« EMT! | Main | Zend PHP conference »

Rainfall + cable TV = autism?

| 8 Comments | 1 TrackBack

A trio of economists at Cornell and Indiana University-Purdue have done an as-yet-unpublished study that makes the following claims:

  • As indicated by the Bureau of Labor Statistics’ American Time Use Survey, there's a correlation between local rainfall levels and amount of TV a young child watches.
  • Comparing autism data and rainfall data in California, Oregon, and Washington (where rainfall levels vary substantially) reveals a correlation between autism rates and rainfall levels.
  • Autism rates also correlate with "percentage of households that subscribe to cable television."
  • Therefore, it's possible that watching a lot of TV before age 3 is "an important trigger for the onset of autism."

Gregg Easterbook of Slate wrote an article about this study a week ago, titled "TV Really Might Cause Autism." Turns out that Easterbrook had speculated that TV might cause autism back in early September; that piece was, as Easterbrook notes in the later article, "sheer speculation, since [he] knew of no researchers pursuing the question." So Easterbrook can be pardoned for being particularly interested in the Cornell study.

Apparently the second Easterbrook piece caused quite a stir online, though I missed that. Then, on October 20, along came a Time article: "A Bizarre Study Suggests That Watching TV Causes Autism." As you might expect from the title, Claudia Wallis (the article's author) doesn't take the Cornell study very seriously.

Wallis provides some useful background about the spike in reported autism rates in the past 35 years:

Some of the spike can be reasonably attributed to a new, broader definition of the disorder, better detection, mandatory reporting by schools and greater awareness of autism[...]. Still, there's a nagging sense among many experts that some mysterious X-factor or factors in the environment tip genetically susceptible kids into autism, though efforts to pin it on childhood vaccines, mercury or other toxins haven't panned out. Genes alone can't explain it; the identical twin of a child with autism has only a 70% to 90% chance of being similarly afflicted.

(There are still plenty of people who claim that the cause is in fact vaccines and/or mercury, but according to the CDC, "The weight of currently available scientific evidence does not support the hypothesis that vaccines cause autism.")

The Time article goes on to take the new study to task for making such statements as "Our precipitation tests indicate that just under forty percent of autism diagnoses in the three states studied is the result of television watching due to precipitation" (in the study's abstract)--and I agree that using phrases like "result of" seems (as Wallis puts it) "oddly definitive." But the word "indicate" in the original sentence (left out by Wallis) makes it sound less definitive, and the study's actual conclusion says, among other things:

Although our findings are consistent with our hypothesis, we do not believe our findings represent definitive evidence for our hypothesis.

Wallis goes on to quote Vanderbilt University geneticist Pat Levitt as suggesting that rainfall could be producing other environmental factors, such as mold or mildew; the implication is that the study's authors didn't even think of that. In fact, though, the study does partly address that possibility:

[...S]ince precipitation is likely correlated with young children spending more time indoors generally, not just young children watching more television, our first main finding could be due to any indoor toxin. Therefore, we also employ a second instrumental variable or natural experiment, that is correlated with early childhood television watching but unlikely to be substantially correlated with time spent indoors.

If I'm understanding the study right (I haven't read the whole thing in detail), that sentence is talking about the cable-TV subscription levels part of the study. The authors later add:

[...I]t is theoretically possible that autism is positively correlated with cable subscription rates because young children in families with cable spend more time indoors and there is a toxin where exposure is higher for indoor activities than outdoor activities. But, although [that's] theoretically possible, we do not believe the effect of cable on time spent indoors by young children is likely to be large enough to [cause greater exposure to such a toxin].

Anyway, so the Time piece goes further than I think is warranted in attacking the study--when I started writing this entry, I was all set to say "Go read this great article in Time debunking this silly study," but then I looked at the study itself and saw that the authors are taking a much more reasonable and scientific approach than the Time article suggests. But the Time piece does end with some interesting stuff about relevant genetic issues, and notes that "[t]here are probably many routes to the disorder, involving diverse combinations of genes and noxious environmental influences."

Time has also published a brief response from Gregg Easterbrook to the Time article, in which Easterbrook notes that the Cornell study may result in someone doing a more definitive study that measures the issues directly rather than indirectly. (Side note: Easterbrook goes on to vaguely imply that, because Time Warner owns the Cartoon Network, Time may have had ulterior motives in "denouncing this research." I am very skeptical about this kind of conspiracy-theory finger-pointing; Easterbrook's use of this tactic makes me less inclined to trust him.)

Another interesting aspect of all this is that attributing autism to TV watching can be seen as yet another in a series of claims that one variety or another of bad parenting is the cause of autism. Kendra Pettengill of Oregon, a parent of an autistic child, discusses the "bad parenting" issue and provides some further criticisms of the study; I haven't read enough of the study to have an opinion on whether the authors anticipated those criticisms or not.

Do I have a point in all this? Not really. I think the study provides some very interesting correlations; I think some of the study's critics are underestimating and/or misunderstanding the study's authors; I think the study doesn't come close to addressing all the causes of autism (nor does it claim to); I think the study probably does oversimplify some things and ignore some relevant issues. And I should reiterate that the study has not yet been published, so I'm guessing it hasn't yet been peer-reviewed. But mostly, I just thought all this was an interesting topic and an interesting discussion.

(And since pretty much everyone I've seen comment on this study has their own axe to grind one way or another, I may as well grind my own, which is that claims that initially seem obviously silly, stupid, or counterintuitive often turn out to be much more reasonable when looked at in detail, beyond the surface level. On the other hand, sometimes such claims turn out to actually be silly or stupid when examined in detail; certainly I'm not claiming that silly-looking claims are always reasonable. (That would itself be both silly and unreasonable.) I'm just making my usual suggestion that people look beyond the surface before coming to conclusions.)


It just seems to me that an awful lot of junk science is based around correlations of these kinds. I agree that one shouldn't reject claims out-of-hand based on their content ("TV causes autism! That's ridiculous!") but I think it's fair to reject "correlation implies causation" claims pretty much out-of-hand. It's fairly interesting that they've identified some things that seem to correlate with the rise in autism, but it's pretty easy to point to a lot of other possible mechanisms than direct causation:

1) some third factor is actually independently correlated with both items (e.g. the Time article suggests that increased rate of cable TV infrastructure could easily correlate with increased rate of sophistication of pediatricians, and thus increased diagnosis of autism)

2)items actually are correlated, but mechanism of causation is different (e.g. what if autism is caused in part by a mother's lack of exercise during pregnancy, and increased TV watching correlates with that? or, as is suggested, indoor pathogens such as mold, etc.)

If the latter is the case, then these correlation results could inspire some really interesting research, but by suggesting that they've identified the causative mechanism already, the authors are undercutting that option.

The authors seem to be taking a kind of arrogant approach of saying, hey, we don't know anything about medicine, but we know numbers, and the numbers don't lie. I honestly don't care if they put in some nice "of course this isn't proof" comments; I still think it's an irresponsible article. Why not present the correlations without implying causation, as a way to propose funding for the large longitudinal study that could actually produce some hard answers?

I read that article and several others on the study, and I find that the results don't quite pass my "does that make sense to me" test. I'm mildly autistic myself, and I have a good idea why (a family history and some other factors). Obviously one exception does not disprove a rule, but in general the people I know who are autistic come from families where you can see how magnifying certain traits would result in autistic behaviour. I don't meet many high-functioning autistic people for whom there is no explanation whatsoever of where that came from.

And there is one problem with the study that I didn't see addressed: being autistic might make you more likely to watch TV in general. TVs are fascinating. All glowing and mechanical and when you look at them you can take the light apart and see the dots moving. You don't even need to watch the program that is on. And outside? Outside is terrifying, full of people and human interactions and strange things. So it seems reasonable that in a household with an autistic child, the preferred rainy-day activity might be TV watching, and the family might even get cable service specifically for that kid (because keeping the kids out of trouble is that much harder when one is autistic).

(We played with Legos on rainy days when I was a kid.)

Also, what if being autistic causes it to rain more?


Part of what I'm saying here is that I don't think the study makes strong claims of causation, nor does it say that (in general) correlation alone implies causality. I agree that it overstates the implications, but only (as far as I saw) in a couple of sentences in the abstract; the rest of the study that I read (and even most of the abstract) claims correlation but does not claim causality.

The Time article makes the causality claims sound stronger and more prominent than they are in the actual study. If you haven't read the study itself, I recommend doing so; please don't judge the study based on the Time article. If you did read the study itself, then I think you and I interpreted various things differently, or focused on different parts of it.

You wrote: Why not present the correlations without implying causation, as a way to propose funding for the large longitudinal study that could actually produce some hard answers?

My impression of the study is that that's pretty much what they did (except that in a couple of places they mildly imply causation). They say they found a particular kind of correlation (after controlling for a variety of other factors); they say the correlation strongly implies but does not prove their causation hypothesis; they call for a bigger better study that could provide clearer answers.

They also discuss alternative possibilities that could also explain the correlation, and to some extent they address those possibilities, so it's not just a case of their saying "well, there's a correlation, so therefore there must be causality." In my interpretation, they're saying "There's a strong correlation, and we've eliminated most of the alternative explanations we can think of, so we think there's a good chance of causality, but further study would be a good idea."

In general, when I encounter claims from a study that don't make sense to me, my first reaction (like that of pretty much everyone I've seen talking about this study) tends to be to attack the methodology, and to point out alternative answers that might give the same results. That's fine--but it's a good idea to check the study itself to see whether they addressed those concerns before saying that those concerns point out a clear flaw in the study. In this case, I think the people who did the study were a lot smarter about it than most of their detractors are giving them credit for.


The study does address the question of whether they've got the causality backward. I recommend taking a look at the study, if you haven't done so; most of the articles about the study seem to be determined to show how foolish the study is.

Here's a quote from the study:

[...A]n advantage of our study is that by using an instrumental variables approach we avoid a problem of reverse causality. That is, a direct finding that young children who watch more television are more likely to develop autism could be due to young children vulnerable to developing autism having a prediliction for watching a large amount of television. But a finding that precipitation is positively correlated with autism is not subject to this criticism--we can be quite certain that autism does not cause precipitation.

(I know Ben beat me to saying that, but I wrote this comment a couple days ago; just didn't get a chance to post it 'til now.)

Also note that the study doesn't claim that TV is the single cause of all cases of autism; it claims to have found a correlation between early TV watching and some percentage of the increase in autism. It doesn't say that there are no genetic factors, nor that there are no other environmental factors; I think there's plenty of room within the study's hypothesis for family history and other factors.

I'm certainly not saying that I think the study is definitely right in its hypothesis about causation. It may well be totally wrong, for all I know. And yes, there are issues about the methodology; and yes, there are areas that the study doesn't address; and yes, there's plenty of room for further study, some of which might easily disprove the original hypothesis. And yes, the study is as-yet unpublished. There's lots of possible ground for criticism.

And for that matter, for all I know the people who are claiming that this study is part of a conspiracy by Big Pharmacy to cover up the evils of vaccines may be right. Hell, for all I know the study may be a hoax or a joke, designed to show the kinds of results you can come up with if you apply economic methods to medical issues. I don't think either of the ideas in this paragraph are at all likely, but I don't know for sure.

But I do think this study is getting undeserved short shrift. The authors have apparently found a statistically significant correlation between various factors, and they've attempted to address some of the alternative possible explanations for that correlation. So I would rather see people attack the study based on what's in the study, rather than what's in the articles about the study.

Earlier, I linked to the main web page for the study; here, in case people would prefer it, is a direct link to the PDF of the study.

I wrote a critique that does address the data presented in the paper. See here.

Joseph: Thanks! Good critique. Others who may be interested: Joseph makes some interesting points about correlation among population density, rainfall, and television. The comments section of his entry is particularly interesting, providing some followup correspondence with the authors of the rainfall/TV study.

I wish I had time to read the article, but I barely have time to read this, with students breathing down my neck for their case study grades...

So, without reading it, I really don't buy it at all. Your take, Jed, that"
"In my interpretation, they're saying 'There's a strong correlation, and we've eliminated most of the alternative explanations we can think of, so we think there's a good chance of causality, but further study would be a good idea.'"
to me just shows the limitations of that "alternate explanations that we can think of" part. There's almost certainly a very very large number of explanations that no one has thought of, and from the clips here they didn't even do a very good job on the ones they did think of.

(eg, more rain + more cable access correlate to more autism.... The idea that autistic kids would much prefer to look at the TV than many other activities makes worlds of sense to me. The fact that reverse causality is clearly false with regards to precipitation doesn't mean that it's also therefore automatically false for the TV link. Just because two things correlate to a third doesn't mean that they're intrinsically linked to each other.)

The main problem I have (which may have been addressed in the article; it's not in any clips here) is that the ages don't work. Part of why there's such a persistent belief in a connection between MMR and autism is the timing -- MMR is first given at 15 mo, and autistic symptoms first become noticeable (sometimes dramatically) between 15 and 18 mo (potentially resulting in a correlation so emotionally charged that the parents are understandably unable to accept volumes of epidemiological data disproving the causality). Kids, in general, don't watch a lot of TV before the age of two. That's why Barney was invented (and why it's so loathesome and noxious to adults) -- the creators wanted something which would hold their two-year-olds' attention for 30 min so mothers could do other things. By the time a kid really likes to watch TV, autism would already be evident; if you have a child under age two who does sit and stare at the TV, it's indicative that he's outside the norm already. Sure, there could be a slight predisposition and vulnerability, and exposing such a brain to the flickery lights might potentiate the problem, but that explanation is on the extreme edge of hypothetical, and the current data dosen't do much to bring it into theory range for me, even.

But then, I'm also annoyed by the physicist "proving" mathematically that vampires don't exist (if they eat/convert one human a month, the whole planet would be vamps within a few years), since it ignores a) sometimes you eat without making your meal into a vampire and b) vampires can get killed, too. Sheesh, you'd think the guy would know these basics if he's doing "research" on the topic!

So I must admit to finding the Cornell article a little speculative. However, it occurred to me that one could construct a falsifiable story surrounding the connection between television viewing and autism if one accepts the premise that certain TV programing can induce seizers as well as the assumption by some that undocumented childhood seizers play a role in triggering autism. If these assumptions are correct, increased television exposure might provide a trigger for a larger number of undocumented minor seizures in children which in turn triggers increased instances of autism. With any luck, someone will actually go about doing direct observations.

Post a comment