Robert Abel
May 1996
INTRODUCTION
`Swing' is a term used to describe a quality which is widely recognised as basic to the perception and performance of jazz, an `essential element' (Stewart 1982) since its formative years. The term has, however, so far resisted concise definition. Schüller (1989) implies that there is more than one way to swing, but states that "swing occurs when a listener inadvertently starts tapping his feet, snapping his fingers, moving his body or head to the beat of the music". Jazz musicians and listeners are usually enthusiastically united in opinion when something is `swinging', and are painfully aware when it is not.
Despite this common code shared between player and listener (probably learned by aural processes), musicians and commentators alike seem to have some difficulty in verbalising this quality, usually describing it in general terms, often employing vague metaphors or mysticism:
"We lock onto that big pulse. It's just like the air- we can't see it but without it we cannot survive." The drummer Abdul Rahman Qadir, 1989. (Prögler 1995).
Schüller (1989) writes that these rhythmic impulses must arise "from one's instincts and natural, at times even unconscious, impulses and feelings. When swing occurs it is innate, not studied."
The choice of the title `swing' and its other common synonyms (`pulse',
`foot-tapping') indicates a close association between this quality and
metre/rhythm. Stewart (1982) transcribed a solo by the trumpeter Clifford
Brown, noting how he used rhythmic displacement " to create the aesthetic
sensation known as `swing'". Tanner and Gerow (1977) described how jazz
musicians play slightly ahead or behind the prevailing tempo as an expressive
device. Gridley (1985) states that the tension created by this practice is
essential to the feel of swing. The performance practice that these
investigators describe could be termed `metrical displacement' and can be
defined thus:-
"deviation by a given amount and in a given direction from the attack times of
notes that would be expected if the melody was synchronised exactly with the
prevailing tempo; and, by extension, from other members of the performing
group."
Musicians have often described this phenomenon when referring to `laying back on the beat' or `pushing ahead'. It seems also that professional rhythm section players are able to describe with accuracy the relative direction of `participatory discrepancies' between the section elements (Prögler 1995). `Metrical displacement' has been the main area of study in attempts to describe swing quantitatively. At the present time, however, no large-scale formal investigation has taken place.
Previous informal, explorative studies have found a range of values of metrical displacement in jazz performance. Although these studies have not attempted to test any proposed interaction between amount and/or direction of displacement and `swing feel', they have posited a close association between quantitative and qualitative elements in the music. The MIDI programmer Michael Stewart's `feel-spectrum' (in Prögler 1995), for instance, illustrates his prescriptions for generating an assortment of rhythmic feels at 130 bpm for use with synthesisers and drum machines. For example, 5ms after the beat is cited for `groove' and just over 20 ms ahead is cited for `drive'.
Taking these investigator's methodology into account, however, their prescriptions should perhaps be taken with caution. Milton Stewart does not state how he arrived at the displacement values in his investigation of Clifford Brown, which suggests that they were done by ear. The implication is that Michael Stewart arrived at his feel-spectrum by trial and error. Indeed, Prögler (1995) has pointed out that what musicians think and say they are playing is not, as objective analysis by computer has found, always what they are playing.
It is clear that metrical displacement is not the only aspect of performance which may have an effect on swing feel. Tempo (Prögler 1995, Ellis 1992, Collier 1994), timbre/ articulation (Stewart 1982) , jazz style phrasing incorporating `swing quavers' (Stewart 1982, Ellis 1991), and syntactical dimensions have all been put forward as contributing factors. Stewart (1982) wrote that Clifford Brown creates swing by `alternation of displacement with conformity during certain measures, depending on the structural function of the notes he was playing.' An oppositional stance has been taken up between the relative merits of metrical displacement due to syntax (according to structural function of notes) and metrical displacement due to process (constant) in influencing swing feel. Schüller (1989) gives higher status to syntactical aspects. Keil (1966) takes an opposing stance: he seeks to show that swing and groove are not embedded in syntax but occur in process. Prögler (1995) takes the conciliatory point of view, that " groove and swing are engendered processes that are from time to time affected by the syntactical dimensions of music; in a sort of mutually dependent balance". He goes no further, however, towards explaining how this process might work!
To what degree does each of these variables contribute to the creation of swing feel and in what way do they interact? In short- how does swing work ? This study is part of the ongoing research attempting to answer this question. Its purpose is to apply a formal experimental method to test if the effects of two of these performance practices, suggested by previous experimenters to be the two dominant variables in swing feel, are statistically significant. The experiment therefore looks at the effect of amount and direction of metrical displacement and level of phrasing in a jazz melody, and the relationship between them, against a standard accompaniment.
In view of the considerable difficulties in isolating, controlling and measuring these variables in existing performances by live musicians `artificial' performances were created using electro-acoustic sound generating equipment. This allowed absolute control over performance variables such as dynamics, duration and attack. A number of informed subjects were asked under experimental conditions to rate these `artificial' performances (comprising different combinations of the variables' conditions) for `amount of swing' using a given scale.
Hypothesis
The main effect of the well-formed performance mode, with swing rating as the dependent value, will be significantly different to the main effect of both straight and perverse modes:
the well-formed performance mode will be preferred above both straight and perverse. Metrical displacement in general will be preferred to none (`on-beat'), and there will be preference for amounts and direction. If Stewart's prescriptions are correct then the results may show that there are limits of acceptability for metrical displacement (`cut-off points). He gives the value of >20 ms delay to produce a `dragging' feeling and a value of approximately 35ms ahead to produce a `nervous' feeling. Considering that these values are given for a tempo of 130 bpm and that Ellis found, by trend analysis, that delays of attacks increased in a linear fashion with tempo for all subjects, the limits of acceptability for metrical displacement might be expected to be marginally lower at 120 bpm.
METHOD
Subjects:
Twenty Cambridge university students of mixed discipline, aged 19 to 22. The criterion for choosing the subjects was a self- classification as listeners and/or players of jazz. Their average listening/playing experience was 12.95 years, with a range of 10 to 17 years. Five of the subjects were studying music as an academic degree.
Materials:
All the stimuli were produced using Mark of the Unicorn's `Performer' (5.02) sequencing programme on a Power Macintosh 7500/100 computer. A Yamaha SY99 digital synthesiser and E-MU Proteus FX sound module were used as sound generating equipment and the stimulus sequence was recorded onto Sony Type II audio cassette using a Nakamichi 582Z tape deck.
The stimuli consisted of fifteen performances of the same jazz melody with a constant, unchanging accompaniment of piano chords, drums and bass.
The following patches were used:
| piano melody | proteus `7th grand' |
| accompaniment piano | proteus `concert grand' |
| bass | SY99 `fingered bass' |
| drums | SY99 `drum kit' |
The melody chosen (see Example 1) was a transcription of a melodic improvisation by the pianist Bill Evans (on chords based on `How Deep is the Ocean') by Johnson-Laird (1991). All the elements of the accompaniment were composed by the experimenter based on the chords given, following guidelines put forward by Johnson-Laird tempered by experience gained in ten years playing and listening to jazz. Performer's `Input Quantize' option was used when recording the accompaniment to ensure that all the attacks occurred exactly `on the beat'.
Prögler (1995) and Collier (1994) have inferred that swing feel is more associated with medium and fast tempos. 120 bpm was chosen as a tempo value that fell suitably within this range.
Piano was chosen as the solo instrument because it is possible that articulation and timbre are significant variables in engendering a swing feel. Other traditional jazz solo instruments (trumpet, alto saxophone), due to their physical and acoustical makeup, produce a far greater range of variation in these areas. In spite of the sophisticated sound generating equipment used the patches for these other solo instruments were far less convincing
Description of the two independent variables and their respective conditions:
1) Performance mode (3 conditions):
* Straight
The melody was performed exactly as notated; dynamic was constant (84 on a
scale of 1 to 127), and every example of a particular note value (crotchet,
quaver, etc.) had the same duration
* Well formed
The preceding straight melody was altered according to two rules. First, the
general rules governing expressive deviation to produce a musically competent
performance proposed by Sundberg, Friberg and Fryden (1991). Second, according
to the experimenter's own sense of jazz performance style for this particular
tempo and genre.
Main alterations included:
- swing quavers replaced straight quavers. The actual duration ratio of swing
quavers is debatable and seems to be, in part, an element of personal style. In
his experiment on swing quaver ratios played by three experienced jazz
saxophonists, Ellis (1991) found significant differences for subjects (p<
.007), tempo (p< .0001) and the interaction between these ( p< .0001).
The ratio of 2:1 was used in this experiment, as this is the default ratio most
commonly employed by jazz educators.
-notes falling on the second and fourth beats of every bar were accented at +11
attack velocity, the accent on 2 and 4 being a fundamental element of jazz
style.
* Perverse
The rules applied within the well formed category were here applied in
reverse. The first and third beats of the bar were accented and the swing
quaver ratio was reversed. (1:2 instead of 2:1).
2) Metrical displacement (5 conditions):
Metrical displacement was applied uniformly to all attack times for a
particular melody,
| Condition of variable | Attack shift (ms) |
|---|---|
| on beat | 0 |
| delay-1 | +10 |
| delay-2 | +30 |
| ahead-1 | -10 |
| ahead-2 | -30 |
Procedure
The subjects were given a response sheet which contained the
following text:
"You will hear the same jazz melody performed fifteen times with accompaniment.
Please rate each separate performance for how much you think it `swings' (this
is sometimes alternatively expressed as `groove', `jazz-feel', or `whether it
makes me tap my foot'!). Give each performance a score from 1 to 7, according
to the following: 1= absolutely no swing to 7= swinging very much."
The subjects were asked to read it through to ensure that they understood what
was asked of them. Any questions to the experimenter were answered. The tape
was then run, uninterrupted, from beginning to end. The fifteen melodies were
presented in two different computer generated random sequences to control for
any effects of order. One half of the subjects heard the first sequence and the
other half the second. There were four bars of silence between each
performance- the whole session lasting about eleven minutes.
RESULTS
A two way analysis of variance (related) with subject the source of error (see Table 1), indicated significant differences for performance mode ( p<.0001) with an overall preference for well-formed over straight and perverse (see Fig. 2), metrical displacement (p< .0033), and the interaction between performance mode and metrical displacement. (p< .0005)
The least squares means table for performance mode effect (Table 2) showed a significant difference for straight versus well-formed ( p<.0009) and for well-formed versus perverse (p<.0001). There was no significant difference, however, between straight and perverse.
The least square means table for metrical displacement effect (see Table 3) shows significant differences for ahead 2 versus all other categories of metrical displacement. There are no significant differences between all other combinations of categories. Therefore there is no indication of a preference for metrical displacement, contrary to the hypothesis.
DISCUSSION
The significant difference between ahead-2 and all other categories of metrical displacement (Fig. 3) is perhaps evidence to suggest a `cut-off point' for metrical displacement in the ahead direction, as predicted. For this particular tempo it seems to lie somewhere between 10 to 30 ms. There is no significant difference between all other categories, however, suggesting that the value of any existing `cut-off point' in the delay direction is larger than in the ahead direction.
These results seem therefore to refute the prescriptions given in Stewart's `feel-spectrum', where the biggest acceptable value for metrical displacement is smaller in the delay direction than in the ahead direction.
Clearly, further research is needed, perhaps using smaller increments of metrical displacement and to more extreme values in both directions. The results of this research might suggest a formal model for acceptable values of metrical displacement at a particular tempo; introducing tempo as a third independent variable might reveal trends in the interaction of metrical displacement with tempo.
Interestingly, there is no significant difference between `on-beat' and `delay-1', `delay-2', and `ahead-1', shedding doubt on the assumptions of previous studies that metrical displacement is required to produce the `tension' seen as essential to the swing feel.
To shed light on the question "How does swing work?" the most revealing results are shown in Fig. 4. At delay-2 there is no significant difference between any of the three performance modes. In other words, at this particular value of metrical displacement it does not matter whether the melody is performed with jazz phrasing, perverse phrasing, or no phrasing at all: the condition of the performance mode variable has no significant effect. Perhaps this indicates a dominance of the effect of metrical displacement over that of melodic phrasing at +30 ms.
It is possible that this relationship is operating, to a lesser extent, at `delay-1'. There is no significant difference here between `straight' and `well-formed' . In the `on-beat' category the main effect of performance mode is clearly present (contrast this with Fig. 1). When metrical displacement (excluding `on-beat') is present, however, the interaction between the independent variables is more complex and the relationship operating for delay and ahead categories seems to be different.
Certainly, at -30ms (`ahead-2') - past the proposed cut-off point- the effects of the respective performance modes are clearly differentiated. In other words, perhaps the effect of performance mode is `revealed' when the dominant ability of metrical displacement to affect swing rating is `reduced'.
In general, for this experiment, at this particular tempo of 120 bpm, it appears that swing rating is more dependent on the amount/direction of metrical displacement. However, the effect of metrical displacement on swing rating is limited by specific cut-off points that enclose a range of values, and this appears to extend further on the delay side.
It seems that the use of expressive deviations in jazz to produce a musically acceptable performance (with one swing feel) is different to their analogous use in classical music. In jazz, the implication is that musical acceptability is more dependent on a constant (processural) relationship of the melody to the prevailing tempo. In classical performance, musical acceptability is more dependent on the syntactical dimensions of expressive deviations.
Clarke, E. 1988 " Generative Principles in Music Performance" in Generative Processes in Music Ed. Sloboda, Oxford: Clarendon Press p1-26
Collier, Geoffrey L. 1994 " An Exploration of the Use of Tempo in Jazz" Music Perception Vol.11, No. 3, p. 219-242
Ellis, Mark C. 1991 "An analysis of `Swing' Subdivision and Asynchronisation in Three Jazz Saxophonists" Perceptual and Motor Skills, No. 73, p. 707-713
Gridley, M.C. 1985 Jazz styles: History and Analysis. Englewood Cliffs, NJ: Prentice-Hall
Jarvinen, Topi 1995 " Tonal Hierarchies in Jazz Improvisation " Music Perception Vol.12, No. 4, p. 415-437
Johnson-Laird, P.N 1991 " Jazz Improvisation : A Theory at the `Computational Level'" in Representing Musical Structure ed. Howell, Cross, West. Academic Press
Keil, Charles 1966 " Motion and Feeling through Music" Journal of Aesthetics and Art Criticism No. 24(3), p.337-349
Prögler, J.A 1995 " Searching for Swing: Participatory Discrepancies in the Jazz Rhythm Section" Ethnomusicology Vol.39, No. 1, p. 21-55
Schüller, Gunther 1989 The Swing Era: the development of jazz, 1930-1945 New York: Oxford University Press
Stewart, Milton L. 1982 " Grid Notation: A Notational System for Jazz Transcription" Annual Review of Jazz Studies Vol. 1 , p.3-12
Sundberg, John; Friberg, Anders; Frydén, Lars 1991 " Common Secrets of Musicians and Listeners: An analysis-by-synthesis study of Musical Performance" Representing Musical Structure ed. Howell, West, Cross. Academic Press
Tanner, P.O., and Gerow,M. 1977 A Study of Jazz Dubuque, IA: Brown.
Todd, N 1985 " A model of expressive timing in tonal music" Music Perception Vol.3, p. 33-59