You can use multiple negative lookbehinds to do what you want. Although I imagine fellow monks will probably have a more elegant solution.

use strict;
use warnings;

for ( <DATA> )
{
        my $variable = qr/(?<!mild\s)(?<!moderate\s)active/;
        print $_ if ( m/$variable/ );
}

__DATA__
foo active
mild active
moderate active
active
[download]

Outputs

foo active
active
[download]

Dear Perl Monks - Thank you so much for your help!!!

The problem is that you have .? after the negative lookahead, which means the regex can: NOT match either of your two words with 'ild ' or 'oderate ', then ignore the . because it is optional and then match 'active'.

Now, you might think that just removing the ? would force it to work by matching the space (I did initially), but: when the preceding word is 'mild', it can match that by NOT matching the word moderate; and vice versa, so still every instance of active will match.

This works (though it will also (for example) not match 'elaborate active' should that appear):

my $re = qr[(?<!mild|rate)\s+active];
[download]

Negative lookbehinds have to be fixed length (hence the truncation of moderate).

Update: rather than risking mismatches caused by the truncation of the longer word, you could pad the shorter word:

my $re = qr[(?<!....mild|moderate)\s+active];
[download]

if you think that will be less likely to cause mismatches.

#!/usr/bin/perl
use warnings;
use strict;
use feature qw(say);


my @w = ('active', 'not active', 'mild active', 'moderate active');

for (@w) {
    my $r = reverse;
    say "$_: ", $r =~ /evitca(?! dlim| etaredom)/ ? 1 : 0, '.';
}
[download]

If you have another look at perlre you'll see that ?!(mild)) is a look ahead assertion and you need look behind ?<!(mild).

But, at least on my version of perl 5.16 the docs say

 "(?<!pattern)"
               A zero-width negative look-behind assertion.  For examp
+le "/(?<!bar)foo/" matches any 
               occurrence of "foo" that does not follow "bar".  Works 
+only for fixed-width look-behind.
[download]

So you probably need to try a different approach, Try doing your test in 2 phases 1. match lines with active and capture the previous word then 2. skip any line where that word is 'mild' or 'moderate'.

If you show a few example lines of your data then someone here may have a better suggestion.