Puzzling NOT EXISTS Behavior

CASE CLOSED: removing the FROM clause noted below handles the problem ....I've used "NOT EXISTS" in inserts before and thought I understood how they work, but I'm puzzled why I get three rows from the following sub-query, which I want to use to prevent errors from duplicate keys. Since it gives me three rows, it will actually try to insert three duplicate records and cause a primary-key fault, which is a twin to the very thing I'm trying to avoid. (of course, there are three records already in the table, none having the key of 20050810)The sub-query is acting like it's using "WHERE DemoDate <> 20050810" instead of a "NOT EXISTS".Attempts to use this statement causes a termination with no records inserted.DemoDate is the primary key and an INT field in SQL Server. Four other int columns for this table have default values.<code>INSERT INTO DemoStats (DemoDate)SELECT     20050810 AS Expr1FROM         DemoStats         <-- remove this line so the effective table has only one row, when the NOT EXISTS is TRUEWHERE     (NOT EXISTS                          (SELECT     *                            FROM          DemoStats                            WHERE      DemoDate = 20050810))</code>