Voila: "The generic vote is not a particularly good predictor of midterm House outcomes, particularly this early in the race. Democrats almost always hold a lead over Republicans in the generic vote of registered voters, with low turnout often a factor in the outcomes. In fact, in 1994, when the Republicans captured the House, Democrats held a narrow lead (+6 in Gallup Polls) in the spring 1994 generic vote. The Republicans took only a modest lead (+7 among likely voters) in the generic poll vote late in the campaign after they introduced the Contract with America."
But as I said, what interests me is the likely voter versus voter gap.