It seems like it's a no-brainer that against ranged attacks, prone creatures would not provide cover to allies, but I don't see any specific rules about it.
Also I guess you could argue that larger creatures might provide cover to smaller creatures when prone, but that's never stated anywhere.
The PHB says that the Prone condition (p277) makes you grant combat advantage to melee attacks from enemies, and gives you +2 bonus to all defenses to ranged attacks from non-adjacent enemies.
Also, it says on p280 that "[w]hen you make a ranged attack against an enemy and other enemies are in the way, your target has cover".
My ruling, not supported by the rules strictly, is that the prone opponent gives up his/her ability to provide cover to allies by getting the +2 bonus to defense against range from being prone (that is, the prone person is spending time kissing the ground and specifically not interposing him/herself along a line of sight to an ally behind him/her).
It all depends on what the rules mean by "in the way" which is not strictly described. The rules for determining line of sight and effect to a square use corners of tactical squares on a 2D plane, so it's hard to translate what "in the way" means in the third dimension.
But I'd say that prone allies get that +2 for lying down and this means they can't lend cover to allies behind them.
Technically, this is something that sort of falls under "DM's discretion".
If you want to go strictly by "Rules As Written" then the answer is yes. The pertinent clause, found in PHB 1 page 280, is this:
Determining Cover: To determine if a target has cover, choose a corner of a square you occupy (or a corner of your attack’s origin square) and trace imaginary lines from that corner to every corner of any one square the target occupies. If one or two of those lines are blocked by an obstacle or an enemy, the target has cover.
There are no rules that specify any general exceptions to this one, for prone creatures.
That being said, the DM is always the final arbiter at the table and is in most cases free to make rulings that are contrary to RAW, should they feel that it is better for the game.
I'd say the rule book is more or less correct, on page 280, but in this case only if the character behind the prone character is also prone, and only if that character is close behind the one providing the cover.
This provision of cover comes at a cost (at least in my ruling) that about half of the miss percentage provided by the prone character providing cover means that that character is hit.
To explain: if I'm trying to shoot an arrow at a prone target, my vertical miss chance is obviously higher (thus the +2 to defense against ranged attacks). If I'm trying to shoot an arrow at a prone target hiding immediately behind a log, a given percentage (depending on the percentage of cover provided by the log) of my shots that would have hit my target will hit the log instead. In this case, the front prone character is the log.
I usually make it a spot ruling, based on the size of the front prone character and the distance to the ranged attacker, as the greater the distance, the lower incident angle of the attack's trajectory. As well, the larger the front prone character, the bigger the "log".
