Cognitive flexibility is the ability to shift between mental sets in response to changing goals or environments, and it is known to decline across the adult lifespan. Competing accounts attribute this decline either to weaknesses in specific abilities such as inhibition or memory, or to more global challenges in managing the interaction between them. Efforts to disentangle contributing processes through targeted, single-function tasks have produced mixed findings, potentially due to confounding variability in stimuli and task demands.The current work contributes to the field's understanding of cognitive flexibility decline by focusing on \emph{inhibition} as both a subcomponent and as an independent process, given its central role in both specific-deficit and integrative theoretical accounts. In particular, we designed a unified task that engages both 1) cognitive flexibility through shifting requirements to encode task-relevant mental sets and inhibit them once they become outdated; and 2) discrete item-level inhibition of perceptual distractors. We tested 86 adults aged 18–73. Accuracy analyses revealed asymmetric interaction effects between aging and these two operations: while learning after a rule-switch worsened as age increased, inhibition of perceptual distractors inhibition improved. We then applied a computational model to quantify how memory precision and the inhibition of previously-relevant sets each contributed to performance. Model fits indicated that age-related decline in learning across rule-switches were driven primarily by reduced memory precision, not by deficits in inhibition.

Widespread cognitive decline in older adults has been hypothesized to stem from a fundamental deficit in inhibition, or the ability to ignore goal-irrelevant information. The extent to which inhibition operates across different timescales, however, has been under-explored. We introduce a novel cognitive task designed to assess both between- and within-trial inhibition using a common set of stimuli. Behavioral results from younger, middle-aged, and older adults (N=100; age range: 18-73) reveal significant age-related differences in between-trial inhibition, with older adults showing less efficient adaptation to rule changes compared to younger adults. Within-trial inhibition, requiring suppression of distractors within the current visual environment, appears to remain intact alongside normal aging. These findings will support the development of tools for the early detection of age-related cognitive decline, prior to subjective awareness of impaired daily functioning.

Evidence-accumulation models (EAMs) are powerful tools for making sense of human and animal decision-making behavior. EAMs have generated significant theoretical advances in psychology, behavioral economics, and cognitive neuroscience and are increasingly used as a measurement tool in clinical research and other applied settings. Obtaining valid and reliable inferences from EAMs depends on knowing how to establish a close match between model assumptions and features of the task/data to which the model is applied. However, this knowledge is rarely articulated in the EAM literature, leaving beginners to rely on the private advice of mentors and colleagues and inefficient trial-and-error learning. In this article, we provide practical guidance for designing tasks appropriate for EAMs, relating experimental manipulations to EAM parameters, planning appropriate sample sizes, and preparing data and conducting an EAM analysis. Our advice is based on prior methodological studies and the our substantial collective experience with EAMs. By encouraging good task-design practices and warning of potential pitfalls, we hope to improve the quality and trustworthiness of future EAM research and applications.

Subjective confidence is an important factor in our decision making, but how confidence arises is a matter of debate. A number of computational models have been proposed that integrate confidence into sequential sampling models of decision making, in which evidence accumulates across time to a threshold. An influential example of this approach is the relative balance of evidence hypothesis, in which confidence is determined by the amount of evidence for the choice that was made compared to the evidence for all possible choices. Here, we modify this approach by mapping distance from a decision threshold to confidence via a sigmoid function. This allows for individual differences in bias toward lower or higher levels of confidence, as well as sensitivity to differences in evidence between choices. We apply several variants of the model to assess potential age differences between young and older adults in source memory decision making in an existing data set (Dodson, Bawa, & Slotnick, 2007). We compare our model to the relative balance of evidence approach, and the results indicate that the sigmoidal method substantially improves model fit. We also consider models in which memory errors can arise from a misrecollection process that involves associating items with the incorrect source, a process that has been proposed to account for age differences in source memory confidence and accuracy, but find no evidence that misrecollection is necessary to account for the results. This work provides a viable model of subjective confidence that is integrated with well-established models of decision making and provides insights into effects of aging on source memory decisions. Public Significance Statement This study presents a theory of how subjective levels of confidence in memory decisions are determined in young and older adults by comparing evidence for different choice options. The findings suggest that the way confidence is calculated differs between people and may change across the lifespan

Widespread cognitive decline in older adults has been hypothesized to stem from a fundamental deficit in inhibition, or the ability to ignore goal-irrelevant information. However, previous efforts to measure inhibition in aging have been hindered by the task impurity problem, where multiple cognitive processes contribute to performance and make it difficult to isolate specific mechanisms. Here, we introduce the Global-Local AtteNtional Control Evaluation (GLANCE), a novel framework designed to 1) assess both global attentional control (between-trial inhibition) and local attentional control (within-trial inhibition) in a unified task; and 2) address the task impurity problem via an accompanying computational model. Behavioral results from younger, middle-aged, and older adults (N=100; age range: 18-73) reveal significant age-related differences in global attentional control, with older adults showing less efficient adaptation to rule changes compared to younger adults. Parameter estimates from fitting the computational model to subject-level data further identified reduced error-related memory suppression as the specific mechanism that accounts for age-related differences in behavior.

Recent advances in neuroscience and psychology show that the brain has access to timelines of both the past and the future. Spiking across populations of neurons in many regions of the mammalian brain maintains a robust temporal memory, a neural timeline of the recent past. Behavioral results demonstrate that people can estimate an extended temporal model of the future, suggesting that the neural timeline of the past could extend through the present into the future. This paper presents a mathematical framework for learning and expressing relationships between events in continuous time. We assume that the brain has access to a temporal memory in the form of the real Laplace transform of the recent past. Hebbian associations with a diversity of synaptic time scales are formed between the past and the present that record the temporal relationships between events. Knowing the temporal relationships between the past and the present allows one to predict relationships between the present and the future, thus constructing an extended temporal prediction for the future. Both memory for the past and the predicted future are represented as the real Laplace transform, expressed as the firing rate over populations of neurons indexed by different rate constants s. The diversity of synaptic timescales allows for a temporal record over the much larger time scale of trial history. In this framework, temporal credit assignment can be assessed via a Laplace temporal difference. The Laplace temporal difference compares the future that actually follows a stimulus to the future predicted just before the stimulus was observed. This computational framework makes a number of specific neurophysiological predictions and, taken together, could provide the basis for a future iteration of RL that incorporates temporal memory as a fundamental building block.

Frontal midline theta oscillatory dynamics have been implicated as an important neural signature of inhibitory control. However, most proactive cognitive control studies rely on behavioral tasks where individual differences are inferred through button presses. We applied computational modeling to further refine our understanding of theta dynamics in a cued anti-saccade task with gaze-contingent eye tracking. Using a drift diffusion model, increased frontal midline theta power during high-conflict, relative to low-conflict, trials predicted a more conservative style of responding through the starting point (bias). During both high- and low-conflict trials, increases in frontal midline theta also predicted improvements in response efficiency (drift rate). Regression analyses provided support for the importance of the starting point bias, which was associated with frontal midline theta over the course of the task above-and-beyond both drift rate and mean reaction time. Our findings provide a more thorough understanding of proactive gaze control by linking trial-by-trial increases of frontal midline theta to a shift in starting point bias facilitating a more neutral style of responding.

Computational models of episodic memory provide tools to better understand the latent neurocognitive processes underlying retention of information about specific events from one’s life. This chapter discusses the representations, associations, and dynamics of influential models of episodic memory, with particular emphasis on models of recognition and free recall tasks. In-depth discussion and model-fitting results of four models – the retrieving effectively from memory (REM) model, the bind cue decide model of episodic memory (BCDMEM), the search of associative memory (SAM) model, and the temporal context model (TCM) – are provided to facilitate understanding of these models, as well as similarities and differences between them. Alternative modeling frameworks, including neural network models, are discussed. Throughout, the importance of context in models of episodic memory is emphasized, particularly for free recall tasks.

Humans use log-compressed number lines to represent different quantities, including elapsed time, traveled distance, numerosity, sound frequency, etc. Inspired by recent cognitive science and computational neuroscience work, we developed a neural network that learns to construct log-compressed number lines. The network computes a discrete approximation of a real-domain Laplace transform using an RNN with analytically derived weights giving rise to a log-compressed timeline of the past. The network learns to extract latent variables from the input and uses them for global modulation of the recurrent weights turning a timeline into a number line over relevant dimensions. The number line representation greatly simplifies learning on a set of problems that require learning associations in different spaces - problems that humans can typically solve easily. This approach illustrates how combining deep learning with cognitive models can result in systems that learn to represent latent variables in a brain-like manner and exhibit human-like behavior manifested through Weber-Fechner law.

Memory reconsolidation refers to the phenomenon whereby a previously consolidated memory, i.e., one that is resistant to interference or disruption, becomes labile due to reactivation, initiating a short window during which that memory can be modified. With a wide range of potential clinical and educational applications, reconsolidation has been demonstrated across multiple domains and timescales, including fear, motor, and episodic memory. This chapter seeks to clarify the psychological processes involved in reconsolidation, making connections to underlying physiological mechanisms, with the goal of providing a framework for understanding why and when reconsolidation takes place. Drawing on reviews of both human and relevant animal studies, this chapter highlights the importance of both context and predictions for determining whether new experiences give rise to new learning or the updating of previously-learned associations.

AbstractDecision-making is an essential cognitive process by which we interact with the external world. However, attempts to understand the neural mechanisms of decision-making are limited by the current available animal models and the technologies that can be applied to them. Here, we build on the renewed interest in using tree shrews (Tupaia belangeri) in vision research and provide strong support for them as a model for studying visual perceptual decision-making. Tree shrews learned very quickly to perform a two-alternative forced choice contrast discrimination task, and they exhibited differences in response time distributions depending on the reward and punishment structure of the task. Specifically, they made occasional fast guesses when incorrect responses are punished by a constant increase in the interval between trials. This behavior was suppressed when faster incorrect responses were discouraged by longer intertrial intervals. By fitting the behavioral data with two variants of racing diffusion decision models, we found that the between-trial delay affected decision-making by modulating the drift rate of a time accumulator. Our results thus provide support for the existence of an internal process that is independent of the evidence accumulation in decision-making and lay a foundation for future mechanistic studies of perceptual decision-making using tree shrews.

The ability to bind, or link, different aspects of an experience in memory undergoes protracted development across childhood. Most studies of memory binding development have assessed extraobject binding between an object and some external element such as another object, whereas little work has examined the development of intraobject binding, such as between shape and color features within the same object. In this work, we investigate the development of intra- and extraobject memory binding in five-year-olds, eight-year-olds, and young adults with a memory interference paradigm. Between two experiments, we manipulate whether stimuli are presented as coherent objects (Experiment 1: n5-year-olds = 32, 19 males, 13 females; n8-year-olds = 30, 15 males, 15 females; nadults = 30, 15 males, 15 females), requiring intraobject binding between shape and color features, or as spatially separated features (Experiment 2: n5-year-olds = 24, 16 males, 8 females; n8-year-olds = 41, 19 males, 22 females; nadults = 31, 13 males, 18 females), requiring extraobject binding. To estimate the contributions of different binding structures to performance, we present a novel computational model that mathematically instantiates the memory binding, forgetting, and retrieval processes we hypothesize to underlie performance on the task. The results provide evidence of substantial developmental improvements in both intraobject and extraobject binding of shape and color features between 5 and 8 years of age, as well as stronger intraobject compared with extraobject binding of features in all age groups. These findings provide key insights into memory binding across early development. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Healthy aging is associated with deficits in performance on episodic memory tasks. Popular verbal theories of the mechanisms underlying this decrement have primarily focused on inferred changes in associative memory. However, performance on any task is the result of interactions between different neurocognitive mechanisms, such as perceptuomotor, memory, and decision-making processes. As a result, age-related differences in performance could arise from multiple processes, which could lead to incomplete or incorrect conclusions about the sources of aging effects. In addition, standard statistical comparisons of group-level summary statistics, such as mean accuracy, may not provide sufficient information to allow detailed mechanistic explanations of age-related change. We argue that these and other drawbacks of relying exclusively on verbal theories can hamper replicability, transparency, and scientific progress in aging research and psychological science more generally, and that computational modeling is a tool that can address many of these limitations. Computational models make mathematically transparent claims about how latent processes give rise to observed behavior and decompose an individual's performance into model parameters governing hypothesized mechanisms. In this work, we present a short memory task designed for and analyzed with mechanistic model-based approaches. We provide an example of a computational model and fit the model to data from young and older adults with hierarchical Bayesian techniques in order to (a) detect differences in latent cognitive processes between young and older adults (as well as individual participants), (b) quantitatively compare models to assess different processes that could underlie performance, and (c) simulate data to make predictions for future experiments based on model mechanisms. We argue that computational modeling is a powerful tool to examine age differences in latent processes, make theories more transparent, and facilitate discovery in cognitive aging research. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Intracortical brain-machine interfaces decode motor commands from neural signals and translate them into actions, enabling movement for paralysed individuals. The subjective sense of agency associated with actions generated via intracortical brain-machine interfaces, the neural mechanisms involved and its clinical relevance are currently unknown. By experimentally manipulating the coherence between decoded motor commands and sensory feedback in a tetraplegic individual using a brain-machine interface, we provide evidence that primary motor cortex processes sensory feedback, sensorimotor conflicts and subjective states of actions generated via the brain-machine interface. Neural signals processing the sense of agency affected the proficiency of the brain-machine interface, underlining the clinical potential of the present approach. These findings show that primary motor cortex encodes information related to action and sensing, but also sensorimotor and subjective agency signals, which in turn are relevant for clinical applications of brain-machine interfaces.

Human learners can readily understand speech, or a melody, when it is presented slower or faster than usual. Although deep convolutional neural networks (CNNs) are extremely powerful in extracting information from time series, they require explicit training to generalize to different time scales. This paper presents a deep CNN that incorporates a temporal representation inspired by recent findings from neuroscience. In the mammalian brain, time is represented by populations of neurons with temporal receptive fields. Critically, the peaks of the receptive fields form a geometric series, such that the population codes a set of temporal basis functions over log time. Because memory for the recent past is a function of log time, rescaling the input results in translation of the memory. The Scale-Invariant Temporal History Convolution network (SITHCon) builds a convolutional layer over this logarithmically-distributed temporal memory. A max-pool operation results in a network that is invariant to rescalings of time modulo edge effects. We compare performance of SITHCon to a Temporal Convolution Network (TCN). Although both networks can learn classification and regression problems on both univariate and multivariate time series f(t), only SITHCon generalizes to rescalings f(at). This property, inspired by findings from contemporary neuroscience and consistent with findings from cognitive psychology, may enable networks that learn with fewer training examples, fewer weights and that generalize more robustly to out of sample data.

Brain Computer Interface (BCI) applications employ machine learning to decode neural signals through time to generate actions. One issue facing such machine learning algorithms is how much of the past they need to decode the present. DeepSITH (Deep Scale-Invariant Temporal History), is a deep neural network with layers inspired by how the mammalian brain represents recent vs. less-recent experience. A single SITH layer maintains a log-compressed representation of the past that becomes less accurate with older events, unlike other approaches that maintain a perfect copy of events regardless of how far in the past they occurred. By stacking layers of this compressed representation, we hypothesized that DeepSITH would be able to decode patterns of neural activity from farther in the past and combine them efficiently to guide the BCI in the present. We tested our approach with the Kaggle "Grasp and Lift challenge" dataset. This motor movement dataset has 12 subjects, 10 series of 30 grasp and lift trials per subject, with 6 classes of events to decode. We benchmark DeepSITH performances on this dataset against another common machine learning technique for integrating features over extended time scales, long short-term memory (LSTM). DeepSITH reproducibly achieves higher accuracy in predicting motor movement events than LSTM, and also takes significantly fewer epochs and less memory to train, in comparison to LSTM. In summary, DeepSITH can efficiently process more data, with increased prediction accuracy and learning speed. This result shows that DeepSITH is an advantageous model to consider when developing BCI technologies.

Brain Computer Interfaces (BCI) decode electroencephalography (EEG) data collected from the human brain to predict subsequent behavior. While this technology has promising applications, successfully implementing a model is challenging. The typical BCI control application requires many hours of training data from each individual to make predictions of intended activity specific to that individual. Moreover, there are individual differences in the organization of brain activity and low signal-to-noise ratios in noninvasive measurement techniques such as EEG. There is a fundamental bias-variance trade-off between developing a single model for all human brains vs. an individual model for each specific human brain. The Robust Shared Response Model (RSRM) attempts to resolve this tradeoff by leveraging both the homogeneity and heterogeneity of brain signals across people. RSRM extracts components that are common and shared across individual brains, while simultaneously learning unique representations between individual brains. By learning a latent shared space in conjunction with subject-specific representations, RSRM tends to result in better predictive performance on functional magnetic resonance imaging (fMRI) data relative to other common dimension reduction techniques. To our knowledge, we are the first research team attempting to expand the domain of RSRM by applying this technique to controlled experimental EEG data in a BCI setting. Using the openly available Motor Movement/ Imagery dataset, the decoding accuracy of RSRM exceeded models whose input was reduced by Principal Component Analysis (PCA), Independent Component Analysis (ICA), and subject-specific PCA. The results of our experiments suggest that RSRM can recover distributed latent brain signals and improve decoding accuracy of BCI tasks when dimension reduction is implemented as a feature engineering step. Future directions of this work include augmenting state-of-the art BCI with efficient reduced representations extracted by RSRM. This could enhance the utility of BCI technology in the real world. Furthermore, RSRM could have wide-ranging applications across other machine-learning applications that require classification of naturalistic data using reduced representations.

Although there have been major strides toward uncovering the neurobehavioral mechanisms involved in cognitive functions like memory and decision making, methods for measuring behavior and accessing latent processes through computational means remain limited. To this end, we have created SUPREME (Sensing to Understanding and Prediction Realized via an Experiment and Modeling Ecosystem): a toolbox for comprehensive cognitive assessment, provided by a combination of construct-targeted tasks and corresponding computational models. SUPREME includes four tasks, each developed symbiotically with a mechanistic model, which together provide quantified assessments of perception, cognitive control, declarative memory, reward valuation, and frustrative nonreward. In this study, we provide validation analyses for each task using two sessions of data from a cohort of cognitively normal participants (N = 65). Measures of test-retest reliability (r: 0.58–0.75), stability of individual differences (ρ: 0.56–0.70), and internal consistency (α: 0.80–0.86) support the validity of our tasks. After fitting the models to data from individual subjects, we demonstrate each model’s ability to capture observed patterns of behavioral results across task conditions. Our computational approaches allow us to decompose behavior into cognitively interpretable subprocesses, which we can compare both within and between participants. We discuss potential future applications of SUPREME, including clinical assessments, longitudinal tracking of cognitive functions, and insight into compensatory mechanisms.

In machine learning, convolutional neural networks (CNNs) have been extremely influential in both computer vision and in recognizing patterns extended over time. In computer vision, part of the flexibility arises from the use of max-pooling operations over the convolutions to attain translation invariance. In the mammalian brain, neural representations of time use a set of temporal basis functions. Critically, these basis functions appear to be arranged in a geometric series such that the basis set is evenly distributed over logarithmic time. This paper introduces a Scale-Invariant Temporal History Convolution network (SITHCon) that uses a logarithmically-distributed temporal memory. A max-pool over a logarithmicallydistributed temporal memory results in scale-invariance in time. We compare performance of SITHCon to a Temporal Convolution Network (TCN) and demonstrate that, although both networks can learn classification and regression problems on both univariate and multivariate time series f(t), only SITHCon has the property that it generalizes without retraining to rescaled versions of the input f(at). This property, inspired by findings from neuroscience and psychology, could lead to large-scale networks with dramatically different capabilities, including faster training and greater generalizability, even with significantly fewer free parameters.