A friend of mine and I had a really interesting discussion the other day about how difficult it is for a reasonable, mediating, middle-ground type of theory to survive in the scientific landscape. Within linguistics, there are a lot of debates I think a middle ground position serves the most efficient explanation.
However, there are several things holding back the success of such views. For one thing, extremes are usually louder. People are more vehement when defending/attacking a radical viewpoint. A mediating voice may simply be passed over as "not getting the whole picture" by the extremist views, instead of as a "reasonable middle ground." Both sides then view the middle ground as too much like the other side to be accepted.
As my friend pointed out, it's also the nature of statistics and experimental design to not support middle ground viewpoints. Outside of the rarely used Bayesian stats, hypothesis testing forces a binary distinction of confirmed or denied. That is, scientific methods promote extremist viewpoints.
If a hypothesis is tested and reveals a confirming answer, it is usually taken to mean that it is right and that opposing views are not — even if opposing views also receive confirming experimentation. What ensues is usually an unreconcilable clash of arguments, where neither side can see the validity of the other. Can't two rights exist at the same time?
Moreover, competing views often legitimately belong to differing paradigms of thinking. New paradigms are often asking new and different questions than the old ones, and thus providing new answers. Often, this involves chucking the old paradigm. However, do paradigm shifts have to throw out both the baby and the bathwater?
Perhaps Sinfest has the right idea...