Single-Trial Event-Related Potential Correlates of Belief Updating

Abstract

Belief updating—the process by which an agent alters an internal model of its environment—is a core function of the CNS. Recent theory has proposed broad principles by which belief updating might operate, but more precise details of its implementation in the human brain remain unclear. In order to address this question, we studied how two components of the human event-related potential encoded different aspects of belief updating. Participants completed a novel perceptual learning task while electroencephalography was recorded. Participants learned the mapping between the contrast of a dynamic visual stimulus and a monetary reward and updated their beliefs about a target contrast on each trial. A Bayesian computational model was formulated to estimate belief states at each trial and was used to quantify the following two variables: belief update size and belief uncertainty. Robust single-trial regression was used to assess how these model-derived variables were related to the amplitudes of the P3 and the stimulus-preceding negativity (SPN), respectively. Results showed a positive relationship between belief update size and P3 amplitude at one fronto-central electrode, and a negative relationship between SPN amplitude and belief uncertainty at a left central and a right parietal electrode. These results provide evidence that belief update size and belief uncertainty have distinct neural signatures that can be tracked in single trials in specific ERP components. This, in turn, provides evidence that the cognitive mechanisms underlying belief updating in humans can be described well within a Bayesian framework.

P300 amplitude variations, prior probabilities, and likelihoods: A Bayesian ERP study

“…These broad anatomical distinctions are in accordance with our findings if one assumes that frontal sources contribute to the generation of scalprecorded P3a waves and that parietal sources contribute to the generation of scalp-recorded P3b waves (Polich, 2007). The P3a finding needs further specification because previous ERP studies revealed an association between Bayesian updating and P3a amplitudes rather than between prior probabilities and P3a amplitudes (Bennett et al, 2015;Kolossa et al, 2015). The P3a waves in this study were more anteriorly distributed than the P3a waves in the previous studies, but this issue clearly needs to be further examined by future studies.…”

“…Kolossa, Fingscheidt, Wessel, and Kopp (2013) modeled the surprise-sensitivity of trial-by-trial P3b amplitude variations as originating from the operation of multiple digital filters in the brain (i.e., two parallel first-order infinite impulse response low-pass filters and an additional fourth-order finite impulse response high-pass filter). Kolossa, Kopp, and Fingscheidt (2015) replicated the surprise-sensitivity of P3b amplitude variations and showed that the anteriorly distributed P3a encodes Bayesian updating, a finding that was also reported by Bennett, Murawski, and Bode (2015).…”

Monetary feedback modulates performance and electrophysiological indices of belief updating in reward learning

“…A simple Bayesian model provided dynamic estimates of the belief states of each participant on each trial, which allowed for trial‐by‐trial quantification of (a) the uncertainty of beliefs, and (b) the magnitude of belief updates. Importantly, these estimates of belief state were found to correlate with the amplitude of components of the ERP, including a positive relationship between P3 amplitudes and belief update magnitude that was not accounted for by simpler models (Bennett et al, ). This finding was consistent with earlier theories and other empirical findings linking the P3 and Bayesian belief updating (Kolossa et al, ; Kopp, ).…”

“…This suggests that discrepancies between Bayesian models and human behavior may be in part motivational: since full Bayesian inference can be computationally demanding, participants' motivation to engage in effortful cognition may moderate their use of strategies that resemble Bayesian belief updating (Lieder & Griffiths, ; Shenhav, Botvinick, & Cohen, ). Since an important factor that directly modulates motivation is the availability of reward, this motivational account predicts that the use of monetary performance incentives in learning tasks (as in the study by Bennett et al, ) may affect the behavioral strategies employed by participants. Monetary incentive has been linked with improved performance in various learning paradigms (e.g., Bonner & Sprinkle, ; Kleih, Nijboer, Halder, & Kübler, ), but the effect of monetary incentives on participants' use of Bayesian belief updating is an open question.…”

Testosterone and Cortisol Jointly Predict the Ambiguity Premium in an Ellsberg-Urns Experiment

“…Afterward, we estimated the normalized entropy of the individual histograms, according to Shannon's formula (Shannon, 1948; Shannon and Weaver, 1949; cf. also Bennett et al, 2015): $\begin{array}{c}left\mathrm{\text{H }}=-\frac{{\sum }_{i=1}^{11}true(p_i\times {log}_2{p}_{i}true)}{{log}_{2}11}\end{array}$ where p i is the relative frequency at bin i . The normalized entropy reflects the degree of belief uncertainty regarding the bag composition.…”