According to my Exam Prep CEH study guide one way of detecting a Honeypot is by testing to see if all the services that appear to be open actually are. Services using SSL in particular should be checked like HTTPS or SMTPS etc.
Other ways of detecting Honeypots/Honeynets include checking the MAC addresses on the network, as has already been mentioned here. A badly configured Honeynet will have the same MAC on all the NIC's.
Many Honeypots have been set up as spam traps by BL's, and so a quick check to see if any mail has been sent from the mail account or if any legitimate-looking mail has arrived to the mail account, could also show if you're on a spam trap or not.
A Honeypot is a bare-bones system, usually with nothing more on it than the OS. It serves as bait for hackers.
I must disagree with Oyle. A Honeypot with nothing more on it than the bare-bones system is just a pot with no honey. So where's the bait? Unless the hacker is intending to use the honeypot as a zombie or stepping stone, they'll be gone in minutes if there's nothing there to keep them there. OK so you may have the logs to analyse how they got on to your Honeypot, but you won't be able to learn anything else.
I found a whitepaper on the Honeynet website called
Detecting Honeypots and other suspicious environments which may give some of you a more definative answer.
BTW, a quick perusal of the Honeynet projects Alumnii, shows a certain Edward Skoudis. A search for the word "Honeypot" on the SANS website comes up with the name Marcus J. Ranum numerous times. Both of these people are members/contribute either here at EH-Net or at CSP Mag. Perhaps Don can convince these experts to add some light on the matter.