Processing strongly reduced forms in casual speech