indo-European languages, resulted in a null finding for Chinese production. Electrophysiological results revealed that semantic relatedness induced significant effects of ERPs after stimuli presentation: a widely distributed positivity in the 400- to 600-ms interval, while marginally significant effects were observed for phonological relatedness in the time interval of 600~700 ms in the right middle region. Furthermore, a negative correlation between the difference of translation latencies (semantically related minus semantically unrelated) and the difference of mean amplitudes (semantically related minus semantically unrelated) approached significance in the 400~600 ms time window in the middle posterior region, suggesting that more positive mean amplitudes were associated with shorter translation latencies.
Although speakers present a weak but reliable neural activation, we suggest that phonological overlap between context pictures and target words had no impact on the translation processing in behavioral. That is, the non-target lemma did not activate their phonological node, and multiple phonological activation was absent in Chinese spoken production. Meanwhile, the semantic information of context pictures was indeed activated, and according to the temporal course of word translation, the time window of 400~600 ms was estimated for conceptual preparation when Chinese-English bilinguals completed a word translation task, although this activation was not transmitted from semantic level to phonological level. Overall, the present findings support a serial discrete model rather than an interactive model in Chinese spoken word production.

Key words: spoken word production, multiple phonological activation, serial discrete models, interactive models