Abstract—Earlier studies seem to indicate that harmony plays a relatively unimportant role in music identification tasks. However, identifying music from its harmony, although a difficult task, can inform our understanding of how listeners mentally represent and remember harmony. The present study is the first to examine the ability of listeners with varying musical expertise to identify well- known classical and pop/rock pieces from their chords. In this study, well-known classical and pop/rock pieces selected in a pilot were used instead of traditional tunes because their chord progressions tend to be more distinctive and their different renditions often preserve the original harmony with little or no modifications. One hundred participants, divided into four groups according to musical expertise, listened to chord progressions of representative passages from 12 pieces and were asked to identify them. Two ways of modifying pitch information in the chord progressions were explored: (1) The progressions were played with piano tones and shepard tones, the latter diminishing the effect of melodic cues. (2) The progressions were played on six transpositional levels. In the second part of the experiment, the participants heard commercial recordings of the same pieces and were asked to identify them. The study showed that music identification from harmony in an open-set task is possible, even when melodic and rhythmic cues are largely missing. Additionally, both the timbre with which the chord progressions were played and the expertise of the participants were statistically significant factors for identification; the professionals being best. We found that stimuli using shepard tones were more difficult to identify than those using piano tones, and that the major part of the difference was explained by the downgrading of melodic cues, not by timbral associations. The differences between transpositions were not statistically significant. Keywords—harmony, memory for harmony, music memory, tune identification. I. INTRODUCTION HEREAS the effects of pitch and rhythmic information on tune identification have been studied extensively [1], the effect of harmony for tune identification has seldom been examined [2], [3]. There are several reasons that could explain why the effect of harmony on tune identification has not been studied more often. First, listeners can easily identify a song or piece of music based on the pitch and rhythmic features of its melody alone, without hearing its harmonic accompaniment [1]. Second, the chords that accompany a well-known melody can be changed without altering the identity of a tune [2]. Third, many tunes imply or are typically accompanied by the same or similar chord progressions, and this one-to-many correspondence can weaken the association between a chord Ivan Jimenez is with the Department of Music, University of Pittsburgh, Pittsburgh, PA 15260 USA and the Sibelius Academy, University of the Arts, Helsinki, Finland (e-mail: iej2@pitt.edu). Tuire Kuusi is with the Sibelius Academy, University of the Arts, Helsinki, Finland (e-mail: tuire.kuusi@uniarts.fi). progression and a specific tune [4], [5]. Additionally, it is difficult to measure the effect of harmony independent of the effect of melody and rhythm since a chord progression cannot be instantiated without also creating some degree of melodic motion and rhythmic activity. Reference [3] found that adding chords to a single-pitch rhythmic version of well-known tunes made identification more difficult, and it is likely that the melodic line created by the highest notes of each of the chords in the progression provided participants with a wrong melody that misled them in their process of tune identification. One way to avoid the problem of incorrect melodic cues is to voice the chords in such a way that the highest note of each chord in the progression corresponds to the most representative pitch of the melody during that time-span. The use of this type of harmonic reductions to summarize the underlying structure of music dates at least as far back as the mid-19 th century [6], and formal musical training usually involves the study of harmonic reductions [7], [8]. Empirical evidence also suggests that listeners can distinguish between correct and incorrect harmonic reductions of a musical passage [9], [10]. Until the present study, however, the ability to identify music from harmonic reductions has not been empirically tested. II. PILOT STUDY Forty pieces, 20 popular songs and 20 pieces of classical music, were selected for the pilot based on their inclusion in recent publications: studies on tune identification [11]-[13], a study on implicit absolute pitch [14], a corpus analysis [15], a CD compilation of popular classical pieces [16], as well as because of their popularity among undergraduate music theory students at the University of Pittsburgh, who were surveyed between the years 2010 and 2014. Classical and pop/rock pieces were used instead of traditional tunes because their chord progressions tend to be more distinctive and their different renditions often preserve the original harmony with little or no modifications. In the pilot study, we tested 18 participants’ ability to identify the 40 pieces from harmonic reductions played with a digital grand piano sound. Harmonic reductions consisted of six, seven, or eight chords from the initial phrase, initial period, or another representative section of the pieces and every chord lasted 2 seconds. Most harmonic reductions had four voices, and all voices had only one note per chord. Since the excerpts chosen for this pilot were the initial phrases of relatively simple pieces and songs, durational accents provided a straightforward criterion for choosing most of the notes. Harmonic reductions preserved both the contour of the different voices and chordal inversions from the original excerpts. Pitches were verified using Sonic Visualiser software. The test was administered online. After listening to each harmonic reduction, listeners were asked to select the name of the piece from a list of the 40 Music Identification from Harmony Ivan Jimenez, Tuire Kuusi W ISBN 1-876346-65-5 © ICMPC14 ICMPC14, July 5–9, 2016, San Francisco, USA 208 Table of Contents for this manuscript