The Anthology of Mondegreens?

« previous post |

Paul Devlin has a fascinating series of articles at Slate on transcription errors in the recently-published Anthology of Rap.  Well, the first one starts out as a review of the book, but after the first paragraph or so, it's all about the Mondegreens: "Fact-Check the Rhyme (The Anthology of Rap is rife with transcription errors. Why is it so hard to get rap lyrics right?)", 11/4/2010; "It Was Written (Why are there so many errors in The Anthology of Rap? The editors respond)", 11/10/2010; "Stakes Is High (Members of the Anthology of Rap's advisory board speak out about the book's errors. Plus: Grandmaster Caz lists the mistakes in his lyrics.)", 10/19/2010.

Some of the cited errors are more consequential than others:

1. At the 1:58 mark, the anthology transcription reads "against the very best." Caz told me it should be "we rock the very best."

2. At 2:03, the anthology has "And you'll be so impressed." Caz said it should be "And baby I want your address."

As Devlin admits,

Transcription of rap lyrics is excruciatingly difficult, due to speed of delivery, slang, purposeful mispronunciation, and the problem of the beat sometimes momentarily drowning out or obscuring the lyrics.

We can add the problem of background noise in live recordings, and  the difficulty of decoding local or personal allusions. But there seem to be a lot of errors that are just careless, and Devlin points out several places where unlikely mis-hearings seem to have been copied from on-line collections of lyrics, contradicting the anthology editors' claims of a painstaking seven-step transcription methodology.   One particularly egregious case:

Here is an example in which it seems like nobody bothered to even listen to the song, but instead must have relied on some incorrect transcript. On the 1979 song "Superrappin'," Melle Mel of the Furious Five does not say "1-2-3-4-5-6-7/ rap like hell make it sound like heaven/ 7-6-5-4-3-2-1 …" He says "one, 23, 45, 67/ rap it like hell make it sound like heaven/ seven, 65, 43, 21 …"

But ironically, Devlin himself is rather careless here, at least by the standards of scholarship that he (appropriately) would like to see The Rap Anthology uphold. In the first place, he slightly misquotes TRA's transcription.  Here's an image of the relevant part of p. 72, courtesy of

In his account of TRA's transcription, Devlin omits "You say" from the 1-2-3-4-5-6-7 line, and the "and" from the "Rap like hell (and) make it sound like heaven" line. He also leaves out an initial "fifty" in the second next-to-last line, which I transcribe (see and hear below) as "fifty-seven sixty-five forty-three twenty-one".

As for what Melle Mel really sang in the "hell and heaven line, it was this:

which (after listening carefully several times with headphones) I transcribe as

rappin' like hell to make it sound like heaven

There are a couple of plausible differences of opinion here. There's definitely a sung syllable between "rap" and "like", and another between "hell" and "heaven". TRA omits the first; Devlin omits the second. I hear the first one as the participial ending "-in" , while Devlin hears it as "it"; I hear the second one as "to", while TRA hears it as "and".

Devlin is absolutely right to excoriate TRA for its transcription of the number-sequence lines. Here's the first one:

I transcribe this as

((I)) say one twenty three forty five sixty seven

TRA's version is so completely wrong that Devlin wonders if they even bothered to listen, rather than just copying some website's version of the lyrics:

You say 1-2-3-4-5-6-7

I wonder whether this might have been botched by a copy editor, who saw something like "1 23 45 67″ and figured it had been badly typed by the transcriber. Whatever the reason, it's a bad mistake. (On the other hand, Devlin didn't listen very carefully either, since he leaves out the initial "fifty" two lines later.)

Devlin notes that the editors promise to correct some of their errors in future editions.

While we're on the subject of possible improvements in future editions of The Rap Anthology, I'd like to note that a critical aspect of these lyrics is systematically missing from the kind of transcripts that this anthology provides, no matter how textually accurate they might become. I'm talking about the relationship of the lyrics to the metrical background.  This is implicit in the original recordings, of course — but so are the words themselves.

As an example, here's the whole passage whose TRA transcription is shown above:

Here's my transcription:

uh comin' up and I got ta step above the rest
((a guy that uses)) that ladder they call success
((I)) say one twenty-three forty-five sixty-seven
rappin' like hell to make it sound like heaven
fifty-seven sixty-five forty-three twenty-one
come on Rahiem uh come and get some

(There are lots of things I'm not sure about. For example, is the syllable before "comin' up" a version of "I" or "I'm", or (as I've transcribed it) just "uh"? This is a good reason, as Devlin suggests, to ask the authors for corrections while it's possible to do so.)

The basic rhythmic analysis of this piece is fairly easy, since its minimal time unit (of a bit more than 1/8 of a second) is pretty clear throughout, either in the rapping or in the musical background. (That doesn't guarantee that I made no mistakes in the analysis below — but it should make it relatively easy to settle any disagreements.)

These minimal units are grouped four to the beat — i.e. every fourth unit is nominally stronger — and there are typically four beats per line, both in the metrical background and in the lyrics. Generally, the stressed syllables in the lyrics align with the strong beats in the background meter, except in the last position in lines with masculine rhymes, where the last stressed syllable generally anticipates the musical beat by one minimal unit. However, sometimes (as in the last line of these three couplets) there is more extensive "syncopation". This line has only three strong syllables in the lyric,  and the last of them is only one that aligns with a strong beat in the metrical background!

(As always in such cases, there's an interesting question whether this should be thought of as a superficial deviation from an underlyingly square rhythm, or rather as a different draw from a set of available polyrhythmic patterns. For some more discussion, see e.g. "Rock syncopation: Stress shifts or polyrhythms?", 11/26/2007. Note in any case that the mixture of four-beat and three-beat (lyric) lines evokes the traditional English ballad meter, whatever we're to make of the variations in alignment.)

In the metrical transcription below, I've used lower-case letters for off-beat minimal time elements, and upper-case letters for the time-points corresponding to the "beat", with X and x for time-points aligned with syllables in the lyric, and O and o for time-points without a corresponding syllable. The audio clip is repeated here for convenience:

This notation is somewhat clunky, and it would be easy to do better, say with some system of diacritics (for indicating the beat alignment of syllables) and interpolated symbols (for indicating beats without aligned syllables). But however we notate it, it seems to me that anyone who cares about rap or hiphop as poetry ought to be interested in its meter. The fact that European and American art poetry has largely been largely unmetered for the past century shouldn't prejudice the analysis of the genuinely sung lyrics of the same period, where the musical background and its relationship to the words is isomorphic to traditional notions of metrical structure in accentual-syllabic verse during the previous five centuries or so.

The metrics of hiphop and rap are particularly rich and interesting, in my opinion, and many scholarly books and papers will no doubt be written on this subject eventually.  In retrospect, TRA's lack of any substantive discussion of rap meter will (I predict) seem even stranger than the scholarly carelessness of its transcriptions.

[There shouldn't be any connection, really, but the transcriptional carelessness of both TRA and Devlin reminds me of the scandalously bad quality of transcription in direct quotes from public figures in main-stream journalism. Whenever I've had occasion to check a published quote against an available recording, I've been shocked by how bad the transcription is. (A few examples among many are here, here, here, …)]


  1. grackle said,

    December 4, 2010 @ 7:11 pm · Edit

    I don't hear the fifty starting the fifth line at all. Are you sure you're not imagining it?

    [(myl) Yes.]

  2. majolo said,

    December 4, 2010 @ 7:39 pm · Edit

    I'm with Grackle, no hint of a fifty. It also breaks the symmetry, for what that's worth:

    [(myl) But LISTEN.]

  3. Twitter Trackbacks for Language Log » The Anthology of Mondegreens? [] on said,

    December 4, 2010 @ 7:39 pm · Edit

    […] Language Log » The Anthology of Mondegreens? – view page – cached December 4, 2010 @ 7:01 pm · Filed by Mark Liberman under Language and music Tweets about this link […]

  4. X said,

    December 4, 2010 @ 7:40 pm · Edit

    I hear "Rappin' like hell, I think it sounds like heaven."

    If I listen to it on a loop until my brain melts, I hear "Rap it like Hell; my finger sounds like Kevin.", which I imagine is from the hip-hop musical version of The Shining.

  5. Rose Fox said,

    December 4, 2010 @ 7:49 pm · Edit

    And I don't hear the "fifty" either.

  6. dfan said,

    December 4, 2010 @ 7:50 pm · Edit

    I also don't hear any voice at all (not even something else mumbled) during the purported "fifty".

    By the way, I assume 'and another between "hell" and "heaven"' should be 'and another between "hell" and "make"'.

  7. Lance said,

    December 4, 2010 @ 7:51 pm · Edit

    I'm with Mark. Symmetry be damned, I'm hearing something before the "seven", and I'd agree that it's "fifty".

  8. D.O. said,

    December 4, 2010 @ 7:56 pm · Edit

    FWIW, I cannot hear "fifty" either.

    [(myl) Again, please LISTEN.]

  9. Amy Stoller said,

    December 4, 2010 @ 8:02 pm · Edit

    The "fifty" is definitely there.

    Really good transcription is an art as much as a science, but there is no excuse for not factoring prosody in when transcribing lyrics.

    One pet peeve of mine is transcribing (or even just writing) numbers as numerals instead of spelling them out, if the work in question is intended to be sung or said aloud, rather than merely read to oneself. How, pray tell, is one meant to pronounce 101, absent more information?

    I'm also puzzled by an anthology for which the editors don't bother to get the authorized version of the lyrics from those who crafted them in the first place.

  10. Cameron said,

    December 4, 2010 @ 8:03 pm · Edit

    I do not here "fifty." There is a sound there, but I think it's just an intake of breath. It's definitely not a vocalized word.

  11. majolo said,

    December 4, 2010 @ 8:08 pm · Edit

    I've LISTENED about 26 times more, and I hear nothing other than a breath intake between heaven and seven.

    [(myl) Please either use a program that lets you zero in on the relevant section, or listen to this excerpt ("… sound like heaven / ((fifty)) seven sixty five …") And use headphones.]

  12. Tora said,

    December 4, 2010 @ 8:12 pm · Edit

    What I hear where his transcription has "fifty" is not "fifty" or any other word, but a fast intake of breath.

  13. Scott Underwood said,

    December 4, 2010 @ 8:12 pm · Edit

    It might be a sharp intake of breath before "seven."

  14. Chris said,

    December 4, 2010 @ 8:13 pm · Edit

    I'm also unable to hear any "fifty" (or any other vocal) between the "heaven" and "seven".

    [(myl) OK, enough with the "fifty". Either you're not listening carefully to the critical stretch, or I'm imagining things. But we know that the lyrics in music like this are hard to transcribe, so let's accept that this is an uncertain case, and move on.]

  15. Peter said,

    December 4, 2010 @ 8:41 pm · Edit

    Listening carefully, with headphones, including to a slowed-down version, I can make myself hear a fifty it I try to, but it comes across more plausibly to me as a sharp intake of breath, with a click of some sort (which I’d place as a percussion instrument of some kind) about halfway through — so the click becomes the t, while the sibilance of the breath on either side becomes the f’s. I don’t claim 57-ers are imagining it entirely, but it’s not that us 7-ers aren’t listening carefully: as you say, it’s one of those spots where transcription is genuinely difficult.

    Are there not accepted ways (looking at the spectrograms, or something) to investigate questions like this less subjectively? One could imagine contexts (transcription of evidence in court, for example) where this sort of thing could become very important! And it would certainly nice to have an option somewhere between “I hear this”/“I hear that”/“I hear this, with knobs on!” and just giving up on it as ineffable.

    [(myl) Even relatively clear recordings, without background music and so on, are full of cases where different people hear different things — depending on material, transcription practices, training, time spent in transcription, etc., we can expect maybe 3-5% of the words to differ between two independent transcriptions. (If the transcribers are journalists, we can expect 40-60% of the words to differ ;-)) Recordings of poorer quality may have much higher inter-transcriber disagreement rates, and often pose issues that are eternally unresolved, like Neil Armstrong's "One small step for ((a)) man" (see here and here, etc.) — but even a better recording might leave similar issues in controversy.

    I think that commenters in this case are focusing on the wrong thing. There are some transcriptions in The Rap Anthology that are clearly, intersubjectively, flat wrong. These should have been fixed. There are others that are uncertain. These should either have been checked with the authors, or indicated as uncertain with the conventional transcription signal of (( … )), or footnoted, or all of the above. We can all hope that future editions will do better.

    Meanwhile, I at least also hope that future editions will consider the problem of meter. If nothing else, a careful attempt to transcribe the meter-to-text alignment will fix the obvious transcriptional mistakes. But more important, it makes no sense to try to document rap as poetry without trying to understand its meter.]

  16. James said,

    December 4, 2010 @ 8:50 pm · Edit

    Weird. The 'fifty' is very distinct to me, not at all subtle. Maybe others aren't listening in the right place? The line break isn't marked in any obvious way in the rap itself.

  17. Rubrick said,

    December 4, 2010 @ 8:54 pm · Edit

    All I can hear is "Dah dah, bah, bah, gah gah… I buried McGurk…"

  18. J. Goard said,

    December 4, 2010 @ 9:08 pm · Edit

    Wow, a truly flagship LL post disguised as another mondegreen list!

    Here in Korea, where everybody sings in noraebang 'singing-rooms', I've become intimately familiar with rap mondegreens. May have picked up a few of them myself :-(, but at least I know that Eminem said

    and these times are so hard
    and it's getting even harder
    trying to feed and water my seed
    plus teeter-totter
    caught up between being a father and a primadonna

    rather than the incongruously self-deprecating

    plus see dishonor

    What's interesting is that whoever provided the noraebang lyrics got many difficult sections right, but garbled some fairly easy ones.

  19. Brian said,

    December 4, 2010 @ 9:23 pm · Edit

    Wow. I've listened to the thing a dozen times now, and it boggles my mind that anyone can hear a "fifty-". I do hear the intake of breath, but I can't make myself hear it as a two syllables. It goes by so fast that there's barely enough room for an inhale. Were it not for the other posters who claimed to hear it clearly, I would have thought it was an elaborate joke.

    [(myl) Did you listen to the whole thing, or to the crucial passage here (which I transcribe as "…sound like heaven / fifty seven sixty five …)? As for length, the period of time between the end of "heaven" and the start of "seven" is about 270 msec. long, which is just about exactly the same as the duration of "heaven" and the duration of "fifty", and just a bit more than we expect for two minimal time-units in this section of the rap, where I count 28 minimal time-units in 3.644 seconds between the start of "one" and the start of "heaven", for an average of 130 milliseconds. But anyhow, as too often happens, commenters have zeroed in with laser-like intensity on a footnote to a footnote. Let's drop the whole "fifty-or-not" business, OK?]

  20. Mfahie said,

    December 5, 2010 @ 12:06 am · Edit

    I wonder if it didn't occur to you that music notation would be the best tool, or if you discarded it as your notation for some reason. I could re-notate this tomorrow using standard music notation if you're interested, although I'm not 100% sure that I could post it.

    [(myl) You could represent each line as four bars of 4/4 time, but I don't think it would add anything. On the contrary, it would obscure the relationship to the meter of traditional English accentual-syllabic verse, and it would be harder to grasp for people who haven't learned to read musical notation. It would be easy enough to do from a technical point of view, using LilyPond or Finale or whatever, and the result could straightforwardly be posted as an image. But as I said, I think that in this case it would be a step backwards.

    For more complex meters, such as those discussed here, here, and here, the use of traditional musical notation can actually be a hindrance to understanding even for those who can read music well.]

  21. matt said,

    December 5, 2010 @ 12:09 am · Edit

    Maybe I missed it in one of the comments, but did anyone else notice that the book transcribes the second sequence of numbers not as 7-6-5-4-3-2-1 but as the even more wrong 7-6-4-5-3-2-1?

  22. kenny said,

    December 5, 2010 @ 12:21 am · Edit

    enough about the "fifty."

    I hear (and I feel that I've listened very carefully) "rappin like hell if it could sound like heaven."

  23. tpr said,

    December 5, 2010 @ 6:53 am · Edit

    Interesting the difference in what people hear. For what it's worth, this is what I heard:

    comin' up and I gotta step above the rest
    'cause uh, 'cause that ladder thing called success
    y'say one twenty-three forty-five sixty-seven
    rappin' like hell, I think it sounds like heaven
    seven sixty-five forty-three twenty-one
    come on, Rahiem, a-come 'n get some

RSS feed for comments on this post · TrackBack URI

Leave a Comment

Logged in as Mark Liberman. Logout »