Mechanisms that trigger emotional response to music

This topic submitted by Lydia, Chris, Michael, Montra ( MgMush08@aol.com ) on 12/15/03 .

Natural Systems 1 Syllabus---Western Program---Miami University

Abstract
Our experiment statistically revisited the now questionable Theory of Affects, which among other things states that songs played in minor keys are associated with negative emotions, while songs played in major keys are associated with positive emotions. We hypothesized that there would be no statistical correlation to the Theory of Affects. We administered surveys to people that asked the emotional tone of eight Baroque song clips on a scale of one (very negative) to seven (very positive). After analyzing the responses using an ANOVA and post hoc tests, our results showed a weak pattern supporting the Theory of Affects. Further research on the topic would be advisable since it would be useful in both the media and music therapy.

I. Introduction
Music is said to soothe the beast within. Music can affect the emotions greatly. This is seen mostly in arts and entertainment. Eerie music fills the theater when the stranded couple enters the haunted house. Composers may try to make music to recreate their feelings or tell some tale. Music is greatly intertwined with culture in general (Meyer). Exactly how music affects the emotions, or at least how Western music does this, is still up to much debate.
A systematic understanding of musicÕs effects could be put to many uses. It could be applied in the arts and in entertainment, and advertising would probably exploit musicÕs effects for marketing purposes. Such an understanding could also result in more positive things. Music therapist could create certain music to positively affect people with depression, anxiety, or everyday stress among other things. Since overall emotional health is shown to have a positive affect on physical health, prolonged use of positive music could be used to help a patient maintain a positive outlook through out treatments, chronic disease, etc. Simply put, well-constructed music could help people stay well balanced.

II. Background
Baroque music is a style of European music that flourished from 1600 to 1750 (Palisca, 2003). Based off the rhetoric theories of the Greeks and Romans, Baroque composers attempted to direct the emotions, or affects, of the listener. These affects were rationalized emotional states. The composers attempted to create a sense of unity in the work, and usually focused on a certain affect (Buelow, 2001).
Although this practice was common earlier, no official theory was developed till about the mid-17th century (Buelow, 2001). At this time theorists such as Mersenne, Kircher, Werchmeister, Pritz, Mattheson, Marpurp, Scheibe, and Quantz began to categorize the different affects. One focus of the studies were the connotations certain aspects of music such as scales, and rhythms. Eventually these theories were compiled by German musicologists Kretzschmars, Goldschmidt and Schering as the ÒTheory of the Affects.Ó It is from this theory that the notion that positive emotions are communicated by major keys and negative emotions are communicated with minor keys.
Keys are derived from musical modes. Modes have three meanings in Western music. Concerning keys, mode denotes the relationship between note values brevis and longa (short and long). In major key, or modus perfectus, the relationship is three to one respectively. In minor key, modus imperfectus, the relationship is two to one (Powers, 2003).
The Theory of Affects fell out of favor after the Baroque period however. More recently the Theory of Affects has been considered by some to be a conditioned response (Boge, 10/8/2003). According to this notion, the only reason why a person would associate major key with positive emotions is because he or she was somehow introduced to the idea that major key denotes positive emotions. Likewise a person who has never heard of the concept that minor key denotes negative emotions probably would not have a negative emotional response to a minor key.
Very recently, music theorists have been revisiting older theories. There is after all debate over how music arouses emotion (Madell, 2002). A recent study published in the Psychology of Music revisits the ancient Greek theory of harmonics, or the relationships to mathematics and music (Crickmore, 2003). Our project attempts to follow this precedent.
The project aims to revisit the Theory of Affects. Using statistics, we hope to test whether or not the response people have to music supports the Theory of Affects.

III. Materials and Methods
The basic materials for this test are the music selections we will use, the equipment used to copy and play it, the survey to collect data, and the rooms in which we will administer the survey. We also used people to take the survey. In addition, after the data was collected, we used a number of computer programs to organize and analyze the data.
Most of our music selections are samples from the website of the Baroque Music Club (http://www.baroquecds.com/musamples.html). They are all clips from longer pieces; they are not the whole work. We have used them with permission.

The music selections from this source are as follows:

Fantasia in c-minor (BWV 906) by Johann Sebastian Bach (from BMC 24)
Fughetta in C Major Ð BWV 872a (Book II, "48" Preludes & Fugues) by Bach (from Bach 732)
Italian Concerto in F Opening (BWV 971) by Bach (BMC 24)
Partita 4 in D major Ð Gigue (BWV 828) by Bach ( Bach 705)
Prelude and Fugue in a minor (BWV 897) by Bach (Bach 732)
Parthie in A major Ð Gigue by Johann Josef Fux (BMC 28)
Suite 3 in d-minor Prelude by George Frideric Handel (BMC 23)
In the above list, the first set of parentheses (BWV #) gives the identification number of the specific song according to a system used to identify each of BachÕs works. The clips each come from a CD that the Baroque Music Club has for sale, and the second set of parentheses tells which CD the sample came from.
One of our pieces comes from a collection of MIDI files of Book II of the Well-tempered Clavier by Bach edited by Dr. Yo Tomita. This piece is not an actual recording of a performance, but it is configured to sound like a harpsichord, and so it will be usable for our purposes. It also is used by permission. This piece is:
Fugue 12 in f minor (BWV 881) by Bach.

These pieces were recorded onto a CD and we then played them using a laptop computer.

We have attempted to use songs that would be generally unfamiliar to people in order to eliminate emotional impressions of a song due to the context of memory. If on the survey a person answered that they do have memories of a certain song, we did not use the data from that song. In this way we hoped to eliminate the problem of impressions coming from the setting in which a song could be remembered.
Once we collected the data, we entered the information into tables using Microsoft Word (see partial table from the in-class test run on next page). This allowed us to organize the data by any category we needed and aided our analysis later on. We could make graphs of this data in order to help predict and understand the results we get with our statistical testing.
The analysis of our data actually included a number separate steps or analyses, which was roughly categorized as control analyses and comparison analyses. The control analyses established the usability of our data by trying to determine whether or not variables other than key influenced our results. Once we established what portion of our data is comparable, we used that to conduct the comparison analyses.
Our control analyses tried to determine whether or not a listenerÕs mood or feelings toward Baroque music influenced their impressions of each sample. In analyzing both of these variables, we looked at the data for each song individually. In the case of mood, we compared the emotional description assigned to a song by listeners who said they were in a good mood to the description assigned by listeners who said in a bad mood, and likewise we compared the data from those in both a good mood and a bad mood to those who said they were in a neutral mood. In order to do this we did a series of three t-tests. We used a similar system for comparing responses based on feelings toward Baroque music. If these tests showed no statistical difference between these groups, we moved on to use the data having established that our results were not be based on these variables. If the tests showed that these variables do, in fact, impact the responses to the songs themselves, we did at that time choose to narrow or split groups so that we were always working with comparable data.
We analyzed the working of the songsÕ key on the results by the use of two comparison analyses. For the first analysis, we used a t-test to compare the response to each song written in major mode to the others, and likewise with those written in minor mode. If we found that there is a statistically significant difference between the response to one song in a major mode and the response to another song in a major mode (or minor mode different from another in minor mode), we would have concluded that the fact that both songs are in major or minor keys does not mean that they will be similar. This would have brought us to our answer in the inverse: we are looking for whether or not the key determines the difference; this says whether or not the key determines sameness.
For the second comparison analysis we used a t-test to compare the response to each song written in major mode and the response to each song written in minor mode. This would bring us to the answer of whether or not people respond differently to songs in major or minor keys. If there was a statistically significant difference in response between each set of major and minor songs, we would have conclude that the key did in fact make a difference in the response to major and minor keys. If there was not a statistically significant difference in response between the sets of major and minor songs, that would demonstrate that the key did not make a difference.
There are two or three ways in which we could have include the class in our project. The first thing we did with the class is perform a preliminary run of our test in order to smooth out any problems our design may have. We also set up a test time at the end of a lecture day, just in case our original tests did not yield enough data. This was a back-up day, and if we got enough data, we would not need it. Finally, we had the class help take the data from our surveys and input it into a computer spreadsheet. This saved us time, as the input process was a rather long one. These are the ways we used our class day.

IV. Results
The ratings individuals gave to each song varied greatly. As shown in Figures 1 and 2, each song received ratings in at least four out of the seven categories, and oftentimes more. The number of times each rating was assigned, however, showed some patterns.

Song 1 received scores from two up through seven, with ten of these falling in the range of two through four, and seventeen of these representing scores of five to seven. Song 4 was rated anywhere between two and six, though most often it was rated either three or four. Song 6 also received scores in a range from two and six, with most of these being scores of three, four, or five. Song 7 received ratings in categories one through six, although the vast majority of these were ratings of one, two, or three. For the most part these songs, which were all written in minor keys, tended to receive scores concentrated in the lower or middle ranges of the scale.
The songs written in major keys did not show quite the same tendency. These scores tended toward the higher ranges, as distinctly illustrated by Song 2. The scores for Song 2 all fell on a range of four through seven, with six being its most common score, followed by five. Song 3 received ratings anywhere between two and six, with most of these being ratings of five, followed by an equal number of ratings four and six. Song 5 received scores ranging from three through seven, with its two highest categories being five and six. Song 8 received scores in all of the categories from two through seven, with the largest number of these ratings being in the categories of five and six. Overall the songs in major keys tended to be
concentrated on the higher end of the scoring range.

Within the frequency of song ratings there emerged a pattern: songs written in a minor key were rated on the middle to lower ends of the scale, and songs written in a major key were rated closer to the higher end of the scale. This pattern came out even more clearly in the means of the scores that each song received. As shown in figure three, when the songs were placed in order from the song with the lowest mean score to the song with the highest mean score, all of the songs written in minor keys ended up together with the four low mean scores, while all of the songs written in major keys ended up together with the four high mean scores.

While these trends seemed to indicate a correlation between the rating a song received and the key in which it was written, there was often very little difference between the mean ratings of two songs. Indeed, as shown in Figure 4, the difference between the mean of Song 1 Ð the song in a minor key with the highest mean score Ð and Song 3 Ð the song in a major key with the lowest mean score Ð is only .16, the second lowest difference between any two songs.
In order to determine which differences between the song means were significant enough to consider, we ran an ANOVA test that compared the data from each of our songs to each of the others. The P-value we received from this test was <.0001, from which we could determine that somewhere between the ratings given the songs there was a statistically significant difference. We then ran both ScheffeÕs and FisherÕs PLSD posthoc tests to find out between which songs the statistical differences were. It was found that two songs considered statistically different in ScheffeÕs posthoc test were always statistically different in FisherÕs PLSD posthoc test, but songs considered statistically different in FisherÕs PLSD test did not necessarily show up as different according to ScheffeÕs test.
Figure 4 displays the means of the different songs and the statistical significance of the difference between each. As illustrated there, with a mean of 2.39, Song 7 has by far the lowest mean score. Indeed, the mean of Song 7 differs from the song mean closest to it by 1.18, the largest difference between any two song means. Both the ScheffeÕs and FisherÕs posthoc tests found there to be a statistically significant difference between the mean score of Song 7 and the means of all the rest. So there was a statistically significant difference between this lowest rating of the songs written in minor keys and all the other songs in minor keys, as well as a difference between it and the songs written in major keys.

At 3.57, the mean of Song 4 was not significantly different from the 3.78 mean of Song 6, and only the FisherÕs PLSD test considered the difference between Song 1Õs mean of 4.59 and Song 4Õs mean to be statistically significant. There was a statistically significant difference between the mean of Song 4 and the means of all the songs written in major keys.
Song 6 showed no statistically significant difference from Song 4. In the ScheffeÕs test, it also showed no significant difference from Songs 1 and 3, though the FisherÕs PLSD test did show a difference in those cases. There was therefore a definite statistically significant difference between the 3.78 mean of Song 6 and the three highest means (beginning at 5.14) of the songs written in major keys, and perhaps also a statistically significant difference between it and the means of the last song in major key and the highest rated song in a minor key.
The mean of Song 1, according to the ScheffeÕs posthoc test, showed no statistically significant difference from any of the songs except for Song 7. The difference between itÕs mean of 4.59 and the means of all the songs in major keys and two of the other songs in minor keys would have been statistically insignificant. FisherÕs PLSD test gave slightly different results, showing a statistically significant difference between Song 1 and Songs 4, 6, and 2. In this case there would be a statistically significant difference between the mean of Song 1, which was in a minor key, and all of the other songs in minor keys, as well as Song 2, which had the highest mean of the songs in major keys.
With a mean of 4.76 Ð the lowest mean out of the songs written in major keys Ð Song 3 showed a statistically significant difference from Songs 7 and 4, which were both in minor keys. According to FisherÕs PLSD test, there was also a statistically significant difference between the means of Song 3 and Songs 5 (in a minor key) and 2 (in a major key). There was therefore a significant difference between the mean of Song 3 and the means of two or three of the songs in minor keys, as well as perhaps the mean of one of the songs in a major key.
The 5.14 mean of Song 8 showed a statistically significant difference from Songs 7, 4, and 6, the three songs with the lowest means among those in minor keys. There were no other statistically significant differences. Therefore, the differences between the mean of Song 8 and all of the other songs in major keys were statistically insignificant, as well as the difference between the mean of Song 8 and the mean of Song 1, which was in a minor key.
Song 5, with a 5.17 only .03 different from Song 8, showed the same statistically significant differences as Song 8 exactly.
According to ScheffeÕs test, Song 2Õs mean score of 5.67 showed the same statistically significant differences as Songs 8 and 5. FisherÕs PLSD test, however, showed a statistically significant difference between the mean of Song 2 and those of Songs 1 and 3. According to FisherÕs test, there would be a statistically significant difference between the mean of Song 2 and those of all the songs in minor keys, as well as the mean of one of the songs in a major key.
Discussion
Although the results we got from the surveys taken agreed with the Òtheory of affectsÓ they did not support our hypothesis. From the surveys we got, most of the songs that in average were positive, were the songs that were in a major key, and most of the songs that were negative in average, were the songs in minor keys.
Over all, the number of surveys taken were only 30, which didnÕt give us accurate results, although even more number of surveys could have resulted in the same results that we got. One of our problems with getting more surveys was gathering people to play the music. Playing the music for individual people would have taken a long time, so we tried to gather as many people at one time, which was the reason why we had difficulties getting more surveys.
The method of measurement that we used was very subjective, which could have obscured the results. We started with two surveys first to try to figure out which survey worked better. With the better survey we still think that our method was subjective. We could have had our scale of 1-7 for the answers to be larger like a scale from 1-15, which would have given us more accurate data, but yet would have been much more difficult to work with. Another way that we could have gone about our research is that we could have picked songs that were the same but played in two different keys. This would have given us much more accurate data. We could have done that by picking a song and recording the song while someone played it on an instrument in two different keys. Because of our limits we were not able to do that.
We also think that the reason why we got the results that we did was because the people that we surveyed were not necessarily taught to think that a song in a major key is emotionally happy, but because of the environment around them that they were raised in, they were conditioned into thinking that way. For example in movies the songs played in a sad scene are usually in minor keys, and in very happy scenes, in major keys. We believe the best way to get accurate results would be to test a group of people who were raised in a jungle with no music experience, and somehow explained what we were doing and surveyed them. That case would give us accurate results because the people would not have been exposed to music.
Besides all the points that we made about how our research could have been better, we still believe that there is much more about our topic that could be researched about.

Works Cited

ÒAPA Style Made Easy.Ó [WWW document http://apastyle.net] Perry: askSam Systems. Accessed October 28, 2003.
-citation help

Boge, Claire. Personal Interview. October 8, 2003
-information of Theory of the Affects

Buelow, George J. ÒAffects, Theory of.Ó The New Grove Dictionary of Music and Muscians. Oxford: Oxford University Press. [WWW. Document http://www.grovemusic.com/shared/views/article.html?section=music.00253&anchor=music.00253] Accessed October 22, 2003
-Theory of the Affects

Crickmore, Leon (2003). ÒA re-valuation of the ancient science of harmonics.Ó Psychology of Music. London: SAGE

Publications. Vol. 31 No. 4 pg. 391-403
-revisits harmonics, sets precedent for expirement

Madell, Geoffrey (2002). Philosphy, Music and Emotion. Edinburgh: Edinburgh University Press Ltd.
-debate over musicÕs arousal of emotions

Meacock, Michael, ed. A Baroque Music Sampler. Baroque Music Club. [WWW. Document http://www.baroquecds.com/ musamples.html]
- Fantasia in c-minor (BWV 906) by Johann Sebastian Bach (from BMC 24)
Fughetta in C Major Ð BWV 872a (Book II, "48" Preludes & Fugues) by Bach (from Bach 732)
Italian Concerto in F Opening (BWV 971) by Bach (BMC 24)
Partita 4 in D major Ð Gigue (BWV 828) by Bach ( Bach 705)
Prelude and Fugue in a minor (BWV 897) by Bach (Bach 732)
Parthie in A major Ð Gigue by Johann Josef Fux (BMC 28)
Suite 3 in d-minor Prelude by George Frideric Handel (BMC 23)

Meyer, Leonard B. (1956). Emotions and Meaning of Music. Chicago: University of Chicago Press.
-relationship of music and emotions

Meyer, Leonard B. (1967). Music: The Arts and Ideas. Chicago: The University of Chicago Press.
-relationship of music and culture

Palisca, Claude V. (2003). ÒBaroque: Etymology and early usage.Ó The New Grove Dictionary of Music and Muscians. Oxford: Oxford University Press. [WWW. Document http://www.grovemusic.com/shared/views/article.html?section=music.02097.1] Accessed October 22, 2003.
-information on Baroque music

Powers, Harold S., et al. (2003). ÒMode.Ó The New Grove Dictionary of Music and Muscians. Oxford: Oxford University Press. [WWW. Document http://grovemusic.com/shared/views/article.html?from=search&session_search_id=3777336694&hitnum=1section=music.43718] Accessed October 22, 2003
-information on major/minor keys

Tomita, Yo (1998). The Second Book of the Õ48.Õ [WWW. Document http://www.music.qub.ac.uk/~tomita/midi.html]
- Fugue 12 in f minor (BWV 881) by Bach

Next Article
Previous Article
Return to the Topic Menu