A sample screenshot showing emotion descriptors, subject selection, and ratings strength
Emotion in Music Experiments
Nick Knouf
We have recently begun a series of experiments into various methods of measuring emotional responses to music. Our first pass involved having listeners select a subset of emotional descriptors (chosen from prior studies of musical emotion) and assign a strength rating to each selected word. We then used unsupervised machine learning techniques to examine the distribution of responses with comparisons to various characteristics of the participants in the experiment. Early results have shown good correspondence with prior work. Further experiments and analysis is underway.
Sample distribution of subject ratings across a set of descriptors and musical excerpts. Yellow colors indicate higher strength ratings.
However, we have to consider that using techniques such as this will not illuminate the answers we seek. For subjects may not have ready access to descriptions of their emotional experience, and no matter how many words we allow subjects to select, the distribution of words will not reflect the underlying experience. We are exploring non-verbal "reactions" to music as an alternative means of measuring emotional responses. These experiments are in the early stages but ongoing.
Feel free to contact us for an early draft of a paper describing this work.
Current Directions in Music, Mind, and Health:
A focus on Alzheimer's DiseaseA focus on Autism
A focus on Emotion
A focus on Performance
Selected Projects in Music, Mind and Health:
- Experimental Understanding of Emotional Responses
- Hyperscore in the Hospitals
- Embedding Cognitive Evaluation in New Music Technologies
- Mathematical Models of Musical Performance