The Limits of Reductionism

I had an op-ed in the LA Times on Sunday. It's about reductionism and the brain:

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 mind is like music. While neuroscience accurately describes our brain in terms of its material facts -- we are nothing but a loom of electricity and enzymes -- this isn't how we experience the world. Our consciousness, at least when felt from the inside, feels like more than the sum of its cells. The truth of the matter is that we feel like the ghost, not like the machine.

I think it's important to be clear about what argument I'm not making. I'm not rehashing Keats' criticism of Newton: I'm all in favor of unweaving the rainbow. I think reductionism can be startlingly beautiful and will always be our primary method of understanding everything. But I think it's important to note that reductionism is not our only method. There are some questions, and these questions happen to include the grandest questions of neuroscience, that can't be answered in such strict and narrow terms.

Tear my argument apart in the comments.

More like this

Sine-Wave speech is a wonderful example of the importance of patterns when it comes to our sense of sound. When people first hear a sentence that's been artificially degraded, the sentence sounds like a sequence of "simultaneous whistles, or science fiction sounds." However, when people are first…
The latest Seed has a very interesting article on the complicated geometry underlying Western music, and the intuitive mathematical understanding demonstrated by composers: The shapes of the space of chords we have described also reveal deep connections between a wide range of musical genres. It…
Richard Dawkin's Unweaving the Rainbow: Science Delusion and the Appetite for Wonder is on my active reading docket. The book has been around for a while (published in 1998), but it's proving to be a most enjoyable discovery as I continue to read it. So far, I concur with complete reviews' take…
A new study of brain responses to music has found a striking difference in brain activity when a symphonic movement ends and the next one begins, compared to other parts of the musical work. A team led by Vinod Menon (and including This Is Your Brain on Music author Daniel Levitin) played excerpts…

>Tear my argument apart in the comments.

I imagine that must get old :D

I have nothing to tear apart, though once again Stephen Wolfram comes to mind. As far as practical (pragmatic) solutions for the anti-reductionist, complexity theory is a nice start. Wolfram reduces things, but his reduced "rules" are not meant to be interpreted literally - their utility lies in what they produce/mimic.

I would suggest that the utility of your arguments gets better as you provide concrete examples of method - introducing a view of the third/fourth culture is nice, but better if it has implications for how science operates. So, for the neuro-scientific method, we have the measure, block and mimic method. I'll take a stab at an anti-reductionist method... contextualize, parallelize, perturb. Place a phenomenon into its global context, draw parallels to systems with similar abstractions, and interact with the system through perturbation of its whole (not its individual reduced components).

I think it's time to see some reduction of limitism.

By Tegumai Bopsul… (not verified) on 21 Jan 2008 #permalink

There are some questions, and these questions happen to include the grandest questions of neuroscience, that can't be answered in such strict and narrow terms.

For example? What are these supposed questions that don't lend themselves to a reductionist approach?

Every time I see one of these arguments that reductionism fails when it comes to certain areas of knowledge, is that the author must be using some definition of reductionism that I am not aware of. When we seek to explain any phenomenon, we do it by describing how it works in terms of other other phenomenon; usually ones that are either simpler, or more fundamental, or better understood. I don't think that there is any other way.

Just to look at your example about music and sound waves: The problem here is that you're trying to jump down too many levels at once. Musicologists certainly do reduce the experience of a piece of music into more basic components such as melody and harmony and rhythm, and cultural references to other music pieces, etc. Each of these in turn could be reduced to other more basic things. Eventually you do get down to sound waves and other physical phenomenon, but you don't try to do it in one step (unless you're doing it solely for rhetorical purposes).

As far as explaining beauty and aesthetics in reductionist terms, you are undoubtedly familiar with the work of V.S. Ramachandran in this area, no?

Beautiful notion as thought of through the metaphor of music. Though the ingredients hardly stop at wavelengths. They interact with timbre, with space, with aspects of performance, with the listener's prior associations, with their context as music, and most importantly, with time. I suppose I'm no longer reducing them, so I'm fundamentally agreeing with you. I'll try to read your full op-ed piece later today. I felt your writing on cooking in your book offered a strong metaphor for our experience of our thoughts - how fat, onion and a little glutamate end up as delicious sauce. In fact it's even the same word - reduction.

Oddly, I'm musing on music and the brain myself today at my place as I react Oliver Sacks' latest.

I agree with Kurt here. Most arguments against reductionism seem to turn out to be against a straw man. What Kurt is talking about is basically what Dawkins called "hierarchical reductionism" in The Blind Watchmaker, which is the idea that phenomena at one level are described in terms of interactions at an appropriate level just below it. So you don't try to explain how a car works in terms of subatomic particles, but in terms of pistons and spark plugs.

I'd agree in general that neuroscience is providing a constant stream of very low-level data that currently doesn't seem very useful because there are not many comprehensive, viable theories of cognition at slightly higher levels of description in which to analyze that data.

I'm currently pursuing a PhD in Cognitive Science, and the philosophical approach of CogSci is to tackle issues of the mind with evidence from multiple subdisciplines working at various levels of description. I think that's the right approach, though your article makes it sound like most researchers studying brain and mind are neuroscientists drilling down to the level of neurons and synapses, unable to see the forest for the trees, and that seems like an awfully unfair and restricted view of the field.

Musicians do reduce music to sound waves on a regular basis. You figure out how you want a phrase to work, experiment with the details of sound production that will produce the effect you want. There is an input (the physical motions which produce an acoustic effect) and the output (the desired turn of the phrase). Listen to a Baroque violinist play a Bach partita and compare it to running the score through a midi synthesizer. Music benefits from reduction to vibrating sound. The human brain is not affected by signals dancing on an oscilliscope the same way it is affected by hearing the same sounds aloud, but the fact that conductors can translate written music directly into the music in their head indicates means that it is not something fundamental about the physical sound. A human can be trained to get the same effect via another medium, and if you play Schoenberg for those unused to it, you will quickly find that sound may not be a universal means of transmission as well.

There is probably something objective (is there a human on Earth who can listen to Debussy's 'La fille aux cheveux lins' without pleasure?) but it's a question of our physical apparatus, not acoustics.

There is an error often conflated with reductionism. Knowledge of the behavior of all component parts at some level of reduction doesn't imply any knowledge of the whole. There isn't enough computing power in the universe to calculate nontrivial, macroscopic phenomenon from first principles from the standard model of particle physics. This is a serious error to make, but it is not equivalent to reductionism.

I would be very interested to see an example where we actually understood less about how a phenomenon proceeds after reduction, but I have grade doubts of its existence.

Your point about feeling like the ghost, not the machine is unfortunately not reliable for two reasons. First, it may be purely cultural. In many Asian cultures, the major symptoms of psychiatric illness are physical, not mental. High level martial artists regard their body as very much themselves. Gurdjieff described man as a "three brained being" --- which then is the ghost? The other problem is that even if it isn't cultural, we may have evolved to regard ourselves as ghost in the machine. Let us say it's not cultural for humans, but we encounter an alien race for whom the Cartesian separation of mind and physical being is completely incomprehensible. What then? Interestingly, it is this dichotomy that the phenomenologists (in philosophy) attempted to bypass, with fair success (see Sokolowski's 'Introduction to Phenomenology').

I think the problem arises from what reductionism can and cannot add to an explanation. Reductionism takes a given problem and looks at it at different levels (typically, finer and finer levels of analysis, but the same process can go either way; to keep with the metaphor of music, a fourier analysis can be used both to break down a complex sound into component wavelengths, and to synthesize a complex sound from component wavelengths). But it is always answering the "what" and "how" questions, and not the "why", and our curiosity is often pointed to the "why".

If you try to answer "why does this chord sequence move me?" with a reductionistic appeal to connections from the cochlea to the temporal lobe, with associations to frontal, to amygdala, and all that amazing incredible wonderful stuff... you are not answering "why", but rather "by what mechanism", or "how". It is a great question to answer, but it is not "why", and so ultimately is unsatisfying, if "why" is what we asked.

I suspect that the "why" answers, or perhaps the satisfying ones (or perhaps just the ones that I myself find satisfying), are found at the same level of analysis as the question is originally asked. For me, that means that a question about human experience, say, must be answered in terms of the person in his or her environment (broadly defined, including private, public, social, and verbal stimuli as well as the "environment" that is more narrowly assumed).

If I could grab another example, look at evolution. The reductionist explanations look at DNA, RNA, protein synthesis, folding, mutation, how a particular bit of this molecule comes to express a physical or behavioral characteristic... these are all the "how" of reproduction, mutation, selection. The "why"s are all in the environment. What food sources are there? What predators? What weather? And these "why" bits are all at the level of the organism (and population), where the observable change that we were interested in is happening. And in the experience of the individual human... we cannot answer "why am I depressed?" with an appeal to dopamine, or serotonin, or prefrontal cortex, or any of those--those are answers to "how am I depressed?" (They are quite useful answers--there may well be many different physiological reductions of depression, with different mechanisms, and each may point to a different intervention.) To see why one is depressed (or happy, or in love, or hungry, or hearing Beethoven), one must look at the environment the individual is in. Causes come from without; answers from within will be inadequate... unless all you are interested in is "how".

Now... let us look to your environment to see why you keep looking for reductionistic explanations in the first place, when they are so unsatisfying. There is a clue in your OpEd. Your education was influenced greatly by the "cognitive revolution"; I note that you quote Chomsky, when the person who spoke most eloquently on the pitfalls of reductionism is Skinner. My guess (based on your age, your topic, your degree--and I would be quite happy to be wrong) is that your understanding of Radical Behaviorism is one that comes mostly from non-behaviorists. You might be surprised at how much RB agrees with your view, and how much it could contribute to it.

We are increasingly rejecting Cartesian dualism (at least, the scientific community is), but I fear we are replacing it with a material quasi-dualism that treats brain and body as separate, rather than looking at a whole organism. We are the products of our past, and our scientific exploration is too much influenced by a prescientific vocabulary of "mind" and "consciousness". It is as if we take it as a given that "what appears, is", and it is our job to explain that... and so we look to explain why the sun climbs in the sky over an immobile earth. Sunrises are illusory. That does not mean they do not exist, but rather that the true explanation is not what it appears. The same might easily be said for "mind" and "consciousness", but even top neuroscientists still use phrases like "gives rise to consciousness", as if presupposing the qualitatively different nature of part of our experience.

The good news, of course, is that sunrises, sunsets, rainbows and eclipses are still stunningly beautiful, even after we know more about the "how" behind them. For myself, I think they are even more beautiful when we have a reductionistic explanation to go along with a non-reduced experience. I am confident that an understanding of what we call "consciousness" or "mind" will not sully the experience of Beethoven, or fresh strawberries, or love. Even if, as I strongly suspect, we find "consciousness" to be illusory.

So what if consciousness, if mind, is not what it appears to be? It is what it is, and understanding it does not kill it (this is only a metaphorical dissection, after all). Removing illusion from understanding is like removing tarnish from silver; what remains gleams more brightly. If it was all tarnish... if it was all illusion... then we have lost nothing of value.

I agree with the "strawman" commenters. Your own example of Beethoven is a case in point. In general, we can't take our analysis straight down to acoustics and physics in one step, but no one said we could. We start by going down one level to his abuse of the sonata form - grossly elongated and even structured development section, brief returns to the exposition to trick you into thinking it's over, etc. Then we can go down to the level of harmony - how he modulates to the mediant like no other. And rhythm - the entire Fifth Symphony revolves around a pattern that's more accurately called a rhythmic motif than any kind of melodic theme.

Then, if you really want to get down to acoustics, you can think about dissonant chords (complicated frequency ratios) and how the listener's mind wants them to resolve to consonant intervals (simple integer ratios). That underlies the entire concept of harmony, where the goal is to move from the dissonant to the consonant in the most roundabout way possible, and one aspect of Beethoven's genius was to apply the same idea to musical form.

All of this is reductionism. What is a holistic analysis of Beethoven?

These are all insightful criticisms. Thanks. I'd take issue with the "strawman" comments. I was discussing reductionism in the context of a scientific explanation of consciousness. Currently, the main neuroscientific attempt revolves the neural correlates approach (NCC) which expressly tries to translate first-person subjective experience into a discrete collection of necessary neural networks. Christof Koch defines the NCC as the minimal set of neuronal events that gives rise to a specific aspect of a conscious percept. In this case, I'd argue that scientists are trying to take the "music" of consciousness directly down into a very low level component.

My main point, though, is that such a reductionist approach is incoherent. It tries to explain our self-conscious experience at the wrong level of analysis.

But thanks for all your comments. Much food for thought.

Some readers here might also enjoy Oliver Sacks' review in the New York Review of Books, from January 2004. Unfortunately, to read it on line, you'll need a subscription; but many libraries have on line access, so perhaps one could read it that way digitally without incurring any expense or the bother, possibly, of searching it out in the flesh? :-)) It's entitled:

In the River of Consciousness, found in:

Volume 51, Number 1 · January 15, 2004

There's a follow up exchange with him & ... :

Volume 51, Number 6 · April 8, 2004

'In the River of Consciousness': An Exchange
By Benjamin Libet, David Ingle, Steve J. Heims, Reply by Oliver Sacks

In response to In the River of Consciousness (January 15, 2004)

The Principles of Psychology
by William James

Dover, two volumes, $16.95 each (paper)
Creative Evolution
by Henri Bergson

Dover, 432 pp., $14.95 (paper)
The Organization of Behavior: A Neuropsychological Theory
by Donald Hebb

Erlbaum, 368 pp., $45.00
Neural Darwinism: The Theory of Neuronal Group Selection
by Gerald M. Edelman

Basic Books, 371 pp. (out of print)
Wider Than the Sky: The Phenomenal Gift of Consciousness
by Gerald M. Edelman

Basic Books, 371 pp. (out of print)
The Physiology of Truth: Neuroscience and Human Knowledge
by Jean-Pierre Changeux

Harvard University Press, 336 pp., $45.00 (to be published in April 2004)
The Astonishing Hypothesis: The Scientific Search for the Soul
by Francis Crick

Scribner, 336 pp., $15.00 (paper)
The Quest for Consciousness: A Neurobiological Approach
by Christof Koch, foreword by Francis Crick.

Roberts & Company Publishers, 448 pp., $45
A Natural History of Vision
by Nicholas J. Wade

MIT Press, 486 pp., $85.00; $37.00 (paper)

The respective URLs are:

By Joseph Urban (not verified) on 21 Jan 2008 #permalink

I wish your comment in the blog that "I'm all in favor of unweaving the rainbow" was part in your op-ed piece. I came away from reading that thinking you were really missing the point of science in general.

My big worry is not for you, but my fellow newspaper readers who come away with the impression that all scientists are so blind they think sound waves alone will explain what makes music beautiful.

Your point (not in the op-ed) that "reductionism is not our only method" is a really important point, and completely missing from what the newspaper reader will expect of you and your book.

Hey, was that the plan? Make people think you hate science because it is so silly, then they buy the book to read more "crazy scientist" stories and they end up learning a thing or two? That would be clever.


Hi Jonah.

I feel I must immediately own up to being merely an interested, lower-than-lay reader of all things scientific, especially neuroscientific. So I hesitate to offer an opinion in such specialised company. But seeing as you seem so willing to consider consciousness in more than simply reductionist [scientific?] terms, I was wondering whether you'd ever read anything by Ken Wilber? He also feels that our phenomenological, subjective awareness cannot be reduced to its physical, objective correlates, and has developed an 'integral' psychological theory which contains a multi-perspectival view of consciousness. It aint neuroscience though. Damn, I would be so interested to know what you thought of his stuff, but I'm not sure it's really appropriate for your site.

Anyway, thanks for your great blog. A recent post and all the comments that followed - You Are an Illusion - was fantastically interesting for someone like myself who is approaching this stuff from a purely existential[ly traumatised :-)] angle. Simon

I realize the thrust of this discussion is about reductionism with an example taken from music - but, at the risk of hijacking the topic, the musically-oriented comments have reminded me that, though I've read (very reductionist) articles about neural processing of sound from the ear through the auditory cortex, and other articles about how brain imaging shows activity in the pleasure centers when people are experiencing their "favorite" music, I still don't know the "in between" - i.e. what are the mental mechanisms by which those captured auditory images are transformed into pleasure. I'd love to find some blog (or article) that addresses this issue. Psychologists have theories about our response to music at various levels ... but the ones that are most fleshed out empirically are the ones most objectively testable - which are at the level of perception of elementary rhythms, phrase contours, discrimination of different timbres in a musical stream. An ungodly amount of time is spent on perfect pitch. What I'd like to see is more emphasis on musical attention, expectation, Western art music aesthetic values, such as the feeling of unity within the context of an individual piece, the perception of novelty within the context of a musical genre - in other words, the squishy stuff. Books have been written on these topics - good books - but it would be a good test of reductionism to see if our response to music could be dissected at this level using neuroscience, instead of at the perceptual level as is most often encountered.

I hear quite some critic about reductionism that it wouldn't be sufficient for really complex systems like the brain. I don't agree, but it may be because of the definition of "reductionism".
I'll point my thoughts out here.
I think the most important fact about reductionism is, that for complex systems (like the brain) one has not to forget the hierarchical and fractal nature of the system when reducing it to simple "rules".
considering the example of the beethoven-score: what the ears will receive is in fact the low-level information of frequencies and phases over time, were in the brain they will be categorized into a collection of patterns which were received the most. a purly reductionist approach, but one has not to forget to apply that same approach into these extracted patterns over and over again, to receive higher order (that means more complex, but also more abstract) patterns. in my understanding, the number of these "iterations of abstraction" and the number of categories define the "intelligence" of that system, since more complex and distant events can be correlated. further more there is a recursive component: the current state of the system is also part of the pattern. so depending on the system's own state, the response to the same external stimulus might be different for 2 different internal states (for example i may respond differently on the same situation depending if i'm in a "good mood" or in a "bad mood"). on the other hand the external stimuli may alter the internal state of the system. this gives a "cylic" or "recursive" dependency, which may even yield "conscious" systems. to the external observer the system responds only in a pseudo-coherent way, because part of the information it uses to produce a response is (or may be) hidden to the external world, since the response may depend purly on the current internal state.
it still is all reductionism, but in a hierarchical and recursive way which makes it really powerful and sofar i couldn't find any good example of proving me wrong. i like to hear, if you got one ;)

As I read this old post, it interests me that you had to finish with that last line. This is what I find so annoying about talking to reductionists. They seem unable to give any credibility to the subjective thoughts and feelings. I suspect they do this because they find them somehow wholly untrustworthy and unenlightening or undermining of the scientific method. In any event, it can be quite irritating and tells me a lot that you cannot even express such views in the current intellectual climate without some trepidation.

Like Kurt and Derek have pointed out, this argument ends up being little more than rhetoric because its version of reductionism is incomplete and therefore is just a straw man. A full reductionist description of music includes many layers of explanation. "Intangible beauty, the visceral emotion, the entire reason we listen in the first place" are responses of the human brain to patterns in pressure waves detected by the auditory system. A reductionist description does not miss these. It recognizes the many layers of the human person from nuclei to atoms to synapses to cells to connections across the brain to conscious perception to social interpretation, while recognizing that it is not yet fully successful in connecting all these layers.

The problem with the reductionist explanation is simply that we can't explain consciousness and cognition on the basis of lower level function. In one sense it is a simple unsolved scientific problem, but in another sense it is the central problem because the human mind is our means of creating our reductionist explanations. All the rhetoric about 'limits of reductionism' misses the point. Our universe is incredibly complicated and so the full reductionist description is incredibly complicated. That is not a limit on reductionism. When you find an emergent property that violates some conservation law obeyed by the component pieces, then we can start talking about limits to reductionism. Until then, you're just selling rhetoric.