Response Bias Reflects Individual Differences in Sensory Encoding

A hierarchical signal detection model with unequal variance for binary responses

Gaussian signal detection models with equal variance are commonly used in simple yes-no detection and discrimination tasks whereas more flexible models with unequal variance require additional information. Here, a hierarchical Bayesian model with equal variance is extended to an unequal-variance model by exploiting variability of hit and false-alarm rates in a random sample of participants. This hierarchical model is investigated analytically, in simulations and in applications to existing data sets. The results suggest that signal variance and other parameters can be accurately estimated if plausible assumptions are met. It is concluded that the model provides a promising alternative to the ubiquitous equal-variance model for binary data.

Reinforcement learning profiles and negative symptoms across chronic and clinical high-risk phases of psychotic illness

Negative symptoms are prominent in individuals with schizophrenia (SZ) and youth at clinical high-risk for psychosis (CHR). In SZ, negative symptoms are linked to reinforcement learning (RL) dysfunction; however, previous research suggests implicit RL remains intact. It is unknown whether implicit RL is preserved in the CHR phase where negative symptom mechanisms are unclear, knowledge of which may assist in developing early identification and prevention methods. Participants from two studies completed an implicit RL task: Study 1 included 53 SZ individuals and 54 healthy controls (HC); Study 2 included 26 CHR youth and 23 HCs. Bias trajectories reflecting implicit RL were compared between groups and correlations with negative symptoms were examined. Cluster analysis investigated RL profiles across the combined samples. Implicit RL was comparable between HC and their corresponding SZ and CHR groups. However, cluster analysis was able to parse performance heterogeneity across diagnostic boundaries into two distinct RL profiles: a Positive/Early Learning cluster (65% of participants) with positive bias scores increasing from the first to second task block, and a Negative/Late Learning cluster (35% of participants) with negative bias scores increasing from the second to third block. Clusters did not differ in the proportion of CHR vs. SZ cases; however, the Negative/Late Learning cluster had more severe negative symptoms. Although implicit RL is intact in CHR similar to SZ, distinct implicit RL phenotypic profiles with elevated negative symptoms were identified trans-phasically, suggesting distinct reward-processing mechanisms can contribute to negative symptoms independent of phases of illness.

Quantifying the contribution of subject and group factors in brain activation

Research in neuroscience often assumes universal neural mechanisms, but increasing evidence points toward sizeable individual differences in brain activations. What remains unclear is the extent of the idiosyncrasy and whether different types of analyses are associated with different levels of idiosyncrasy. Here we develop a new method for addressing these questions. The method consists of computing the within-subject reliability and subject-to-group similarity of brain activations and submitting these values to a computational model that quantifies the relative strength of group- and subject-level factors. We apply this method to a perceptual decision-making task (n = 50) and find that activations related to task, reaction time, and confidence are influenced equally strongly by group- and subject-level factors. Both group- and subject-level factors are dwarfed by a noise factor, though higher levels of smoothing increases their contributions relative to noise. Overall, our method allows for the quantification of group- and subject-level factors of brain activations and thus provides a more detailed understanding of the idiosyncrasy levels in brain activations.

Revealing the influence of bias in a letter acuity identification task: A noisy template model

In clinical testing of visual acuity, it is often assumed that performance reflects sensory abilities and observers do not exhibit strong biases for or against specific letters, but this assumption has not been extensively tested. We re-analyzed single-letter identification data as a function of letter size, spanning the resolution threshold, for 10 Sloan letters at central and paracentral visual field locations. Individual observers showed consistent letter biases across letter sizes. Preferred letters were named much more often and others less often than expected (group averages ranged from 4% to 20% across letters, where the unbiased rate was 10%). In the framework of signal detection theory, we devised a noisy template model to distinguish biases from differences in sensitivity. When bias varied across letter templates the model fitted very well - much better than when sensitivity varied without bias. The best model combined both, having substantial biases and small variations in sensitivity across letters. The over- and under-calling decreased at larger letter sizes, but this was well-predicted by template responses that had the same additive bias for all letter sizes: with stronger inputs (larger letters) there was less opportunity for bias to influence which template gave the biggest response. The neural basis for such letter bias is not known, but a plausible candidate is the letter-recognition machinery of the left temporal lobe. Future work could assess whether such biases affect clinical measures of visual performance. Our analyses so far suggest very small effects in most settings.

Re-evaluating frontopolar and temporoparietal contributions to detection and discrimination confidence

Previously, we identified a subset of regions where the relation between decision confidence and univariate functional magnetic resonance imaging (fMRI) activity was quadratic, with stronger activation for both high and low compared with intermediate levels of confidence. We further showed that, in a subset of these regions, this quadratic modulation appeared only for confidence in detection decisions about the presence or absence of a stimulus, and not for confidence in discrimination decisions about stimulus identity (Mazor et al. 2021). Here, in a pre-registered follow-up experiment, we sought to replicate our original findings and identify the origins of putative detection-specific confidence signals by introducing a novel asymmetric-discrimination condition. The new condition required discriminating two alternatives but was engineered such that the distribution of perceptual evidence was asymmetric, just as in yes/no detection. We successfully replicated the quadratic modulation of subjective confidence in prefrontal, parietal and temporal cortices. However, in contrast with our original report, this quadratic effect was similar in detection and discrimination responses, but stronger in the novel asymmetric-discrimination condition. We interpret our findings as weighing against the detection-specificity of confidence signatures and speculate about possible alternative origins of a quadratic modulation of decision confidence.

Examining the robustness of the relationship between metacognitive efficiency and metacognitive bias

We recently found a positive relationship between estimates of metacognitive efficiency and metacognitive bias. However, this relationship was only examined on a within-subject level and required binarizing the confidence scale, a technique that introduces methodological difficulties. Here we examined the robustness of the positive relationship between estimates of metacognitive efficiency and metacognitive bias by conducting two different types of analyses. First, we developed a new within-subject analysis technique where the original n-point confidence scale is transformed into two different (n-1)-point scales in a way that mimics a naturalistic change in confidence. Second, we examined the across-subject correlation between metacognitive efficiency and metacognitive bias. Importantly, for both types of analyses, we not only established the direction of the effect but also computed effect sizes. We applied both techniques to the data from three tasks from the Confidence Database (N > 400 in each). We found that both approaches revealed a small to medium positive relationship between metacognitive efficiency and metacognitive bias. These results demonstrate that the positive relationship between metacognitive efficiency and metacognitive bias is robust across several analysis techniques and datasets, and have important implications for future research.