Emotion Content of raag

I am fundamentally interested in how music evokes emotions. How can something so abstract and seemingly unconnected with adaptive behavior, induce such profound, visceral states. My research focuses on expectation evoked-emotion, such as the tension and resolution that occurs when music satisfies or defies expectations, as well as innate auditory cues.

Survey

We recently undertook a web-based survey in which participants were asked to listen to ten raag excerpts and rate the extent to which each of twelve emotions ('happy', 'sad', 'angry/aggressive', 'awakening/arousing', 'calming/soothing', 'devotional', 'emotional/passionate', 'loving/romantic', 'peaceful', 'powerful/strong', 'restless', 'tender/soft') were evoked by the excerpt. To ensure generalization, several excerpts were used for each raag, and one was chosen randomly for each subject.

A total of 553 subjects participated, making a total 21,377 judgments, where each judgment was the assignment of a score to a particular emotion for a particular raag excerpt. We also collected some demographic data specifically, age, sex, years of musical training, and overall familiarity with North Indian Classical Music (NICM). 40.0% of subjects reported having 'no' familiarity with North Indian classical music, 30.7% 'a little', 15.4% were 'somewhat' familiar, and 9.95% were very familiar.

I presented the results of this experiment at the International Conference for Music Perception and Cognition (2008) in Sapporo, Japan.

Survey results

It is clear from the data that there are significant differences between the emotions that raags elicit. Broadly, it seems that raags cluster according to valence. Informally, there seems to be a clear correspondence, at least for this set of raags, between scale degrees used and valence. Raags that use 'minor' notes such as the m2 and m6 scale degrees are considered negatively valenced, while raags that use 'major' notes tend to be positively valenced.

An ANOVA analysis showed that raag was a significant factor in emotion responses (p < .001). Using confidence intervals adjusted for multiple comparisons (p < .05), we found that, for many of the emotions, a clear distinction could be made between many raags. Across many emotions, a clustering pattern was evident, with a group of positively valenced raags, Khamaj, Desh and Bageshri, a medium one consisting of Bhimpalasi and Yaman, and a negatively valenced group with Gujari Todi, Marwa, Shree and Darbari. Certain emotions were found to apply more or less equally to all raags. Almost all raags were considered not 'angry/aggressive', while at the same time most raags were rated highly 'emotional/touching'.

Additionally, we found that familiarity with NICM was a significant main effect for many emotions. The general pattern was that 'very' experienced listeners tended to rate the excerpts as more intensely emotional (e.g. more sad or more peaceful) as compared with listeners with little or no prior exposure.

We also found that the highest scoring emotions for a given raag conformed well to traditional descriptions, whether or not the listener was at all familiar with Indian music. For example, Khamaj and Desh were 'happy', 'peaceful' and 'calming', whereas raags such as Darbari, Shree and Marwa were 'powerful', 'sad', 'strong', and 'restless'.

Visualize survey data

We are making the main survey results available to you. Using the controls below, you can call up charts of the results. You can sort by raag or by emotion, and filter the results by various demographic variables. This will generate a chart showing the average values given the constraints you specify. You may add any number of charts to be able to easily compare them on the same page.