Framing the Debate

Hey, Neuroblog readers! As someone on the outside of the connectomics debate looking in, I thought I’d more or less figured out what everyone was talking about, and what the major arguments were. But as it turns out, a recent experience with someone who’s closer to the discussion taught me not to take my assumptions for granted, and got me thinking about why the debate might look the way it does now, and whether it ought to look differently.

Framing the Debate

A funny thing happened to me at a recent NeuWrite-West workshop. Until then, I’d had a set of preconceived notions about what the whole Connectomics debate was about—I had assumed, for example, that the question about connectomics was about relative utility. Simply put, I’d figured that the real issue was one of resource allocation: should we spend money on connectomics that could be spent elsewhere? Is the utility we’d get out of a targeted investment worth what we’d put in, and are there situations that would give us more or less return than others? Do we know what the kind of data we’d want would look like, and how we’d handle it, and what sort of technologies we’d want to acquire it? I assumed that everyone more or less agreed that knowing which cells are connected to which cells would be useful, and the disagreement was about where to apply it and how useful it would be, or at least if it’d be useful enough to fund a scale-up. I also assumed that even the most dyed-in-the-wool pro-connectome person didn’t believe that connections were the only component you needed to understand how the brain works, and that the “who” in “who is connected to who” involved a functional definition that included neural activity and computational properties. Anyone who claimed that pure connections were all you needed was either taking an exaggerated point of view for the purpose of an argument, or was Sebastian Seung (kidding!). And anyone who claimed that connection information on a small scale would be entirely useless was similarly taking an extreme position for the sake of argument.

And then I got to the Neuwrite workshop, and I found out I was wrong. Or, at least, I discovered that not everyone was having the same debate I thought they were.

In short, that day’s presenter (who I’ll leave anonymous), had written a piece in defense of connectomics, or at least in defense of a targeted connectomics approach. This person may end up publishing their work later, so I won’t steal any thunder, but suffice it to say that, upon a first reading, the thesis felt profoundly modest to me: that understanding precise neural connectivity at a microscale could be useful and informative on a computational level, particularly in certain cases. From what I could tell, the piece was designed to fend off more extreme arguments that such knowledge would be useless for understanding the nervous system, the old “we’ve known the C. elegans connectome for years and where has it gotten us” train of thought, combined with a dash of “if we know in general, on average how things are connected, there’s nothing that more precise information could add to our understanding.”

To me, the piece seemed to be refuting incredibly exaggerated arguments. At one point during the workshop, I even started to push the presenter, telling him he needed a bolder thesis—weren’t these trivial arguments? Of course this info would be useful! Why not go further and say that knowing connections is necessary to a full understanding of the brain, or that there are fundamental things we’d miss without that information? Wasn’t it a matter of justifying that it was worth a substantial investment?

The answer I got, from this presenter who’s been a bit closer to the trenches of this particular fight, went something like this: “Wouldn’t it be nice if that was the debate we were having? I wish that was the debate—I sincerely hope that these are a vocal minority, and that somewhere out there are sensible people having a modest argument about resource allocation. But that’s not what I’m hearing.” Insofar as I can tell, the presenter genuinely believed that a connectomics approach would be necessary to answer many questions about how the nervous system works, but his experiences with his colleagues had convinced him that he had to be very careful about the way his arguments were put forward. What I had thought was an attempt to defend against one straw man argument (“It’s ridiculous to believe that connections are the only important thing”) by building up, and then tearing down, a straw man on the opposite side (“It’s ridiculous to believe that information about connections is entirely useless”), turned out to be an impassioned defense against very real opinions, with the aim of softly asserting the usefulness of a connectomics approach and letting people draw their own conclusions. I would never begrudge those who feel the need to defend themselves against such arguments, but to me, the most important thing we can do is to make sure we’re having the right conversation to begin with.

Where might this other, more extreme debate come from? One possibility is that the tight funding situation has everyone on edge—in a sense, funding is a zero-sum game where every dollar that goes to someone else is a dollar that isn’t going to your own work. This means that when money goes towards developing techniques to precisely probe neural connections, a lack of confidence in the premise can morph into outrage. As Leslie Voshall once tweeted, $300 million for the BRAIN initiative, if it comes out of an unchanged NIH budget, is enough to deprive 750 investigators of an R01-level grant that could keep their lab afloat for one year. After all, some might say, why are we wasting our time trying to gather information when we’re not exactly sure of what the endpoint will look like, or how it’ll be used? There are experimentally tractable questions happening right now, in need of funding and not getting nearly the same level of attention!

This isn’t a new problem: funding agencies have been trying to balance incremental and transformative research for decades, and framing it in terms of useful vs useless doesn’t add much to the discussion. Better to approach this as what we are: scientists! What do we actually know about what techniques are within reach, what are the hypotheses, and what are the risks? Rather than talking about things in generalities, are there specific cases where precise, micro-scale connectivity would provide the largest benefit with the least investment? According to Cori Bargmann, who is in part running the NIH working group on this very topic, no one is (or should be) proposing that we de-fund the laboratory work of the most strident Connectomics advocates—Jeff Lichtman, Sebastian Seung, Winifred Denk. The question is whether the techniques being developed in these labs should be optimized and mass-produced and adapted to larger and larger volumes of brain tissue. What we’re seeing in the debate, unfortunately, is a side effect of the tense climate in which science gets done in this country.

There’s another, more pervasive brand of myopia that factors in as well: everyone has their preferred scale at which they like to think about the brain. Some people like genes and molecules. Others prefer single cells, or circuits, and yet others prefer working with large populations of cells, or computational abstractions. From a general perspective, this is ideal: someone should be studying the brain at all of those levels, and this specialization allows for depth of research without too much duplication of effort. What makes this a problem is when some people can’t see the value of a scale that’s too many steps away from the one they like to use. These are the curmudgeons I’ve seen at seminars throughout academic science: computational neuroscientists squirming their way through molecular talks, or molecular biologists who tune out the speaker as soon as they start talking about dimensionality reduction. It’s easy to imagine that your way of thinking about the brain is the only productive one, and disregard the rest. But it’s just as disingenuous, and intellectually dishonest, to think that connectomics will be either an utterly unproductive waste of time with no redeeming merits, as it is to think that connectomics will be a magic bullet that can solve all the mysteries of the brain.

I’m not going to take a stand on one side of the connectomics debate today. But I will come down on what the debate itself ought to be: given that knowing precise patterns of connectivity will be useful, and will definitely inform our understanding of how the brain works, we don’t know how useful they would be, and whether the enhancement of our understanding is worth a substantial investment of time and money. I think that’s the debate that’s productive, and the one that’s worth having. Maybe this seems like a trivial point to you, dear readers, as it did to me, but it’s our duty as the quiet, reasonable majority to make sure it gets heard.