The rhythm interacts with melody and accent structure during the word “railroad.” The higher note sounds like it is accented by virtue of its position in the rhythm, falling on a strong beat (the “one” in a four-beat measure of music) while the lower note falls on a weaker beat (the “three”). In this way the rhythmic structure implies a word whose accent is on the first syllable, like “ráil-road,” as opposed to say, “gui-tár,” whose accent is on the second syllable.
Clichés, or less pejoratively, common word combinations, also help us to remember lyrics. Expressions such as “I’ll love you until the end of time” or “letting the cat out of the bag” are so common that if we hear (or recall) only a few of the words, the rest of them follow. For example, you might remember only the first four words of this lyrical phrase: “We used to fight like cats and ——.” Even a child could fill in the missing lyric, because this phrase occurs so often in ordinary speech. In fact, this exact phrase “fight like cats and dogs” is found in dozens of pop lyrics including songs by Dolly Parton (“Fight and Scratch”), Paul McCartney (“Ballroom Dancing”), Harry Chapin (“Stranger with the Melodies”), Tom Waits, Nanci Griffith, and Indigo Girls (“Please Call Me, Baby”), and Phil Vassar (“Joe and Rosalita”). Another common idiom, the phrase “spill the beans,” shows up in songs by six artists that couldn’t be more diverse: Yes (“Hold On”), Lard (“Pineapple Face”), DJ Jazzy Jeff & the Fresh Prince (“I’m All That”), Boomtown Rats (“Tonight”), Carly Simon (“We You Dearest Friends”), and Smash Mouth (“Padrino”). In all these cases, songwriters are taking advantage of the brain’s ability to retrieve sequences of pre-stored information when cued with only a small piece.
Poetic features of lyrics, such as assonance or alliteration, also help to bring up words from memory; if we remember that the song has these features, we again don’t necessarily have to store every word. Great feats of memory are all about parsimony, cognitive economy—using rules that will generate dozens or hundreds of correct answers on the spot rather than memorizing and having to recollect each item.
Some songwriters flout these customary principles, and this itself can become a memory aid. When Paul McCartney sings “Hey Jude/Don’t make it bad/Take a sad song . . . ,” each word falls right on a melody note in perfect time, just as you would expect. But on the final line of that first verse, he makes a “mistake,” one that sounds odd, singing: “. . . and make it bet-ter-er-er,” stretching the second syllable of the word “better” out over four notes. On first listening, it is jarring. But we remember it for its distinctiveness. Even if you forget the word “better” (or fail to encode it, although neither possibility is likely, because the very distinctiveness of this compositional move virtually guarantees it will become solidly encoded in memory), you can re-create the word just by remembering that there was something funny going on there, a two-syllable word stretched out to four syllables. Given the semantic constraints of the text before, there just aren’t that many words that can fit in that final slot. (Paul uses the same technique later in the song, of course, stretching out the word “be-gi-in” to three syllables.)
The individual effects of rhyme, rhythm, accent structure, melody, cliché, and poetic device can be subtle. The idea of mutually reinforcing constraints is that the effects are additive—no single effect is always enough to help us generate a missing word, but together they can turn an incomplete memory into a near perfect performance. The interaction of these different cues allows the lyrics of ballads and other knowledge songs to remain relatively stable over centuries.
In Wallace and Rubin’s studies of expert ballad singers, they found that even when the singers made word errors, they seldom misremembered the structure of the song—the rules describing its invariant features. The end-rhyme sound, the number of beats per line, and the number of lines per verse remain relatively constant within a particular ballad, and the singers studied rarely made errors about these. Ballads in general are stabilized by the common characteristics they share—well-known, stylistic norms within a given tradition whether it is Yugoslavian, Golan, Indonesian, North Carolinian, or Ancient Greek. These common characteristics are so embedded in the idiom that when singers of a given tradition are asked to write a new ballad, encoding a new event, they tend to employ all of the same tools and incorporate the same structural characteristics of that form.
In one song Wallace and Rubin studied extensively, “The Wreck of the Old 97,” the form of the ballad was clearly serving as a memory aid. “Words and music,” they write, “are intertwined. The words have a metrical pattern, which must correspond to the rhythmical pattern, the beat structure, and the time signature of the music. . . . In this ballad, the meter consists of two unstressed syllables followed by a stressed syllable. Usually the unstressed syllables are shorter in length than the stressed syllable. . . . The meter and rhythm are closely related in that the number of stressed syllables equals the number of beats in the music. . . . Thus, the music and words constrain each other.”
If these multiple and reinforcing structural constraints are such an important ingredient in remembering song lyrics, it follows that a song with fewer of them would be recalled with more errors. Wallace and Rubin created just that situation for an ingenious experiment—they changed twenty-four words in “The Wreck” to eliminate assonance, alliteration, and rhyme. Notice that they specifically made changes that affected poetic characteristics of the words, the weakest of the structural constraints I listed above. (No end rhymes were altered, and the number of syllables in each word and the stress patterns were maintained.) The unaltered and the new version of the song were then taught to people who had never heard it before. Analyzing the words that had been changed, Wallace and Rubin found that singers were more than twice as likely to recall a word verbatim if it possessed these poetic elements (the original version) than if it did not (the altered version). This shows just how constraining—or helpful—the relatively weak factor of poetics is.
As of this writing, a popular television show in the United States features hapless contestants misremembering song lyrics, which on the surface may seem to undermine my arguments. First, what makes the show funny is that millions of viewers remember the lyrics that the contestants forgot and can’t understand how they could have forgotten them. Second, understand that network television is entertainment, and that the producers undoubtedly go out of their way to pre-select people with a bad lyrical memory; they may represent only a minority of the population, but they are the ones that make for the most entertaining spectacle. Third, recall that the average person today is exposed to tens of thousands of songs. The average oral historian may have only been responsible for memorizing and retelling thirty or forty in the days before radio. Finally, our ability to recognize that we’ve heard songs and to repeat portions of them should not be confused with the potential ability we all may have to truly commit songs to memory when we focus our attention and energy on doing so; asking people to sing back songs that they may have only ever heard in the background at a shopping mall or on a car radio is not the most rigorous test of their abilities (and not a scientific experiment!).
Now here comes the amazing part of Wallace and Rubin’s study (and why they are my heroes). It’s based on the principle of constructive memory—that we don’t actually remember all the details we think we do, we fill in many of them subconsciously by making plausible inferences. Wallace and Rubin looked at the errors made by those singers who were asked to learn the altered version of the song (and heard only that version). Many of them spontaneously, and without prior knowledge, recovered the poetic words that had been removed by the experimenters. In other words, the stylistic norms of assonance and alliteration—the tendency for them to be prominent in these sorts of epics and ballads—was something that the subjects in their experiment had encoded in their long-term memory. When they tried to remember the song—involving a process of reconstructing the words—they “remembered” the words that origi
nally belonged in the song. For example, if the original song contained the alliterative phrase “real rough road,” and the altered version was rewritten and learned as “real tough road,” some singers (wrongly?) sang it back as “real rough road.”
In family and tribal histories, ballads commemorating important battles, and so on, minor errors in wording are probably so insignificant as to be not worth mentioning. There are at least two cases, however, where errors are completely unacceptable. One is those knowledge songs that were created to preserve precise, practical information about how to do something, such as make a watertight raft, prepare food or medicine that might otherwise be poisonous, or accomplish some other action where both the order of the steps and the details of the procedure are critical. A second case is those knowledge songs that are intended to convey religious information, where every single word must be preserved verbatim because it is considered sacred and divinely given—there are strong cultural pressures to recall such material accurately or not at all. In cases like these, humans exhibit a remarkable ability for virtually perfect recall. How could this be?
When people accurately recall texts of great length, they are typically texts set to music—songs. It is far rarer for such prodigious memory to be demonstrated with straight, musicless text, and the reason for that appears to be the theory of multiple, reinforcing constraints. Insight into the matter comes from another very clever experiment by Rubin.
Rubin asked fifty people to recall the words of the Preamble to the United States Constitution (try it yourself now before reading any further; the words are in an endnote at the back of the book). This of course does not have music, but that’s the point—the kinds of errors that people made here were profoundly different than those made by people trying to recollect songs. One group of people remembered the first three words (“We the people”) and then . . . stopped. You hardly ever see this in song recollection. When they forget the words, people singing a song typically continue on humming or singing la-la-la or something else, along with the melody that continues, uninterrupted, to play in their heads, and they pick up words here and there along the way, sometimes recalling entire stretches of the song at a later point down the line. Another group of Rubin’s subjects recalled the first seven words (“We the people of the United States”) and then quit. Nearly every one of his fifty subjects stopped cold when they failed to recall the next word in the sequence, whatever that word was.
The places in the text where people got stuck were far from arbitrary: 94 percent tended to get stuck at phrase boundaries, natural breaks where one would pause to take a breath. Rubin replicated these findings with the Gettysburg Address. Songs, because of their rhythmic momentum, are far more easy to “keep on going” in our heads without the words; this gives our brains a chance to jump in whenever a word or word fragment becomes available again, making lyric recall in songs typically better than lyric recall without them. This is one reason why the average person probably has a more intimate and emotional connection to music than to poetry—because he or she can recall more of it, and more effortlessly.
It’s important to look in more detail at how rhythm plays its role in aiding recollection of lyrics. When we recall a song, rhythm provides an internally consistent hierarchy of temporal units—syllables form poetic feet, which in turn form lines, which form phrases, verses, or stanzas—and these rhythmic units usually coincide with the units of meaning in oral traditions. These rhythmic units with their accent structure create positions in the music where words must go—they don’t permit us to omit part of a rhythmic unit, thus conspiring to preserve the integrity of lines and verses. (If any omission is to occur, it would generally be at the level of a large-scale repeating unit such as a verse, and then different mnemonic processes come into play to maintain its integrity.)
Recall from Chapter 1 that a mark of most effective poetry is that it has its own rhythm, a music about it that we hear when we recite it out loud. If poetic rhythm, combined with the sorts of other poetic devices we looked at in Chapter 1, really helps us to remember poetic text, you might expect that people would have better memory for poems than straight text, and moreover, that their errors would be more similar to those we find in music recollection. Both of these turn out to be true. David Rubin asked undergraduates to recite from memory the 23rd Psalm, which contains many poetic elements but (in the English version) lacks rhyme. Here, most undergraduates started up again at a later point in the psalm after they stopped (and it was usually at the beginning of a new section that they came in). It seems as though the internal rhythms of the psalm kept on playing in the students’ heads, and they jumped in whenever they could recall a word. (Don’t forget that the Psalms, as originally composed, were set to music.)
Additional evidence for the notion that people “play back” such recollections in their heads comes from my own experiments with Princeton computer scientist Perry Cook. We found that when college undergraduates sang their favorite songs from memory, they tended to sing them at almost exactly the right tempo. If they forgot words, they kept going and jumped in later, again at what would have been the right place if they’d kept on going, as though the band kept playing and they simply came in at the next appropriate moment. And from their own subjective reports, all of them had vivid mental imagery of the music. They weren’t so much trying to reproduce the song from memory as singing along with a track in their heads.
Getting back to the memorization of text, when very long word sequences are involved, we typically resort to two tried-and-true mnemonic techniques: rote memorization and chunking. Rote memorization is simply reciting a sequence back over and over again (often in the quiet privacy of our own minds) until we’ve got it. This is how most of us learned our multiplication tables in grammar school, the Pledge of Allegiance, the Gettysburg Address, or the Preamble to the Constitution. The interesting thing here, though, is that not all words are created equal in rote memorization. Some words take on more importance because of the expressive emphasis we are taught to give them, some because they evoke particularly pleasant or vivid imagery, and some because they contain certain (preplanned, on the part of their writers) poetic qualities. Internal rhymes, assonance, and alliteration of a Cole Porterish quality help to reinforce the Gettysburg Address for example (maybe because assonance makes the heart grow fonder): Four score and seven years ago
Our fathers brought forth
We see the repetition of the f sound here acting as a mnemonic, as well as the repetition of the long o sound in four, score, ago, forth.
When the text we’re trying to memorize is more than a dozen or so words long, we tend very naturally and without coaching to break it up into bite-sized, more readily memorized, sanitized and organized units, or chunks, and then stitch these chunks together later. This is also how musicians memorize pieces they perform. With the exception of special neurological cases of people with photographic memory (or the auditory equivalent, what I call pho nographic memory), most musicians, dancers, actors, and other performance artists, young and old, do not sit down and learn a new piece from front to back all at once. They concentrate on getting a small part of it just right and then they learn another small part. They then spend some time learning the transitions from one part to another. The evidence for this process remains long after the piece has been committed to memory and its performance is flawless: Actors who have to redo a take often ask to go back to the beginning of the line, paragraph, or scene. Musicians return to the beginning of the phrase. Bring out the score, point to an arbitrary note where you’d like them to start in a memorized piece, and most musicians will ask to start somewhere else, at the beginning of some chunk that they learned. When musicians make errors in prepared pieces, these errors provide additional clues to the way the piece was originally learned. It is more common for a musician to skip an entire section (a failure to remember how the chunks were stitched together) than for a musician to skip a note or short group of notes within a secti
on. Between-section errors are far more common than within-section errors. Not all notes or words are equally salient.
The Gettysburg Address is substantially easier to memorize than multiplication tables, which typically require pure rote memorization. Personally, I remember having great difficulty with these in grammar school. I made a little card with them written down, and while I walked to school or had a spare minute, I would glance at the card and test my memory. With no rhythm or melody to attach the numbers to, pure brute force of repetition was required: two times two is four; two times three is six; two times four is eight. I got up to the sixes without too much difficulty, but well into high school I couldn’t remember the “twelves” unless I started from a part of the tables that had personal meaning for me, the one that represented my height in inches at the time I was originally trying to memorize them: Twelve times four is forty-eight. If I wanted to recall twelve times eight, I’d have to start my “poem” at this salient point and work my way up: twelve times four is forty-eight; twelve times five is sixty; twelve times six is seventy-two; twelve times seven is eighty-four; twelve times eight is ninety-six. Because of incessant teasing from my neighbor in third grade, Billy Latham (an excellent drummer at the time, by the way), who had learned his tables all the way up to twelve times twelve, I had also memorized the one that Billy taunted and drilled me with most frequently: twelve times nine.
The World in Six Songs: How the Musical Brain Created Human Nature Page 16