An Equilibrium of Predictability in Song Melodies

Daniel Dudney, Texas Tech University

In his 2006 Sweet Anticipation, David Huron conjectures that one's auditory expectation is derived predominantly from exposure-based statistical learning. Huron draws from a corpus of German folk melodies to compute two first-order tendencies in detail: scale-degree succession and rhythmic attack point succession. He then posits this data as a template by which to derive a predictability value for each succession.

This study analyzes examples from the genre of song through the lens of Huron's melodic and metric models, uncovering dynamic interactions between these two different types of predictability. Through this process, I extract and illuminate a pronounced tendency that may be pervasive across the genre: the predictability values for pitch and event onset often exist in an inverse relationship. That is, when a less predictable scale-degree occurs, it is often placed at a fairly predictable metric position, and vice versa. To demonstrate this regularity as one that is ubiquitous in song, I explore excerpts from a wide range of cultures, time periods, and styles.

Indeed, the nature of song gives rise to a wealth of unique characteristics and tendencies. The inverse predictability trend may have arisen from compositional considerations such as a desire to effectively communicate the text, and a regard for the human physiology. This relationship seems to manifest, on a deeper level, an ongoing pursuit for perceptual equilibrium -- seldom reaching it until, perhaps, the conclusion of the song.