The computational approach of autocorrelation relies on recurrent patterns within a musical signal to identify and analyze the meter of musical passages. This paper suggests that the autocorrelation process can act as a computational proxy for the act of period extraction, a crucial aspect of the cognition of musical meter, by identifying periodicities with which similar events tend to occur within a musical signal. Three analytical vignettes highlight three aspects of the identified patterns: (1) that the similarities between manifestations of the same patterns are often inexact, (2) that these patterns have ambiguous boundaries, and (3) that many more patterns exist on the musical surface than contribute to the passage's notated/felt meter, each of which overlaps with observations from music theory and behavioral research. An Online Supplement at chriswmwhite.com/autocorrelation contains accompanying data.

自相关的计算方法依赖于音乐信号中的重复模式来识别和分析音乐段落的韵律。本文提出，自相关过程可以作为周期提取行为的计算代理，周期提取是音乐表认知的一个关键方面，通过识别音乐信号中类似事件倾向于发生的周期性。三个分析小插曲突出了已识别模式的三个方面：(1) 相同模式的表现形式之间的相似性通常不准确，(2) 这些模式的边界不明确，以及 (3) 音乐表面上存在更多模式而不是对段落的记谱/感觉计做出贡献，每一个都与音乐理论和行为研究的观察重叠。