Depends on the frequency of particular nonalpha characters. If the
nonalphas were random with few if any multiple instances in any cell,
then your approach may be faster. But if there were multiple instances
of particular characters, getting rid of all of them (and shortening
the string in the process) could be faster.

There is nothing like knowing the data you are working with when designing
an algorithm.

Also, you don't need to insert " " for instances of " " in mystr, so
make the second Like pattern "[!A-Za-z ]".

LOL... Good catch... It always surprises me how easy it is to overlook the
obvious.

Rick