November 26, 2007

Rock syncopation: stress shifts or polyrhythms?

Here's something that's been on my to-blog list for a long time. The topic is the basic nature of American popular music. The illustrative examples are taken from a talk I gave at Stanford in January of 2005, as part of a workshop on Language and Poetic Form, and the problem is one that I've been thinking about, on and off, since I was in graduate school. I don't expect to have a chance to write it up any time soon, so I'm using this blog post as a way to get some of the issues on the table for discussion with anyone who may be interested.

A good way to explain English accentual-syllabic verse is to compare the metrical scansion of a line to the alignment of a song's lyrics with the meter of the music. In an earlier post ("An internet pilgrim's guide to accentual-syllabic verse", 7/6/2004) I recycled a pedagogical example that I've been using for a long time -- the fact that American pre-school children know, without being told, how to set the many verses of Skip to my Lou, e.g.

 E               C               E               G               (pitches)
 X               X               X               X               (1/4 notes)
 X       X       X       X       X       X       X       X       (1/8 notes)
 X   X   X   X   X   X   X   X   X   X   X   X   X   X   X   X   (1/16 notes)
little  red     wa-     gon     pain-   ted    blue 
cat      in the but-ter milk    lapping up    cream
rab-bit  in the corn    field   big as  a      mule
hogs in the  po-ta- to  patch   rooting up     corn
dad's    old    hat      and    ma- ma's old   shoe 

It's not just in traditional children's songs that the musical meter and the meter of the lyrics line up, at least to the extent that the strong syllables of the verse generally align with the musical beat.. This was the standard pattern in English-language art songs, popular songs and folk songs alike for hundreds of years; and famous examples of some popular styles continue the tradition. For example, the words to Hank Williams' I'm so lonesome I could cry are an example of the traditional ballad stanza, with alternating four-stress and three-stress lines; and the strong positions in the lyrics line up with the beats of the music in a simple way:

      |  #          #           #           # 
       Hear   that lone-  some whip-   poorwill,
      | #           #           #           #
    He sounds too blue     to fly.
      | #           #           #           #
   The mid- night train    is whin-   in'  low,
      | #           #           #           #
I'm so lone-   some I   could cry. 

The music has four beats to the bar, each subdivided into three by the strumming guitar -- this might be notated in 12/8 time, or in 4/4 time with 8th-note triplets. And as usual in traditional sung ballads, each of the strong syllables in the lyrics is aligned with one of the musical beats, leaving one empty musical beat at the end of the even-numbered lines.

In the first two lines, Williams delays the weak syllables so that he can draw out the strong syllables like that lonesome train whistle. (I've marked the musical (and lyrical) beat with #, and guitar-strumming pulse with %.)

   #                       #                       #                       #
   %       %       %       %       %       %       %       %       %       %       %
   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x
  Hear               that lone-              some whip-               poorwill,

In the last two lines of the stanza, the off-beat words align with the musical meter in a more standard sort of way, lining up with the third of the ternary subdivisions of the beat:

           x                       x                       x                       x
   x       x       x       x       x       x       x       x       x       x       x    
   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x   x
  I'm so  lone-           some     I             could    cry

An 18th-century audience would have been puzzled by the reference to a "midnight train", and no doubt by other aspects of the content, but in general, this style of setting words to music would have been familiar to them.

But the situation is very different in many genres of American popular music -- a different sort of tune-text alignment starts to appear in the early years of the 20th century, and becomes dominant by the century's end. Here's a stanza from James Taylor's 1971 lullaby You can close your eyes:

 #      . |  #   .  #  .    #  .  #  .  | #  .  #  .  #  . #  .  | #    # # # | # # # 
Well, the   sun  is   sure-      ly sink-          in'       down,

 #      . |  #   .  #  .    #  .  #  .  | #  .  #  .  #  . #  .  | #    # # # | # # # 
but   the   moon is  slow-       ly ris-           in'.

 #      . |  #   .  #  .    #  .  #  .  | #  .  #  .  #  . #  .  | #    # # # | # # # 
       So  this  old  world   must still    be      spinnin'      round.
 #      . |  #   .  #  .    #  .  #  .  | #  .  #  .  #  . #  .  | #    # # # | # # # 
       and   I       still          love          you.

The meter of the lyrics is the same traditional ballad stanza, with its alternation of four-stress and three-stress lines: "Well, the SUN is SUREly SINKin' DOWN / but the MOON is SLOWly RISin'". (It's unusual to rhyme the odd lines rather than the even ones, and the third line has an extra stress group in it, but never mind that for now.) The musical meter is simple 4/4 time, with four beats per bar, subdivided into two. The setting is sparse -- there are a couple of wordless bars after every line, so that the ballad stanza turns into a 16-bar form rather than an 8-bar form. But the key thing here is that only the first strong position of each line is reliably aligned with a musical beat. Of the other 10 metrically-strong syllables, 8 are off the beat.

What's going on? The lyrics are metrically clear enough, and the music is plain vanilla 4/4 -- but the musical meter and the lyrical meter don't line up very well at all.

A theory of such tune-text misalignments can be found in David Temperley, "Syncopation in Rock", Popular Music 18(1), 1999:

The Oxford Companion to Music defines syncopation as 'the displacement of the normal musical accent from a strong beat to a weak one'. While I will propose a slightly different understanding of the term, this definition does point to two important aspects of syncopation. First, syncopation involves a deviation from the 'normal' placement of an accent: usually, accents occur on strong beats, but in a syncopation, a weak beat (or rather an event on a weak beat) is accented. Second, syncopation involves displacement; in a syncopation, an accent that belongs on a particular strong beat is shifted or displaced to a weak one (I will suggest that it is actual events, rather than accents, that are displaced, but the idea of something being displaced from where it belongs is essentially right). The distinction between these two points -- the accenting of weak beats, and the displacement of an accent from one beat to another -- is important. Simply understood as the accenting of weak beats, syncopation is commonplace in many kinds of music, including classical music. [...] However, it is not at all clear that these passages involve displacement; are the accents really heard as belonging on some other beat? In the case of rock, however, the sense of displacement is much more apparent, and there are strong constraints on the ways in which this displacement occurs. In this way, I shall argue, the nature of syncopation in rock is fundamentally different from that in classical music. [...]

The clearest evidence of syncopation in rock is found in the setting of text. [...] It is a well-known fact that text tends to be set to music in a way which matches the stresses of the text with the metrical accents or 'strong beats' of the music. In most pre-twentieth century vocal art music or folk song, it will be found that stressed syllables generally coincide with relatively strong beats. This can be seen in the Handel passage in Example 2a. ["The mighty God, the everlasting father, the Prince of Peace".] In the two Beatles melodies, however, this is clearly not the case. [...] In 'Here Comes The Sun', 'been', 'long', 'lone-' and 'win-' are stressed; the syllables in between are unstressed. But in the music, each of these syllables falls on a weak metrical beat [...] At first thought, this might suggest that the rules for text-setting in rock are fundamentally different from those in classical music. Notice, however, that if certain notes in the melodies are shifted over by a quaver or semiquaver, the metrical grid become nicely aligned with the stress pattern of the words. It seems reasonable to suggest, therefore, that in the internal representation we form when listening to rock music, we are understanding the metric grid and the stress pattern as coinciding. We retain, on principle, the assumption that stressed syllables should occur on strong beats, but we understand certain syllables as 'belonging' on beats other than the ones they fall on.

A similar set of issues arises in purely instrumental music, but tune-text alignment provides an especially clear form of the phenomenon, since the intrinsic stress pattern of the words is understood independent of their setting.

Temperley's examples are typical, his description is correct, and his explanation is compelling. But it seems to me that it's not the whole story, and for some American popular music it might not be the right story. In some cases, at least, there's an alternative account, in which stressed syllables don't stylistically shift off of strong beats, but rather land on strong beats that aren't in the standard place -- because we're dealing with polyrhythmic music. The African roots of American music are elaborately and diversely polyrhythmic; and some of this has remained, and been periodically reinforced by influences from Latin America and from Africa directly.

For example, the "habanera rhythm" (originally contradanza habanera) is 8 beats divided as 3+3+2 or 3+(1+2)+2 or (1+2)+(1+2)+2, and at the same time (at least implicitly) divided into the "square" 4+4 rhythm. In musical notation, the 3+3+2 rhythmic pattern might be rendered something like:


Counting in 16th-notes, these would be something like the two lines below the digit string:

o       o       o       o       o
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 ...
o     o     o   o     o     o   o
o o   o o   o   o o   o o   o   o

According to Steven Loza's article on Latin Caribbean Music in Vol. 3 of the Garland Encycopedia of World Music,

The habanera was the first Cuban style to strongly influence music in the United States and was probably the most influential style throughout the Americas. Besides directly affecting jazz, it was the root of the Argentinian tango and fed Mexican styles that were to travel north. [...]

In the early nineteenth century, the contradanza preceded the popularity of the habanera. According to many sources, the contradanza was brought to the United States by French refugees fleeing the Haitian revolution. The dance was also influenced by certain African forms. In Cuba it developed two closely related time signatures, 6/8 and 2/4, both of which influenced the later dance form. "It was presumably black musicians who began to syncopate the contradanza's rhythm. A so-called ritmo de tango, extremely similar to the Argentinian tango, became a feature of Cuban contradanzas, and spread into many other local forms" (Roberts 1979:5). In addition to the merengue of Haiti and the Dominican Republic, the cinquillo, a five-beat throb cast in duple meter, also became a feature of Cuban contradanzas, most likely via the French contradanse. Both the ritmo de tango and the cinquillo became fundamental to the contradanza habanera by the early nineteenth century. Black musicians further syncopated the ritmo de tango, Africanizing this widespread European dance. Though not purely a Havana style as implied by its name, and never referred to as such by its creators, the habanera was easily absorbed into American music largely because its rhythmic pattern was contained in a single measure of common 4/4 meter, and it was frequently incorporated into the bass line of piano compositions. The earliest known piano version of the habanera was "La Pimienta," written in 1836.

Jelly Roll Morton called such polyrhythms the "Spanish Tinge":

Now in one of my earliest tunes, "New Orleans Blues", you can notice the Spanish tinge. In fact, if you can't manage to put tinges of Spanish in your tunes, you will never be able to get the right seasoning, I call it, for jazz.

In a robotic midi implementation, a habanera rhythm sounds something like this:

A more graceful and elegant performance is the background to Little Richard's Slippin' and a Slidin':

Here's the time waveform for the background segment, showing the 3+3+2 structure:

But the 4+4 rhythm is also implicitly there, and that's the pattern that the lyrics lock onto for the first eight bars -- which is one line repeated twice, as usual for in this song form -- while the instrumental background keeps up the 3+3+2 pattern:

  o          o          o      o          o       o
  1  2   3   4   5  6   7  8   1  2   3   4  5  6 7  8
  o              o             o             o
slippin' and a slid-    in'   peepin' and a hid-  in'

  o          o          o     o          o   o    o
  1  2   3   4  5   6   7  8  1   2   3  4   5  6 7  8
  o             o             o
been  told   a long    time ago

And then in the final four bars of the stanza, the lyrics shift and lock onto the 3+3+2 rhythm:

1  2  3  4  5 6  7 8  1  2  3  4   5 6  7  8  1 2  3  4 5 6 7  8  1 2 3 4 5 6 7 8
o                     o                       o                   o
o        o       o    o        o        o     o       o     o     o
I  been told    baby you been bold   I won't be your fool  no more

You could analyze this as a "rock syncopation" in Temperley's terms -- the strongly-accented words "told", "bold" and "fool" have all shifted from the fifth eighth-note of their bars to the fourth. In the standard, square (4+4) organization of those notes, this is a shift from one of the strongest positions in the meter to one of the weakest. But in the (3+3+2) habanera pattern, the fourth eighth-note is exactly where the ictus belongs.

So was it the "square" setting of the previous line that was actually "syncopated"? Not really -- the (4+4) pattern is also simultaneously available. Neither setting really represents a shift away from the strong positions in the musical rhythm. Instead, the setting is shifting between one polyrhythmic definition of metrical strength and another.

But on the other hand, as John Halle has pointed out to me, the final strong syllable "more" anticipates the downbeat, in a way that is so common in this style of music as to be essentially obligatory -- and this seems much more like one of David Temperley's shifts.

It seems to me to be an open question how much of verse-setting in American popular music should be seen in terms of stresses shifted off the beat, in Temperley's sense, and how much in terms of stresses shifting between polyrhythmic figures, as in the Little Richard song analyzed above. There are many examples that seem clearly to call for the syncopation theory, and many others that seem equally clearly to motivate the polyrhythm theory -- and many where either theory might plausibly be applied. It may well be wrong to aim to resolve the question, at least in any overall way-- perhaps this ambiguity is one of the symptoms of the cross-fertilization of musical traditions in the past century of popular music. The same piece may even have a "European" interpretation for some performers and listeners, and an "African" interpretation for others.

A great deal of evidence is out there, in the patterns of thousands of performances in dozens of styles. There's been relatively little research on this, oddly enough; and the field is open to anyone with a laptop and an interest in the question.

Posted by Mark Liberman at November 26, 2007 06:13 AM