To test the effects of talker variability and explicit instruction on the statistical learning of lexical tone, 80 monolingual English listeners were taught an artificial language that mimicked Mandarin’s asymmetric distribution of syllable-tone co-occurrences. Training stimuli consisted of either speech from one talker or speech from four talkers. Participants were either never instructed or explicitly taught associations between phonemes (CVs), tones, and nonce symbols across four consecutive days. Learning was assessed by the accuracy of mouse clicks and eye movements to visual nonce symbols. Critical trials induced competition between the target symbol, which matched the acoustic input, and a competitor symbol that had a statistically more probable tone (but mismatched the acoustic input). Eye fixations indicated that participants were sensitive to syllable-tone co-occurrence probabilities even without explicit instruction of tone. The degree to which statistical knowledge was used to recognize words appeared to increase when participants processed more variable speech.

中文翻译：

---

多词输入和教学方法对音节组合维度统计学习的影响

为了测试说话者可变性和明确指令对词汇音调统计学习的影响，向 80 名单语英语听众教授了一种模仿普通话音节共现不对称分布的人工语言。训练刺激包括来自一个谈话者的演讲或来自四个谈话者的演讲。在连续四天中，参与者要么从未被指示或明确教授音素 (CV)、音调和随机数符号之间的关联。通过鼠标点击和眼球运动对视觉随机符号的准确性来评估学习。关键试验引发了与声输入匹配的目标符号和具有统计上更可能的音调（但与声输入不匹配）的竞争符号之间的竞争。眼睛注视表明，即使没有明确的语调指示，参与者也对音节音调共现概率敏感。当参与者处理更多可变的语音时，使用统计知识识别单词的程度似乎增加了。