The paper is "Sources of multi-decadal variability in Arctic sea ice extent," J J Day et al 2012 Environ. Res. Lett. 7 034011, and here's a little video clip of the lead author.
They found that the Arctic Oscillation (AO) moves ice around but doesn't diminish it, and that the Atlantic Multidecadal Oscillation (AMO), which is currently in its warm phase, about to turn colder, accounts for 5-30% of the observed decrease in September SIE.
This strongly suggest the Arctic sea ice will not be "coming back" when the AMO turns cold sometime (likely) later this decade. But then, it's been melting since before the AMO turned warm, according to data from Cryosphere Today (below). If there's any left, it only might melt a little slower (or may only cancel out the ice-albedo feedback).
A petition the Heartland Institute has out says "sea ice in the Arctic is rebounding." That, of course, is