Masataka N
Department of Anthropology, Faculty of Science, University of Tokyo, Japan.
Child Dev. 1992 Oct;63(5):1177-85.
Discriminant analysis was used to distinguish statistically between the comfort-state vocalizations uttered by Japanese infants following 5 different types of pitch contours of maternal speech. Ontogenetic changes of their vocal behaviors were investigated during the first 5 months of life. Structural variability in infant vocalizations across variants of maternal speech was found to be characterized by a set of quantifiable physical parameters. However, infant's age when a vocalization was recorded was not an important contributor. Successful attempts at cross-validation, in which the discriminant profiles derived from one sample of vocalizations were used to classify a second set of vocalizations, indicated that the result obtained was not an artifact of using the same data set to derive the profiles to test reclassification accuracy. Proportion of cross-validated vocalizations that were misclassified decreased as age increased. The results of the present study demonstrate that a statistically significant relation exists between acoustic features of maternal speech and those of following infant vocalizations, and that such a relation strengthens over age.
判别分析用于从统计学上区分日本婴儿在听到母亲五种不同音高轮廓的语音后发出的舒适状态发声。对其出生后前五个月的发声行为的个体发育变化进行了研究。发现婴儿发声在母亲语音变体中的结构变异性具有一组可量化的物理参数特征。然而，记录发声时婴儿的年龄并不是一个重要因素。交叉验证的成功尝试表明，从一组发声样本得出的判别特征用于对另一组发声进行分类，结果表明所获得的结果并非使用同一数据集得出特征以测试重新分类准确性的人为产物。随着年龄增长，交叉验证中被错误分类的发声比例降低。本研究结果表明，母亲语音的声学特征与随后婴儿发声的声学特征之间存在统计学上的显著关系，并且这种关系随着年龄增长而增强。
