Unraveling the influence of trial-based motivational changes on performance monitoring stages in a flanker task

Figure 1. (a) Schematic depiction of the Monetary Incentive Flanker Task. Participants were instructed to respond with the left or right button according to the direction of the middle arrow. In the potential gain (PG) context, the 80% fastest correct responses were rewarded (1a), in the loss avoidance context, the 80% fastest correct responses were not punished (2a). Incorrect and the 20% slowest correct responses were either not rewarded (1b), or punished (2b). Logistic regression on accuracy (b) and multiple single-trial regressions on RT (c) were used to assess behavioral effects in the task, controlling for confounding variables and the interdependence of effects. Interference effects are reduced after errors with regard to accuracy (d). Motivational context predicted RT, in that the loss context increased accuracy, as participants only showed post-error increase in accuracy (PIA) within the PG context (e), as well as RT (f). Stimulus incongruence decreased accuracy and increased RT, and interacted with accuracy (g). b, c depict mean within participant t-values, p-values are derived from t-tests of individual regression weights against zero and were Bonferroni corrected. RT is…

Figure 2. (a) Stylized illustration of the winning multi-stage drift diffusion model (DDM) based on simulated data. Depicted are three decision processes: incongruent correct trials (Rinc, green decision process), congruent correct trials (Rcon, blue decision process), and incongruent error trials (Rerr, yellow decision process). The DDM consisted of five separate stages (depicted as gray bars underneath) and included nine free parameters: drift rate (v), variance in drift rates (sv), boundary (a), variance in start points (sz), non-decision time (Ter, reflecting visual processing and motor execution times), variance in Ter (st), bounds collapse (k), flanker weighting (f), and variance in f (sf). Stage 1 represents a pre-stimulus baseline. On each trial the decision process starts with a random start point sz, which is drawn from a uniform distribution. Stage 2 represents the non-decision time (Ter) per trial. Ter varies on each trial depending on st, which is also modelled as a uniform distribution. In this period the decision process randomly drifts away from the start point. Stage 3 reflects a noisy diffusion in flanker direction (sv), which varies from trial-to-trial (sf). Both are…

Figure 3. Time course of stimulus-locked regression effects for significant regressors congruency (incongruent, congruent) and context (loss, gain). The first and third row represent topographical maps of the associations between EEG activity and the respective regressor (first row: congruency; third row: motivational context). Grand average event-related potential (ERP) waveforms at electrodes FCz and Pz are depicted in the second (regressor: congruency; blue lines indicate incongruent trials, orange lines indicate congruent trials) and fourth row (regressor: context; blue lines indicate loss avoidance trials, orange lines indicate potential gain trials), stimulus-locked (at 0 ms) for correct trials. Shades indicate the SEM. Gray shading behind the waveforms indicates significance of regression weights at p < 2.2654e−4 (FCz) and p < 3.2189e−4 (Pz). Note that regular ERPs do not account for other task factors.

Figure 4. Time course of response-locked regression effects for significant regressor accuracy (correct, error). The first row represents topographical maps of the associations between EEG activity and the accuracy regressor. Grand average event-related potential (ERP) waveforms at electrodes FCz and Pz are depicted in the second row, response-locked for incongruent trials. Shades indicate the SEM. Gray shading behind the waveforms indicates significance of regression weights at p < 2.8827e−4 (FCz) and p < 3.4804e−4 (Pz). Note that regular ERPs do not account for other task factors.

Figure 5. Time course of feedback-locked regression effects for significant regressors feedback (positive, negative), context (loss, gain) and the interaction of feedback and context. The first, third and fourth row represent topographical maps of the associations between EEG activity and the respective regressor. Grand average event-related potential (ERP) waveforms of the feedback regressor at electrodes FCz and Pz are depicted in the second row, feedback-locked for correct trials. Shades indicate the SEM. Gray shading behind the waveforms indicates significance of regression weights at p < 3.7708e−4 (FCz) and p < 2.7407e−4 (Pz). Note that regular ERPs do not account for other task factors.

Figure 6. Time course of feedback-locked regression effects for significant regressor context (loss, gain) within correct trials with negative feedback and correct trials with positive feedback. The first and third row represent topographical maps of the associations between EEG activity and the context regressor within the respective trial type. Grand average event-related potential (ERP) waveforms at electrodes FCz and Pz are depicted in the second and fourth row, feedback-locked for correct trials with negative or positive feedback. Shades indicate the SEM. Gray shading behind the waveforms indicates significance of regression weights at p < 2.2664e-4 (FCz) and p < 2.1000e-5 (Pz) for negative feedback, and p < 1.9288e-4 (FCz) and p < 4.2192e-4 (Pz) for positive feedback. Note that regular ERPs do not account for other task factors.

Figure 7. Difference in feedback-locked time–frequency power (in dB) between contexts (potential gain minus loss avoidance) within correct trials with negative feedback and correct trials with positive feedback. Time courses of feedback-locked time–frequency power within correct trials with negative feedback (left column) and correct trials with positive feedback (right column) for the difference between potential gain and loss avoidance context are depicted for electrodes FCz and Pz. Black lines indicate areas of significant difference at p < 4.6630e−4 (FCz) and p < 1.1885e−7 (Pz) for negative feedback and at p < 1.9453e−4 (FCz) and p < 6.6510e−5 (Pz) for positive feedback. Topographical maps for theta band power difference (4–8 Hz) between 250 to 300 ms and delta band power difference (1–4 Hz) between 400 to 600 ms are depicted in the bottom panels.