This is extremely important. The danger comes from a thief replaying the transaction you made on the new chain. If he replays it on the original Bitcoin chain, then he can claim the coins in the transaction from your Bitcoin wallet. If the wallet is empty, then he can't do this.
Most of the answers in this thread should have been really helpfull for the OP, i didn't see many mistakes being made by other posters. I did, however, wanted to elaborate on this statement.
If, for example, bitcoin forked into replayattackbitcoin (a fork without replay protection) and you have 1 unspent output funding address 1MyoriGinalAddress before the fork, what could happen if you spent this unspent output on the replayattackbitcoin network?
1) You create a transaction using the unspent output to fund 1AddressGeneratedbyAseller
2) you broadcast the transaction on replayattackbitcoin's network
3) the attacker saves this transaction and broadcasts it on bitcoin's network
4) the transaction gets included into a block on both the original and the new network's blockchain
The end result is that address 1AddressGeneratedbyAseller will become funded on BOTH chains. The attacker cannot use your transaction funding 1AddressGeneratedbyAseller to fund a thirth address owned by the attacker.
In other words, the attacker has to convice you to fund his own address on chain A, then replay the transaction on chain B, otherwise he'll gain nothing from the attack (he can still perform the attack to annoy you, but he won't have any financial motivation to do so).
I actually made this diagram a long time ago (copyright: me). It shows a replay attack and a simple way to protect yourself from such an attack.