Speeding Up, Slowing Down
I had just decided to study music for real when I began to study bass with Ron McClure. A great bassist and pianist, McClure would play piano in our lessons. His first assignment; learn to internalize tempo. When we tried to play a tune (it was “Alice in Wonderland,” a waltz by Sammy Fain) I clearly did not know how to handle a simple 1-2-3, 1-2-3 count-off. “You gotta be able to take a count-off!” McClure rightly pointed out. I have since learned this skill, more or less, but the complexity of the task is notable.
Keeping a simple beat is at once the most elementary of jobs and a window into the only vaguely understood cognitive process of musical timekeeping. A standard jazz count-off, half notes verbally counted 1, 2; then quarter notes 1, 2, 3, 4, is two measures long. At faster tempi this is sometimes counted as whole notes 1, 2; then half notes 1, 2, then quarter notes 1, 2, 3, 4, and totals four measures. These count-off patterns are notable first because of their brevity; musicians are expected to adjust their sense of musical time for long repeated forms (jazz pieces are often more than 300 measures long) based on a count-off unit which is only 1/16th the length of a standard AABA form, and a tiny fraction of the whole piece. In this context tempo, such a key aspect of a piece, is (in part at least) determined by a count-off so short it can be compared to measuring a football field in inches rather than yards. There is also the quickness with which musicians internalize and express these brief count-offs as temporally regulated music. Even if a musician ends up playing a bit faster or slower than the count-off tempo, they have still accomplished a remarkable feat of cognition by taking a 2 or 3 second count-off and turning it into a real time reflection of the counter’s intended speed.
One analogy for this is the carpenter’s level. A short six inch level (like a short count-off) doesn’t give an accurate sense of a surface’s relationship to the “level” ground below. Carpenters use 3 foot or longer levels to see if something is straight. While the ability to maintain constant tempo based on prompts like the count-off is considered a basic skill for musicians, it’s pretty unusual for a count-off to be as long in time as an accurate level is in length, though it seems such a practice would result in a better, more “level” relationship between intended and performed tempi.
This analogy overlooks an important point: The carpenter’s level is a mechanical device used to show the relationship between an object’s surface and gravitational pull (a fixed quantity), whereas human musical time is a neurological process which is expressed using mechanical means such as instruments and the voice. There is no ideal fixed (“level”) quantity in musical timekeeping. While it is possible to compare a live performance to a metronome (this will happen later in the post), real timekeeping in music isn’t directly comparable to mechanical time; it’s apples and oranges. The human ability to keep time is a phenomenon influenced by personal, social, and cultural rhythmic preferences, not mechanical forces such as gravity or clock-like ticking. While the metronome is definitely useful and has been around since Beethoven’s time, it is a tool for practice and reference, not suitable for deep analysis of musicians’ in-the-moment rhythmic expression. But there must be some way to understand why some performances are so engaging rhythmically. We can all think of music that is rhythmically “good,” and musicians often describe another player’s pleasing rhythmic skills as “good time.” What is “good time?”
This question is often mixed-up with a related but different question something like “does [person] have good [i.e. metronomic] time?” While the ability to keep time with clock-like accuracy, especially in improvisational or groove-based music, is a sign of exceptional musicianship, this ability is only related to the idea of good timekeeping in music. Saxophonist Bob Parsons, another teacher of mine, once described “good time” as a player’s ability to rhythmically invite both listeners and other players into the musical texture. While there must be a basic level of rhythmic continuity (principal beats more or less equally spaced in time) for a groove to be perceived, it is the intent of the performed rhythm, not its relationship to mechanical time, that really makes for “good time.”
As I compile ideas and guidelines for developing an organic drum machine, it’s significant to realize that “good time” and metronomic time are two rather different concepts. While all computers parse time in a clocklike fashion, it must be possible to create data sets and algorithms that go beyond standard drum machine swing and randomness settings (which use deviation from and return to metronomic time in an effort to emulate human rhythm) and generate digital pulse based rhythm without hegemonic reliance on metronomic time. What kinds of information can be compiled or inferred from actual performed rhythm to make a rhythm machine whose fealty is not just to clock time?
Returning to the count-off, it seems unlikely that musicians could take an uttered “1…2…1..2..3..4!” without context and turn it into a groove. The count-off is more of a final declaration of tempo that begins with an acknowledgement of musicians present (“here we are, for whatever reasons, about to play music”), continues to a more specific statement of repertory (“let’s do “The Flintstone’s Theme Song”"), and ends with the count-off (“1…2…1..2..3..4!”). Along the way to the count-off there’s a whole lot of other contextual information that refines each musician’s sense of how to exhibit “good time” in the performed music. So the count-off, interesting as it is phenomenologically, is only a small part of the rhythmic information available to performers when starting a piece. An awareness of these musical and socio-cultural contexts may help to define a new “hyper-metronomic” approach to computerized rhythm.
The following example illustrates the disconnect between “good time” and metronomic time. Pianist Harold Mabern recorded “Seminole” on his 1991 album Straight Street with Ron Carter, bass, and Jack Dejohnette, drums. More than players with good time, this trio consists of genre-defining jazz musicians whose rhythmic concepts have been central to the development of jazz over the past 50 years. Straight Street was well received, and a Critics Pick for top ten jazz album in The Village Voice in 1992. A 64 measure AABA song form based on Ray Noble’s “Cherokee,” “Seminole” speeds up significantly from its original tempo. Here are the first two A sections:
Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.
After a full chorus of melody and three choruses of piano solo the final chorus is reached. Here are the first two A sections of the final chorus:
Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.
From a starting tempo where quarter notes are about 156 per minute the piece speeds up (especially in the first 96 measures) to about 210 quarter notes per minute, and by the second bridge is about 33% faster than at the beginning. While this speeding up is noticeable, and quite apparent at the final chorus, to my ears the music and players have “good time,” and I do feel invited into their musical space. Further, each member of the trio is an authoritative figure in the world of jazz. For some listeners this amount of acceleration could reveal bad time, an inability to maintain metronomic accuracy, and, if the trio were lesser known players I might join (right or wrong) in this judgement. Without a doubt if I were making an album and realized a piece accelerated that much I’d want to re-record it. It’s possible but unlikely that the trio intentionally sped up.
If great musicians speed up (and slow down) great drum machines should too! While the example above is a clear case of speeding up, most (all?) real performances of groove based music deviate in more subtle ways from metronomic time, both within metrical units (between beats and small groups of beats) and at the phrase or form level. Can this aspect of human musical timekeeping be integrated into digital rhythm programming? In the first part of this series I proposed that musicians in groove based music tend to “metronomize” or internalize a rhythmic pattern that’s syncopated in relation to the principal strong beats of their part. How are these internalized patterns influenced by the various contexts (both communally shared and individual) provided by the musical moment, and temporally stretched or compressed in reaction to these contexts?
Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.
“Seminole” Harold Mabern, piano; Ron Carter, bass, Jack Dejohnette, drums.
“Seminole” is a contrafact based on “Cherokee,” a very standard jazz standard. “Cherokee” is attributed to bandleader and composer Ray Noble (although he is known to have bought others’ compositions and put his name on them), and its song structure is notable in the development of bebop as the basis of Charlie Parker’s “Koko” and other jazz improvisational frameworks. “Cherokee” and its contrafacts are well known in jazz for their very quick tempi, often performed at speeds above 300 quarter notes per minute. “Cherokee’s” largely pentatonic melody has lots of whole and half notes, and the melody is usually played by horn players. Mabern’s “Seminole,” on the contrary, is played at a medium tempo and the A section melody is arranged for string bass and piano in unison. These contrasts between “Cherokee” and “Seminole” are contributing factors to the speeding up discernible in Mabern’s recording. I had first hand experience playing “Cherokee” with Harold Mabern at the NYC local 802 jazz jam. He called the tune in an exotic key (E major) and counted off a blistering tempo. Most jazz musicians have played “Cherokee” many times, and usually very quickly. These remembered experiences of playing “Cherokee” fast predispose the trio to push the tempo in their performance of “Seminole.”
The first two A sections of “Seminole” maintain a more or less metronomic tempo, though the passages with triplets do speed up a bit. Carter begins to play time rather than the melody from the last four measures of the second A into B, and this familiar texture initiates a more intense speeding up. The B section (:52- 1:12) is marked by a recurring 2-measure rhythmic displacement, “hits” marked by the whole group. Here the speeding up is apparent; especially at :55-1:01 Mabern’s chords anticipate Dejohnette’s brushes, and by the last A of the opening melody the piece has almost reached its quickest tempo. By the B section of the second chorus “Seminole” has a stable tempo of around 210 quarter notes per minute, which is more or less maintained until the end of the piece.
Chris Dobrian, one of the great professors I studied with at UC Irvine, once mentioned to me that when music speeds up or slows down in this way it’s often because the chosen tempo is not well suited to the composition. Was Mabern’s count-off too slow? It’s hard to say. The initial tempo is a great but difficult “in the pocket” type of medium swing. It seems more likely to me that the trio (especially Carter), perhaps under-rehearsed, was a bit tense and working hard to read through the first two A sections, and when the more open feel leading to B arrived the group (especially Mabern) was glad to release this tension as an increase in tempo. When the rhythmic hits at B arrived Mabern and Carter both ran with the feeling of release, and the performance sped up a lot. It’s notable that Dejohnette compensates for this shift almost instantaneously, maintaining the trio’s good time even as the tempo is really taking off.
This analysis is based not on any firsthand knowledge of the participating musicians’ recording process but is instead my own impression of the way in which the piece speeds up. I am projecting personal assumptions and tendencies onto the rhythmic phenomena found in “Seminole.” While surely inaccurate and more guesswork than fact-based inference, it is just these kinds of fuzzily logical processes that inform musical enactment of rhythm. For the organic drum machine to have good time, to emulate real human rhythmic inflection, it must also make educated guesses about where tempo and inflections should be, just as this top-level trio does in “Seminole,” and I do in analyzing “Seminole.”
Where should I start with creating these kinds of algorithms? When Ron McClure counted off “Alice in Wonderland” and I was not able to effectively take the count-off the problem was partly my novice abilities. Beyond that, however, I did not have the context necessary. Once we were able to play the piece I recalled that it was performed by Bill Evans on the classic album Sunday at the Village Vanguard. I didn’t know the song by its title, but, once I played it, I was able to access my own memory of Evans’ recording of the piece (a memory which I am sure McClure shared). Given that shared context a count-off becomes much less important. I have since played “Alice in Wonderland” dozens of times, and each version’s tempo has probably been within 20% of Evans’ recorded tempo. The organic drum machine should have just this kind of hyper-metronomic contextual information such that, for an extreme example, if you used it to play “Alice in Wonderland,” but indicated a tempo much slower than that of recorded versions of the piece, it might tend to speed up a bit. Further, such a drum machine should maintain a sense of form and exhibit some of the classic human behaviors; sometimes rushing into new sections, tending to pull back or join in when others seem to speed up. These are all rather advanced programming concepts, and, while they tilt at a phrase-level hyper-metronomic system for contextualizing performed rhythm, rely on a rational framework for defining beats and beat subdivision. Part three of this series will discuss what beats are and how to find them using max/msp/jitter further discuss phenomenon related to human execution of rhythm.
