This study is entirely useless because they invalidated themselves from the start by providing the worst possible Perl examples they could. I quote the salient parts:

"participants were given a code sample worksheet for the particular language group they were in. The general idea of the experiment is to give novice users code samples similarly to if a participant was learning to program from home on the Internet.

This code shows one of the code samples provided to participants."

    $x = &z(1, 100, 3);

    sub z{
        $a = $ [0];
        $b = $ [1];
        $c = $ [2];
        $d = 0.0;
        $e = 0.0;
        for ($i = $a; $i <= $b; $i++){
            if ($i % $c == 0) {
                $d = $d + 1;
            }
            else {
                $e = $e + 1;
            }
        }
        if ($d > $e) {
            $d;
        }
        else {
            $e;
        }
    }

This ridiculous thing was supposed to help their test subjects learn Perl.

If all their perl samples looked like that it should not be a surprise to anyone that their test subjects were entirely confounded. What is missing in that paper is a description of how they actually sourced their code samples, as well as a list of all code samples provided.