On Wednesday last week the town where I lived got 20 inches of snow in a twenty-four hour period. I got an unexpected, happy, day off work. Bizarrely, southern regions like DC and Baltimore got much more than northern ones; Baltimore got 40 inches, Philadelphia got about the same. Records were set in both places for the snowiest winters in recent history. People were left wondering and reeling at this capriciousness of the Norse Gods.
So what could be the reason for this sudden onslaught of severe weather? That's akin to asking what could be the reason for cancer suddenly emerging in someone's body or for a particular drug demonstrating a slew of side-effects. The reasons are non-obvious, often non-intuitive, complex, multifactorial and extremely hard to determine. And that is also what one should say if asked to elucidate reasons for a particularly snowy winter.
But human beings don't work that way. Immediately there sprung up a debate about whether global warming could be responsible for the increased snow. Engaging in the common and never-dying fallacy of equating weather with climate, climate change skeptics declared the cold to be a slap in the face of AGW proponents. On the other side, while most climate scientists are pointing out that single weather events have scant connection with global warming, some proponents are also
saying that this is actually a good instance of the effects of global warming, that global warming does predict extreme weather events, that all this is simply part of the connected whole. More vapor in the air, El Nino and other events have been suggested as plausible candidates.
Now let's step back a little and think about this from the educated layperson's perspective. Less snow has been commonly predicted to be a consequence of global warming, but now the same explanation is being provided for lots of snow. The layman should be excused for being skeptical about a model that seems to equally explain diametrically opposite events. Of course, as we mentioned before, more or less snow neither "proves" nor "disproves" global warming. But to me this is yet another reminder of why I don't say much about the topic these days; the whole damn thing has gotten so overly politicized that each side feels compelled to say something non-scientific just to make the other side shut up.
However, from a scientific perspective too this issue illustrates the pitfalls that natural science faces in the twenty first century. When I discussed this issue with my father who is an economics professor the other day, he said "Welcome to the social sciences". Social scientists face such problems all the time. What happens when a model becomes so complex that it can explain virtually any observation you throw at it? (To begin with the model also become so complex that you stop truly understanding it; case in point- derivatives on Wall Street). Surely there seems to be a problem with a model that is invoked to explain both more precipitation and absence of precipitation. That would be akin to a molecular model that predicts the same reason for compounds with both high and low potencies against a protein target.
The hallmark of a judicious model is that it is not so spare as to be useless but also not so full of parameters and variables so as to fit almost any data point. A model that seems to explain everything (as sometimes seems to be the case with global warming) is a bad scientific model because in principle it's hard to see how it could be falsified (Popper again). In addition, you should always ask how many data points are enough to build confidence in a model; statisticians have struggled for decades with this sampling problem, and there is no straightforward general answer. Sadly, it's not climate scientists that have first raised such issues through their model-building. That dubious honor belongs to evolutionary biologists.
Evolutionary biology is notorious for advancing adaptationist explanations which can account for almost any observed trait. For instance, polar bears are white because they are supposed to camouflage well into their surroundings, but penguins are black because their skin should absorb enough sunlight to keep them warm. Now why does the first explanation not apply to the second case and vice versa? Well, in case of evolutionary biology the short answer that is given is "trade offs". Depending on the details of the problem (in this case the species, its body requirements, genetic makeup etc.), in the first case the ability to camouflage won out over the need to remain warm, and vice versa for the second one. But beyond a certain point it can be impossible to actually explain such trade offs since so much in evolution is a matter of contingency. And one can conveniently invoke trade-offs (Shazam!) as a magical whitewashing word for almost any trait. That's hardly an actual explanation.
Nevertheless, such ingenious explanations have often been advocated by evolutionary biologists. In a
classic article, Stephen Jay Gould and Richard Lewontin shot down this relentless urge to wrap everything into an adaptionist program; their main point was that every trait is not the consequence of adaptation and natural selection and some traits can be simply carried along for the ride with others without possessing any evolutionary benefit. The main merit of the adaptationist explanations is their internal logic. However, internal logic by itself, no matter how tempting, does not make an explanation. In the absence of experimental data, such hypotheses about evolutionary adaptations are just that, good hypotheses waiting to be validated by good data. A professor of mine got so fed up with these ingenious evolutionary explanations for everything from homosexuality to sloths coming down from trees to bury their feces that he wrote a highly readable book about it. Again, it's not that these ideas are bad, but in the absence of causal evidence they can only remain respectable armchair speculation. So how then do we come up with explanations?
Sadly, here's when fields like evolution and climate change run into fundamental roadblocks of the kind faced by social scientists; the sheer complexity of the system thwarts attempts at clean experiments. The big problem with fields like psychology, sociology and economics is that it is often difficult or even impossible to perform controlled experiments. Admittedly the situation here is worse than climate science, since the data itself is variable and represents a moving target (was the state of mind of your experimental subjects the same on Monday as on Tuesday?). Consider the hundreds of pop science books on neuroscience claiming that things like fMRI scans can "explain" emotions like hate and jealousy, and even spiritual and religious experiences. Other problems notwithstanding, how on earth do we not know that at the very least, like the perpetual observed-induced reality in quantum mechanics, we do not influence what we want to observe? But even in the apparently more rigorous discipline of climate science, models are the result of data conducted under less than ideal non-isolated laboratory conditions from thousands of places over dozens of years. Who can guarantee that at least some of this data won't even be contradictory, let alone that all of it would consistently be of the same standard? To be fair to climate scientists, they usually perform stringent checks on the validity of their models but no checks can help duplicate fine differences in experimental conditions spread over thousands of data points over long time frames.
Lest one think that only the "softer" sciences face these problems, witness the current debate about string theory. Skeptics say that about the only reason that the framework is so highly regarded is because it seems to be logically internally consistent, is mathematically elegant and seems to tantalizingly "feel right". All these qualities can be respectable, but I suspect that the pioneers of modern science in the eighteenth and nineteenth century would not have been happy with this state of affairs. From what I have read, there seem to be no hard experimental tests that could provide strong support for string theory.
That then is the dilemma the natural sciences find themselves in in my opinion, a dilemma that the social sciences have faced for centuries. In fact one can argue that the dilemma has been caused by the social sciences finally intersecting with the natural science as their integrated whole has become more and more complex and is now tackling extremely convoluted territory like the brain, the climate, the universe, human behavior, the economy, evolution and the mechanisms of drug action and disease. With this kind of complexity, scientists have been resigned to pick between two quite unsastisfactory choices; either no explanation at all, or an "explanation" based on models, internal logical consistency, "aesthetics" and elegance (case in point- string theory) and ingenious sounding armchair explanations. In many cases the underlying systems are simply so dense that scientists are forced to perform extensive parametrization and model building. There is probably an equation somewhere relating excessive parametrization to risk of model failure.
Nonetheless, in the absence of controlled experiments, there is not much that science can lean on at this point. But in fact one can argue that science actually proceeds in this way, by tentatively accepting hypotheses. As long as it's kept in mind that the hypotheses are only hypotheses, we can still have a wall to grasp as we grope around in the dark. If we start regarding the hypotheses as explanations and facts, we will leave the safety of that frail wall to grasp at imaginary will-o-wisps at our own peril.
In the latest issue of J. Med. Chem., researchers from Roche in Basel have a nice analysis of intramolecular hydrogen bonds in druglike molecules. An internal hydrogen bond can intuitively confer an important property on a drug; it can make the drug more lipophilic by shielding the hydrogen bonding groups from solvent. Thus, intramolecular h-bonding has emerged as a useful strategy in improving membrane permeability.