The present paper tests the widely accepted hypothesis that on-line judgment implies functional independence between memory for, and judgment of, verbal stimuli (e.g., Anderson, 1989; Hastie & Park, 1986). In the present study, participants recalled lists of words, after having assessed each for its pleasantness. Presentation position of a negative item within the lists was manipulated. Also, items memorability was manipulated after their presentation – by inserting a filled delay between presentation and the judgment task; in this way, on-line judgment formation was spared. The memory manipulation reduced recall rates for negative items presented in the last position – and their negative influence on pleasantness ratings accordingly. These results contradict the predictions of pure on-line approaches to judgment formation (e.g., Betsch, Plessner, Schwieren, & Gütig, 2001) and suggest that even in on-line judgment tasks, memory plays a role.