Monday, September 19, 2005

Music and Emotion

Dave Munger has a good post up about a study on the effect of tempo and mode (major or minor) on the perception of emotion. I offer some criticism, however. Emotion was limited to happy and sad, which is far simpler than the most-used models of emotion in psychology. This will skew the results when multidimensional responses are forced into a two-choice response.

This study reminds me of a paper given at the Leipzig conference last spring. Emmanuel Bigand and his team decided to investigate how quickly people identify the emotion of a musical work. They made 27 musical excerpts one second in duration, and asked participants to group the works by their emotional similarity. There was a high degree of agreement between subjects and within subjects, even with this very short time period. In fact, the results even agreed with a similar experiment using 25-second excerpts, suggesting that 1 second is enough to convey emotional content. Emboldened by these results, Bigand's team had participants evaluate how emotionally moving a variety of excerpts were on a 10 point subjective scale. These excerpts started as only 250 milliseconds. Next, the participants judged the same excerpts, but they were extended to 500 ms, and then 1, 2, 5, and 20 seconds. Incredibly, some judgements were consistent from the 250 ms duration through the 20 second duration, meaning that the listeners made the judgments of emotional affect in very short time periods.

As a tangent, Emmanuel Bigand has an interesting Shockwave animation of Chopin's Prelude in E major. While it is a serious explication of the cognitive significances of Fred Lerdahl's Tonal Pitch Space, it's also plain old fun to watch. Zoom out to get the big picture, or zoom in to watch the movement among the "moons".

E. Bigand, "The Time course of emotional responses to music," paper presented at The Neurosciences and Music - II conference in Leipzig, Germany on May 7, 2005.


Michael J. West said...

I understand your criticism of the Munger piece, but I dispute it. The models of emotion as "happy" and "sad" are extremely useful in musical analysis because its emotional quality is so subjective to the listener. In fact, it might even be more subjective as the music's complexity is greater.

Put another way, the dimension and details of the "happiness" or "sadness" of the music can only be gauged by what it specifically evokes in the listener. If a Bach piece, or a Dexter Gordon one, reminds me of a specific time or place in my life, that colors the whole perception, doesn't it?

So "happy" and "sad"--possibly the most simple emotional characterization you can work with--has to be kept pretty much to the Least Common Denominator. Trying to give it more depth can only skew the possible responses away from one portion or another of the audience sample.

Scott Spiegelberg said...

First, I should clarify that my criticism is of the research done by Dalla Bella et al, not the post by Dave Munger.

Second, it is the very limiting nature of only two dimensions that make the happy/sad contrast less useful for psychological studies. As you say, music is very complex in emotional content, and a study that only samples a tiny fragment of that emotional content is not giving an accurate picture.

One easy fix would be to include another dimension, activity. Many models of emotion have a Cartesian map of Positive/Negative for one axis and Active/Passive for the other axis. SOme pieces are judged as neutral in terms of happiness or sadness, but can be exciting or sleepy. Others have mixes of the two axes, allowing for a simple way to measure more complex emotional judgements.

Michael J. West said...

I should also clarify: I'm not a psychologist by any means. I'm merely a music appreciator (if such a term exists).

But in my purely lay opinion, it does still seem to me that even the double-axis emotional model would provide heavily subjective responses. I'm thinking particularly of Deryck Cooke's models, with the famous characterization of the augmented fourth as "active anguish in a context of flux."

Just a little too complicated to have any universal meaning, at least from my unschooled perspective.