with all the input from people like you this has become quite an interesting
thread. It takes me back to the thousands of hours I spent programming for
a bank, and looking up the minute details of certain functions and running
dozens of test routines to check behaviour of functions. It's really quite
amazing how when you push something to its limits you find all kinds of
quirky things.

"Harlan Grove" wrote in message
...
"Tom Ogilvy" wrote...
Precedence of operations is stated in the help. As far as I know, there
is
no universally accepted order of precedence that covers all combinations
and
permutations.
...

Somewhat of a tangent. APL and its descendants (APL2, J and K) have
sepaate
tokens for numeric sign and minus operator. The numeric sign character is
part of the number token, so effectively always binds tighter than any
other
token or operator. Evaluation is always right to left, so different
results
are produced when using the sign character vs the minus operator followed
by
a number raised to an even power. Excel treats unary minus like APL's sign
character rather than like APL's unary ('monadic') minus. This is unlike
most other programming languages that use the same character to represent
signs in numeric tokens and unary and 'diadic' minus.