Their numbers and mine differ slightly.  The primary reason for this is that in several states, especially New York, small parties are allowed to run major party candidates on their ballot line.  Many here will be familiar with the Working Families Party and the Liberal Pary who often endorse Democratic Party candidates.  (The Liberal Party also endorsed Rudy Guiliani, which I'm sure many Republicans will be hearing about if he should run for President :D).

Anyway, they count those votes those candidates receive on the minor party ballot lines as votes for those parties.  My argument is that the candidate is running primarily as a Democrat or a Republican, and they serve as a Democrat or Republican, so those votes should count for the Democrat or Republican Party.  And, I'm right on this and they're wrong :D

Anyway, that difference isn't that big a deal.

1990
Total Vote: 62,354,853
Democratic: 32,397,732 52.0%
Republican: 27,402,036 43.9%

1994        
Total Vote: 70,493,648
Democratic: 31,542,823 44.8%
Republican: 36,325,809 51.5%

1998
Total Vote: 66,604,802
Democratic: 31,391,834 47.1%
Republican: 31,983,612 48.0%

2002
Total Vote: 73,280,244  
Democratic: 33,623,365 45.9%
Republican: 37,390,372 51.0%

2006
Total Vote: 80,121,069
Democratic: 42,339,571 52.8%
Republican: 35,938,282 44.9%

As we can see, with the exception of 1998, each mid term cycle has swung from one party to the other with the winning party not winning by switching votes from the other side, but by getting millions of new voters to the polls.

Obviously these voters have come from the pool of voters that voted in the general elections but didn't historically vote in midterms.

A couple observations:
1.I think these results put a lie to the Republican spin that "Republicans lost in 2006 because conservatives stayed home."  In fact, the vast majority of Democratic gains were because they got millions of more votes out this time than in 2002.
Never the less, if Republicans want to believe they lost because they weren't conservative enough, I say we should encourage them in their beliefs.

2.For 2002 and 2006, I think there are two theories that explain the increase in votes and partisanship.

Theory A: The return of 'serious issues' (I.E Iraq).  Unlike in 1998 when the concerns were about Clinton's personal infidelities, the war is an issue that genuinely matters to people and drove them to the polls.

Theory B: The God-help-me 'Robert Novak theory': Negative advertising, contrary to the conventional belief, drives up voter turnout. His view is that the U.S is so partisan that negative advertising gins up party supporters, turning it into a 'my guy versus your guy kind of thing' and gets people to the polls.  

I imagine there are other theories as well.  It is interesting though, and obviously hard to argue given the numbers, that despite the increase in the negativity of negative ads (if not the number of negative ads themselves), that midterm turnout has increased substantantially over the last 16 years.