Neurophysiology of Human Perceptual Decision-Making.

The discovery of neural signals that reflect the dynamics of perceptual decision formation has had a considerable impact. Not only do such signals enable detailed investigations of the neural implementation of the decision-making process but they also can expose key elements of the brain's decision algorithms. For a long time, such signals were only accessible through direct animal brain recordings, and progress in human neuroscience was hampered by the limitations of noninvasive recording techniques. However, recent methodological advances are increasingly enabling the study of human brain signals that finely trace the dynamics of the unfolding decision process. In this review, we highlight how human neurophysiological data are now being leveraged to furnish new insights into the multiple processing levels involved in forming decisions, to inform the construction and evaluation of mathematical models that can explain intra- and interindividual differences, and to examine how key ancillary processes interact with core decision circuits.

Locus Coeruleus Integrity Is Linked to Response Inhibition Deficits in Parkinson's Disease and Progressive Supranuclear Palsy.

Parkinson's disease (PD) and progressive supranuclear palsy (PSP) both impair response inhibition, exacerbating impulsivity. Inhibitory control deficits vary across individuals and are linked with worse prognosis, and lack improvement on dopaminergic therapy. Motor and cognitive control are associated with noradrenergic innervation of the cortex, arising from the locus coeruleus (LC) noradrenergic system. Here we test the hypothesis that structural variation of the LC explains response inhibition deficits in PSP and PD. Twenty-four people with idiopathic PD, 14 with PSP-Richardson's syndrome, and 24 age- and sex-matched controls undertook a stop-signal task and ultrahigh field 7T magnetization-transfer-weighted imaging of the LC. Parameters of "race models" of go- versus stop-decisions were estimated using hierarchical Bayesian methods to quantify the cognitive processes of response inhibition. We tested the multivariate relationship between LC integrity and model parameters using partial least squares. Both disorders impaired response inhibition at the group level. PSP caused a distinct pattern of abnormalities in inhibitory control with a paradoxically reduced threshold for go responses, but longer nondecision times, and more lapses of attention. The variation in response inhibition correlated with the variability of LC integrity across participants in both clinical groups. Structural imaging of the LC, coupled with behavioral modeling in parkinsonian disorders, confirms that LC integrity is associated with response inhibition and LC degeneration contributes to neurobehavioral changes. The noradrenergic system is therefore a promising target to treat impulsivity in these conditions. The optimization of noradrenergic treatment is likely to benefit from stratification according to LC integrity.SIGNIFICANCE STATEMENT Response inhibition deficits contribute to clinical symptoms and poor outcomes in people with Parkinson's disease and progressive supranuclear palsy. We used cognitive modeling of performance of a response inhibition task to identify disease-specific mechanisms of abnormal inhibitory control. Response inhibition in both patient groups was associated with the integrity of the noradrenergic locus coeruleus, which we measured in vivo using ultra-high field MRI. We propose that the imaging biomarker of locus coeruleus integrity provides a trans-diagnostic tool to explain individual differences in response inhibition ability beyond the classic nosological borders and diagnostic criteria. Our data suggest a potential new stratified treatment approach for Parkinson's disease and progressive supranuclear palsy.

Variation in gaze following across the life span: A process-level perspective.

Following eye gaze is fundamental for many social-cognitive abilities, for example, when judging what another agent can or cannot know. While the emergence of gaze following has been thoroughly studied on a group level, we know little about (a) the developmental trajectory beyond infancy and (b) the sources of individual differences. In Study 1, we examined gaze following across the lifespan (N = 478 3- to 19-year-olds from Leipzig, Germany; and N = 240 20- to 80-year-old international, remotely tested adults). We found a steep performance improvement during preschool years, in which children became more precise in locating the attentional focus of an agent. Precision levels then stayed comparably stable throughout adulthood with a minor decline toward old age. In Study 2, we formalized the process of gaze following in a computational cognitive model that allowed us to conceptualize individual differences in a psychologically meaningful way (N = 60 3- to 5-year-olds, 50 adults). According to our model, participants estimate pupil angles with varying levels of precision based on observing the pupil location within the agent's eyes. In Study 3, we empirically tested how gaze following relates to vector following in non-social settings and perspective-taking abilities (N = 102 4- to 5-year-olds). We found that gaze following is associated with both of these abilities but less so with other Theory of Mind tasks. This work illustrates how the combination of reliable measurement instruments and formal theoretical models allows us to explore the in(ter)dependence of core social-cognitive processes in greater detail. RESEARCH HIGHLIGHTS: Gaze following develops beyond infancy. The highest precision levels in localizing attentional foci are reached in young adulthood with a slight decrease towards old age. We present a computational model that describes gaze following as a process of estimating pupil angles and the corresponding gaze vectors. The model explains individual differences and recovers signature patterns in the data. To estimate the relation between gaze- and vector following, we designed a non-social vector following task. We found substantial correlations between gaze following and vector following, as well as Level 2 perspective-taking. Other Theory of Mind tasks did not correlate.

Active maintenance in working memory reinforces bindings for future retrieval from episodic long-term memory.

Many theories assume that actively maintaining information in working memory (WM) predicts its retention in episodic long-term memory (LTM), as revealed by the beneficial effects of more WM time. In four experiments, we examined whether affording more time for intentional WM maintenance does indeed drive LTM. Sequences of four words were presented during trials of simple span (short time), slow span (long time), and complex span (long time with distraction; Experiments 1-2). Long time intervals entailed a pause of equivalent duration between the words that presented a blank screen (slow span) or an arithmetic problem to read aloud and solve (complex span). In Experiments 1-3, participants either serially recalled the words (intentional encoding) or completed a no-recall task (incidental encoding). In Experiment 4, all participants were instructed to intentionally encode the words, with the trials randomly ending in the serial-recall or no-recall task. To ensure similar processing of the words between encoding groups, participants silently decided whether each word was a living or nonliving thing via key press (i.e., an animacy judgment; Experiments 1 and 3-4) or read the words aloud and then pressed the space bar (Experiment 2). A surprise delayed memory test at the end of the experiment assessed LTM. Applying Bayesian cognitive models to disambiguate binding and item memory revealed consistent benefits of free time to binding memory that were specific to intentional encoding in WM. This suggests that time spent intentionally keeping information in WM is special for LTM because WM is a system that maintains bindings.

Cognitive modeling of the Mnemonic Similarity Task as a digital biomarker for Alzheimer's disease.

BACKGROUND: The Mnemonic Similarity Task (MST) is a popular memory task designed to assess hippocampal integrity. We assessed whether analyzing MST performance using a multinomial processing tree (MPT) cognitive model could detect individuals with elevated Alzheimer's disease (AD) biomarker status prior to cognitive decline. METHOD: We analyzed MST data from >200 individuals (young, cognitively healthy older adults and individuals with mild cognitive impairment [MCI]), a subset of which also had existing cerebrospinal fluid (CSF) amyloid beta (Aß) and phosphorylated tau (pTau) data using both traditional and model-derived approaches. We assessed how well each could predict age group, memory ability, MCI status, Aß, and pTau status using receiver operating characteristic analyses. RESULTS: Both approaches predicted age group membership equally, but MPT-derived metrics exceeded traditional metrics in all other comparisons. DISCUSSION: A MPT model of the MST can detect individuals with AD prior to cognitive decline, making it a potentially useful tool for screening and monitoring older adults during the asymptomatic phase of AD. HIGHLIGHTS: The MST, along with cognitive modeling, identifies individuals with memory deficits and cognitive impairment. Cognitive modeling of the MST identifies individuals with increased AD biomarkers prior to changes in cognitive function. The MST is a digital biomarker that identifies individuals at high risk of AD.

Connecting process models to response times through Bayesian hierarchical regression analysis.

Process models specify a series of mental operations necessary to complete a task. We demonstrate how to use process models to analyze response-time data and obtain parameter estimates that have a clear psychological interpretation. A prerequisite for our analysis is a process model that generates a count of elementary information processing steps (EIP steps) for each trial of an experiment. We can estimate the duration of an EIP step by assuming that every EIP step is of random duration, modeled as draws from a gamma distribution. A natural effect of summing several random EIP steps is that the expected spread of the overall response time increases with a higher EIP step count. With modern probabilistic programming tools, it becomes relatively easy to fit Bayesian hierarchical models to data and thus estimate the duration of a step for each individual participant. We present two examples in this paper: The first example is children's performance on simple addition tasks, where the response time is often well predicted by the smaller of the two addends. The second example is response times in a Sudoku task. Here, the process model contains some random decisions and the EIP step count thus becomes latent. We show how our EIP regression model can be extended to such a case. We believe this approach can be used to bridge the gap between classical cognitive modeling and statistical inference and will be easily applicable to many use cases.

A tutorial on fitting joint models of M/EEG and behavior to understand cognition.

We present motivation and practical steps necessary to find parameter estimates of joint models of behavior and neural electrophysiological data. This tutorial is written for researchers wishing to build joint models of human behavior and scalp and intracranial electroencephalographic (EEG) or magnetoencephalographic (MEG) data, and more specifically those researchers who seek to understand human cognition. Although these techniques could easily be applied to animal models, the focus of this tutorial is on human participants. Joint modeling of M/EEG and behavior requires some knowledge of existing computational and cognitive theories, M/EEG artifact correction, M/EEG analysis techniques, cognitive modeling, and programming for statistical modeling implementation. This paper seeks to give an introduction to these techniques as they apply to estimating parameters from neurocognitive models of M/EEG and human behavior, and to evaluate model results and compare models. Due to our research and knowledge on the subject matter, our examples in this paper will focus on testing specific hypotheses in human decision-making theory. However, most of the motivation and discussion of this paper applies across many modeling procedures and applications. We provide Python (and linked R) code examples in the tutorial and appendix. Readers are encouraged to try the exercises at the end of the document.

Heterogeneity of rules in Bayesian reasoning: A toolbox analysis.

How do people infer the Bayesian posterior probability from stated base rate, hit rate, and false alarm rate? This question is not only of theoretical relevance but also of practical relevance in medical and legal settings. We test two competing theoretical views: single-process theories versus toolbox theories. Single-process theories assume that a single process explains people's inferences and have indeed been observed to fit people's inferences well. Examples are Bayes's rule, the representativeness heuristic, and a weighing-and-adding model. Their assumed process homogeneity implies unimodal response distributions. Toolbox theories, in contrast, assume process heterogeneity, implying multimodal response distributions. After analyzing response distributions in studies with laypeople and professionals, we find little support for the single-process theories tested. Using simulations, we find that a single process, the weighing-and-adding model, nevertheless can best fit the aggregate data and, surprisingly, also achieve the best out-of-sample prediction even though it fails to predict any single respondent's inferences. To identify the potential toolbox of rules, we test how well candidate rules predict a set of over 10,000 inferences (culled from the literature) from 4,188 participants and 106 different Bayesian tasks. A toolbox of five non-Bayesian rules plus Bayes's rule captures 64% of inferences. Finally, we validate the Five-Plus toolbox in three experiments that measure response times, self-reports, and strategy use. The most important conclusion from these analyses is that the fitting of single-process theories to aggregate data risks misidentifying the cognitive process. Antidotes to that risk are careful analyses of process and rule heterogeneity across people.

An individual differences perspective on pragmatic abilities in the preschool years.

Pragmatic abilities are fundamental to successful language use and learning. Individual differences studies contribute to understanding the psychological processes involved in pragmatic reasoning. Small sample sizes, insufficient measurement tools, and a lack of theoretical precision have hindered progress, however. Three studies addressed these challenges in three- to 5-year-old German-speaking children (N = 228, 121 female). Studies 1 and 2 assessed the psychometric properties of six pragmatics tasks. Study 3 investigated relations among pragmatics tasks and between pragmatics and other cognitive abilities. The tasks were found to measure stable variation between individuals. Via a computational cognitive model, individual differences were traced back to a latent pragmatics construct. This presents the basis for understanding the relations between pragmatics and other cognitive abilities. RESEARCH HIGHLIGHTS: Individual differences in pragmatic abilities are important to understanding variation in language development. Research in this domain lacks a precise theoretical framework and psychometrically high-quality measures. We present six tasks capturing a wide range of pragmatic abilities with excellent re-test reliability. We use a computational cognitive model to provide a substantive theory of individual differences in pragmatic abilities.

A linear threshold model for optimal stopping behavior.

In many real-life decisions, options are distributed in space and time, making it necessary to search sequentially through them, often without a chance to return to a rejected option. The optimal strategy in these tasks is to choose the first option that is above a threshold that depends on the current position in the sequence. The implicit decision-making strategies by humans vary but largely diverge from this optimal strategy. The reasons for this divergence remain unknown. We present a model of human stopping decisions in sequential decision-making tasks based on a linear threshold heuristic. The first two studies demonstrate that the linear threshold model accounts better for sequential decision making than existing models. Moreover, we show that the model accurately predicts participants' search behavior in different environments. In the third study, we confirm that the model generalizes to a real-world problem, thus providing an important step toward understanding human sequential decision making.

Meta-Learned Models of Cognition.

Psychologists and neuroscientists extensively rely on computational models for studying and analyzing the human mind. Traditionally, such computational models have been hand-designed by expert researchers. Two prominent examples are cognitive architectures and Bayesian models of cognition. While the former requires the specification of a fixed set of computational structures and a definition of how these structures interact with each other, the latter necessitates the commitment to a particular prior and a likelihood function which - in combination with Bayes' rule - determine the model's behavior. In recent years, a new framework has established itself as a promising tool for building models of human cognition: the framework of meta-learning. In contrast to the previously mentioned model classes, meta-learned models acquire their inductive biases from experience, i.e., by repeatedly interacting with an environment. However, a coherent research program around meta-learned models of cognition is still missing to this day. The purpose of this article is to synthesize previous work in this field and establish such a research program. We accomplish this by pointing out that meta-learning can be used to construct Bayes-optimal learning algorithms, allowing us to draw strong connections to the rational analysis of cognition. We then discuss several advantages of the meta-learning framework over traditional methods and reexamine prior work in the context of these new insights.

Strategy selection in decisions from givens: Deciding at a glance?

People deciding between alternatives have at their disposal a toolbox containing both compensatory strategies, which take into account all available attributes of those alternatives, and noncompensatory strategies, which consider only some of the attributes. It is commonly assumed that noncompensatory strategies play only a minor role in decisions from givens, where attribute information is openly presented, because all attributes can be processed automatically "at a glance." Based on a literature review, however, I establish that previous studies on strategy selection in decisions from givens have yielded highly heterogeneous findings, including evidence of widespread use of noncompensatory strategies. Drawing on insights from visual attention research on subitizing, I argue that this heterogeneity might be due to differences across studies in the number of attributes and in whether the same or different symbols are used to represent high/low attribute values across attributes. I tested the impact of these factors in two experiments with decisions from givens in which both the number of attributes shown for each alternative and the coding of attribute values was manipulated. An analysis of participants' strategy use with a Bayesian multimethod approach (taking into account both decisions and response-time patterns) showed that a noncompensatory strategy was more frequently selected in conditions with a higher number of attributes; the type of attribute coding scheme did not affect strategy selection. Using a compensatory strategy in the conditions with eight (vs. four) attributes was associated with rather long response times and a high rate of strategy execution errors. The results suggest that decisions from givens can incur cognitive costs that prohibit reliance on automatic compensatory decision making and that can favor the adaptive selection of a noncompensatory strategy.

Modeling Human Steering Behavior in Haptic Shared Control of Autonomy-Enabled Unmanned Ground Vehicles.

OBJECTIVE: A human steering model for teleoperated driving is extended to capture the human steering behavior in haptic shared control of autonomy-enabled Unmanned Ground Vehicles (UGVs). BACKGROUND: Prior studies presented human steering models for teleoperation of a passenger-sized Unmanned Ground Vehicle, where a human is fully in charge of driving. However, these models are not applicable when a human needs to interact with autonomy in haptic shared control of autonomy-enabled UGVs. How a human operator reacts to the presence of autonomy needs to be studied and mathematically encapsulated in a module to capture the collaboration between human and autonomy. METHOD: Human subject tests are conducted to collect data in haptic shared control for model development and validation. The ACT-R architecture and two-point steering model used in the previous literature are adopted to predict the operator's desired steering angle. A torque conversion module is developed to convert the steering command from the ACT-R model to human torque input, thus enabling haptic shared control with autonomy. A parameterization strategy is described to find the set of model parameters that optimize the haptic shared control performance in terms of minimum average lane keeping error (ALKE). RESULTS: The model predicts the minimum ALKE human subjects achieve in shared control. CONCLUSIONS: The extended model can successfully predict the best haptic shared control performance as measured by ALKE. APPLICATION: This model can be used in place of human operators, enabling fully simulation-based engineering, in the development and evaluation of haptic shared control technologies for autonomy-enabled UGVs, including control negotiation strategies and autonomy capabilities.

Multitasking While Driving: Central Bottleneck or Problem State Interference?

OBJECTIVE: The objective of this work was to investigate if visuospatial attention and working memory load interact at a central control resource or at a task-specific, information processing resource during driving. BACKGROUND: In previous multitasking driving experiments, interactions between different cognitive concepts (e.g., attention and working memory) have been found. These interactions have been attributed to a central bottleneck or to the so-called problem-state bottleneck, related to working memory usage. METHOD: We developed two different cognitive models in the cognitive architecture ACT-R, which implement the central vs. problem-state bottleneck. The models performed a driving task, during which we varied visuospatial attention and working memory load. We evaluated the model by conducting an experiment with human participants and compared the behavioral data to the model's behavior. RESULTS: The problem-state-bottleneck model could account for decreased driving performance due to working memory load as well as increased visuospatial attentional demands as compared to the central-bottleneck model, which could not account for effects of increased working memory load. CONCLUSION: The interaction between working memory and visuospatial attention in our dual tasking experiment can be best characterized by a bottleneck in the working memory. The model results suggest that as working memory load becomes higher, drivers manage to perform fewer control actions, which leads to decreasing driving performance. APPLICATION: Predictions about the effect of different mental loads can be used to quantify the contribution of each subtask allowing for precise assessments of the current overall mental load, which automated driving systems may adapt to.

An exploration of error-driven learning in simple two-layer networks from a discriminative learning perspective.

Error-driven learning algorithms, which iteratively adjust expectations based on prediction error, are the basis for a vast array of computational models in the brain and cognitive sciences that often differ widely in their precise form and application: they range from simple models in psychology and cybernetics to current complex deep learning models dominating discussions in machine learning and artificial intelligence. However, despite the ubiquity of this mechanism, detailed analyses of its basic workings uninfluenced by existing theories or specific research goals are rare in the literature. To address this, we present an exposition of error-driven learning - focusing on its simplest form for clarity - and relate this to the historical development of error-driven learning models in the cognitive sciences. Although historically error-driven models have been thought of as associative, such that learning is thought to combine preexisting elemental representations, our analysis will highlight the discriminative nature of learning in these models and the implications of this for the way how learning is conceptualized. We complement our theoretical introduction to error-driven learning with a practical guide to the application of simple error-driven learning models in which we discuss a number of example simulations, that are also presented in detail in an accompanying tutorial.

Distributional social semantics: Inferring word meanings from communication patterns.

Distributional models of lexical semantics have proven to be powerful accounts of how word meanings are acquired from the natural language environment (Günther, Rinaldi, & Marelli, 2019; Kumar, 2020). Standard models of this type acquire the meaning of words through the learning of word co-occurrence statistics across large corpora. However, these models ignore social and communicative aspects of language processing, which is considered central to usage-based and adaptive theories of language (Tomasello, 2003; Beckner et al., 2009). Johns (2021) recently demonstrated that integrating social and communicative information into a lexical strength measure allowed for benchmark fits to be attained for lexical organization data, indicating that the social world contains important statistical information for language learning and processing. Through the analysis of the communication patterns of over 330,000 individuals on the online forum Reddit, totaling approximately 55 billion words of text, the findings of the current article demonstrates that social information about word usage allows for unique aspects of a word's meaning to be acquired, providing a new pathway for distributional model development.

Data-driven experimental design and model development using Gaussian process with active learning.

Interest in computational modeling of cognition and behavior continues to grow. To be most productive, modelers should be equipped with tools that ensure optimal efficiency in data collection and in the integrity of inference about the phenomenon of interest. Traditionally, models in cognitive science have been parametric, which are particularly susceptible to model misspecification because their strong assumptions (e.g. parameterization, functional form) may introduce unjustified biases in data collection and inference. To address this issue, we propose a data-driven nonparametric framework for model development, one that also includes optimal experimental design as a goal. It combines Gaussian Processes, a stochastic process often used for regression and classification, with active learning, from machine learning, to iteratively fit the model and use it to optimize the design selection throughout the experiment. The approach, dubbed Gaussian process with active learning (GPAL), is an extension of the parametric, adaptive design optimization (ADO) framework (Cavagnaro, Myung, Pitt, & Kujala, 2010). We demonstrate the application and features of GPAL in a delay discounting task and compare its performance to ADO in two experiments. The results show that GPAL is a viable modeling framework that is noteworthy for its high sensitivity to individual differences, identifying novel patterns in the data that were missed by the model-constrained ADO. This investigation represents a first step towards the development of a data-driven cognitive modeling framework that serves as a middle ground between raw data, which can be difficult to interpret, and parametric models, which rely on strong assumptions.

The Natural Stories corpus: a reading-time corpus of English texts containing rare syntactic constructions.

It is now a common practice to compare models of human language processing by comparing how well they predict behavioral and neural measures of processing difficulty, such as reading times, on corpora of rich naturalistic linguistic materials. However, many of these corpora, which are based on naturally-occurring text, do not contain many of the low-frequency syntactic constructions that are often required to distinguish between processing theories. Here we describe a new corpus consisting of English texts edited to contain many low-frequency syntactic constructions while still sounding fluent to native speakers. The corpus is annotated with hand-corrected Penn Treebank-style parse trees and includes self-paced reading time data and aligned audio recordings. We give an overview of the content of the corpus, review recent work using the corpus, and release the data.

The simultaneous recognition of multiple words: A process analysis.

In everyday life, recognition decisions often have to be made for multiple objects simultaneously. In contrast, research on recognition memory has predominantly relied on single-item recognition paradigms. We present a first systematic investigation into the cognitive processes that differ between single-word and paired-word tests of recognition memory. In a single-word test, participants categorize previously presented words and new words as having been studied before (old) or not (new). In a paired-word test, however, the test words are randomly paired, and participants provide joint old-new categorizations of both words for each pair. Across two experiments (N = 170), we found better memory performance for words tested singly rather than in pairs and, more importantly, dependencies between the two single-word decisions implied by the paired-word test. We extended two popular model classes of single-item recognition to paired-word recognition, a discrete-state model and a continuous model. Both models attribute performance differences between single-word and paired-word recognition to differences in memory-evidence strength. Discrete-state models account for the dependencies in paired-word decisions in terms of dependencies in guessing. In contrast, continuous models map the dependencies on mnemonic (Experiment 1 & 2) as well as on decisional processes (Experiment 2). However, in both experiments, model comparison favored the discrete-state model, indicating that memory decisions for word pairs seem to be mediated by discrete states. Our work suggests that individuals tackle multiple-item recognition fundamentally differently from single-item recognition, and it provides both a behavioral and model-based paradigm for studying multiple-item recognition.

A Review of Individual Operational Cognitive Readiness: Theory Development and Future Directions.

OBJECTIVE: The aim of this paper is to provide a comprehensive and original review of the theoretical development of the individual operational cognitive readiness (OCR) theory. BACKGROUND: Cognitive readiness (CR) is a concept that has the potential to predict the performance of human individuals and teams prior to engaging in complex, dynamic, and resource-limited task environments. However, the current state of the literature is confusing and laborious, with heterogeneous views regarding the theoretical frameworks among leading researchers. METHOD: This review (1) undertakes a systematic approach toward categorizing published CR literature into theoretical and measurement contributions across the different levels of CR, (2) carries a critical evaluation of the CR and OCR theoretical frameworks, and (3) provides directions for future research guided by gaps identified during the review process and other published literatures. RESULTS: Results from the categorization of published CR literature provide a new, valuable, synthesized CR library for researchers to consult to streamline their CR literature review process. Critical examination of the CR and OCR theoretical frameworks leads to positing that new components should be explored for OCR. CONCLUSION: There are many possible directions for future research including evaluating domain-independent components of OCR and evaluating the relationship between biofeedback measures and performance in CR models. APPLICATION: The Defense domain continues to be the focal application of CR. However, CR could be used by other application domains, such as sports and emergency services, that require their working personnel to engage in complex, uncertain, and dynamic task environments.