November 10, 2006

"When I tell him I think he's the end, he giggles a lot with his friend"

Not so tall, not so tan, but still lovely—it's "The Girl From Ipanema," as processed through a simulated neural map and projected onto an unfolded torus representing the gamut of Western tonal centers. I think. You should proabably ask Petr Janata, whose lab did this research. (Here's a partial explanation.) I know it has to do with trying to localize the sensation of tonal harmony in the prefrontal cortex, but Science will only let me look at the abstract without paying. Nevertheless, Dr. Janata has posted a movie (very big file, but worth it) of the whole song that, well, moves so cool and sways so gently. Perfect for a Friday.

And now that the song is stuck in your head, here's 54 covers of it to tide you over.

The title comes from Mary Rodgers and Stephen Sondheim's lethal parody "The Boy From..." (which has been hilariously translated into American Sign Language—fingerspelling the name of the town every time must bring down the house).


Scott Spiegelberg said...

I got in a big argument with Petr about this mapping at a confence last year. Much of his research is fine, but the actual torus is not based on good cognitive theory.

Matthew said...

It seems a little too good to be true. One of the things I was imagining was whether you could do a similar map for a 12-tone piece, using row forms and transpositions instead of tonal centers. It would probably come out similarly persuasive, and similarly disconnected from the way at least I experience listening. As an analytical tool for visualizing tonal space I think it's interesting—similar to the sort of thing Dmitri Tymoczko has been working on.

Petr said...
This comment has been removed by the author.
Petr said...

@Scott - I didn't think it was a big argument; rather a spirited discussion. But, I am puzzled why you state the torus isn't based on good cognitive theory. The torus is a parsimonious mapping of the pitch statistics (and associated percepts) of the major and minor keys in Western tonal music, as first shown by Krumhansl in the 1980s, and corroborated by probe tone experiments from many labs. So, if one converges on the tonal representation on a torus from actual behavioral data, how is the torus not a model that is reasonably isomorphic with cognitive representation?