I always knew you could specify positions in ORDER BY, but I did not know about GROUP BY. This is very useful for exploring data. Thanks. I often have long complicated expressions for GROUP BY. I vaguely recall there is a reason why this is not implemented in, say, Oracle, for example, that there is some ambiguity that arises, but thinking about it now I cannot see any problem with it. By experiment, it appears I can also use a column alias in the GROUP BY that is defined in the SELECT. Again, I vaguely recall there is a reason why this is not implemented in many SQL engines. In this case, I am pretty sure ambiguities can arise. It's a question of scope. For example, I can swap the names of two columns using something like: SELECT A AS B, B AS A FROM .....; In this case, does GROUP BY A refer to the original column name or the alias? In the case of ORDER BY, I think the alias has precedence. Anyway, the position notation should be documented in GROUP BY. Should I try to write an edit for the documentation?