非同时期普通话语音对说话人辨别及自评置信度的影响

Effects of Non-Contemporaneous Mandarin Speech on Talker Discrimination and Self-Reported Confidence

摘要: 说话人辨别是听者基于听觉线索判断两段语音是否来源于同一说话人的过程。本研究在同异判断(same-different) AX范式下，构建1年和16年时间间隔的非同时期普通话语音材料，控制语料长度和说话人性别，招募42名男性听音被试完成同异判断及置信度评分，采用广义线性混合模型进行分析。结果表明，时间跨度显著降低辨别准确率；语料长度呈“长句 > 短句 > 词语”的非显著提升趋势；说话人性别对辨别准确率无显著影响；自评置信度能够显著正向预测判断准确率。信号检测分析结果显示，被试整体具备一定感知敏感性且个体差异明显，判别标准整体接近中性。研究表明，时间跨度会削弱普通话语音的可辨识性，自评置信度可作为判断可靠性的有效参考指标。

Abstract: Talker discrimination refers to the process by which listeners judge whether two speech utterances originate from the same talker based on auditory cues. In this study, non-simultaneous Mandarin speech materials with time intervals of 1 year and 16 years were constructed under the AX same-different judgment paradigm. Speech length and talker gender were controlled, and 42 male listeners were recruited to perform same-different judgments and confidence ratings. Generalized linear mixed models were used for data analysis. The results showed that time span significantly reduced discrimination accuracy; speech length exhibited a non-significant increasing trend of “long sentences > short sentences > words”; talker gender had no significant effect on discrimination accuracy; and self-rated confidence significantly and positively predicted judgment accuracy. Signal detection analysis revealed that listeners generally possessed certain perceptual sensitivity with obvious individual differences, and the overall decision criterion was close to neutral. The study indicates that time span weakens the identifiability of Mandarin speech, and self-rated confidence can serve as an effective indicator for judging the reliability of discrimination.