It isn't true, however, that pure polarization is inevitable. Chait says:
The creation of more ideologically-coherent parties is a natural trend, and it's hard to see how or why it will stop. Northern moderates vote for Susan Collins and Olympia Snowe, but they're going to join a Republican filibuster to stop even a health care bill they voted for. Arkansans who think Barack Obama is a dangerous socialist probably aren't going to want to keep Blanche Lincoln in the Senate any more. To the extent that moderates do win Senate seats these days, it's largely by obscuring the political choices before their electorate.The thing is that northern moderates may stop voting for Collins and Snow -- the way they stopped voting for all of those Republican Members of the House such as Nancy Johnson and Chris Shays -- if they don't maneuver to make the median voter in their districts happy. If Republicans nominate only Jim Inhofes and Sam Brownbacks, they'll have ideological coherence but very few seats. And while Arkansans who think Obama is a socialist might not vote for Lincoln, there are plenty of Arkansans who are willing to vote for Lincoln and Pryor as long as they clearly separate themselves from at least some of Obama's positions on public policy issues.
If parties had no influence at all on candidate ideology, and ideology drove voting, then we could expect (per Anthony Downs) close races everywhere, with the Democratic candidate just to the left of the Republicans in each district -- and plenty of New England, west coast, and perhaps even inner city Republicans who would be far to the left of bunches of Democrats from the south, from places such as Kansas and Nebraska, and from Utah. Since neither of those conditions holds (parties do influence candidate ideology, and many things other than ideology -- such as group membership -- drive partisanship which then drives voting), we can't quite expect that result. But some of it should be expected. What's probably really going on is that ideological conservative activists have captured Republican nominations to such a large extent that Republicans start off with a large disadvantage, but those Republicans who do win in moderate or liberal districts are pretty far to the right of their constituencies.
It is not at all inevitable that this will continue, however. It's possible that liberals will eventually learn to copy Republican activists and better control nominations for purist candidates -- and it's also possible that Republicans may choose to nominate candidates more in line with their districts in order to maximize total seats in Congress. We're never going to return to the mid-20th century lowpoint of polarization, but there's no reason to expect that polarization at quite this historically high level is inevitable, either. As usual, I recommend scrolling down to the bottom of this page and watching the pretty pictures (courtesy of Poole, Rosenthal, and McCarty), and then explore their other links, if you want a better long-term idea of polarization in Congressional voting.