Maybe it would be easier to express the distinction as spatial versus temporal scoping, which is an approach that I think makes sense at least until you start throwing threads and processes into the mix.

Let's take this example:

print "Just before start of scope\n";
{
  print "Just after start of scope\n";
  my $a = 1;
  local $b = 1;
  print "a=$a, b=$b\n";
  foo();
  print "Just before end of scope\n";
}
print "Just after end of scope\n";
[download]

Now, in this example, the line 'my $a = 1;' introduces a lexical variable. Any reference to $a in the statements following that introduction until the end of the scope refer to this newly introduced variable - and only those lines. This is a spatial scope: if the foo subroutine refers to a $a variable, it won't see this $a, because foo is not defined within this scope.

The line 'local $b = 1;' introduces a localisation - it tells the interpreter to save the current copy of the (package) variable $b, to be restored when execution reaches the end of the scope. Now it is a matter of time: after this localisation has been executed, anyone that looks at this package variable $b before the end of this scope is reached will see the localised version. If the foo subroutine refers to the package variable $b, it will see the localised version (for this call, at least).

Does that help at all?