Skip to content

Musical groove: functions, mechanisms, and origins

Research project Groove is the quality of music that makes people tap their feet, rock their head, and get up and dance. Why and how this occurres is studied in a series of experiments where listeners’ experience and movements are set in relation to rhythmical and other properties of the sound signal.

Specifically, measures of rhythmical properties are obtained from the music by means of numerical computer modelling methods, and their relations to listeners’ behaviour and experience are used to form specific hypotheses. These hypotheses are tested by (1) systematic selections of real music, (2) manipulation of real music, and (3) synthesized sound examples with systematically varied properties. The team members are Guy Madison, expert in human timing and synchronization, Björn Merker, neuroscientist working on principles of brain organization and on the biology of music, and Fabien Gouyon, expert on music content processing.

Project overview

Project period

2009-01-01 2011-12-31


Finansår , 2009, 2010, 2011

huvudman: Guy Madison, finansiar: Riksbankens Jubileumsfond, y2009: 551, y2010: 932, y2011: 932,

Research subject


Project description

Research program
Our perspective has identified a number of testable hypotheses, as outlined above, as well as the need for an exploratory approach towards increasing our understanding of the signal properties associated with the experience of groove. We combine these in the first two experiments.

1. Do we like music because it has groove or do we think it has groove because we like it?
The possible adaptive value of entrainment implies that the experience of groove be associated with a hedonic value, feelings of pleasure or positive appreciation, an hypothesis supported by correlations between “having groove” and “good” in our previous rating work (Madison, 2006). However, correlations are not informative about causality, so the reverse might in fact be the case: If for any reason (even an entirely culturally constructed one) we have learned to associate groove with something pleasurable, then we might tend to rate music that we like (for any reason) higher on groove.

To discriminate between these possibilities, we will use listeners whose style preferences we already know before the rating experiment. Prospective participants will first rate their preference for a number of styles, such as jazz, punk, samba, reggae, and so forth. Only those whose preference – in relation to the average rating – is both higher for one style and lower for another style will be selected. This way a number of opposite style pair groups are obtained: One group likes samba and dislikes punk, and another group dislikes samba and likes punk, and so forth. In a subsequent rating experiment, all groups rate a number of music examples from each style that differ in their amount of groove. If preference causes groove we will find that those who like punk will also rate punk as having more groove, to a similar extent as the opposite group rates another genre high on groove. If groove causes preference, we will instead find that style preference will have a relatively small effect, but that amount of groove becomes a major determinant of ratings largely independently of genre.

The exploratory part of this design consists of computing low-level rhythmic descriptors for all music examples, as described above, and assessing how well they capture the listeners’ experience in terms of correlations between each descriptor and listeners’ ratings.

2. Do the same or different properties of the sound signal induce the experience of groove among music styles?
Madison (2006) showed that there was substantial consistency in ratings of groove among listeners across 64 music examples from a wide range of genres and countries of origin (not limited to Europe and North America). This means that the phenomenon is recognisable across music genres with otherwise quite disparate structural characteristics, which suggests that the phenomenon of groove possesses psychological reality and specificity in its own right.

This in turn encourages a search for possible differences among cultural traditions, since it is clear that music from different cultures and geographical areas differ in its structural features.
Important questions are therefore to what extent they differ in inducing groove and in how they induce groove. It is possible that some structural properties serve the same function in different musical styles. Another possibility is that groove can be induced by alternative structural means. Conceivably, more similar properties can be expected among music intended for dancing. We have designed a first experiment to explore these issues, in which five different styles are compared: Greek, jazz, samba, classical Indian, and West African. For each style there are 20 typical examples of that tradition, which makes 100 examples in all. Main effects of musical style might indicate that styles differ in their propensity to induce groove.

Again, we will explore the correlations between ratings of groove and some 25 acoustical descriptors, to see whether some descriptors are more correlated with the ratings for certain styles, which would indicate that different means are employed to achieve the same end.

3. Is groove related to entrainment performance?
Synthesised sound examples are required to address this question, because musical structure induces systematic temporal deviations in participants’ responses that we cannot entirely predict (Repp, 1998a; 1998b). The general method will be sensorimotor synchronisation, which we already have extensive experience of (Madison & Merker, 2004; 2005; 2008a; 2008b). Two aspects of performance will be considered: local (beat-for-beat) temporal precision monitored with millisecond resolution while participants beat the isochronous pulse, and gross response pattern correctness and precision of participants’ ability to reproduce prescribed rhythmical patterns.

When experiments according to approaches 1 and 2 above have provided more detailed hypotheses about signal properties, these properties will be manipulated (turned on/off, or where feasible varied in magnitude) and their effects on synchronisation performance measured as described in the following. Pending that, our theoretical perspective already allows us to predict that the amount of temporal information in the sound pattern will be correlated with good synchronisation performance. Thus, we will compare a simple metronome-like isochronous click sequence with approximately 500 ms intervals (poor information) with a structured pattern that contains many isochronous levels typical of music: whole notes, half notes, quavers, crotchets, and so forth (rich information). In the response pattern correctness conditions, the prescribed patterns will occur a number of times as a sounding model, after which the participant should continue to beat that model synchronised with the same elements in the poor or rich sound pattern.

The same design will then be applied for other candidate sound signal properties, which will be synthesised as they emerge in our analyses.

4. What are the properties of the sound signal that induce the experience of groove?
This central issue will be pursued by the just mentioned combination of computational modelling analysis and sound synthesis. However, we will also put the synthetic sound patterns to the test by comparing them with real music examples known to be rated high on groove. In this way we may be able to confirm the adequacy of the properties identified and also gauge their relative contribution to groove. For this we will employ both listening experiments, in which ratings and movements are obtained, and sensorimotor synchronisation experiments.

The descriptors that we have already developed will serve as initial models for the sound synthesis. However, new descriptors will continually be developed to replace less successful ones, based on the repeated application of the newer descriptors to the “old” music examples and their corresponding ratings, and on the application of all these descriptors to new music examples in current and subsequent experiments. The population of descriptors will thus undergo an “evolution” based on their respective success and redundancy in terms of accounting for the variance in ratings of groove, determined by multiple regression. In this way we can gradually fine-tune the synthetic stimuli in quest for one or perhaps several fundamentally different sound patterns but equally effective for inducing groove.

5. Does groove correspond to temporal information redundancy?
As mentioned, several aspects of the suggested studies will elucidate the possibility that groove characterises sound patterns that are optimal for synchronisation and co-ordination. A possibility not covered by the studies described above is that groove corresponds to higher information redundancy, regardless of whether it actually leads to better synchronisation performance under optimal, laboratory conditions. This hypothesis will be addressed specifically in synchronisation experiments in which the same stimuli as before are partially masked by distracting noise or distorted by, for example, deviations in pulse rate.

6. To what extent is groove related to motivational/emotional factors?
Though the association between groove and rhythmic movement would seem to be a basic and perhaps primary one, its role in aesthetic appreciation should not be neglected, as mentioned in the section on theoretical perspective, and as supported by intricate covariance patterns among the many different rating words used in some of our prior experiments.

We hope to elucidate the role of motivational/emotional factors by comparing ratings of groove, goodness, and “movement inducibility” from the listening experiments above with measures of listeners’ movements, and with results from the sensorimotor synchronisation experiments. If, for example, synchronisation performance is unrelated to groove while the tendency to move increases with groove, this would indicate that the effect of groove is primarily a motivational one. Should results favour such an interpretation, we might find it appropriate to follow that up by obtaining free verbal responses or ratings of a considerably extended set of rating words for selected sets of stimuli.

7. How general is the experience of groove?
Although we have found that groove is to some extent a general experience across listeners, this project will extend the scope of this question to different behaviours and different musical styles, and even to non-musical sound patterns. The ensuing patterns of results will most likely shed light on the generality of different functions of groove.