Perception of affect in unfamiliar musical chords

This study investigates the role of extrinsic and intrinsic predictors in the perception of affect in mostly unfamiliar musical chords from the Bohlen-Pierce microtonal tuning system. Extrinsic predictors are derived, in part, from long-term statistical regularities in music; for example, the prevalence of a chord in a corpus of music that is relevant to a participant. Conversely, intrinsic predictors make no use of long-term statistical regularities in music; for example, psychoacoustic features inherent in the music, such as roughness. Two types of affect were measured for each chord: pleasantness/unpleasantness and happiness/sadness. We modelled the data with a number of novel and well-established intrinsic predictors, namely roughness, harmonicity, spectral entropy and average pitch height; and a single extrinsic predictor, 12-TET Dissimilarity, which was estimated by the chord’s smallest distance to any 12-tone equally tempered chord. Musical sophistication was modelled as a potential moderator of the above predictors. Two experiments were conducted, each using slightly different tunings of the Bohlen-Pierce musical system: a just intonation version and an equal-tempered version. It was found that, across both tunings and across both affective responses, all the tested intrinsic features and 12-TET Dissimilarity have consistent influences in the expected direction. These results contrast with much current music perception research, which tends to assume the dominance of extrinsic over intrinsic predictors. This study highlights the importance of both intrinsic characteristics of the acoustic signal itself, as well as extrinsic factors, such as 12-TET Dissimilarity, on perception of affect in music.

本研究探究了外在与内在预测因子在博伦-皮尔斯微音调律系统（Bohlen-Pierce microtonal tuning system）下，多数陌生和弦的情感感知中所发挥的作用。外在预测因子部分源自音乐中的长期统计规律，例如与被试相关的音乐语料库中某和弦的流行程度。与之相对，内在预测因子无需借助音乐中的长期统计规律，例如音乐固有的心理声学特征，如粗糙度（roughness）。本研究针对每个和弦测量了两类情感维度：愉悦度/不愉悦度，以及快乐度/悲伤度。我们采用多种新颖且成熟的内在预测因子对数据进行建模，具体包括粗糙度、和谐度（harmonicity）、频谱熵（spectral entropy）与平均音高；同时采用单一外在预测因子——12平均律差异度（12-TET Dissimilarity），该指标通过计算目标和弦与任意12音平均律和弦的最小距离进行估算。音乐素养被建模为上述预测因子的潜在调节变量。本研究开展了两项实验，二者分别采用博伦-皮尔斯音律系统的两种变体：纯律（just intonation）版本与平均律版本。研究发现，在两种音律版本与两类情感反应下，所有测试的内在特征与12平均律差异度均以预期方向产生一致影响。该结果与当前多数音乐感知研究形成鲜明反差，后者往往默认外在预测因子相较于内在预测因子占据主导地位。本研究凸显了声学信号本身的内在特征，以及诸如12平均律差异度这类外在因素，在音乐情感感知中的双重重要性。