July 06, 2004

An internet pilgrim's guide to accentual-syllabic verse

Yesterday I cited evidence that some English professors may be a little shaky about verse scansion. In the comments, Charles Hartman raised a valid point about the (in)appropriateness of the classical taxonomy of foot types as a basis for metrical description, and also asked what sort of linguistic analysis might be genuinely useful in analyzing poetic rhythm.

I'm going to try to answer Charles' questions in three stages.

First, in this post, I'll lay out a way of thinking about accentual-syllabic verse. This point of view is by no means original to me, but I'll present it without references and in a somewhat idiosyncratic form, for ease of assimilation. Second, in a later post, I'll take up the question of "feet" and the classical bestiary from amphibrach to trochee, and discuss Charles Hartman's implication that this framework can be confusing and even misleading if taken too seriously as a foundation for metrical analysis. In a third post, I'll offer a few suggestions about the basic aspects of language sound structure that (I think) anyone who cares about poetic meter -- and rhythmic patterns in poetry more generally -- should understand and know how to apply descriptively.

I hope that all this will be helpful for readers who are uncertain about how metered verse in English works. I recognize that there are other readers who know much more than I do about some or all aspects of these phenomena. I hope that they'll find a few interesting things here, if only by association or in reaction, and I expect to learn from their comments and objections.

The rest of this post is a mildly-edited version of the metered-verse section of my notes for lecture #16, "Linguistic Form in Art, Ritual and Play", from Linguistics 001, the introductory linguistics course at Penn.

Tune-text alignment in English

Consider the first verse of the simple song Skip to my Lou, as presented in Ruth Crawford Seeger's American Folk Songs for Children (Doubleday, 1948).

In this verse, as throughout the song, a single line is repeated three times, against a simple melody that sketches a major triad in the tonic, the dominant, and then again the tonic.The verse ends with the invariant line "skip-a to my lou, my darling."

The songbook gives a couple of dozen other verses. Each has the same structure -- a single line repeated three times, and the invariant ending "skip-a to my lou, my darling." Thus the problem of fitting words to music can be reduced for each verse to the problem of
fitting a single line to the first two bars of the melody -- everything else is just repetition.

This is about as simple as songs get. Nevertheless, a four- or five-year-old learning new verses has to solve a non-trivial problem.

One way to look at the problem is to line a few verses up against a depiction of the metrical structure of the first two bars of the song. These two bars contain four "quarter notes". The metronome marking at the top of the music says that the quarter note equals 132,  i.e. 132 quarter notes per minute, or a little more than two quarter notes per second.

Standard western musical structure assumes a regular hierarchical subdivision of  time. In this case, each quarter note can be divided into two eighth notes, each eighth note into two sixteenth notes, etc. At each level, the first of the subdividing notes is "stronger" than the other -- it is the "downbeat."

Three levels are enough for this musical example. As for the alignment with the melody, the song provides a separate pitch for each quarter note. If that note is subdivided by the syllables of the verse, then the subdividing syllables just repeat the same note.

Here is the first verse -- this is just a schematic presentation of exactly the information provided by the musical notation above:

 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

Here are some other verses, aligned under another copy of the same melodic and metrical schema:

 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)
pig      in the par-    lour    what'll  I      do
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
 
There are many other verses -- Seeger provides a couple of dozen in the publication cited, and says "this song has hundreds of stanzas and is always picking up new ones. One collector alone gives 150, from which the above 22 were selected as encouragement to further improvisation."

The  samples given above are enough to give us a guess about the principles involved. For a start, we can say something about what the principles are NOT:

  • There is not a  fixed number of syllables in a line -- the samples have between 8 and 11 syllables.
  • Even for a given number of syllables, the alignment with the melody and rhythm of the song is not fixed by syllable order, but depends on the stress and structure of the words.

For instance, both "little red wagon painted blue" and "dad's old hat and mama's old shoe" have eight syllables, but if we used the syllable-by-syllable alignment of the first line for the second line, we'd get the impossible pattern:

  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)
dad's old hat     and     ma-    ma's     old   shoe

which gives the impression of stressing the line as "DAD's-old-hat AND maMA's old SHOE" (where the capitalized syllables correspond to the quarter-note beat of the song, and also to the points of pitch change).

No self-respecting American nursery school graduate would ever think to sing the line that way -- except perhaps as a joke.

[as a party trick, Haj Ross used to sing "Take me out to the ball game" with the words shifted by half a beat relative to the music. It's hard to learn to do, and curiously hilarious to hear.]

The principles of tune-text alignment for this song seem to be:

  • Each of the four quarter notes is always aligned with a stressed syllable
  • There is always at least one syllable in between each pair of quarter notes, so that the intervening eighth-note positions are always filled, but these intervening syllables are not always stressed
  • Syllables aligned with sixteenth notes  may be added to taste.

This implies that the minimum plausible line of "skip to my lou" might be seven rather than eight syllables. For instance, "Jane's old hat and Jim's old shoe" might be OK.

We can make some other observations, such as this one:

  • The eighth sixteenth-note position seems to be avoided (the "upbeat" to the second measure); this corresponds to a tendency to break the line into two parts at this point
  • The tenth sixteenth-note position  is often filled, perhaps to emphasize the continuity of the second measure.

You can verify for yourself that the rest of Ruth Crawford Seeger's cited lines follow the same pattern:

    Pull her up and down in the little red wagon
    Rats in the bread tray, how they chew
    Chickens in the garden, shoo shoo shoo
    Cow in the kitchen, moo cow moo
    Going to market two by two
    Back from market, what did you do?
    Had a glass of buttermilk, one and two
    Skip skip skip-a to my lou
    Skip a little faster, that won't do
    Going to Texas, come along too
    Lost my partner, what'll I do
    I'll get another one prettier than you
    Catch that red bird, skip to my lou
    If you can't get a red bird, take a blue
    If you can't get a blue bird, black bird'll do

The only real novelty in these additional examples is in the line ending with wagon, where there is an extra syllable aligned after the fourth quarter-note.

We can rephrase our observations by saying that Skip to my lou has a four-beat line, where the beats correspond to the quarter notes of the first two bars of the song, and where one to four additional syllables occur between each adjacent pair of beats.

I have seen four- and five-year-old children making up new verses to this song. No one has to teach them the rules -- they figure them out easily enough by themselves.

Most songs are more complicated than this one, but the basic principles of tune-text alignment in English remain the same: syllables are aligned with notes so that the stress pattern of the text and the rhythmic structure of the tune are congruent. If you have some familarity with designing computer algorithms, you might see if you can design one that will correctly specify the tune-text alignment for a simple song like this one.

To make up new verses -- or to sing old ones correctly -- you have to understand, implicitly, the metrical hierarchy of the music, the stress pattern of the text, and the way that they can be aligned. This understanding comes effortlessly to young children, providing  more evidence of the psychological reality (and naturalness) of the linguistic (and musical) concepts involved.

Why should there be something natural about the process of aligning two structures so as to make them rhythmically congruent? One plausible hypothesis is that this is the basis of coordination among speech articulators in ordinary talking. On this view, singing is just a kind of regularized and stylized form of speaking. In both cases, rhythmic structures are serving a coordinative function.

Accentual/syllabic verse in English

The principles of tune-text association for Skip to my Lou are basically the same as the principles that underlie most metered verse in English.

This is especially clear if we look at verse with a very clear rhythmic pattern, like Mother Goose rhymes, or Lewis Carroll's The Hunting of the Snark, or Robert W. Service's ballad The shooting of Dan McGrew, or Run-DMC/Aerosmith's Walk this way.

Let's take a look at how McGrew works. The poem has 58 lines, of which the first six are given below.

    A bunch of the boys were whooping it up in the Malamute saloon;
    The kid that handles the music-box was hitting a rag-time tune;
    Back of the bar, in a solo game, sat Dangerous Dan McGrew,
    And watching his luck was his light-o'-love, the lady that's known as Lou.
    When out of the night, which was fifty below, and into the din and the glare,
    There stumbled a miner fresh from the creeks, dog-dirty, and loaded for bear.

If you read these lines out loud, you can hardly avoid getting an impression of the intended rhythm. It's a seven-beat line, with either one or two additional syllables between each pair of adjacent beats. The beginning of the line can start with zero, one or two "upbeat" syllables. There is always a phrasal break between the fourth and fifth beats of each line, and occasionally there is no intervening syllable at this point (as if it were a line break).

We can annotate the rhythmic structure of the next six lines of the poem by using a sharp sign (#) for each "beat", a period for additional syllables, and a slash (/) for the phrase break:
 

     .  #      .   .  #   .   .  #   .    .   #  /.     #    .   .    #     .  .  #
    He looked like a man with a foot in the grave and scarcely the strength of a louse,
     .   .  #  .  .  #   .   #   .    .  # / .    .  #      .    #     .    .  #
    Yet he tilted a poke of dust on the bar, and he called for drinks for the house.
      .    .   #    .      #     .    #  .     #  /   .     .  #       .   #      .  .   #
    There was none could place the stranger's face, though we searched ourselves for a clue;
     .   .   #    .   #      .     .  #    .   #  / .   #  . .    #   .  #
    But we drank his health, and the last to drink was Dangerous Dan McGrew.
      .      #    .   #   .   .     #   .   #   / .    #     .   #    .   .   #
    There's men that somehow just grip your eyes, and hold them hard like a spell;
    .    #    .   #  .    .  #      .  # / .   .  #    .  .   #    .   #
    And such was he, and he looked to me  like a man who had lived in hell;
     .   .  #    .    #    .     .   #  .   #  / .  .  #    .    #  .   #
    With a face most hair, and the dreary stare  of a dog whose day is done,
    .   .  # .      .   #     .   #   .    #  / .     .   #    .   #    . #
    As he watered the green stuff in his glass, and the drops fell one by one.

This kind of annotation of the rhythmic structure of a verse is called scansion, and the basic rhythmic pattern of a poem (if it has one) is called its meter. The scansion shows us how the underlying pattern (here a seven-beat line with one or two intervening syllables) is realized in each line of the poem.

There is quite a bit to say about meter and scansion, even of metrically simple poems (some might even say doggerel) like McGrew. The point that we want to draw out here is that the basic principles are the same as those that applied in the case of Skip to my lou -- a certain number of beats per line, with variable (but constrained) numbers of syllables between the beats, and a regular break in a certain position.

Theorists distinguish among various kinds of poetic meter. The word meter means measure, and in each case, something is being measured or counted. In syllabic meters (as in French poetry), the only thing that matters is the number of syllables per line. In some languages (Classical Greek, Latin, Arabic, Hausa), the pattern of "long" and "short" syllables is regulated. In accentual meters, what is counted is accents -- or more properly beat-aligned accents. Most English metered verse is accentual-syllabic -- each line has a given number of "beats", but there are also more or less strong restrictions of the number of intervening syllables.

It is important to remember that poetic meter is an abtract pattern, a kind of grid against which the poet arranges his or her lines according to some general principles of congruence. How the congruence is defined depends on the poetic style, but also very much on the sound structure of the language that the poetry is written in. For metered verse to be a living form -- as it has been in many cultures around the world, both ancient and modern -- its patterns have to be defined in terms of phonological categories whose patterns poets and their audience can hear and feel.

Metrical feet

In the notation we've been using, The shooting of Dan McGrew is written in a fairly even mixture of . # and . . # rhythmic elements (225 . # and 198 . . # to be precise). The distinction doesn't really seem to matter to its form, which we described simply as seven beats with one or two intervening syllables, divided into two half-lines of four beats and three beats. The lyrics to the 1986 hit Walk this way (by Run-DMC and Aerosmith) have the same basic pattern: seven beats, divided as four plus three. However, now there can be as many as three weak syllables between each pair of strong syllables (where "strong" means "aligned with the beat"):

 . .  #   .  .    #    .    .  #      .    #   /  .   .  #  .   .  .   #  .  .   #
So I took a big chance at the high school dance, With a missy who was ready to play.

 . .    #   .  .   #  .     .      .   #    .    .  .   # . /   .    .  #    .  .   .  #     .   #
Wasn't me she was foolin' 'cause she knew what she was doin', when she told me how to walk this way. 

Thus Walk this way has exactly the same meter and rhyme scheme as The shooting of Dan McGrew, except for a slight relaxation of the meter: instead of one or two weak syllables between beats, Aerosmith's song has one, two or three.

Lewis Carroll's mock epic The Hunting of the Snark also has the same basic meter as The shooting of Dan McGrew: Here are the first two stanzas:

              .     .   #    .  .   #       .  #   .    #
            "Just the place for a Snark!" the Bellman cried,
                .   .  #  .   .    #   .    #
                As he landed his crew with care;
             .  #  .   .     #  .    .  #  .    .  #
            Supporting each man on the top of the tide
                 . .  #  .  .   #    .   .   #
                By a finger entwined in his hair.
 
              .     .   #    .  .   #     .  .    #   .    #
            "Just the place for a Snark!  I have said it twice:
                  .  . #     .    .  #  .     .   #
                That alone should encourage the crew.
             .     .   #    .  .   #     .  .    #   .     #
            Just the place for a Snark!  I have said it thrice:
                  .  .  #    .     .   #    .    #
                What I tell you three times is true."

Snark has alternate lines of four and three beats -- corresponding to the four/three division of the seven-beat line in McGrew. With the promotion of the half lines to full lines, additional rhymes are added (here cried/tide and twice/thrice) to reinforce the stanzaic form, but the meter is basically identical.

In Snark, however, the balance between . # and . . # shifts dramatically towards . . # 

There are 1754 . . # sequences, to only 251 . # sequences, for a ratio of about seven to one, while the . # sequences that do occur are essentially all at the beginning or the end of a line. Thus Snark is moving in the direction of fixing not only the number of "beats" -- of strong syllables in the line -- but also the number and placement of weak syllables.

In order to characterize poetic forms in which the arrangement of strong and weak positions is regulated in this way, poets and critics have borrowed the terminology of Greek (and Latin) metrics. The Greek metrical system was based on patterns of totally different units -- their meters did not care about the location of accented syllables, but rather regulated the pattern of long and short syllables. They then established a congruence between long and short syllables and patterns of long and short time-units in the musical meters of the period. Metrical systems that depend on syllable-length in this way are called quantitative. By contrast, English lyric poets rely on a congruence between stress patterns and the beat structure of our music, resulting in a metrical system that is called accentual or accentual-syllabic.

These different choices of basic poetic stuff are not arbitrary. The (classical) Greek language made a systematic distinction between long and short vowels, whereas English does not; English word-stress organizes the rhythm of English speech in a way that Greek accent did not.

Nevertheless, all poetic forms are based on analogies among different sorts of patterns, and it is easy enough to make an analogy between the Greeks' patterns of long and short syllables, and our patterns of strong and weak syllables. Thus we can borrow the Greek term iamb -- applied to the Greek pattern "short long" -- and apply it to the English pattern "weak strong." The Greeks called these basic patterns "feet" (actually of course they called them the equivalent in Greek). Here are some of the commoner foot names, represented with the typographically convenient (but non-standard!) notation of "." for short positions and "#" for long ones:
 
 

iamb . #
anapest . . #
trochee # .
dactyl # . .
spondee # #

In discussing classical (Greek and Latin) metrics, it's more common to see the macron used for long positions and the breve for short ones, with a vertically-stacked combination of the two symbols used for "common" positions that might be either long or short, thus:

For English accentual/syllabic verse, we are dealing with patterns of stressed and unstressed (rather than long and short syllables), and the usual notation is something like acute accents over stressed syllables with breves over unstressed ones, as exemplified in this page. Other explanations of English verse scansion use more convenient typography, such the symbols / and u as substitutes for the acute accent and breve respectively. As you've seen, we've used '#' and '.' -- changing the notation doesn't change the ideas, or shouldn't do so, anyhow.

Using whatever notation, the meters we've been examining (The Hunting of the Snark, The Shooting of Dan McGrew, Walk This Way) combine iambic and anapestic rhythms, with alternating lines of four and three feet. This ballad stanza is a common form in English folk poetry.

The Greeks (and their Roman students) identified types of poetic lines in terms of the type of rhythmic pattern (foot) used, and the number of repetitions of the pattern. Thus a pattern consisting of five iambs would be an iambic pentameter; a pattern consisting of six dactyls would be a dactylic hexameter; and so on.

In this way of talking, the ballad stanza alternates tetrameters (four-foot lines) with trimeters (three-foot lines). A limerick is typically two lines of anapestic trimeter, followed by two lines of anapestic dimeter, followed by a final anapestic trimeter:

    . #     .   .    #    .   .  #
    I used to think math was no fun
      .    .  #   .    .   #  .   .   #
    'Cause I couldn't see how it was done
     .  #  .     .  #
    Now Euler's my hero
     .  .  #   .    .  #
    For I now see why zero
    .  #  .   .  # .   .  #
    Is e to the pi i plus 1.


Iambic pentameter

When we look at the scansion of Skip to my lou or The shooting of Dan McGrew, we see that there is a very good correlation between strong positions in the meter ("beats") and main-stressed syllables of content words.

This correlation is not perfect. There are a few cases where stressed syllables of content words are in weak positions. Here is a line with four examples (in bold face):

    .  #    .     #   .  .   #     .    #   /   .    #   .    .  #    .      #
    A half-dead thing in a stark, dead world, clean mad for the muck called gold
     

The rhythm of this line remains clear, however, since in each case there are adjacent stressed words that are naturally more prominent, and so a completely ordinary reading of the line still gives a direct expression of the "swing" of the meter.

What never happens -- in such verse -- is for the main-stressed syllable of a polysyllabic word to be scanned in a weak position.

There are also examples of a "beat" position occupied by a word -- such as a function word -- whose natural degree of prominence is weak. However, in nearly all of these cases, there are even weaker words adjacent, so that the natural rhythm of the phrase still expresses the meter clearly:

      .  #   .   #  .  .  #    .   #
    That one of you is a hound of hell . . .

As a result, it is impossible to read the poem without intuitively grasping its meter -- whether you want to or not!

In most metered verse in English, the meter is not as obvious. Nevertheless, the basic rules of meter and scansion remain the same: there are a certain number of strong ("beat") syllables per line, with a specified number of intervening syllables permitted between the beats. The main-stressed syllable of a polysyllabic word is never allowed to occur in one of the intervening weak (non-beat) positions in the meter. Naturally strong monosyllables may occur in metrically weak positions.

Perhaps the biggest difference between Service's handling of meter and that of subtler English poets is the treatment of metrically strong positions. In poems where the meter is obvious or even insistent, like limericks or The shooting of Dan McGrew or The hunting of the Snark, metrically strong positions are usually occupied by strongly stressed syllables, with weaker syllables around them. In subtler metrical styles, this correlation is relaxed, so that weak monosyllables often appear in strong positions in the meter.

The result is a flexible sort of meter, able to vary between clear and obvious rhythmicity and more prose-like patterns.
 
Like much English-language art poetry from Shakespeare to Auden, Alexander Pope's An Essay on Criticism is written in iambic pentameter, that is, in lines of five repetitions of the iambic rhythm /. #/

Sometimes his lines express the meter as directly as any limerick:

    .   #     .   #    .    #     .    #  .   #
    In search of wit these lose their common sense

Each strong position in the previous line is occupied by the stressed syllable of a content word; each weak position is occupied by a prosodically-weak function word, or by the unstressed syllable of a polysyllabic word. In the line given below, both stressed syllables of imagination are used in metrically strong positions, but otherwise the situation is the same:

      .    #    .   #   . # . # .     #
    Where beams of warm imagination play

However, it is easy to find lines in the same poem where unstressed and prosodically weak syllables are put in metrically strong positions, like is and the in the couplet below:

    .  #   .  #   .   #  .  #    .     #
    A little learning is a dang'rous thing;
      .    #    .   #     .    #  .#  .     #
    Drink deep, or taste not the Piërian spring:

 As a result of this (and other aspects of poetic practice, such as frequent inversion of the first foot of a line), there may be a complex relationship between the abstract metrical pattern of iambic pentameter and its linguistic rhythms. This somewhat indirect relationship between the metrical pattern and its linguistic relationship is typical, although the details differ from meter to meter and even from poet to poet.

Both the ballad meter and the iambic pentameter (in English) can be seen as involving sequences of alternating weak and strong positions:

. . . w s . . .

The ballad stanza involves 4 + 3 + 4 + 3 of these /w s/ units, while couplets of iambic pentameter (as in Pope's poem) involve 5 + 5. The two different forms also typically exhibit different principles about the instantiation of these weak and strong metrical positions.

In the English ballad meter, the basic idea seems to be that "strong" positions in the meter should coincide with single syllables that are "peaks" of linguistic stress, in the sense that they are naturally more prominent than the syllables around them. The weak positions in this meter are relatively unconstrained, and in particular may correspond to different numbers of syllables with different stress properties, depending on the poet (or the poem). The result is verse in which the natural rhythm of linguistic performance strongly evokes the metrical form.

In English iambic pentameter, on the other hand, the basic constraints seem to be that both strong and weak positions in the meter should correspond to single syllables, and that "weak" positions in the meter should not coincide with stress peaks (that is, syllables that are naturally more prominent than those around them). The "strong" positions are relatively unconstrained. The result is verse in which the natural rhythm of linguistic performance, while metrically constrained, need not evoke the regular alternation of the metrical form very strongly.

[There is a great deal more to be said about all this: what counts as a syllable; where word boundaries are permitted, required or forbidden; the special treatment of positions at the edges of metrical units, and the reasons for this; and so on.]

As the comparison between ballad meter and iambic pentameter illustrates, different poetic styles may constrain the relationship between metrical patterns and the rhythms of language in quite different ways, even within the same language. However, the relationship is usually not a matter of completely arbitrary conventions -- say, "you can add extra syllables if they contain the letter 'x'". Rather, like children's language games and the relation of lyrics to music in songs, it is rooted in the sound structure of the language. Metrics is applied phonology.

 

[Note: if you have comments, please feel free to send them to me by email.]

Posted by Mark Liberman at July 6, 2004 11:42 PM