Public understanding of autism research suffers from two perennial complications: correlational studies and press releases. The report by Waldman, Nicholson and Adilov (“Profs: TV May Trigger Autism,” October 27) is, I fear, the latest example of both these foibles. The controversy surrounding Waldman et al.’s unpublished and incompletely substantiated study illustrates the danger that arises when the conduct of science by peer review is supplanted with science by press release.
Waldman et al.’s curious hypothesis that television viewing may trigger autism stems from four items of backgroun evidence: a reported correlation between television viewing and later diagnosis of Attention Deficit Hyperactivity Disorder (ADHD), an abnormal slowness to disengage attention in children whose family histories put them at risk for later diagnosis of autism, large increases during the late twentieth century in both autism and television viewing and a low incidence of autism amongst the Amish, who do not use electricity and therefore do not watch television. These separate indications, Waldman et al. claim, add up to plausibility for the idea that television viewing may trigger autism in certain susceptible children.
However, nascent ADHD seems more likely to spur television viewing than vice versa, abnormalities of attention might predispose to autism without any involvement of television and many factors other than television habits distinguish both the late twentieth century and the Amish.
This argument from television is cast further into question when one recognizes that Waldman et al. haven’t actually studied any children with autism. Their claim rests not on individual data but on group data, measurements from demographic and geographic databases that sample the general population. Waldman et al. present a roundabout argument in which correlations between autism incidence and precipitation rates and, separately, between precipitation and television viewing, amount to a causal effect of television on autism. Such a correlational argument is as apt to establish, say, that ice cream sales cause homicides, since both peak at the same time of year. Many environmental variables in addition to television viewing likely correlate with precipitation — for instance, agricultural run-off, exposure to indoor toxins, or even the geographical factor of proximity to an autism diagnostic clinic (since coastal cities tend to contain more of these).
In addition, in a set of as many variables and comparisons as Waldman et al. apparently have examined, one is certain to find correlations that arise simply by chance. Verifying such correlations demands that the test replicate in separate samples. Waldman et al.’s report does suggest a correlation between autism and precipitation in data from Oregon and Washington. In California, though, the greatest incidence of autism arises in Los Angeles and other relatively dry counties in the south. In addition, Waldman et al.’s data on ethnicity effects seem so equivocal as to argue against any robust relation between autism and television: an effect is found in the case of the native American population, but not in the cases of black and Hispanic populations.
Paradoxically, Waldman et al. argue that their indirect finding that autism is related to television through precipitation is less assailable than a direct correlation between television and autism would be, since although autism might cause increased television viewing, autism cannot of course cause increased precipitation. Even assuming that the precipitation-autism connection is real, this argument ignores the strong possibility that precipitation could cause both autism and television viewing, separately. Further doubt is sown by the model’s lack of biological plausibility: indications from studies of early brain growth are that autism is determined by the end of the first year of life, before the time at which a great deal of television exposure would have occurred. Although it is true that autism’s social and communicative deficits aren’t reliably diagnosed until toddlerhood, other, more subtle abnormalities of neurophysiology and behaviour do exist early on.
Every few years, a press release touts a breakthrough in autism research or treatment. The list of these discredited causes or treatments of autism includes secretin infusion, the MMR vaccine, chelation therapy and facilitated communication. Each of them has given parents false hope, or worse, false guilt. The researchers behind these press releases suffer from the most ethical of motivations: they sincerely believe that their conclusions are sound, they perceive a prejudice of the scientific establishment against these conclusions, and they sense an urgency to communicate their findings to the broadest possible audience of scholars and, more significantly, parents, so as to prevent autism’s ongoing tragedy. This well-motivated urgency often results in circumvention of peer review, the careful process by which scientific findings are vetted and judged worthy of presentation to the broader community. Despite its many faults and delays, peer review remains essential to maintaining public trust in science.
Perhaps the worst consequence of circumventing peer review is that those elements of the study that might have withstood scrutiny end up written off along with all the rest. Waldman et al.’s correlation between autism and precipitation would, if verified, constitute an interesting result on the possible role of environmental toxins. But its association with the autism-and-television conclusion may render it less likely to be followed up. (A similar blanket dismissal occurred in the case of the very flawed technique of facilitated communication, certain elements of which are in fact compatible with what’s known about autistic attention and motor planning.)
Waldman et al. are sensitive to the urgent need for intervention in autism’s growing public health crisis, and for this sincere motivation they deserve our respect. They’re correct in asserting that their hypothesis about autism and television is easy enough to test prospectively. I encourage them to do so, and to report their findings in a peer-reviewed publication.
Matthew Belmonte is an assistant professor at the Department of Human Development at Cornell University. He can be contacted at email@example.com. Guest Room appears periodically.