Yuri Broze
Cognitive and Systematic Musicology Laboratory
Edward F. Hayes Graduate Research Forum
Ohio State University
Feb 24, 2012
Johann Sebastian Bach (1685-1750)
Music is commonly described using emotional terms, such as "happy," or "sad."
However, music is not an animate agent, and is incapable of experiencing emotional states.
Therefore, it is problematic to claim that music expresses emotions in the same way that your friend Janet or cousin Louie might.
Instead of believing that music expresses emotions, one might instead say that music is "expressive of emotions" (Kivy 1980).
If a visual percept has features which resemble a human emotional expression, we might perceive it as expressing an emotional state (Davies, 2010).
Recognition of certain facial expressions of emotion appears to be cross-cultural.
Contour theory would predict that these six "basic emotions" might be perceived in artworks which show resemblance to such expressions.
However, music doesn't have a face, postures, or visible shapes: it is heard, not seen.
Juslin and Laukka (2003) found that acoustic cues for basic emotions are very similar in speech and music. (Slow speeds, quiet sounds, low pitch level, falling pitch contours, etc.).
These acoustic cues for basic emotions also appear to be valid cross-culturally (Balkwill et al 2004, Fritz et al 2009, Mohn et al 2011).
Music appears to be expressive of the so-called "basic" emotions at least in part because it mimics prosodic and intonational patterns of emotional human speech.
For example, Hevner (1936, 1937) defined clusters for "exciting," "serene," "vigorous," "dignified," "dreamy," and "graceful" as well as "happy" and "sad."
Polyphonic music includes several melodies at once, and therefore does not resemble a single human voice.
In the visual domain, Heider and Simmel (1944) demonstrated that moving geometric shapes can spontaneously be perceived as animate agents with goals and feeling states.
Participants spontaneously described not only "basic" emotions, but also more complex affective states such as frustration, rage, hesitation, love, and vengeance.
These emotions might be more complex than the "basic" emotions described by straightforward contour theory.
In particular, perception of "social" emotions such as love, jealousy, pride, shame, or loneliness might depend on the presence or absence of a polyphonic texture.
Listeners might use information about the number of musical voices present --- the music's voice multiplicity --- to infer the expression of certain emotions.
We conjectured that polyphonic music might be expressive of different emotions than monophonic music.
In particular, listeners might be aware of the number of "musical lines" present in a musical texture: the music's voice multiplicity.
They might then use this information to make judgments as to the emotional content of the music.
This might be particularly relevant in detecting a "social" emotion such as loneliness.
However, this should only be true to the extent that listeners are correctly able to perceive the number of musical lines in an excerpt.
Huron (1989) showed that while listeners can easily identify 1-voice and 2-voice textures by sound, they easily confuse 3-voice textures with 4-voice textures.
That is, perceptual categories for voice multiplicity appear to be "One," "Two," or "More."
This "signature pattern" can be used to design a test for the specific effect of voice multiplicity per se on loneliness perception.
Thus, listeners should rate 1-voice music as lonelier than 2-voice music, and 2-voice music as lonelier than 3-voice music. However, 3-voice and 4-voice should sound equally lonely.
We asked OSU School of Music Undergraduates (N=25) to rate brief (5s) musical excerpts for expression of happiness, sadness, loneliness, and pride.
These emotions were chosen to provide two "basic" and two "complex" emotions, with one positive and one negative emotion per type.
We controlled for musical style and mode by excerpting all stimuli from minor-mode fugues from J.S. Bach's Well-Tempered Clavier.
Four stimuli of voice multiplicity levels 1, 2, 3, and 4 were sampled from each fugue's exposition section. These formed sets of four, controlling for the effect of tempo, melodic content, and motivic content. In all, 36 stimuli were generated.
Recordings of harpsichord performances were used, due to the instrument's lack of dynamic control. Furthermore, a single human performer was responsible for all textures, regardless of musical voice multiplicity.
For each participant, all 144 stimulus/emotion combinations were presented in fully randomized order to be rated using an unlabeled slider in a computerized interface.
Following the experiment, each participant was asked follow-up questions to check for possible demand characteristics, and debriefed.
As a statistical test of Hypothesis 1, we conducted a one-sided t-test comparing the difference in loneliness rating per person per fugue between voice multiplicities of one and four, and found that this difference is statistically significant (t192 = 12.97, p < .001).
Multiplicity effects for loneliness were consistent with the pattern predicted by limitations in stream segregation. A one-sided paired t-test indicates that the predicted ‘elbow’ in loneliness ratings does occur (t190 = 2.49, p = 0.007). This result is consistent with the idea that that the effect is due to voice multiplicity per se, and not other concomitant changes in the musical surface.
The two "complex" emotions appear to show greater sensitivity to changes in voice multiplicity than the "basic" emotions, consistent with our conjectures.
Surprisingly, all emotion ratings are much stronger for monophonic music than for polyphonic music. Perhaps emotion expression cues are more easily perceived when only a single melody is present.
No emotion ratings display significant differences between 3-voice and 4-voice multiplicities. This suggests the perception of harmonic textures is qualitatively different than those with one or two voices alone.
Special thanks to my collaborator Brandon Paul, and advisor David Huron.
