Disentangling the origins of confidence in speeded perceptual judgments through multimodal imaging

Abstract

<jats:title>Abstract</jats:title><jats:p>The human capacity to compute the likelihood that a decision is correct - known as metacognition - has proven difficult to study in isolation as it usually co-occurs with decision-making. Here, we isolated post-decisional from decisional contributions to metacognition by combining a novel paradigm with multimodal imaging. Healthy volunteers reported their confidence in the accuracy of decisions they made or decisions they observed. We found better metacognitive performance for committed vs. observed decisions, indicating that committing to a decision informs confidence. Relying on concurrent electroencephalography and hemodynamic recordings, we found a common correlate of confidence following committed and observed decisions in the inferior frontal gyrus, and a dissociation in the anterior prefrontal cortex and anterior insula. We discuss these results in light of decisional and post-decisional accounts of confidence, and propose a generative model of confidence in which metacognitive performance naturally improves when evidence accumulation is constrained upon committing a decision.</jats:p>

Response-related signals increase confidence but not metacognitive performance

“…The past years have seen a growing interest in elucidating the sources of information that contribute to confidence judgments, as a window into potential computational processes that allow the brain to monitor itself. Converging evidence from very different experimental paradigms suggested that confidence is modulated by motor information concurrent with the first-order response (Faivre et al, 2017;Fleming et al, 2015;Patel et al, 2012;Pereira et al, 2018); for a review see (Anzulewicz, Hobot, Siedlecka, & Wierzchoń, 2019). Here, we set out to directly investigate this possibility.…”

“…Other studies have addressed the same question by using different experimental manipulations, that can be broadly grouped as following one of three approaches. A first set of studies have asked participants to rate confidence of observed, rather than committed, actions, whilst letting participants observe only first-order RTs (Patel et al, 2012;Vuillaume, Martin, Sackur, & Cleeremans, 2019) or both RTs and stimuli (Pereira et al, 2018) before making the confidence judgement. A second group of studies have instead manipulated the timing of the confidence judgement relative to that of the first-order response (Siedlecka, Paulewicz, & Wierzchoń, 2016;Wokke, Achoui, & Cleeremans, 2019).…”

“…Other experimental manipulations produced the converse effect, namely a modulation of metacognitive performance with no change in confidence bias. This includes the comparison of confidence in committed vs. observed decisions (Pereira et al, 2018), and confidence under high or low sensorimotor conflicts . Using a similar design comparing prospective and retrospective confidence judgments, Siedlecka and colleagues (Siedlecka et al, 2016) found that both confidence bias and metacognitive performance increased in presence of actionrelated signals.…”

“…Beyond behaviour alone, correlational analyses showed that confidence is higher in the presence of subthreshold motor activity prior to first-order responses (Gajdos, Fleming, Saez Garcia, Weindel, & Davranche, 2019). Using electroencephalography, we recently showed that a signature of motor preparation prior to first-order response correlates with confidence over different perceptual tasks (Faivre et al, 2017) and that metacognitive performance for decisions that are committed with a key press is better than that to equivalent decisions that are observed (Pereira et al, 2018). A drift-diffusion model explained this effect by showing that the accumulation of perceptual evidence is constrained by firstorder decisions.…”

Confidence in perceptual decision-making is preserved in schizophrenia

“…Namely, we continuously tracked the position and kinematics of the mouse that participants used to indicate their first-order response during a motion discrimination task (Dotan, Meyniel, & Dehaene, 2018;Dotan, Pinheiro-Chagas, Al Roumi, & Dehaene, 2019) . In addition, we modeled first and second-order responses as derived from a bounded evidence accumulation process starting when participants initiated a mouse movement (Pereira et al, 2018;Pleskac, Busemeyer, & others, 2010;Resulaj, Kiani, Wolpert, & Shadlen, 2009;Van Den Berg et al, 2016) . Together, these two approaches following a pre-registered plan allowed us to finely characterize decision-making and metacognitive monitoring in schizophrenia in relation to clinical traits while avoiding the typical confounds that may have contaminated previous results in the field.…”