This study examines the difference between prominent and non-prominent lines (e.g., melodies and accompaniments). After reviewing research suggesting that lines with few repeating patterns would readily capture attention, the hypothesis that prominent lines tend to be less repetitive is tested. Various probabilistic models are used to quantify the repetitiveness of lines from three musical corpora—two containing Classical string quartets, one including songs by the Beatles. The results suggest that notes from prominent lines tend to have lower probability. This trend, along with others found in the corpora, is consistent with the hypothesis that prominent lines are generally less repetitive.

This study uses a corpus of excerpts from eighteenth- and early nineteenth-century string quartets to examine how four acoustic cues—onset and offset synchrony, pitch comodulation, and spectral overlap—help to afford the perception of auditory streams. Two types of streams are dealt with: textural streams , which house individual string parts or groups of them that function as single musical units; and music streams , which typically house the music as a whole and distinguish it from other simultaneous sounds of music. The corpus contained real excerpts from classical string quartets as well as synthesized excerpts in which lines from two different quartets were combined. Both the author and ten survey respondents analyzed the corpus, identifying likely textural streams. Each of the four acoustic cues was modeled computationally, in order to assess its prevalence in textural and music streams found in the corpus. The results suggested that some cues are more important than others in establishing textural streams, music streams, or both.