The scientific literature on readability does not back this up. Results show large individual variability with a small overall preference for dark-on-light.
Some old papers cited here: https://www.joedolson.com/2006/08/on-the-readability-of-inve...
EDIT: Also here is a good example of how reading white on black is hard: http://www.ironicsans.com/owmyeyes/
After coming back to HN from that site, I've been seeing text superimposed everywhere.
Furthermore, white-on-black isn't really the best. It's just the inverse of what we see on most sites/devices. Try green, or red on black. It has a very different effect as it is better tuned to the biology of the human eye, which is why it works better in low-light conditions such as driving.
This is an unfortunately common misconception. Sometimes yes, sometimes not. It is highly dependent on condition, in particular with what part of the field is obscured/impaired. I lack around 90% of my FOV - no central vision - resulting in ~20/200 acuity and while I find black on white to be incredibly discomforting I also don't get on with many 'high contrast' dark schemes. At the same time many of my friends with similar acuity but different conditions would be driven screaming from the room by some of my color scheme choices.
The contrast/color sweet spot varies (anecdotally, at least) greatly across both conditions and individuals.
It took some googling because I didn't know the terminology, but this is what I am talking about: A "Pivot" screen for the Mac by Radius.
http://lowendmac.com/wp-content/uploads/radius-pivot-display...
Another example of this is street signs. The oldest sign standards (speed limits, street names) are black text on white backgrounds. But the newer ones (highway information signs) are normally white on a dark green background, which is much easier to read at night.