Error awareness and salience processing in the oddball task: shared neural mechanisms

Abstract

A body of work suggests similarities in the way we become aware of an error and process motivationally salient events. Yet, evidence for a shared neural mechanism has not been provided. A within subject investigation of the brain regions involved in error awareness and salience processing has not been reported. While the neural response to motivationally salient events is classically studied during target detection after longer target-to-target intervals in an oddball task and engages a widespread insula-thalamo-cortical brain network, error awareness has recently been linked to, most prominently, anterior insula cortex. Here we explore whether the anterior insula activation for error awareness is related to salience processing, by testing for activation overlap in subjects undergoing two different task settings. Using a within subjects design, we show activation overlap in six major brain areas during aware errors in an antisaccade task and during target detection after longer target-to-target intervals in an oddball task: anterior insula, anterior cingulate, supplementary motor area, thalamus, brainstem, and parietal lobe. Within subject analyses shows that the insula is engaged in both error awareness and the processing of salience, and that the anterior insula is more involved in both processes than the posterior insula. The results of a fine-grained spatial pattern overlap analysis between active clusters in the same subjects indicates that even if the anterior insula is activated for both error awareness and salience processing, the two types of processes might tend to activate non-identical neural ensembles on a finer-grained spatial level. Together, these outcomes suggest a similar functional phenomenon in the two different task settings. Error awareness and salience processing share a functional anatomy, with a tendency toward subregional dorsal and ventral specialization within the anterior insula.

Keywords: anterior insula; error awareness; eyetracking; magnetic resonance imaging; oddball processing; salience.

Figure 1

Oddball task. A series of non-target (0) and target (X) images was presented against a gray background. The target- and non-target ranged from 3 to 17 non-targets between targets. Participants were to react as quickly and accurately as possible to targets only by pressing a button with their index finger.

Figure 2

Antisaccade task: participants were instructed to fixate on a central target and generate an immediate eye movement away from an abrupt peripheral target to its mirror location on the opposite side of the screen without making an eye movement to the peripheral target itself. After the response a cross appeared in the correct square indicating the correct gaze direction. Participants were to evaluate their performance by pressing one of two buttons of an fMRI-compatible response box. An initial eye movement toward the peripheral target was classified as an error. The erroneous responses participants had rated as incorrect were classified as aware errors and erroneous responses rated as correct were classified as unaware errors.

Figure 3

(A) Aware error: statistical parametrical map of difference in BOLD activation between aware and unaware errors. Red and yellow voxels represent clusters of significant BOLD signal increase. (B) Salience processing: statistical parametrical map of difference in BOLD activation for the parametrical oddball. Red and yellow voxels represent clusters of significant BOLD signal increase. Renderings (on MNI stereotactic space) are thresholded at z = 2.3 and p = 0.05. (C) Overlap: plotted overlap between BOLD activation in the same subjects and the same scan session during aware errors and during the salience processing Note: R, right; ACC, anterior cingulate cortex; AIC, anterior insula cortex; FEF, frontal eyefields; S1, somatosensory cortex.

Figure 4

Spatial overlap map of clusters of activation on group-level that survived, within each participant’s native functional space, both the threshold for the awareness contrast and the threshold for salience processing. Analyses were we constrained by creating for each individual an “error awareness brain mask” within which activation was sensitive to salience processing. Renderings (on MNI stereotactic space) are thresholded at z = 2.3 and p = 0.05. Note: ACC, anterior cingulate cortex; SMA, supplementary motor area.

Figure 5

(A) Mean percent BOLD signal change within subjects across tasks (for the contrast aware errors as compared to unaware errors; and for the contrast of salience processing, i.e., linear signal increase across inter-target interval) in anterior insula and posterior insula (thresholded at z = 2.3 and p = 0.05). Participants showed during both error awareness and salience processing a significantly higher percent signal change in the anterior than in the posterior insula. The main effect of task indicated no differences in percent signal change in the insula between error awareness and the processing of motivationally significant events (B): plotted overlap in the anterior insula within subjects across tasks (for the contrast aware errors as compared to unaware errors; and for the contrast of salience processing, i.e., linear signal increase across inter-target interval).

Figure A1

Neural activation on aware errors. Statistical parametrical map of difference in BOLD activation between aware and unaware errors. Red and yellow voxels represent clusters of significant BOLD signal increase across all subjects. For a full list of activated regions (z > 2.3, whole-brain cluster-corrected, p < 0.05), see Table A1.

Figure A2

Illustration of brain areas showing increasing amplitude of the hemodynamic response to target stimuli with longer target interval. Target interval effects were found in numerous brain structures, including bilateral thalamus, bilateral anterior insula, dorsal anterior cingulate, supplementary motor area, dorsolateral prefrontal cortex, bilateral middle temporal gyri, bilateral pre- and postcentral gyri (somatosensory cortex), bilateral inferior and superior parietal lobules, parietal occipital junction, superior/middle and inferior frontal gyrus, precuneus, and bilateral cerebellum. The legend shows z-score value associated with the color map. The statistical parametric map has a threshold of z > 2.6; p < 0.05 (cluster-corrected). For a full list of activated regions, see Table A2.

Figure A3

Statistical parametrical map of hemodynamic response varying in each individual with aware errors and with the interval effect on target detection. Red and yellow voxels represent clusters of significant BOLD signal which passed the thresholding in the target interval contrast (ITI3–ITI1) and also survived thresholding in the awareness contrast (aware versus unaware errors). Four major brain areas were involved in both contrasts: bilateral thalamus, supplementary motor area, rostral cingulate, and in bilateral parietal lobule. Furthermore, overlapping activations were found in the precuneus and lateral occipital gyrus.