I don't think it's entirely clear from the article what 'cycle' should mean in this context. But 1/50 of a degree matches relatively closely to the traditional Snellen (as in the guy who made the eye charts) definition of normal vision (20/20 or 6/6 in metric countries http://en.wikipedia.org/wiki/Visual_acuity#Normal_vision) as being able to discern letters whose features were 1 minute of arc(e.g. 1/60 of a degree). At 12 inches, the angle subtended by a pixel (which is, I think, the corresponding minimum feature size) is arcsine( (1 in /326)/12 in) is 0.88 minutes (that is, less than the 1 minute Snellen definition of normal vision, which is in turn smaller than the 1/50 definition that this guy gives).

So I think Soneira guy is off base. But I think the much bigger problem is that both Soneira and the 1 minute definition are talking about the acuity of a 'normal' person, and seem to be largely ignoring the issue of a significant variation in the population. My question is, for what portion of the population does this display exceed the limits of their visual acuity?

The definition in that Wikipedia article says that 20/20 vision means being able to resolve two points separated by one arc-minute. Again, on the output side that would mean being able to display two points at that separation with something contrasting in between.

On the other hand, elsewhere Wikipedia says that it means being able to distinguish Snellen optotypes whose total size is one arc-minute. On the face of it, that implies more resolution than two pixels per arc-minute. (Hand-wavily, maybe it's about the same: if you can resolve 2x2 pixels in a square of side 1 arcminute, then for high-contrast Snellen-type images that maybe suggests that you ought to be able to distinguish about 2^(2x2) = 16 different such images, and a Snellen chart actually uses 10 or 12 different optotypes, depending on whether it's one of Snellen's original ones or a modern variant. That's pretty close to 16.)

Phil Plait's blog entry that someone else linked to talks about the difference between "normal" and "ideal", and concludes that an average person looking at a new iPhone at 1' distance will indeed not be able to resolve the pixels (though not by much).

Actually, wrt the Snellen optotypes, the entire optotype subtends 5 minutes (I couldn't find the 'elsewhere' you're talking about), but distinguishing them requires that you be able to resolve features 1 minute in size. In fact, on Snellen charts, the letters are carefully designed to reflect this. For instance, on the 'E', the width of each bar of the 'E' is equal to the width of the white space between bars. http://en.wikipedia.org/wiki/Snellen_chart#.2220.2F20.22_.28...

So yeah, if you drew out a tiny "E" (or "P" or "F", I don't think it matters) on a 326 ppi screen a bit more than 12 inches from the viewer's eyes, where the width of each bar was 1 pixel and the space between bars was 1 pixel, then I think that would match up closely with the standard for normal vision, at least in terms of visual acuity for one eye.

Actually, now I can't find the "elsewhere" either and I wonder whether I misread. Having looked again, I agree with you: standard-according-to-Snellen vision means being able to resolve features corresponding to (e.g.) an "E" on a 5x5 pixel grid with pixels of size 1 arc-minute.

Matching that up with the resolution of a display device is still a bit subtle. For instance, suppose you're trying to display an "E" of that size on the display, but it's offset by half a pixel vertically. Result: you get a grey rectangle that's a bit darker along the left edge. :-)

(I think my conclusion from all this is: what Apple are claiming about the iPhone 4 display is about as close to the truth as it's reasonable to expect in marketing materials. That is: everything they've said is at least defensible, but they've put a very positive spin on everything. Seems fair enough to me. And as a pixel-freak who isn't currently a smartphone user, I'm awfully tempted by the new iPhone...)