What comes after one? Usually four.

July 29, 2011 10:09 AM

A corpus analysis of rock harmony [PDF] - The analyses were encoded using a recursive notation, similar to a context-free grammar, allowing repeating sections to be encoded succinctly. The aggregate data was then subjected to a variety of statistical analyses. We examined the frequency of different chords and chord transitions ... Other results concern the frequency of different root motions, patterns of co-occurrence between chords, and changes in harmonic practice across time. More information, analysis, and explanation here.
posted by Wolfdog (33 comments total)

This is considerably over my head, both in terms of analysis and music theory. Is there a layman's version?

This seems like it should be interesting, and I'd love to read an very dumbed down summary.
posted by Admiral Haddock at 10:15 AM on July 29, 2011 [1 favorite]

gabba gabba hey
posted by stupidsexyFlanders at 10:21 AM on July 29, 2011

"I'd love to read an very dumbed down summary."

Wolfdog edited it out of the FPP quote from the abstract. "The results showed that IV is the most common chord after I and is especially common preceding the tonic." Which, if you have spent any time flogging barre chords, is pretty obvious.
posted by Ardiril at 10:31 AM on July 29, 2011 [3 favorites]

"The results showed that IV is the most common chord after I and is especially common preceding the tonic."

Louie, Lou-why?
posted by joe lisboa at 10:34 AM on July 29, 2011

"The analyses were encoded using a recursive notation, similar to a context-free grammar, allowing repeating sections to be encoded succinctly."

This reminds me of when I used to reduce Soul Coughing songs to either all numbers or all place names.

Los Angeles, Los Angeles, Los Angeles, Reseda, Ohio...
posted by Eideteker at 10:35 AM on July 29, 2011

This is the reason I fell asleep consistantly in Music Theory I - IV. Plate of beans anyone?
posted by cmdnc0 at 10:38 AM on July 29, 2011

Rock 101: "I, IV, ..., IV, I"

or, Lynyrd Skynyrd, if you know the Freebird joke.
posted by Ardiril at 10:40 AM on July 29, 2011

This is considerably over my head, both in terms of analysis and music theory. Is there a layman's version?

This seems like it should be interesting, and I'd love to read an very dumbed down summary.

Try this:

Their primary conclusion is that the I > IV chord sequence is the most common in pop music (based on their data set).

I = "tonic": the chord based on the key the song is in.
IV = "subdominant"
V = "dominant"

If the song is in A, thats: A, D, and E respectively.

Most songs begin on the I chord, and almost alway end there (resolve) unless there's a 'false ending' or something else more interesting happens along the way.

The easiest way to hear this is to consider a blues-based song, like "Crossroads" or "Red House".

Hear it in your head, or pull out your records and listen along:

================================

A [I]
I went down to the crossroads, fell down on my knees
There's a red house over yonder, that's where my baby stay

D [IV]
I went down to the crossroads, fell down on my
There's a red house over yonder, that's where my baby

A [I]
knees
stay

E [V]
Begged the Lord above for mercy,
I ain't been home to see my baby

D [IV]
help me if you
in 99 and one half

A [I]
please
days
================================

So that chord sequence between the first two lines of each verse is what they're talking about. It's in a lot of songs. A lot of songs, and not just blues based songs. Seems obvious, but they've proven it statistically.

The obvious next question is: "What's the second most common chord sequence in pop music?"

lather rinse repeat. . .
 
posted by Herodios at 10:50 AM on July 29, 2011 [3 favorites]

I wouldn't call that "an very dumbed down summary", H. ;-P
posted by Ardiril at 10:53 AM on July 29, 2011

Looking at the methodology, what they've actually proved is that Rolling Stone magazine really likes songs with these harmonic features.
posted by hilker at 10:56 AM on July 29, 2011 [1 favorite]

This is in direct relationship to a question that is intriguing me for years: How do harmonic complexity of rock/pop and classical music compare? My intuition as a somewhat-of-a-musician tells me that harmonics of rock/pop are poor, seen from a classical point of view. But has there been done any "corpus analysis" of classics? From, say, Monteverdi to, at least, Schoenberg? It is no surprise that any connections of I-IV, V-I ans so on are most frequent. But what about the more complex progressions? Does anyone know of a comparative analysis?
posted by megob at 11:01 AM on July 29, 2011

How do harmonic complexity of rock/pop and classical music compare? . . . has there been done any "corpus analysis" of classics?


From the linked article:

. . . statistical corpus methods have been a part of music research for several decades at least (see, for example, Budge (1943)[1], the last decade has seen greatly increased activity in this area – no doubt due to the general rise of scientific, empirical approaches in music research and also to the new possibilities afforded by computer technology. Especially notable is the work of David Huron (2006)[2].

. . . With regard to patterns of harmonic progression, the strong asymmetries of root motion found in common-practice music are notably absent in rock.

[1] Budge, H. 1943. ‘A study of chord frequencies based on the music of representative composers of the eighteenth and nineteenth centuries’, PhD thesis (New York, Columbia University)

[2] Huron, D. 2006. Sweet Anticipation: Music and the Psychology of Expectation (Cambridge, MIT Press)

posted by Herodios at 11:15 AM on July 29, 2011

How many bass players does it take to...?

I. V. I. V. I. V.
posted by jfuller at 11:16 AM on July 29, 2011 [1 favorite]

i really wish they hadn't stopped at 2000, as it's my subjective hearing that the IIb chord has become much more frequent than it used to be, and not just in heavy metal music, but mainstream pop and r&b

My intuition as a somewhat-of-a-musician tells me that harmonics of rock/pop are poor, seen from a classical point of view.

which period of western art music are we talking about, though? - the tendency has been towards more complexity as time goes on, until in the 20th century, new principles of harmonic construction were explored

but i'd agree - rock music is less harmonically complex, but more complex as far as timbre and rhythm is concerned - it's been my experience as a composer of rock music that it's really hard to increase harmonic complexity without sacrificing rhythm and timbre, although western art music is another story

this was a very interesting paper - and i wonder what a comparision to jazz or "the great american songbook" would reveal
posted by pyramid termite at 11:27 AM on July 29, 2011 [1 favorite]

pyramid termite: In a nutshell (i.e. oversimplified), ii - V - I
posted by Greg_Ace at 11:40 AM on July 29, 2011

Thanks Herodios! That was very helpful. I've just started guitar lessons again (though I've played for nearly 20 years) and am working on some theory. Your explanation was just what the doctor ordered.
posted by Admiral Haddock at 11:46 AM on July 29, 2011

Chord progressions 101.
posted by swift at 12:02 PM on July 29, 2011 [1 favorite]

More thanks to Herodios. I knew the expectation / anticipation theory from L. Meyer, but didn't not about it's newer developments. Never heard of Budge as well ... it seems that he did more or less what I was looking for, although in a probably less formal way than the paper from the FPP.
Your quotation seems to support the observation that in rock/pop there is nearly no harmonic modulation at all. No wonder that a song is hardly longer than 3-5 minutes, as all the larger architectural structures based on harmonic base shifts are absent. Difficult to hold the tension for al longer tim then.
Pyramide termite: - yes, we should talk about a wide choice of historic periods. But it seems to me that the main principles of the harmonic system were well developed by the time of Bach, and remained valid at least till Mahler (with extensions and elaborations). It would be interesting to have a statistical analysis of these 200 (roughly) years.
Interesting that your experience in composing leeds you to the idea, if I understand well, of a trade-off between harmonic complexity and complexity of rhythm and timbre. I hope for the day in the evolution of music where they will be integrated again in more complex works of music. They would forcefully be longer, and hopefully intellectually and emotionally more moving than most of what we have today.
posted by megob at 12:52 PM on July 29, 2011

If you had the whole list of trigram probabilities — and a similar set of probabilities for melodic moves, I guess — then you could write Markov-chain rock songs.

Well, not "songs," really, since most real songs repeat the same set of changes over and over, and a randomly generated set of changes wouldn't repeat. More like Markov-chain senselessly meandering prog rock epics.

Come to think of it, maybe it's a good thing that they didn't give all the trigram probabilities.
posted by nebulawindphone at 1:30 PM on July 29, 2011 [4 favorites]

| I ii | IV | V | ... | is about as British Invasion as you can get. We really should spell it "The KiiNKS".
posted by Ardiril at 1:41 PM on July 29, 2011

"then you could write Markov-chain rock songs" - Yes
posted by Ardiril at 1:42 PM on July 29, 2011 [2 favorites]

or ELP even
posted by thelonius at 2:23 PM on July 29, 2011 [1 favorite]

"then you could write Markov-chain rock songs

Are you trying to summon Cortex or something?
posted by TwoWordReview at 3:56 PM on July 29, 2011 [3 favorites]

One song, Public Enemy’s ‘Bring the Noise’, was judged by both of us not to contain any triadic harmony, and was therefore not included in any further analyses.

Heh. That was the only one in that list of 500 with no harmony?
posted by spitefulcrow at 7:12 PM on July 29, 2011

Also, yeah, this:

Looking at the methodology, what they've actually proved is that Rolling Stone magazine really likes songs with these harmonic features.

I want to see this done at the very least for a corpus of progressive rock.
posted by spitefulcrow at 7:25 PM on July 29, 2011

"Unwearied then were Markov's folk/Beneath the vocals, keyboards woke"
posted by thelonius at 9:20 PM on July 29, 2011

Wow ! Thanks a lot.
posted by nicolin at 2:58 AM on July 30, 2011

If your're interested in chord progressions, take a look a Chordspace (vsti) and Tonespace (standalone), a couple of free programs for exploration. (Get SAVIHOST to run Chordspace as a standalone.)

There is also a novel live performance program called Chordspace Playa for £35.00 worth looking at as well.
posted by Enron Hubbard at 7:17 AM on July 30, 2011 [1 favorite]

"Unwearied then were Markov's folk/Beneath the vocals, keyboards woke"

When I read this it was immediately obvious that this is sung to the tune of "Hallelujah". (The Leonard Cohen song, not the chorus from Messiah.)

It seems like the actual source wouldn't fit that melody, though. Oh well.
posted by madcaptenor at 7:59 AM on July 30, 2011

If you had the whole list of trigram probabilities — and a similar set of probabilities for melodic moves, I guess — then you could write Markov-chain rock songs.

You could indeed. Really, this is to some extent covered ground; you look at stuff like Band in a Box that's been around for years and years and, on the stochastically-generated chordal sequence front you're already there. See also, infamously, Microsoft's Songsmith, based as far as I could discern on BiaB via license even.

There's been a lot of academic work on these general ideas, though I've mostly only skimmed over that stuff occasionally because a lot of it's over my head and a lot more I haven't really had access to in any case.

I actually took a (very, very simplistic) stab at doing some Markov-based digestion and reiteration of Mozart in college. I didn't spend enough time with it to produce anything particularly interesting—you need to do a lot of prep work to assemble a properly digestable corpus of music, one of the things that makes this paper interesting to me—but even at a very low level it's immediately obvious that there's potential for that sort of thing to be interesting.

Well, not "songs," really, since most real songs repeat the same set of changes over and over, and a randomly generated set of changes wouldn't repeat. More like Markov-chain senselessly meandering prog rock epics.

Assuming you wanted to produce something other than meandering prog, using the Markov model to create substructures (verse, chorus, bridge, vamp, etc) and then some other process to manipulate those generated substructures into a complete three or four minute song would be a pretty good method. Emphasize creativity in the substructures, emphasize familiar tropes of repetition and transfiguration in the overall structure to make it poppy and listenable. A weird figure played once just sounds weird; a weird figure repeated in an anticipatable fashion is a hook.
posted by cortex at 8:38 AM on July 30, 2011 [1 favorite]

A weird figure played once just sounds weird; a weird figure repeated in an anticipatable fashion is a hook.

Every improvising musician knows, "if you make a mistake, play it twice".
posted by Herodios at 8:42 AM on July 30, 2011 [4 favorites]

"Unwearied then were Markov's folk/Beneath the vocals, keyboards woke"

Tenacious D/Dio?
posted by ersatz at 1:09 AM on July 31, 2011 [1 favorite]

I think that this is putting the pattern to the test:

YouTube: "The Axis of Awesome 4 Chords"
posted by kmartino at 4:01 AM on July 31, 2011