In a previous study on Evaluative Conditioning it was found that unlike signal-learning, evaluative learning is resistant to extinction. In the present experiment we tried to replicate and corroborate this finding. In addition, we investigated the effectiveness of a counterconditioning procedure to alter acquired evaluative value. On a within-subject base (n = 30), it was found that the mere contingent presentation of neutral with (dis)liked pictures of human faces was sufficient to change the affective-evaluative tone of the originally neutral stimuli in (negative) positive direction. Next, we replicated the finding that an extinction procedure does not have any influence on the acquired evaluative value of the originally neutral stimuli, and this even in subjects fully aware of the extinction presentations. We demonstrated the effectiveness of a counterconditioning procedure to alter acquired evaluative value, and in addition, found some evidence that awareness of the counterconditioning manipulations is not a prerequisite for its effectiveness. The theoretical consequences of these findings are phrased in terms of a fundamental distinction between signal-learning and evaluative learning, and some important implications for the behavioral treatment of affective-emotional disorders are suggested.