If your underlying database supports sub queries, you can build a query with a TOP N that returns ObjectIds (or any other field) and use that in your definition query.
OBJECTID IN (*query here*)
For example, here's a SQL Server query for a TOP 20, random selection:
The above query will return 20 different rows each time the map or attribute table is refreshed. This also could come in handy for an ArcMap practical joke for less experienced users (April Fools is less than a month away).
Of course, the sub query doesn't have to be a TOP N query nor do you need to use a unique id. Maybe work up a query to find duplicates:
FIRE_ID IN (SELECT FIRE_ID
GROUP BY FIRE_ID
HAVING COUNT(FIRE_ID) > 1)