I found a paper which kind of explains the physiology behind PenTile:

http://www.cvs.rochester.edu/william...da99nature.pdf

The original PenTile layout had five subpixels in a group (thus the name - penta-tile), and the results show why: the occurrence of "blue" cones is so much lower that using less of them makes sense. The low rate is also consistent.

For red and green receptors, their ratio seems highly variant so making a case for skimping on red subpixels seems more difficult.

Green has one advantage, it generates responses in all three types of receptors, so for pure brightness information, resolution would be the greatest here. This looks like a possibility to trade off color reproduction accuracy for brightness and perceived "crispness" (as opposed to actual high resolution), and this is what's probably done in AMOLED PenTile displays.

Oh yeah, there's one problem with the study: The sample size is pretty much completely ridiculous (two). This paper is from 1999, so maybe there just isn't any more recent, comprehensive research. I guess PenTile was driven foremost by cost considerations and previous experience from video compression.

I admit that I lack real insight on how perception of contrasts and colors interrelate so maybe there is an opportunity for enhanced subpixel layouts that I can't see. Right now I think whether PenTile is better or worse doesn't have many hard and fast arguments either way...