That's the result of an analysis in Seismological Research Letters by Robert Thériault et al (Nature summary here); the Smithsonian has a video of earthquake lights in Sichuan in 2008. Rifts are defects in plates far from the edges of plates - so in other words, not where two plates are grinding alongside each other (like in California) or where one is going under another (like in Japan or the Pacific Northwest). The unexpected and massive New Madrid quake in Missouri in 1811 was a rift quake.
(Added later: a well-known skeptics' blog weighs in as well, pointing out that these theories are actually not new.)
I've written about earthquake lights here before, related to a recent quake in Peru; one theory is there is ultra low frequency ULF) EM generated by these events, and that the lights (but not the quakes) are reproduced by ULF usage elsewhere. (See the Vogel study at the link, which investigated lights on a reservation near Yakima.) That's not the theory that Theriault is advancing, which is that oxygen ion release is in the causal pathway. (What would be really interesting is an experiment showing similar ionization with ULF.) Earthquake lights have been reported before earthquakes for centuries - surprisingly, these articles don't mention the glow that one swimmer saw at Ocean Beach, San Francisco after a pre-dawn swim on a certain April morning in 1906, a few hours before the big one - but it wasn't until someone actually filmed them in Japan in the 1960s that people started taking them seriously.