Anyone care to do the math? I'm not convinced. Yes, angular separation is higher the smaller you go. At the same time, so too are the number of teeth that must actually be cleared in order to actually remove the chain reduced. Imagine if you will a cog that's so small that it has only 4 teeth. If I can get the chain off of 2 of them, I have cleared an entire semicircle. On the other hand if we consider instead a 68 tooth ring as at

http://www.bikecult.com/works/chainr...sinTTcrank.jpg

it takes far more than 2 teeth worth of slack.

I don't think it's reasonable to compare with belts and pulleys because a chain drive flexes a great deal less than a rubber belt. So while the increase stretch available along the length of a large belt is very appreciable that's much less true for a strong steel chain.

What is required is that a certain amount of play be available to sufficiently separate the chain from the chainwheel or cog. If I have 1/2" of slack in my chain, that equates to 4 teeth worth of play (what that actually amounts to in terms of clearing the teeth and moving laterally off the wheel is a harder question to answer since it requires knowing things about the lateral rigidity of the chain). Obviously on larger rings, as cited above, that amount of slack will clear less of the drivetrain than on smaller rings.