Quote from dmo:
This is very easy to explain if you look at the relationship between volatility and the probability distribution - that is, how likely it is that at expiration the underlying will be ATM, and how likely is that at expiration the underlying will be OTM.
It should be intuitively obvious that the higher the volatility, the LESS likely it is that at expiration the underlying will be ATM. At the same time, the MORE likely it is that the underlying will be OTM.
You can imagine how the bell-shaped probability curve (actually it's a lognormal distribution curve) changes as volatility increases. The mean (middle, or ATM) probability goes down, while the tails (OTM) thicken, or get higher.
The gamma curve follows that probability curve. There's your exact mathematical explanation. If this is difficult for anyone to picture, I can dig out a graphic and post it.