Jonah Lehrer talks about his op-ed in the LA Times regarding the limits of reductionism.

I commented on his blog, but I thought I'd post about it here as well.

Early on he says:

The success of modern neuroscience represents the triumph of a method: reductionism. The premise of reductionism is that the best way to solve a complex problem -- and the brain is the most complicated object in the known universe -- is to study its most basic parts. The mind, in other words, is just a particular trick of matter, reducible to the callous laws of physics.

First of all, reductionism isn't just a triumph in neuroscience, it's been wildly successful as the primary scientific approach to understanding the universe. But he mischaracterizes the concept.

The approach is not to study the most basic parts of a complicated object. The idea is to explain how things work at a given level of complexity in terms of the interactions going on at an appropriate level of complexity just below the current one.

For example, if car mechanics showed up for a class on engine repair, it would be more than a bit silly to try to explain the operation of the engine in terms of the interaction of atoms. The appropriate level of description is in terms of parts (like pistons and spark plugs) whose interactions describe phenomena at the next level. Richard Dawkins called this "hierarchical reductionism."

Lehrer makes the mistake of confusing extreme reductionism with hierarchical reductionism, here in the case of music:

But the reductionist method, although undeniably successful, has very real limitations. Not everything benefits from being broken down into tiny pieces. Look, for example, at a Beethoven symphony. If the music is reduced to wavelengths of vibrating air -- the simple sum of its physics -- we actually understand less about the music. The intangible beauty, the visceral emotion, the entire reason we listen in the first place -- all is lost when the sound is reduced into its most elemental details. In other words, reductionism can leave out a lot of reality.

The problem is, this is a straw man. Yes, if you focus on only one, very low level of description, you're going to be leaving out a lot of important interactions and dynamics, but that just isn't the way things are being carried out.

Lehrer's op-ed sounds like an admonishment to researchers who study the brain and mind. He makes it sound like they're drilling down to the level of synapses and neurotransmitters and missing the forest for the trees. But I'm in an interdisciplinary department where people look at issues of cognition from different angles, at various levels of description. I would agree that we tend to have a low of very low-level, and high-level data. The high-level data comes from the multitude of psychological and linguistic research. If anything, I would argue that we need more theories at intermediate levels of description, to bridge the gap.