In Kent Beck's book "Contributing to Eclipse" he writes:

The test-driven development cycle looks like this:

1. Write a test for the next bit of functionality you have in mind. The test should succeed only when the functionality has been implemented correctly.

2. Make the test compile by creating stubs for all the non-existing classes and methods referenced by the test.

3. Run the test. It should fail.

4. Implement just enough functionality to get the test to succeed.

5. Clean up the implementation as much as possible, typically by removing duplication.

I wondered, what does that look like?  So, I built a viewlet showing how to do test driven development using Eclipse.  You can find the tutorial at oneclipse.