Has anyone noticed how weird things are getting? When the controversy over the Dubai port deal was unloosed last week, the partisans ran, as they predictably will, to the corners of the ring, where they assumed the customary fighting stance – dukes up, lip curled, shoulders hunched.

So far, so normal – except for this: They ran to the wrong corners!

The Bush administration and its surrogates, routinely caricatured as right-wing Neanderthals, adopted the left-wing tactic of accusing their critics of racism. The Democrats, widely portrayed as peaceniks and squishes on national security, accused the Bushies of negligence and naivete in fighting a treacherous enemy.

Of course, this confusion might just be in keeping with the times. Nowadays the professional pugilists of partisan politics refuse to agree on anything, while pretending that every dispute is a matter of high principle.

The usual word for this is polarization, and in the U.S. Congress the evidence for it is plain. The explanation for it, on the other hand, and an accounting of why in the last decade polarization has become the default dynamic of every political debate are harder to find.

A lot of reformers and other good-government types thought the answer lay in the way congressional districts are drawn: Thanks to computerized gerrymandering and self-interested state legislators, more districts are designed to be uncompetitive – either purely Democratic or purely Republican.

Without the need to appeal to the center within their constituencies, congressmen are free to pander to their partisan extremes.

This, at any rate, was the most plausible explanation I’d heard, and many column inches in this space have been given to flogging it.

Yet now this redistricting thesis – also known as the “median voter theory,” which says that a competitive district will yield a winner who appeals to the centrist voter – has been thrown into doubt by David C. King, a Harvard University political scientist.

King’s work penetrates to the very wellsprings of partisan psychology.

For political theorists, he notes, competition in elections is a signal virtue. It forces politicians to articulate views clearly, encourages voter participation and moderates the positions and rhetoric of voter and candidate alike.

So the theorists believe, anyway. Yet King argues just the opposite. “Polarization,” he writes, “has risen with the increase of centrist (or competitive) congressional districts.”

How does this happen? Most congressional elections, King points out, are “two-tiered”: A party’s candidate is chosen in a primary and then meets his opposite number in a general election.

Turnout for primaries is typically low – around 17 percent in the 2002 primaries – and those who do vote tend to be the party’s activists, more ideologically extreme than voters in general elections. Primaries thus yield candidates these activists find appealing.

Close competition in a district also means the congressional seat is more likely to change hands. That probability, King writes, likewise “creates a strong incentive for previously non-mobilized potential activists to become involved in primary elections.”

With partisan lines so sharply drawn, centrist voters tend to stay home, and victory falls to whichever candidate can best arouse his partisans.

The final, and unpleasant, irony: “Political extremists in Congress are more likely to arise out of centrist or two-party competitive districts, rather than out of districts dominated by one party or the other.”

King cites as an example Ohio’s Sixth District, an area so mainstream that companies such as Burger King and Procter & Gamble Co. routinely use it as their middle-America test market.

Yet throughout the ’90s, thanks to highly motivated party advocates, its congressional seat was passed from a very liberal Democrat to a very conservative Republican and back again, by margins that seldom were wider than 51 percent to 49 percent.

“A political centrist in one of America’s most ‘average’ districts,” King writes, “might get whiplash from the wild ideological swings.”

From this evidence King tested a hypothesis – “the more competitive parties are in congressional districts, the less well members will represent interests in those districts” – by statistical methods that will be comprehensible to social scientists and almost nobody else.

His conclusion, though, is to the point: “The ideological preferences of politically centrist (and hence competitive) districts are represented, on average, the worst of all.” With few members left to represent the centrists, Congress is inevitably polarized.

As a good political scientist, King declines to make a normative judgment about this state of affairs, much less to offer a means of fixing it.

One solution does suggest itself, though: a return to less populist, more centralized ways of choosing congressional candidates. We might even consider reviving the famous “smoke-filled room” of old, where party leaders picked favorites on criteria beyond ideological purity.

It would make for a nice paradox, if nothing else – curing polarization, which horrifies good-government types, by returning to the smoke-filled room, which horrifies them even more.

But for now, King’s conclusions mean that the electoral system, with primaries choosing candidates and activists dominating primaries, appears to be rigged to reward partisan intensity above all else.

It also means that opinion-slingers like me will have to apologize to all those gerrymandering state legislators. See how weird things are getting?

Mr. Ferguson is a columnist for Bloomberg News.