The neural underpinnings of how reward associations can both guide and misguide attention

Abstract

It is commonly accepted that reward is an effective motivator of behavior, but little is known about potential costs resulting from reward associations. Here, we used functional magnetic resonance imaging (fMRI) to investigate the neural underpinnings of such reward-related performance-disrupting effects in a reward-modulated Stroop task in humans. While reward associations in the task-relevant dimension (i.e., ink color) facilitated performance, behavioral detriments were found when the task-irrelevant dimension (i.e., word meaning) implicitly referred to reward-predictive ink colors. Neurally, only relevant reward associations invoked a typical reward-anticipation response in the nucleus accumbens (NAcc), which was in turn predictive of behavioral facilitation. In contrast, irrelevant reward associations increased activity in a medial prefrontal motor-control-related region, namely the presupplementary motor area (pre-SMA), which likely reflects the preemption and inhibition of automatic response tendencies that are amplified by irrelevant reward-related words. This view was further supported by a positive relationship between pre-SMA activity and pronounced response slowing in trials containing reward-related as compared with reward-unrelated incongruent words. Importantly, the distinct neural processes related to the beneficial and detrimental behavioral effects of reward associations appeared to arise from preferential-coding mechanisms in visual-processing areas that were shared by the two stimulus dimensions, suggesting a transfer of reward-related saliency to the irrelevant dimension, but with highly differential behavioral and neural ramifications. More generally, the data demonstrate that even entirely irrelevant reward associations can influence stimulus-processing and response-selection pathways relatively automatically, thereby representing an important flipside of reward-driven performance enhancements.

Figure 1.

Experimental paradigm and behavioral results. A, Colored words were presented for 600 ms each on a black background, separated by a stimulus onset asynchrony of 1.5–6 s. Participants responded to the word's ink color (task-relevant dimension) while ignoring its semantic meaning (task-irrelevant dimension). B, A subset of ink colors and the respective responses were associated with potential-reward (termed Color$; e.g., green/blue), and the remaining colors were not (termed Color0; e.g., red/yellow). Accordingly, task-irrelevant word meanings could not only be congruent, incongruent-ineligible, or fully incongruent to the ink color, but the latter word category could moreover implicitly refer to the potential-reward (Wi$) or the no-reward (Wi0) ink-color subset. C, Relative RT differences are depicted for all incongruent word categories compared with the congruent ones within both ink-color subsets (Color$: turquoise bars; Color0: orange bars). Error bars represent the SEM for the difference values; asterisks indicate significant paired t tests (***p < 0.001; **p < 0.01; *p < 0.05; above single bar: difference relative to congruent words; between bars: difference between two word categories).

Figure 2.

Cortical activity associated with relevant and irrelevant reward. A, The actual prospect of reward, defined by comparing relevant potential-reward with no-reward trials, engaged various prefrontal regions, including right dlPFC and inferior frontal junction, right IFG, and lateral frontopolar cortex (lf polar), as well as bilateral FG and inferior parietal cortex (IPC). B, Irrelevant reward information, defined by comparing incongruent reward-related with incongruent reward-unrelated words in no-reward trials, was associated with increased activity in dlPFC and FG, as well as in the right middle temporal gyrus (MTG). C, Neural activity associated with relevant reward (white) and irrelevant reward (black) is overlaid to illustrate the activity overlap (gray) within right dlPFC and bilateral FG (display cutoff p < 0.001).

Figure 3.

Differential activity patterns related to relevant and irrelevant reward associations. The NAcc was activated by relevant reward (A), whereas the pre-SMA was activated by irrelevant reward (B). ROI analyses of the extracted signal-change values confirmed that these regions were insensitive to the respective opposite type of reward information as indicated by ns (nonsignificant comparison). Asterisks indicate the reiterations of the voxelwise results (significant at p < 0.01). C, Furthermore, the NAcc activity increase that was selective for relevant reward (depicted as difference between Color$ and Color0) was correlated with response acceleration (depicted as RT difference between Color$ and Color0). D, In contrast, increased pre-SMA activity related to reward-related words (depicted as differences between Color0_Wi$ and Color0_Wi0) was associated with response slowing (depicted as RT difference between Color0_Wi$ and Color0_Wi0).