Göran Krantz & Bjorn Merker & Guy Madison

Melodic intervals and body movement.

Krantz, Göran & Merker, Bjorn & Madison, Guy: "Melodic intervals and body movement", 17th International Congress on Dance Research, Naxos, 22-26/10, 2003.

Abstract

It is often assumed that musical intervals carry meaning beyond that of the simple pitch difference they embody, yet substantive studies of the issue are few and far between. In an earlier study, employing free word responses from a large population of volunteers, we found evidence that musical intervals indeed tend to be perceived in emotional and aesthetic terms with some degree of consistency. The present study employed a different response measure on a smaller population of volunteers: 28 participants were asked to portray their impression of musical intervals by non-verbal movement or gesture. Responses were recorded on video. As a first step towards evaluating the responses, two observers were asked to assess each response on each of 14 rating scales anchored in bipolar word pairs such as up-down, passive-active, gloomy-cheerful and receive-repulse. Fair inter-rater correlations were obtained for some of the scales, and these scales disclosed a number of trends in participant responses to the musical intervals. One of these - conspicuous on scales such as up-down - appears to reflect the size of the pitch difference across intervals. Some intervals, such as second, sixth and seventh tend to break this pattern on some of the scales, suggesting an influence on subject movements of factors other than simple pitch difference, presumably from factors of an emotional or aesthetic nature.

1. Introduction

Throughout the history of music we find a widespread belief that musical intervals carry meaning beyond that of the simple pitch difference they embody. Few systematic investigations of the issue are available, however (Huber, 1923; Maher, 1980; Makeig, 1982; Smith and Williams, 1996; Krantz, 2001). The last mentioned study collected free word reports from 228 participants exposed to the eight diatonic intervals, presented melodically. The results provided evidence that musical intervals indeed tend to be perceived in emotional and aesthetic terms with some degree of consistency, as reflected in the statistics of the choice of verbal descriptors. In an attempt to extend these results with a different response measure, the present study asks whether any consistent patterns of response might emerge when volunteers are asked to perform body movements in response to heard musical intervals. The approach is similar in some ways to that of Truslit (1938), but instead of using melodies and musical phrases the present study employed the musical intervals themselves.

2. Method

Twenty-eight participants - 23 women and 5 men - aged 11 to 59 years were tested.
Participants had no special background in the field of music or dance. They ranged from schoolchildren, over students from different disciplines such as agricultural-, artistic-,
teacher education, to adults active in working life. Stimuli consisted of the ascending (melodic) intervals in c-major, diatonic scale. They were played manually by the first author, from a hidden position, on a normally tuned grand piano in mezzoforte. The sequence was always C1 (for 1 s duration), pause (1 s), second tone (1 s), followed by a longer pause. To control for possible sequence effects in the order of presentation of intervals, the order of stimuli was randomly changed across participants. In what follows, when intervals are identified by number, these are 1 = prime, 2 = major second, 3 = major third, 4 = fourth, 5 = fifth, 6 = major sixth, 7 = major seventh, 8 = octave.

Participants faced a video camera from eight meters distance in a large room offering freedom of movement. They were asked to first listen with concentration to the interval and then to express their inner response through movement. The video-filmed responses were edited by removing dead time and ordering responses according to musical interval. Edited videos were reviewed by two expert observers who rated responses on seven point rating scales representing each of the following bipolar descriptors: down-up, close-distant, inwards-outwards, wide-narrow, closed-open, round-straight, undirected-directed, symmetric-asymmetric, passive-active, relaxation-tension, smooth-jerky, uniform-varied, receive-repulse, and gloomy-cheerful.

Observers rated only the visual sequences, and had no information regarding the musical intervals to which the movements were made. They were given no instructions regarding how to apply the rating scales to the movement sequences beyond that of using them to gauge their impression of the movement.

3. Results

First of all we consider the consistency in ratings across the two expert raters. A common indicator in this situation is the correlation coefficient across all ratings. These were not as high as one generally likes to see (> 0.80) and varied considerably among the 14 bipolar scales. For example, the four scales with the highest inter-rater correlation were down-up (0.75), inwards-outwards (0.64), passive-active (0.65), and gloomy-cheerful (0.60). However, these correlations are based on vectors of data (N = 224) that comprise variability from two independent sources, namely intervals and participants. Possible interactions between these upon the raters judgements may therefore yield a correlation coefficient which underestimates the true consistency pertaining to the intervals alone. We therefore obtained correlations based on the mean ratings for each rater across all participants, that is, for two vectors with only 8 data. These correlations were 0.95, 0.97, 0.93, and 0.83 for the four scales listed above, respectively. This suggests that the raters did indeed pick up some invariant features of the movements that relate strongly to the stimulus intervals. Figure 1 depicts these four rating scales.

Figure 1. Mean ratings of movements in response to each of the eight pitch intervals. Error bars depict 95 percent confidence intervals. Note that each scale is bipolar, such that a rating of 4 is neutral in the sense that it is neither active nor passive, for example, while higher ratings denote increasing activity and lower rating denote increasing passivity.

A conspicuous diagonal orientation of ratings are in evidence above all for the scales up-down, outwards-inwards and active-passive. That is, the mean ratings on these scales tend to reflect or correspond to interval size, from prime successively up to octave. Exceptions to this general trend are in evidence for the second and the sixth, in particular on the up-down and cheerful-gloomy scales, on which the second and sixth receive more extreme ratings than the general trend would indicate. We note also the remarkably low rating variance on the scale cheerful-gloomy for movements performed in the presence of the fifth.

It is to be emphasised that the results presented so far are only for the scales with the highest inter-rater correlation on both correlation measures. Additional significant results are contained on some scales that did not reach the same levels of inter-observer correlation. One example is judgements on the scale receive-repulse, which had an inter-observer correlation of 0.51 based on raw scores, and of 0.92 based on means. This descriptor showed a strong significant effect for the seventh interval, as illustrated in Figure 2.

Figure 2. Means and 95% confidence intervals for judgements of movements on the scale receive-repulse. Note the different scale on the y-axis compared to Figure 1.

A factor analysis was performed in order to test our impression that the ratings suggested a few common dimensions. The analysis was based on raw data rather than means across participants and raters. The reasons for this was first that only the eight intervals would constitute too few cases for the analysis, and second that we wanted to include the inter-rater variability, since the correlation between the raters was only moderate. Principal component extraction yielded four factors with eigenvalues greater than 1, but the fourth factor had small loadings throughout and resisted interpretation. The scree plot also indicated that three factors were more appropriate. The three-factor solution detailed in Table 1 explained 50.4 percent of the variance, which is rather good considering that the raw data include both inter-participant and inter-rater variability. Factor I assumes the lion’s share of the variance (>30%), summoning adjective pairs like down–up, close–distant, inwards–outwards, wide–narrow (negative), closed–open, undirected–directed, passive–active, and gloomy–cheerful. Although the one emotional scale was included, factor I seems essentially to constitute a spatial dimension. Factor II (11%) appears to reflect harmonic tension, and the fact that consonance is found to be smooth and is preferred (received rather than repulsed) over jerky dissonance. Factor III (8%) seems to be concerned with shape, with its high loadings for symmetry. The two raters were quite different in their judgements for the other adjectives with substantial loadings for Factor III, however: Round-straight was more important for Rater 1 but wide-narrow was more important for Rater 2. Both raters produced moderate loadings for undirected-directed, which seems to tally with the description of a shape.

Table 1. Factor loadings for a three-factor solution based on raw data from two raters of the movements of 28 participants upon 14 adjective scales (N = 6272).

4. Discussion

The aim of this study was to examine what patterns - if any - might emerge when volunteers are asked to perform body movements in response to heard musical intervals. As a first approach to this question, two observers were given a set of 14 bipolar rating scales on which to assess the video recordings of participant responses. The fact that there was fair inter-observer agreement in judgements on a number of scales indicates that there were in fact some non-random tendencies in participant responses, and that the conceptual dimensions of the rating scales captured these. The four scales with highest inter-observer correlation on two different correlation measures were up-down, in-out, active-passive, and gloomy-cheerful, suggesting two spatial, one intensity, and one valence/emotional dimension with good inter-observer agreement in ratings.

The results of this study confirm results from earlier research (Truslit, 1938) that the general direction of the body movement to melodies agrees with the pitch direction. This pitch effect explains the main trend in the results obtained for the scales up-down, and inwards-outwards. This pitch effect is global in the sense that the larger the pitch step of the melodic interval the higher was generally the observers’ rating on the up and outwards dimension. This does not mean, however, that participants confined themselves to portraying this rather "physical" stimulus dimension in their movements and gestures. There is an inflection in the general diagonal tendency of the ratings at the second and the sixth intervals. Responses to the second were rated as more extreme than the prime on the up-down and gloomy-cheerful scales (i.e. down and gloomy), while the sixth was rated more extreme than the seventh on the down-up, the inwards-outwards, and the gloomy-cheerful scales as well. Presumably this indicates the influence of a factor other than the physical pitch dimension on subject responses, such as its emotional or aesthetic quality. In a parallel study in which a large population of participants gave free word responses to the same intervals the three most used words for the second were sorrow, sadness, and melancholy while for the sixth they were joy, light (in the sense of contrasting with dark), and upwards. The distribution of words across intervals showed that joy was primarily used as a descriptor for the sixth and sorrow for the second. There are thus some striking correspondences between the present study and the study employing free word generation, but also some differences. The effect of the physical pitch distance of the intervals was not as prominent in the word-generation study as in the present study. It stands to reason that the task of producing physical movements to the intervals would put a premium on stimulus dimensions that are easily pantomimed with the help of body movement, such as the size of the pitch difference composing the interval. The rich conceptual vocabulary embodied by language, in contrast, suggests many additional stimulus dimensions to the imagination of participants. This makes the results for responses performed to the second and sixth all the more interesting: Even when expressing themselves in physical movements, participants are not limited to performing a pantomime of surface characteristics of the stimuli, but may portray additional influences of the stimuli, such as musical or emotional qualities.

Acknowledgements

The authors G.K. and B.M. were supported by a grant from the Signe and Ane Gyllenberg and Kempe-Carlgrenska Foundations. Authors B.M. and G.M. were supported by a grant from the Bank of Sweden Tercentenary Foundation.

References

Huber, K. (1923). Der Ausdruck Musikalischer Elementarmotive. Leipzig: Verlag Barth

Krantz, G. (2002). Responses to melodic intervals. Proceedings of the Seventh International Conference on Music Perception and Cognition, July 17-21 2002, Sydney.

Maher, T. (1980). A rigorous test of the proportion that musical intervals have different psychological effects. American Journal of Psychology, 93, 309-327.

Maher, T. (1982). Verbal and exploratory responses to melodic musical intervals. Psychology of Music,10, 11-27

Makeig, S. (1982). Affective versus analytic perception of musical intervals. In M. Clynes (Ed.). Music, mind and brain, pp. 227-247. New York: Plenum Press.

Smith, L, & Williams, R. (1998). Children’s artistic responses to musical intervals. American Journal of Psychology.

Truslit, A. (1938/1993). Shaping and motion in music. Psychology of Music, 21, 48-72.

Presentation of the authors

Göran Krantz studied eurythmy in Järna, Sweden and recieved his diploma 1979. He is leader of the eurythmytraining at R. Steiner Universitycollege, Järna, Sweden and 1998 he founded the research institute for eurythmy - music, language and movement. International activities in eurytmy, choreography and research on music and movement. Three of his coreographies are presented in 2003 one in England, one in Japan and one in Finland.

Bjorn Merker studied psychology and brain science in the US, receiving his doctorate at the Massachussets Institute of Technology in 1980 for work on midbrain orienting mechanisms. He then worked in visual neurophysiology at UCLA and New York University. An interest in the the origins of language led him to study song development in the singing apes, the gibbons, and eventually to research on the biological roots and evolutionary background of human music. He is currently at the Department of Psychology at Uppsala University in Sweden, where he conducts research with Guy Madison on human rhythmic coordination.

Guy Madison’s main research areas are human timing and music. A special focus is what musical structure and performance can communicate to listeners. His PhD is in experimental psychology, but he has also studied musicology, music psychology, and computer science. He is presently the director of a research project called ”Musical pulse and human coordination” located at at the Department of Psychology, University of Uppsala, Sweden. Having performed and toured with semi-professional groups, he now mostly plays the drums at home in ensemble with his wife. See further a personal research profile at http://www.psyk.uu.se/hemsidor/guy.madison.

Mr. Göran Krantz

Dr. Mr. Björn Merker & Dr. Mr. Guy Madison