Pick a database to learn
First understand joins - thoroughly. Do nothing else until you absolutlely understand joins. To do anything with a relational database without a thorough understanding of joins is like trying to learn to read without learning the alphabet.
Learn how to use a where clause
Read about indexing.
Learn about aggregate function and grouping data.
Now is the time to learn normalization (it will make much more sense if you understand how to query first).
Buy a big fat book on performance tuning for the database backend of your choice and read it cover to cover. You will find plenty of best practices and things to never do. You will also find that you currently do alot of the things in the never do list.
Get another big fat book on refactoring databases. Databases live much much longer than applications and are much harder to refactor if you don't think about how to do this from the start. Here's a good book to read cover to cover:
http://www.amazon.com/Refactoring-Databases-Evolutionary-Database-Design/dp/0321293533/ref=sr_1_1?ie=UTF8&s=books&qid=1268158669&sr=8-1
Databases need to be designed from the start for data integrity (never rely on the application for this, it is one of the top 10 stupid mistakes you can make in database design), performance (this is not premature, there are known performance killers to be avoided from the start that are much harder to refactor later) and security. They also need to be designed to be able to capture audit records and to do refactoring.
Have fun. If you get that far, then consider learning about data warehousing, reporting and ETL processes.