Kiyofumi Miyoshi

This study explores the use of response time (RT) data in type-2 receiver operating characteristic (ROC) analysis, a method traditionally used to examine the relationship between confidence and response correctness. Analyses of 16 perceptual decision-making datasets revealed that: (1) RT data contains roughly two-thirds as much information about response correctness as confidence; (2) RT carries unique predictive power for response correctness, independent of confidence; (3) RT and confidence interact synergistically, with confidence becoming a stronger predictor of response correctness when RT is short; (4) Despite these unique properties of RT, type-2 sensitivity (meta-d′) derived from RT and confidence showed a reasonably high correlation across subjects. These findings carry two key implications. First, in the absence of confidence data, RT can serve as a viable proxy, incurring minimal cognitive and logistical costs. This makes type-2 analysis feasible across various settings, potentially including infant and animal studies. Second, when both RT and confidence data are available, their combined use in type-2 analysis offers complementary insights into the processes underlying subjects’ behavior. To illustrate this, we present a simulation showing how the observed behavioral patterns align with the two-stage dynamic signal detection model. We propose that RT-based type-2 analysis is a valuable tool for researchers, helping to uncover previously underexplored aspects of decision-making behavior.

This study explores the use of response time (RT) data in type-2 receiver operating characteristic (ROC) analysis, a method traditionally used to examine the relationship between confidence and response correctness. Analyses of 16 perceptual decision-making datasets revealed that: (1) RT data contains roughly two-thirds as much information about response correctness as confidence; (2) RT carries unique predictive power for response correctness, independent of confidence; (3) RT and confidence interact synergistically, with confidence becoming a stronger predictor of response correctness when RT is short; (4) Despite these unique properties of RT, type-2 sensitivity (meta-d′) derived from RT and confidence showed a reasonably high correlation across subjects. These findings carry two key implications. First, in the absence of confidence data, RT can serve as a viable proxy, incurring minimal cognitive and logistical costs. This makes type-2 analysis feasible across various settings, potentially including infant and animal studies. Second, when both RT and confidence data are available, their combined use in type-2 analysis offers complementary insights into the processes underlying subjects’ behavior. To illustrate this, we present a simulation showing how the observed behavioral patterns align with the two-stage dynamic signal detection model. We propose that RT-based type-2 analysis is a valuable tool for researchers, helping to uncover previously underexplored aspects of decision-making behavior.

This study examines performance evaluation in perceptual detection tasks using response time (RT) data for signal detection theory (SDT) analysis. A defining feature of detection tasks is the asymmetry between trials with stimulus presence and absence, often reflected in asymmetric type-1 ROC curves. This asymmetry indicates greater signal variability in stimulus-present trials, which contradicts canonical assumptions in equal-variance SDT models. Across multiple datasets, we implemented an unequal-variance SDT model using RT data, and compared it with the traditional confidence-based method. RT-based estimates of SDT parameters, the SD ratio (σ) and mean difference (μ) between stimulus-present and stimulus-absent distributions, aligned closely with confidence-based estimates. The resulting sensitivity measure dₐ—an unequal-variance extension of d′—derived from RT and confidence showed strong consistency across all datasets. Notably, the conventional d′ measure consistently overestimated detection performance compared to the dₐ measures, highlighting the importance of accounting for unequal variance. RT-based SDT analysis offers a cost-effective alternative for robustly quantifying perceptual detection performance, particularly when confidence ratings are impractical or unavailable.

Everyday decision making often involves choosing between multiple alternatives, and developing unified theories for this process would benefit broad disciplines of behavioral science. A key challenge lies in explaining behavioral irrationalities that arise specifically in multialternative decisions. This study, based on preregistered procedures, investigated such nonnormative behaviors in a three-choice dot numerosity discrimination task. We systematically manipulated the value of the lowest numerosity alternative (the dud stimulus) and quantified its effects on choice, response time, and confidence. We then explained observed behaviors with a sequential evidence accumulation model featuring a "max versus next" decision algorithm. This pseudo-optimal decision-making rule naturally accounted for a well-known nonnormative choice pattern: the violation of the independence of irrelevant alternatives. Our model also explained a multichoice metacognitive bias, known as the dud-alternative confidence boost, by identifying two critical time windows in confidence constructions: Confidence is based on sensory evidence collected shortly after stimulus presentation and evidence gathered right after decision making. These findings represent a significant advancement in understanding the dynamics underlying multichoice irrationalities. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Everyday decision making often involves choosing between multiple alternatives, and developing unified theories for this process would benefit broad disciplines of behavioral science. A key challenge lies in explaining behavioral irrationalities that arise specifically in multialternative decisions. This study, based on preregistered procedures, investigated such nonnormative behaviors in a three-choice dot numerosity discrimination task. We systematically manipulated the value of the lowest numerosity alternative (the dud stimulus) and quantified its effects on choice, response time, and confidence. We then explained observed behaviors with a sequential evidence accumulation model featuring a “max versus next” decision algorithm. This pseudo-optimal decision-making rule naturally accounted for a well-known nonnormative choice pattern: the violation of the independence of irrelevant alternatives. Our model also explained a multichoice metacognitive bias, known as the dud-alternative confidence boost, by identifying two critical time windows in confidence constructions: Confidence is based on sensory evidence collected shortly after stimulus presentation and evidence gathered right after decision making. These findings represent a significant advancement in understanding the dynamics underlying multichoice irrationalities.

Humans perceive a vividly colored world coherently across the visual field, even though our peripheral vision has limited color sensitivity compared to central vision. How is this sense of color uniformity achieved? This question can be explored through a phenomenon called the pan-field color illusion, in which observers perceive scene images achromatized in the peripheral region (chimera images) as full-color images. Our previous work demonstrated that inattention to the peripheral visual field contributed to this illusion. Upon this observation, the current study presents signal detection theory modeling to quantify the effects of inattention on internal color signal strength and its variability. The model fitting has revealed that inattention to the peripheral region increases internal signal variability. This inattentional noise increases the occurrence of strong color signals that exceed the internal color detection criterion, thereby intensifying the pan-field color illusion. These results extend previously documented effects of inattentional noise on simple psychophysical tasks (e.g., “subjective inflation” in grating detection and discrimination), accounting for the illusorily enriched peripheral color experience for naturalistic scene images. Our findings suggest that inattentional noise inherent in peripheral vision would be a key factor in achieving the vivid visual phenomenology experienced across the entire visual field.

Human decision behavior entails a graded awareness of its certainty, known as a feeling of confidence. Until now, considerable interest has been paid to behavioral and computational dissociations of decision and confidence, which has raised an urgent need for measurement frameworks that can quantify the efficiency of confidence rating relative to decision accuracy (metacognitive efficiency). As a unique addition to such frameworks, we have developed a new signal detection theory paradigm utilizing the generalized Gaussian distribution (GGSDT). This framework evaluates the observer's metacognitive efficiency and internal standard deviation ratio through shape and scale parameters, respectively. The shape parameter quantifies the kurtosis of internal distributions and can practically be understood in reference to the proportion of the Gaussian ideal observer's confidence being disrupted with random guessing (metacognitive lapse rate). This interpretation holds largely irrespective of the contaminating effects of decision accuracy or operating characteristic asymmetry. Thus, the GGSDT enables hitherto unexplored research protocols (e.g., direct comparison of yes/no vs. forced-choice metacognitive efficiency), expected to find applications in various fields of behavioral science. This article provides a detailed walkthrough of the GGSDT analysis with an accompanying R package (ggsdt). (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Humans are not passive recipients of information from the external world. They are “metacognitive” agents, actively engaged in self-referential monitoring of internally expressed information. This introspective insight in turn enables proactive behavioral adjustments, facilitating efficient learning, preventing erroneous actions, and promoting information exchange with others. The study of metacognition has witnessed a recent expansion, incorporating a fusion of computational modeling, neuroimaging, machine learning techniques, and other interdisciplinary approaches. This chapter serves as an invitation for readers to explore the introspective realm of metacognition, providing an overview of the latest research findings.