The W3C states a minimum of 4.5:1 contrast ratio[0] for body copy, or 7:1 for users with poor eyesight[1]. If Snook's calculator[2] can be relied upon, this makes #444/#fff a contrast of 9.74:1. Even a darker bg of #eee yields 8.39:1.
This also depends on the font weight, stroke, and size.
[0]: http://www.w3.org/TR/WCAG/#visual-audio-contrast-contrast
[1]: http://www.w3.org/TR/UNDERSTANDING-WCAG20/visual-audio-contr...
Better be safe and go with a darker text.
But darker text isn't necessarily safer, because having too much contrast is also a bad idea. That is why these guidelines about not using pure black or pure white get passed around so much in the first place.
Only dumb-ass designers have asserted that grey-text-on-grey-backgrounds is in any way a good thing, and they are wrong. (At least as far as conveying information goes.)
As someone who does read a lot of research about design and usability issues, I am always sceptical of people who write things like "There are hundreds of studies to that effect" yet don't cite a single one. If it were really as simple as you claim, we would all be reading white or yellow text on black screens and no-one would get headaches because their monitor was set too bright.
If you really do believe in looking at empirical data, you might like to read some of the research into how excessive contrast can exacerbate negative conditions, such as the "shaking" effect seen particularly by dyslexics or the "halo" effect seen particularly by older readers.