Pure Pedantry

i-d2d3ed4a35fb751f979d126feafc12bd-scale_of_justice.jpgNYTimes Science section, why do you make me so mad?

Gretchen Reynolds published an article in the Times on cognitive improvements associated with exercise, and I would like to use it to make a point about how science journalism often gets the facts right but the interpretation wrong.

It begins with the following incorrect statement:

The Morris water maze is the rodent equivalent of an I.Q. test: mice are placed in a tank filled with water dyed an opaque color. Beneath a small area of the surface is a platform, which the mice can’t see. Despite what you’ve heard about rodents and sinking ships, mice hate water; those that blunder upon the platform climb onto it immediately. Scientists have long agreed that a mouse’s spatial memory can be inferred by how quickly the animal finds its way in subsequent dunkings. A “smart” mouse remembers the platform and swims right to it. (Emphasis mine.)

While the Morris water maze is indeed a test of spatial memory in mice and rats, spatial memory is not equivalent to intelligence. Tests like the Morris water maze test particular classes of memory. They are not a general measure of how “smart” the mouse is.

Let me explain. If I were to lesion the hippocampus of a mouse or rat (a brain structure involved in spatial memory among other things), it would perform poorly in the Morris Water Maze (MWM). However, if I were to raise the platforms out of the water, so that they were no longer hidden the mouse would perform just fine again. By changing the nature of the task, we have changed what brain regions are required for its successful performance.

What this and other results with the MWM suggest is that particular types of tasks require particular structures. In no sense do these results suggest a general capacity to perform all tasks is encoded in one brain region, nor do they purport to measure general intelligence. Smartness in behavioral tasks is operationalized: you can only make a statement of how smart animal is in reference to their performance on a particular defined task.

In this sense, Reynolds has right from the get-go gotten the facts right and the interpretation wrong. The fact — mice do better in the MWM with exercise — is accurate. The interpretation — the rats are smarter — utterly misses the point.

After beginning on that bright note, Reynolds goes on to accuratively summarize a great deal of research that links exercise with spatial learning and neurogenesis in the hippocampus with exercise. However, in accurately summarizing the research, she exaggerates many conclusions that require qualification.

With respect to the facts, she is on target. For example, exercise improves rats performance on the MWM. (I couldn’t find the papers for mice, but the results are the same.) This is true for both adolescents and aged rats. Also, exercise increases hippocampal neurogenesis.

She goes on to point out that exercise increases blood flow to the hippocampus in humans, and that the hippocampus has been implicated in a lot of diseases that we care about such as Alzheimer’s. Many researchers have used experiments with changes in blood flow to the hippocampus in humans — mostly from fMRI experiments — to infer that the findings in rodents will carry over into humans.

“It’s reasonable to infer, though we’re not yet certain, that neurogenesis was happening in the people’s hippocampi,” says Scott A. Small, an associate professor of neurology at Columbia and the senior author of the study, “and that working out was driving the neurogenesis.”

All of this is well within the bounds of what is known in the field. My problem with this article is not that it says anything that is fundamentally untrue. Every finding she talks about is real. My problem with the interpretation. There are two major issues:

  • 1) She conflates general intelligence with spatial memory in rodents. The hippocampus has been linked much more clearly to what is called declarative memory in humans, so you might have more of a leg to stand on there. (Declarative memory is memory about statements, facts, or events as opposed to procedural memory which is memory about how to perform a task.) However, even declarative memory is not the same as general intelligence.
  • 2) She downplays the fact that we do not mechanistically understand the association of hippocampal neurogenesis and memory. I am certain that a lot of the researchers in this field would like to play up the association; however, the field itself is extraordinarily complicated.

    In brief, we would predict that if the neurogenesis-memory connection is valid ablating new neurons with either radiation or chemotherapy would cause a decline in performance in hippocampal-dependent tasks. This experiment has been done, and in some cases declines have been detected. However, Rene Hen at Columbia has shown that it depends heavily on which tasks you select — some are worse, some are better. Furthermore, I would also argue that radiation and chemotherapy have a lot of non-specific effects that make interpretation of the lesion studies extremely difficult. I remain unconvinced that the effects of the radiation/drugs are on new neurons exclusively.

    This is definitely not discussed in the article. While it is not necessary to go into the nitty-gritty details of hippocampal neurogenesis, I think that a sentence of caveat that said, “The link between exercise, neurogenesis, and memory has not been mechanistically established,” would make this article much more accurate. (The subject of what we know and do not know about neurogenesis and memory is discussed in this review.)

To step back for a minute, I am sensitive to the size constraints that restrict the ability of newspapers to discuss complex scientific issues. Science journalism is reaching out to people without a technical understanding of the subject, and trying to do this means we always have to deal with the issue of how much complexity to include.

However, I do not think it is unreasonable for science journalists to (A) know enough about what they are writing about to know if a metaphor is incorrect and (B) to include caveats about where a lot is controversial even if those caveats are brief.

This article is a great example of science journalism that gets the facts right but not the interpretation. The interpretation is important because interpretation tells the reader where the field is likely to go and how fast. Most of us agree on the fact of global warming temperature data. What people disagree on is the interpretation of that data.

Interpretation matters, and science journalists are dropping the ball.


  1. #1 Alvaro
    August 22, 2007

    Great post. No doubt, physical exercise -as well as stress management and a good nutrition- may contribute to overall cognitive functioning and neurogenesis, but are only indirect ways to improve one’s cognitive and self-regulation skills -i.e., “smarts”. And the article is missing this fundamental difference.

  2. #2 BikeMonkey
    August 22, 2007

    Er, did you get past the first couple of paragraphs, Jake?

    Regarding Morris: Critical as I am of science journalists, c’mon, the over-interpretation of smarts in a rat based on a single test or two is the hallmark of rat/cog studies of all stripes. when you are talking a PNAS or such like…well, the whole point of these is to wildly oversell the relevance. MWM criticism is a bit on the technical side don’t you think?

    Regarding dot-connecting, this article did a fair job, given the available human evidence. did you miss the fact that this got a splash because it had a human study in it? there was a correlation of VO2max improvement with measures of verbal learning. the NYT goes on to review a couple of related prior findings of exercise/cog enhancement. so there’s pretty good within-article support for the conclusions drawn IMO.

    Now, if scientists want to geek away about how it hasn’t been “mechanistically linked”, well, fine, debate away. but it seems a bit much to expect science journalism to convey these particular subtleties.

    On the other coast, did you happen to see the 4-parter from the LA Times on Gary Lynch, LTP and ampakines? They didn’t really connect the dots here either and in fact took the Lynch perspective that there was no prior “convincing link” between LTP and memory. opinions might vary a bit on that and it shows that it really depends on how the given scientist frames the “link” and “mechanistic proof” argument. A detached view lets us see that it is all a continuum of “proof” and a debate about relevance and levels of analysis.

  3. #3 gary
    August 23, 2007

    I have long been disappointed in the NY Times science coverage. Unfortunately, these days it seems to be getting even worse. That’s one of the reasons why I’m glad I have discovered Scienceblogs. Keep up the good work.