Have a single authoritative source for your schema. Everyone should know where the official schema lives. You should be able to walk up to a fresh machine, pull from source control, build, and run a simple tool to set up the database. Ideally the build process handles it automatically.
Version your database. The goal is to propagate changes from development to test to production in a controlled way. You should also be able to recreate the database at any point in time. If someone reports a bug in build 20100612.1, you need to reproduce the exact database state from that build.