Reply

The authors report two studies. In one, they used the standard version of our gambling task and replicated our results. In the other, they used a new task, with a superficial resemblance to the original, but with substantial conceptual differences. They suggest that their results are not compatible with our interpretation of the original gambling task or with the somatic marker hypothesis. We welcome their question but disagree.

The task in their second experiment involves two decks of cards (A & B) for which both rewards and punishments are high, but in which the rewards always outweigh the punishments; playing from these decks offers both immediate and long-term gains. (A & B are 'good decks'.) Decks C & D have smaller rewards and punishments, but still the punishments are four times higher than the rewards; playing from these decks gives low immediate reward and long-term loss ('bad decks'). Compared to the original task, decks A & B are more readily recognizable as preferable (immediate reward is 10 times the value of that in C & D) and ultimately advantageous. (Penalties never cancel the gain, as in decks C & D.) The immediate tendency to prefer the high reward does not need to be opposed in order to achieve. Apparent and ultimate goodness coincide. There is no conflict. Normal subjects should prefer decks A & B.

In the original task, the higher anticipatory SCRs preceded card turns from bad decks; by contrast, in the modified task, higher anticipatory SCRs preceded turns from good decks. Because higher anticipatory SCRs related to decks carrying the immediate higher magnitude of reward or punishment, the authors argue that “...anticipatory SCRs are driven by the immediate act to be performed” and are irrelevant to the final outcome. We agree that SCRs index a process driven by the immediate act to be performed, but this does not mean that anticipatory SCRs are unrelated to long-term decisions.

The authors believe that if somatic markers drive the evaluation of the goodness or badness of the decks, then higher anticipatory SCRs should precede picking from bad decks (decks C & D) in the modified task. They assume that somatic markers can only be negative and only precede options leading to a bad outcome. However, somatic markers can be either positive or negative, and under conflict and uncertainty they help reject or endorse an option of action. This suggests a possible interpretation of the authors' results, namely that their task inverts the marker signal. In the original task, the higher anticipatory SCRs relative to bad decks reflected a negative somatic state that promoted avoidance of bad options. In the modified task, higher anticipatory SCRs to good decks may reflect a positive somatic state that promotes approach. Our own work8,9 with another modified gambling task that preserved conflict and uncertainty, but switched reward and punishment, revealed approach behavior coupled with high anticipatory SCRs. Also, patients with ventromedial prefrontal cortex lesions (which preclude the development of these anticipatory SCRs) performed disadvantageously on that modified task, just as they did on the original task. Although higher anticipatory SCRs probably relate to the magnitudes of reward or punishment hidden in the deck from which subjects are about to select, depending on whether anticipatory SCRs reflect negative or positive somatic states, higher anticipatory SCRs also coincide with the long-term consequences—anticipation of a long-term negative or positive outcome. When anticipatory SCRs do not develop, a support mechanism for making advantageous decisions under conflict and uncertainty falls apart, as was critically demonstrated in patients with prefrontal damage6.

Another explanation for the finding would be that high-magnitude anticipatory SCRs before good decks reflect a non-conscious danger signal related to the likely risk of a large penalty. Given the modified task's design, such signals would not influence behavior because the conscious assessment of the overall goodness of the decks would prevail. Either interpretation is in accord with the somatic marker hypothesis. Somatic markers assist decision-making, covertly or overtly, but are not engaged in every decision, and their 'advice' need not be heeded. We caution against the idea that emotion-based signals 'decide' for us, other than in extreme situations.

See "Do somatic markers mediate decisions on the gambling task?" by Tomb, Hauser, Deldin, and Caramazza.