Normalizing vs de-normalizing

I have a legal database with a [Cases] table and a [Defendants] table and they are both linked together with a [CaseNumber] field.There are usually about 40 defendants or so for every case, and it's a "one to many relationship". Here's the deal though: Sometimes I need to pull up a query that says "show me all casenumbers where the defendant can be (Arco or Shell or Exxon) AND the casenumber MUST have defendant Union76."  

This is a bit awkward for SQL if the Defendants table and Cases table are separate. I, at first, thought about keeping the tables separate and doing a series of tables that list case numbers that build/exclude upon each other for ecah criteria, but this method seems slow performance-wise. Another method I thought of is to DENOMARMALIZE the database and simply list all defendants per case in a single long text field within the [Cases] table with each name separated by a token (in this case "~"). Then I can do all the "ands" and "ors" in that single text field without SQL awkwardness. Performance seems fast too.

Do you think this is the best method or is there a better way?

• Steve