There are four main methods of killing spam. They can be classed as heuristic, bayesian, dictionary, and fingerprinting.
Sorry Maureen! Sometimes when its really early in the morning my fingers fail to find their correct resting places, you know, asdf and jkl; on the keyboard.
What I'm trying to say is that its a typo ;)
Heuristic scans create a profile of an email message from its headers and other core attributes to rate its likelihood of being spam. Bayesian filters use a statistical approach whereby the filtering system is trained to distinguish between spam and legitimate email using an algorithm. Dictionary scans are used to filter against particular words and phrases in the headers or body of an email. You know those words... the ones in the sometimes funny subject lines which can cause embarassment at work. Finally, email fingerprinting is used to create a hash uniquely representing known spam messages, which is a reactive rather than a predictive technique.
Instead of using a fixed set of virus definition files and known spammers blocklists which must be updated, advanced heuristic spam (and virus) detection may explore the message in a "lets try it and see" sandbox. Leting the program do what its going to do inside a "quarantine" environment it can clearly detect wheter it will behave badly if let loose on the rest of your system, without the need to have the most current definition of the virus or spam message. With spam, it explores other things besides executables like third party links, web bugs, malformed headers, spoofed addresses, spoofed URLs, etc.

