You are very close. Try making the regex look like this:

split(/(?<=\S)(?=\s)/);
[download]

That is, use both look-behind and look-ahead assertions, both zero-width. (update: made the above line coincide with the tested code below) I tried the following snippet on the command line to verify that:

$s="AAA    B   X Y Z  W\n";
@a=split(/(?<=\S)(?=\s)/, $s);
print length($s),":$s";
print join(":",length(join("",@a)),scalar(@a),"\n");
print join("\n",@a),"\n";
[download]

Output:

20:AAA    B   X Y Z  W
20:7:
AAA
    B
   X
 Y
 Z
  W
[download]

Note that this treats the final LF (or CRLF, if that's your flavor) as a token -- the second line of output shows that the array got seven elements, while the string had just six non-whitespace tokens followed by LF.