Time-feel
The Analysis, Modeling and Employment of
Sub-notational Rhythmic Expression
Dr. Milton Mermikides
Monday, 28 February 2011
Overview
From a practitioner’s perspective, the analysis and theoretical modeling of
sub-notational rhythmic expression - variously referred to as groove, swing, feel,
expressive micro-timing - and here collectively termed time-feel.
Are they ways that we can identify, quantify and consciously develop and
reapply those aspects of rhythmic expression - particularly associated with
jazz, funk, rock and other styles - that are vital to the musical experience -
but escape easy standard notation and description?
Monday, 28 February 2011
Motivation
Degree from Berklee College of Music. Chasing the pristine experience of
good time-feel - the ‘swing’ - but direct talk of it was incomplete and contradictory,
and obfuscated. Intuitive development, once fleeting and then more secure.
Extensive experience with music technology and editing. Quantisation, and
micro-timing edits, so an absolute position (often not strictly quantised) is chosen
but on what basis? ‘Tight but loose’. Technology’s critical timing.
Scientific/mathematical background and interest (LSE degree) the application of
simple mathematical models to clarify some rhythmic aspects.
Monday, 28 February 2011
Practioner’s perspective
Rhythmic time-feel is the most basic, fundamental element communicated by
the soloist, and appreciated (or criticized) by an audience. The greatest
technique, creativity, melodic accuracy, lyricism, sound, style, etc. matters very
little if the music doesn’t feel good rhythmically, whereas less evolved technique,
ideas, melodic choices, sound etc. can actually sound okay when executed with
rhythmic accuracy (good time-feel) and conviction.
Crook 1991, p 10
Even some of the most articulate jazz critics and chroniclers will avoid a
penetrating discussion of swing and generally back themselves into a corner
when they are asked to engage in one.
Coker 1964, p 45
I don’t think about that stuff.
Silver 1994
Monday, 28 February 2011
Wishart’s Lattice
Wishart 1996, p 26
The representation of standard notation’s limitation as a lattice, due to the ‘notational economy’ of ‘finistic’ division of pitch, timbre and
rhythmic subdivision. Exploration of all 3 dimensions of musical inconveniences and irrationals.
Monday, 28 February 2011
Research Background
Stylistically distant research into notes inégales, ritartendo, classical ensemble asynchrony and
time warping in computer music (Fuller 1980, Desain & Honing 1990, Gabriellson 1988 and
Dannenberg 1997 respectively).
Jazz micro-timing may be broadly categorized into five reasonably separate groups.
1) Musicological, anecdotal and heuristic responses to the experience of hearing and
performing jazz rhythm (Schuller 1968, Werner 1996, Sudnow 1995).
2) Pedagogical material to encourage and guide exploration of sub-notational timing in jazz
practice (Crook 1991, 1996, 1999 and Moore 1995).
3) Audio analysis, measurement and music software applications of swing 8ths (or jazz
quavers) (Cholakis 1995, Friberg & Sundström 2002 and Benadon 2006).
4) Tempo modulation mechanisms including behind-the-beat playing, rubato and
superimposition of differing simultaneous tempi (Prögler 1995, Ashley 2002, Collier & Collier
2002, Folio & Weisberg 2006 and Benadon 2009)
5) Studies of ensemble synchrony, the internal construction of grooves and rhythmic
templates and statistical analysis of rhythmic placement (Cholakis 1992, Millward 2001a,
Millward 2001b, Tait 1995, Butterfield 2006, Gouyon 2007, Hennessy 2009 and Naveda,
Gouyon, Guedes & Leman 2010).
Monday, 28 February 2011
The Aims
Create a model, building on the available research, that is born
of a practitioner’s perspective and that can provide a:
1) Succinct representation and differentiation of a wide-range of
appreciable and stylistically relevant time-feels.
2) Simple enough to be understood by a wide range of
practitioners and analysts alike.
3) Simple enough to be a pedagogical tool and applied in real
time, rather than post-hoc analysis. Alignment with the visceral
experience of performing.
4) Amenable to technological analysis and reapplication.
Monday, 28 February 2011
The SLW model
Illustrated at quaver level
Monday, 28 February 2011
Swing
Often used as an ill-defined
qualitative measure of feel,
but here it is specifically the
asymmetry of the off-beat.
Monday, 28 February 2011
The Swing Continuum
Monday, 28 February 2011
Swing and Technology
Technology’s weakness, and usefulness, is its lack of interpretatio.
The evolution in Logic of computer quantization of swing, from none (left, Logic Notator
2.0,1988), to discrete, but not explicitly defined, values (middle Logic Pro 6, 2004) to continuous
values and advanced options (Logic Pro 9, 2009).
Monday, 28 February 2011
Latency
Mingus, cited in Berliner 1994.
Monday, 28 February 2011
Latency Continuum
Monday, 28 February 2011
Positive Latency
The problem I see in a lot of students is, I think, anxiety-related
when they play and they are actually rushing everything…If
you’re not going to play like a sequencer and be bang on the
note every time, the other thing that can work is actually playing
late, listen to Miles Davis, or something like that, or a lot of the
hip-hop guys doing their thing…it’s often preposterously late*, and
it never sounds like a mistake, it just sounds like they’re cool,
they’re relaxed, they’re not in any hurry to get to the next note,
because the note they are doing right now is so good.
Govan (2010) 4:29 – 4:52
* Hyperlatency = latency that may be easily rewritten notationally
Monday, 28 February 2011
Who’s Late?
•Why is one player described as late, and the other as early?
•Hierarchy of rhythmic establishment, the idea that one
performer holds more (or all) of the authority in the
definition of the time line than others.
•Latency cannot be adequately equated with rubato, latency is
an expressive mechanism concerned with friction against a
relatively rigid time line, not the elasticity of the time-line
itself.
•An individual’s placement against a mutually negotiated
master time-line (defined later)
•Tempo fluctuations: rubato and there are moments in the
jazz (and wider) repertoire when the concept of one master
time-line can no longer be usefully maintained. Ivesian,
xenochrony.
Monday, 28 February 2011
Weighting
Off-beat emphasis
For now a simple definition: the relative volume of the off-
beat to the on-beat but may extended to include parameters
of articulation
Positive weighting is a stronger off-beat.
The 2 and 4 - Jazz walking bass-line at the crotchet level.
(Tomassi’s instruction) and the bop of bebop
Monday, 28 February 2011
3 Separate Elements
Swing and Weighting are ‘private’
Latency is ‘public’
Monday, 28 February 2011
Where’s The One?
" = {1, 2...E} !
Ensemble with E performers
{t 0e ,t1e …t Xe }
! Quaver timings for performer e
Monday, 28 February 2011
Master time-line
{T0 ,T1 …TX } !
Master quaver set
E E
! T =" t *r
X [( x ) x ]
e=1
e e
where "r
e=1
xe = 1!
Determined by weighted
average of all performers
! !
Monday, 28 February 2011
Implications of
Hierarchy
Hierarchical asymmetry provides opportunity for latency
mechanisms
The musical experience of an ensemble rhythmic pattern
depends on context
Monday, 28 February 2011
One duo event: 3 contexts
a) Harmonized melody (Rx1 = 1/2, Rx2 = 1/2)
Monday, 28 February 2011
One duo event: 3 contexts
b) Guitar 1’s solo (Rx1 = 0, Rx2 = 1)
Monday, 28 February 2011
One duo event: 3 contexts
c) Guitar 2’s solo (Rx1 = 1, Rx2 = 0)
Monday, 28 February 2011
One trio event: 3 contexts
c) Guitar 2’s solo (Rx1 = 1, Rx2 = 0)
a) Equal weighting. All performers have equal say
Monday, 28 February 2011
One trio event: 3 contexts
c) Guitar 2’s solo (Rx1 = 1, Rx2 = 0)
b) Oligopoly, performers 2 and 3 share responsibility
Monday, 28 February 2011
One trio event: 3 contexts
c) Guitar 2’s solo (Rx1 = 1, Rx2 = 0)
c) Monopoly where performer 3 has all the metric responsibility.
Monday, 28 February 2011
Calculating Swing
(mean, standard deviation and Ensemble)
Monday, 28 February 2011
Mean Swing of Phrase P
(t1 " t0 ) + (t 3 " t0 ) +!+ (t X "1 " t X "2 )
( t 2 " t 0 ) (t 2 " t 0 ) ( t X " t X "2 )
µs P = !
2X
Swing Standard Deviation of Phrase P
! (tight-loose)
1
"P =
2X
( [ 2 2 2
S0 # µP ) + ( S2 # µP ) +! + ( SX #2 # µP ) ! ]
Monday, 28 February 2011
Ensemble swing for E performers in Phrase P
E
1
µsPE = " µsPe !
E e=1
SD of Ensemble swing for E performers in Phrase P
" SP = !(
1 *## 1
2 X ,+$$ E
&
'
)
&2 ## 1
' $$ E
(
&
'
)
&2
'
## 1
$$ E
(
&
'
)
&2 - !
,%% S01 + S0 2 +!+ S0 E ( ) µ SP ( + %% S21 + S2 2 +!+ S2 E ( ) µ SP ( +!+ %% S X )21 + S X )2 2 +!+ S X )2 E ( ) µ SP ( /
' /.
Monday, 28 February 2011
Swing Friction between performers a and b in Phrase P
sfPab = µsPa " µsPb
Swing Friction between performer c and ensemble
! sfPcE = µsPc " µsPE
Monday, 28 February 2011
Latency Calculation
Monday, 28 February 2011
Mean Latency of Phrase P over X quavers
# (t 0 " T0 ) (t 2 " T2 ) (t X "2 " TX ) &
% + +!+ (
$ (T2 " T0 ) (T4 " T2 ) (TX " TX "2 ) '
µ LP =
2X
SD Latency of Phrase P over X quavers (tight - loose)
!
2
$ (t # T ) '
X /2
*& (T # T ) )
2x 2x
x=0 % 2 x+2 2x (
" LP = !
2X
Monday, 28 February 2011
!
Mean Weighting of Phrase P over X quavers
X /2
1
WP = "
2X x=0
dBt 2 x+1 ,t2 x ! where dBt1 ,t 0 is the dB differential of t1 and t0
!
SD Weighting of Phrase P over X quavers
! X /2
1 2
" WP = $
2 X x=0
(
dBt 2 x+1 ,t2 x #WP ) !
Monday, 28 February 2011
Time-feel Matrix
Phrase P
#µ " SP '
SP
% %
P = $µ L P " LP ( !
% %
µ
& WP " WP )
Monday, 28 February 2011
Time-feel Matrix
Solo J
#µ ' #
" S1 µ S2 " S2 ' # µ Sn " Sn '
S1
% %% % % %
J = $µ L1 " L1 (, $µ L2 " L2 (!$µ Ln " Ln (
% %% % % %
&µW1 " W1 ) &µW2 " W2 ) &µWn " Wn )
Monday, 28 February 2011
Time-feel
dynamics and subsets
Monday, 28 February 2011
Relationships between Parameters
Swing, Latency and Offbeat Placement
OFFBEAT
PLACEMENT
Monday, 28 February 2011
Offbeat placement
Monday, 28 February 2011
Characterisation of quavers
The comparison of 50% latency on four different time-feels.
Only in example a) is a 50% latency exactly equivalent to a rhythmic displacement.
Monday, 28 February 2011
The Identity of Quavers
The significant difference between the transformation of phrase a) through b) 50%
latency and c) quaver displacement.
Hyperlatency and notational implications
Monday, 28 February 2011
Technology and Time-feel
Accuracy, repeatability
Monday, 28 February 2011
Technology and Time-feel
Sonogram analysis
Monday, 28 February 2011
Technology and Time-feel
Statistical analysis
Monday, 28 February 2011
Technology and Time-feel
SLW-Coach
Real-time time-feel training
Monday, 28 February 2011
Time-feel and Perception
Monday, 28 February 2011
Swing with missing data
Monday, 28 February 2011
Case Studies
Monday, 28 February 2011
59% Swing on Swing ’42
Swing '42 (Reinhardt 1949)
Listen to an extract ending with the phrase in question
Monday, 28 February 2011
59% Swing on Swing ’42
Swing '42 (1:04-1:06) (Reinhardt 1949)
Although short, this continuous sequence of quavers is performed along one string and with
alternating down-up picking, so technical considerations are unlikely to influence execution
significantly.
The passage also has unmistakable gypsy-jazz feel, so it is of value to discover the contributing
time-feel components.
The tempo of the quavers allows a wide range of swing values.
Listen to an extract ending with the phrase in question
Monday, 28 February 2011
The mean swing comes out at around 59.3% with a standard deviation of just 1.6%. At
this fast tempo (≈204bpm) each crotchet lasts about 294ms, so there is a mere 50ms
separating offbeat placement for 50% and 66.6% swing. Reinhardt manages to sit tightly
between these extremes, occupying a time zone certainly no greater than 30ms, even
allowing for measurement ambiguity. Can the listener hear the difference between straight
and triplet quavers at this tempo, let alone a value in between?
Listen to 50%, 59%, 66.% and 75% with all articulation, timbral and weighting effects removed
Extract, sequences and clicks
Monday, 28 February 2011
The musical effect of this swing value (coupled with the simple
polymetry in the phrase) is at least as important as the note
choices, but is lost to standard notation.
A transcription may provide the notes, but the musician is left to
absorb this equally important aspect unconsciously.
Monday, 28 February 2011
The relevance of the swing curve, the changing swing values through the phrase, is miniscule
and of little perceptual significance other than perhaps the introduction of an extremely
subtle human looseness.
Audio 1 plays the phrase with 59% swing, followed by a faithful rendering of each swing
value (rounded to the nearest 1%) very subtle more even quaver ‘kick’ at the end of the
phrase
Audio 2 plays the last two quavers at 59% then 54% swing) but this 5% distinction(<15ms) is
difficult to hear in isolation even with directed attention.
Monday, 28 February 2011
This case-study demonstrates a useful strategy of
listening, objective analysis, theoretical modeling,
directed attention and aural testing training of
the ‘theory-filtered’ results.
Analysis divorced from music perception is of
little value, but music technology now allows a
powerful tool to marry theoretical analysis with
subjective experience.
Monday, 28 February 2011
A Little Drag
Swing, latency and hierarchy in Michael Jackson’s The Way You Make Me Feel
. The song is characterised by an offbeat keyboard stab, and Jackson’s direction to the music director and
keyboardist Michael Bearden concerning the placement of this offbeat provides a remarkable insight into
his attention to time-feel and means of communicating it to his band. The language employed is a neat
illustration of the separation of swing and latency outlined in the SLW model.
Monday, 28 February 2011
The song is characterised by an offbeat
keyboard stab, and Jackson’s direction to the
music director and keyboardist Michael
Bearden concerning the placement of this
offbeat provides a remarkable insight into his
attention to time-feel and means of
communicating it to his band. The language
employed is a neat illustration of the
separation of swing and latency outlined in
the SLW model.
Monday, 28 February 2011
Predominately shuffle rhythm, and the introduction, the focus of this study, is
at a slow tempo (≈82bpm).
For the purposes of this study it will describe it as a 2nd quaver with 66.% swing.
The distinction has little bearing on the substance of the analysis, but this
choice has been made to keep terminology consistent.
Monday, 28 February 2011
Extract A
Monday, 28 February 2011
Extract B
Monday, 28 February 2011
Extract A
Approximate calculations
1.3% (≈9ms), 3.3% (≈22ms) and 5.% (≈44ms)
Jackson’s instructions ‘little drag’ and play ‘a little more behind the beat’,
‘like you’re dragging yourself out of bed’ - evocative description of performing
latency
Monday, 28 February 2011
Extract A Perception
Approximate calculations
89bpm, three repetitions of the swing values of 68%, 71% and 73% are played
sequentially (panned right) against the baseline 66.67% triplet quaver (panned
left) while a click (panned centre) marks the pulses.
In the author’s experience the panning effect is barely noticeable at 68% but
clearly identifiable at 71% and above.
Monday, 28 February 2011
Extract A as Latency
Recalculate as latency to align with the vocabulary
used - musically identical with no downbeat
Monday, 28 February 2011
Extract B
Jackson happier with feel, starts singing and playful lateness
Monday, 28 February 2011
The playfully late chord (bar 3, beat 3) seems to cause Moffett to drop back a little tempo
(2bpm). The analytical context is not adequately controlled, the discrepancy small and the
technological limitations far from ideal, so little should be read into the event.
However, if the drummer did in fact feel compelled to accommodate the significant latency
with a reduction in tempo, this would suggest that the governance of tempo is not entirely his
responsibility.
Using the terminology of the SLW model this would imply that master timeline determination
is not a monopoly but an oligopoly:
The latency of Bearden’s (implied) onbeat actually stretched the master time-line, so Moffett
did not accept total responsibility for its placement. In this example the situation is tenuous, but
the calculation may as well be completed: A 2% drop in tempo was caused by a 13% latency, so
for that bar, the hierarchical weighting of master time-line determination would be calculated as
Moffett: ≈85%, Bearden ≈15%.
Monday, 28 February 2011
No-Man’s Land
2-bar sequence of the keyboard part played at 67%, 69%, 71%, 73% and then 75%
offbeat placement.
The 70-73% range would appear to cover Jackson’s desired feel, a no-man’s land (or
obtusely written) area of standard notation, but a perceptual and effective musical
experience nonetheless.
Monday, 28 February 2011
This case study highlights the potential value in the analysis of rehearsal footage (and
multitrack recordings) of particular artists.
Research of this kind is a powerful tool in the identification of the practitioner’s intention -
and perception - of time-feel, and a valuable contribution to our understanding of the
mechanics of rhythmic expression.
Monday, 28 February 2011
Constant Friction
Swing friction in Chuck Berry’s Johnny B. Goode (1958)
Monday, 28 February 2011
Constant Friction
Swing friction in Chuck Berry’s Johnny B. Goode (1958)
The straightness of rhythm guitar creating
a continuous time-feel dissonance with drums and piano.
Listen to extract.
Monday, 28 February 2011
Amalgamated Time-feel Template
Listen:
1) As above 2) All swung 3) All straight
4) Meet in the middle 5) As above
Monday, 28 February 2011
Introduction of weighting and looseness
Monday, 28 February 2011
Theoretical analysis. Technological recreation to test the results divorced from other
aspects of performance.
Analysis and recreation (technologically and live performance) of Ensemble time-feel
Monday, 28 February 2011
Swing Blocks
Swing values as arrangement in Little Wing (1967)
Monday, 28 February 2011
Swing Blocks
Swing values as arrangement in Little Wing (1967)
Seminal electric guitar piece but challenging to reproduce with complete authenticity.
The idea that time-feel can change significantly from phrase to phrase adding another
(generally-missed) structural level.
Monday, 28 February 2011
Multi-level swing analysis
(Semiquaver and quaver)
grouped in 4 levels:
Straight (<53%)
Light (53-56%)
Medium (56-63%)
Heavy (>63%)
Monday, 28 February 2011
4 levels of swing
Straight (50%)
Light (55%)
Medium (62%)
Heavy (69%)
Monday, 28 February 2011
Generally straight quavers,
but bar 7 hints at the
concept of
Swing Telescopy
Monday, 28 February 2011
Semiquaver/Quaver Swing
(60:50)
One rhythmic event
change:
large music difference
Semiquaver/Quaver Swing
(50:60)
Monday, 28 February 2011
The typical viewpoint of swing is as a stylistic characteristic, or
representative of a particular artist. Little Wing however provides a clear
example of widely varied swing values used as a structural mechanism in
performance (a far more sophisticated and intuitive version of the
‘swing-latin-swing’ format found in some jazz arrangements) and explains
part of the virtuosity in its execution and the challenge in its convincing
reproduction.
Monday, 28 February 2011
Push-Pull
Expressive Latency Contours in Comfortably Numb (1979)
Monday, 28 February 2011
The simplicity of arrangement in Comfortably Numb (Pink Floyd 1979) affords ample opportunity for
Nick Mason and Dave Gilmour’s effective use of latency in order to build and release tension.
Nick Mason’s much admired Time-feel- but not explicable by him.
Can a time-feel analysis reveal anything of this ‘knack’?
Monday, 28 February 2011
4 fills exhibiting a push-pull mechanism on last 2 beats.
Time points & latency below:
Monday, 28 February 2011
4 fills sequenced in clicks: first quantized then as in performance.
Monday, 28 February 2011
This analysis goes some way to explaining the mechanics behind this aspect
of Mason’s playing and the emotive effect it has on its listeners. It also
raises the question of the plasticity of the crotchet (and other
subdivisions).
Does a gentle latency contour allow a greater range of values for
aesthetically acceptable latency values? In other words if latency is
increased (or decreased) gradually, can extreme values be tolerated (and
enjoyed) more than random stabs of onsets either side of the beat? One
can visualize this idea as a fabric that can be stretched slowly, but will snap
if pulled too quickly.
Monday, 28 February 2011
Temporal
Plasticity
Pat Martino’s time-feel and M-Space mechanisms.
Monday, 28 February 2011
Just Friends (1962)
Solo Break
Monday, 28 February 2011
Monday, 28 February 2011
Monday, 28 February 2011
Monday, 28 February 2011
Reconstruction using upbeat iso-placement
Generalises some of the details of the passage (in particular the swung and late quaver pair in
bar 5, beat one) but much of the feel is captured
Monday, 28 February 2011
How much of this is known by Martino?
Monday, 28 February 2011
Opportunity for purpose built study with Martino
Known tune ‘Welcome To A Prayer’
Calibrated and well measure audio
Slow tempo so he could exploit all manner
of his well-known rhythmic expressions in ballad format
Melodic transformation with time-feel
Monday, 28 February 2011
Melodic shadowing
Monday, 28 February 2011
Monday, 28 February 2011
Melodic shadowing
Delay (and emphasis) of key notes from the
tacit ‘platonic ideal’
Monday, 28 February 2011
Time-feel moments
Monday, 28 February 2011
Note separation: Temporal plasticity
Monday, 28 February 2011
Latency
Latency contours/temporal plasticity
Monday, 28 February 2011
Note separation against melodic register: Expressive contours
Multi-level nature of music
Monday, 28 February 2011
Reapplication in performance, rehearsal and electronic composition
Omnia 5’58” Mermikides 2007
Monday, 28 February 2011
Avoiding Analysis Pareidolia and Confirmation Bias
Monday, 28 February 2011
Finding the important stuff, and not bean-counting
Monday, 28 February 2011
24/7
Musical expression is found between the lines
Monday, 28 February 2011
Milton Mermikides
[email protected]
http://scribd.com/mmermikides
miltonmermikides.com
PA40
Monday, 28 February 2011
Monday, 28 February 2011
