Analysis of Hit Songs

[AnthonyCeseri]

I've been writing a newsletter extensively analyzing hit songs and why they work. If you're a writer, or just interested in the topic, you'll want to check it out...

You can read past issues and sign up for new ones here:
http://www.successforyoursongs.com/past ... er-issues/

[christophersclaybo]

PLEASE DELETE, Taxi Admin!!!

[mojobone]

I've been writing a blog extensively analyzing hit songs and why they work. If you're a writer, or just interested in the topic, you'll be interested in this...

http://www.anthonyceseri.tumblr.com

Cool blog; some good analysis there.

[anne]

Deconstructing and analysing what you want to achieve is an excellent way of understanding the process. Nice Blog, Anthony! Have you read Robin Frederick's book as well? (Shortcuts to writing a hit song, on Taxi's resources page) -

[AnthonyCeseri]

Hey thanks guys!

@Anne - I haven't read Robin's book (yet). First I plan on reading the TV/Film one she wrote. I'm super interested in that topic : )

[cardell]

I'll check it out.

Stuart

[gongchime]

This is one of my favorite topics. I come at it from a slightly different angle. I researched what expert composers tell computer software to do. And what programmers observe expert composers doing that the expert composers were not conscious of doing and so didn't tell the programmers. The computer programs didn't used to fool anyone that the music was composed by a computer but now a jazz performer can be followed by a computer's reinterpretation of what the live player played in real time, and a blindfolded audience can't tell what is computer and what is real. The reason is that experts have ways of unconsciously shuffling elements around in an intuitive way that works. Transformations have to be musically relevant such as inversion, retrograde, augmentation, diminution etc...(Random permutation sucks) or you can only shuffle the things that you're "allowed" to shuffle (not structural pitches) in order to come up with something based on the original but different enough not to be mistaken for it and not so different as to be gobbledygook while injecting the appropriate amount of new material but not random material. New material has to be from the acceptable database of known probabilities/possibilities. Each style has it's own melodic vocabulary as well as harmonic and rhythmic. The programs/software that do this stuff are called Expert Systems.

It starts with pitches. Analysing music from the western tradition reveals that (if we take C to be the first note of the scale ) C is used more than E and G (interestingly E and G are actually statistically equivalent), C, E and G are used more than D and A (which are also statistically equivalent, while B and F are used the least (also statistically equal to each other in weight). But you can't stop there of course. Music isn't only made up of pitch elements. There are also rhythmic elements.

Rhythms of "popular" melody tend to hover around the eighth note. The farther away you move from eighth notes for a slow song or say quarter notes for a fast song, the less probable they are. It means 32nd notes are rare, though they do occur, and dotted half notes or whole notes, though they occur, are similarly more rare.

Also, important notes of the chord/scale are more likely to occur on the downbeat. The next most important rhythmic position is the third beat in the measure which has more metrical weight than the second or fourth beat. The order of importance of beats in a measure is 1,3,2,4. The offbeats get non chord tones primarily. Next up, is intervallic motion.

Motion by second is more prevalent than motion by third and motion by third is more common than motion by fourth. Motion by fourth is more common than motion by fifth. Motion by minor sixth is the last of the statistically relevant. Motion by major 6th or any kind of seventh is very, very rare.

If you made a computer program that took all of this into account you'd still come up with aimless wandering melodies. There's another element that needs to be considered and that is melodic direction. Melodies don't keep changing direction randomly. This last point ties everything together and makes statistically weighted choices of pitches, intervals and rhythms finally start to be relevant. Good melodies tend to favor both an upper and a lower curve.

For my music I made a deck of cards with the correct number of each pitch, interval, note values and melodic shapes. Then, I take the common bass progressions and choose one. For classical music, I map out what the inversion of the bass would look like in the treble clef with a solid line. Then I draw in guide tones. Next, I choose cards to fill in the quarter notes connecting all the guide tones. I don't agonize over ANY notes. Listen to my Diotima Concierto over in Peer to Peer to hear the effectiveness of this method. It only took me a morning to compose, an afternoon to sequence and the evening to hit record. I wasn't trying to be adventurous with chromaticism and dissonant extended harmony on that tune but I CAN do that if I choose.

I've also got a jazz tune called "Break Time" which reshuffles the notes of "Take Five" while placing it in 4/4. I only reshuffle the notes I'm allowed to reshuffle and voila, awesome tune. My tune Alegna is the melody from the hit TV show Taxi played retrograde. This stuff works maeng.

They've also got something like all this that applies to lyrics but it's not as effective yet because language is more complex since it has to make sense. haha But basically there's only a few rhyme schemes. They're on their way.

I don't feel I'm composing computer music though. They were only teaching computers what expert PEOPLE were doing so it's like I learned from the pros bypassing the computer.

I forgot, check out the awesome tabla programming I did on the tune "Fortuitous Events." On my cards I also have rhythmic patterns such as aaa, aab, aba, abb, abc, aaaa, aaab, aaba etc... To create the tabla part I used what I call meta patterns where the A in ABA would stand for one of the rhythms above that I would choose at random so lets say A actually equals aab and B equals aaba (actually ccdc) Then I would choose a rhythmic subfiguration for each lower case letter e.g. "a" might equal a dotted eighth followed by a sixteenth and "b" equal two eighth notes and "c" an eighth note followed by two sixteenths. Good stuff. It creates sensible rhythmic phrases that we've all heard pros play.

[Len911]

"The soul of a song defies analysis" I am quoting from a 1915 book from Brett Page.

I've taken a different approach also. I have been taking notes from books nearly a century old. I have found more germs of knowledge in just one or two books from that era than a whole stack of "contemporary" books. I know it sounds counter-intuitive, I thought so too, but they are actually more steeped in principle, clear and precise, and don't include a lot of filler material or over-analysis of Beatle tunes,lol!
A few of the gems I have discovered:
---Regularity and precision are absolutely fatal, because it loses expression.
---Rhythmic swing provides the unity, and is the secret to successful songs.
---Words and music must blend so precise and perfect, having a unity so compressed and so compactly lyrical, that to take out one note or lyric, would ruin the total effect.
---Contrast is the key to the successful hook.
---Always put love in your lyrics.
---Narrative may be proper in verses, but chorus should be flashes of emotion.
---Hooks must be the most attractive lines.
---Music and words are inseparable. The lyric must sing the music, and the music sing the words.
---The words designed to convey emotion, do not become lyrics until the emotion is shown.
I won't list all my notes,lol, but I would think if one analyzed songs according to the principles, it might remove a lot of the mystery surrounding why or why don't songs become "hits". It seems that maybe for contrast and balance, non-hits should be analyzed as well. For the sake of argument, there are other variables as well, such as marketing, never listened to by a producer, etc. etc.

[cardell]

Thanks Greg & Len, great info.

Stuart

[gongchime]

Please feature me and my music on your blog. I've been participating on more than 100 blogs trying to be featured. I initially just asked directly but then I found out that almost never works. Blog owners usually want to know you first and learn to trust you. Then if you say you've got something their readers will enjoy, they'll believe you. Maybe your readers will enjoy that post. Feel free to use plop the whole thing right in your blog if you'll site my web addresses.

Thanks,