The Effects of Context and Prosody on the Perception of Verbal Sarcasm

Two experiments were conducted with Chinese materials in the current study.
In the first part of Experiment 1, we invited a 24-year-old Chinese native speaker to record 42 acoustic sentences (21 with sincere prosody and 21 with sarcastic prosody). Applying voice processing software, we extracted or calculated seven acoustic parameters (mean intensity, mean pitch, intensity SD, pitch SD, intensity range, pitch range, tempo) of these 42 sentences (as what the file named Experiment 1 Part 1 shows). The result demonstrated that sarcastic prosody and sincere prosody were distinguishable by seven acoustic features. In the second part of Experiment 1, twenty-eight participants were invited to score the recorded experimental stimuli. After listening to the target sentences, participants were asked to rate on a 7-point Likert scale to what extent they thought the intention of each sound excerpt was sincere or sarcastic (1 = very sincere, 2 = sincere, 3 = a little sincere, 4 = neutral, 5 = a little sarcastic, 6 = sarcastic, 7 = very sarcastic). Forty-two target sentences in acoustic form were played in a random order. The data (as what the file named Experiment 1 Part 2 shows) indicated that the listener could distinguish a literal sincere meaning from an implied sarcastic meaning merely according to the prosody.
Experiment 2 further explored the interplay role of context and prosody. The materials of Experiment 2 were 42 target sentences and 42 contexts written, recorded, and validated in Experiment 1. A single trial of Experiment 2 consisted of a situational context followed by a relevant target sentence. There were 21 contextual scenarios and 21 target sentences, and each contextual scenario included two versions: a negative context (leading to a “sarcastic” perception) and a positive context (leading to a “sincere” perception). Additionally, each target sentence included two prosodies: a negative prosody and a positive prosody. Thus, there were 84 combinations of context and target sentence (21 scenarios × 2 versions × 2 target sentences with different acoustic features). All materials were divided into four item pairs according to the Latin square design, and each participant finished one item pair. Each test consisted of 21 trials with four combined pairs. The four item pairs were combined with contexts (negative and positive) and prosodies (negative and positive). Thus, four item pairs were obtained: NN, PP, NP, and PN.
We obtained a total of 1680 (21 trials × 80 participants) responses including choices and RTs (21 trials × 80 participants). We rejected some invalid responses, remaining 90.48% (1520) of the total responses (as what the file named Experiment 2 shows).

本研究以中文材料为对象，共进行了两项实验。在实验一的第一部分中，我们邀请了一位24岁的母语为中文的参与者录制了42个声学句子（其中21个带有真诚的语调，21个带有讽刺的语调）。通过应用声音处理软件，我们提取或计算了这42个句子的七个声学参数（平均强度、平均音高、强度标准差、音高标准差、强度范围、音高范围、节奏）。研究结果表明，讽刺语调和真诚语调可以通过七个声学特征进行区分。在实验一的第二部分中，我们邀请了28名参与者对录制的实验刺激物进行评分。在听完目标句子后，参与者被要求在7点李克特量表上评估他们认为每个声音片段的意图是真诚还是讽刺的程度（1=非常真诚，2=真诚，3=略真诚，4=中立，5=略讽刺，6=讽刺，7=非常讽刺）。42个目标句子以随机顺序播放。数据（如名为实验一第二部分所示的文件所示）表明，仅根据语调，听众就能区分字面真诚意义和隐含讽刺意义。实验二进一步探讨了语境和语调之间的相互作用。实验二的材料包括42个目标句子和42个语境，这些句子和语境均已在实验一中编写、录制并验证。实验二的每个测试由一个情境语境随后跟随着一个相关目标句子组成。共有21个情境场景和21个目标句子，每个情境场景包括两个版本：一个负面语境（导致“讽刺”感知）和一个正面语境（导致“真诚”感知）。此外，每个目标句子还包括两种语调：负面语调和正面语调。因此，共有84种语境与目标句子（21场景×2版本×2个具有不同声学特征的目标句子）的组合。所有材料根据拉丁方设计分为四组项目对，每位参与者完成一组项目对。每个测试由21个试验和四个组合对组成。四个项目对与语境（负面和正面）和语调（负面和正面）相结合。因此，获得了四个项目对：NN、PP、NP和PN。我们共获得了1680（21试验×80名参与者）的响应，包括选择和反应时间（21试验×80名参与者）。我们排除了一些无效的响应，最终剩余了总响应的90.48%（1520）。