Learning of Chunking Sequences in Cognition and Behavior

Examining the neural basis of unitization: A review

Associative memory refers to the ability to form and remember associations between individual pieces of information rather than memory for a single object or word. Encoding associations in memory tends to be a more difficult task than item (only) encoding, because associative memory requires encoding multiple items as well as the specific links amongst the items. Accordingly, researchers have worked to identify interventions and strategies to reduce the effort and neural resources required for successful associative memory processing. Unitization is one such strategy that has traditionally been defined as the process by which two or more discrete items are processed, or encoded, such that they are perceived as a single ensemble. The current review explores the neural research on unitization while considering the behavioral benefits that accompany the process.

Microstructural dynamics of motor learning and sleep-dependent consolidation: A diffusion imaging study

Memory consolidation can benefit from post-learning sleep, eventually leading to long-term microstructural brain modifications to accommodate new memory representations. Non-invasive diffusion-weighted magnetic resonance imaging (DWI) allows the observation of (micro)structural brain remodeling after time-limited motor learning. Here, we combine conventional diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI) that allows modeling dendritic and axonal complexity in gray matter to investigate with improved specificity the microstructural brain mechanisms underlying time- and sleep-dependent motor memory consolidation dynamics. Sixty-one young healthy adults underwent four DWI sessions, two sequential motor trainings, and a night of total sleep deprivation or regular sleep distributed over five days. We observed rapid-motor-learning-related remodeling in occipitoparietal, temporal, and motor-related subcortical regions, reflecting temporary dynamics in learning-related neuronal brain plasticity processes. Sleep-related consolidation seems not to exert a detectable impact on diffusion parameters, at least on the timescale of a few days.

Does Motor Memory Reactivation through Practice and Post-Learning Sleep Modulate Consolidation?

Retrieving previously stored information makes memory traces labile again and can trigger restabilization in a strengthened or weakened form depending on the reactivation condition. Available evidence for long-term performance changes upon reactivation of motor memories and the effect of post-learning sleep on their consolidation remains scarce, and so does the data on the ways in which subsequent reactivation of motor memories interacts with sleep-related consolidation. Eighty young volunteers learned (Day 1) a 12-element Serial Reaction Time Task (SRTT) before a post-training Regular Sleep (RS) or Sleep Deprivation (SD) night, either followed (Day 2) by morning motor reactivation through a short SRTT testing or no motor activity. Consolidation was assessed after three recovery nights (Day 5). A 2 × 2 ANOVA carried on proportional offline gains did not evidence significant Reactivation (Morning Reactivation/No Morning Reactivation; p = 0.098), post-training Sleep (RS/SD; p = 0.301) or Sleep*Reactivation interaction (p = 0.257) effect. Our results are in line with prior studies suggesting a lack of supplementary performance gains upon reactivation, and other studies that failed to disclose post-learning sleep-related effects on performance improvement. However, lack of overt behavioural effects does not detract from the possibility of sleep- or reconsolidation-related covert neurophysiological changes underlying similar behavioural performance levels.

An easy way to improve scoring of memory span tasks: The edit distance, beyond "correct recall in the correct serial position"

For researchers and psychologists interested in estimating a subject's memory capacity, the current standard for scoring memory span tasks is the partial-credit method: subjects are credited with the number of stimuli that they manage to recall correctly in the correct serial position. A critical issue with this method, however, is that intrusions and omissions can radically change the scores depending on where they occur. For example, when recalling the sequence ABCDE, "ABCD" is worth 4 points but "BCDE" is worth 0 points. This paper presents an improved scoring method based on the edit distance, meaning the number of changes required to edit the recalled sequence into the target. Edit-distance scoring gives results close to partial-credit scoring, but without the corresponding vulnerability to positional shifts. A reanalysis of memory performance in two large datasets (N = 1093 and N = 758) confirms that in addition to being more logically consistent, edit-distance scoring demonstrates similar or better psychometric properties than partial-credit, with comparable validity, a small increase in reliability, and a substantial increase of test information (measurement precision in the context of item response theory). Test information was especially improved for harder items and for subjects with ability in the lower range, whose scores tend to be severely underestimated by partial-credit scoring. Code to compute edit-distance scores with various software is made available at https://osf.io/wdb83/ .

Neural mechanisms for learning hierarchical structures of information

Spatial and temporal information from the environment is often hierarchically organized, so is our knowledge formed about the environment. Identifying the meaningful segments embedded in hierarchically structured information is crucial for cognitive functions, including visual, auditory, motor, memory, and language processing. Segmentation enables the grasping of the links between isolated entities, offering the basis for reasoning and thinking. Importantly, the brain learns such segmentation without external instructions. Here, we review the underlying computational mechanisms implemented at the single-cell and network levels. The network-level mechanism has an interesting similarity to machine-learning methods for graph segmentation. The brain possibly implements methods for the analysis of the hierarchical structures of the environment at multiple levels of its processing hierarchy.

Neuronal Sequence Models for Bayesian Online Inference

Various imaging and electrophysiological studies in a number of different species and brain regions have revealed that neuronal dynamics associated with diverse behavioral patterns and cognitive tasks take on a sequence-like structure, even when encoding stationary concepts. These neuronal sequences are characterized by robust and reproducible spatiotemporal activation patterns. This suggests that the role of neuronal sequences may be much more fundamental for brain function than is commonly believed. Furthermore, the idea that the brain is not simply a passive observer but an active predictor of its sensory input, is supported by an enormous amount of evidence in fields as diverse as human ethology and physiology, besides neuroscience. Hence, a central aspect of this review is to illustrate how neuronal sequences can be understood as critical for probabilistic predictive information processing, and what dynamical principles can be used as generators of neuronal sequences. Moreover, since different lines of evidence from neuroscience and computational modeling suggest that the brain is organized in a functional hierarchy of time scales, we will also review how models based on sequence-generating principles can be embedded in such a hierarchy, to form a generative model for recognition and prediction of sensory input. We shortly introduce the Bayesian brain hypothesis as a prominent mathematical description of how online, i.e., fast, recognition, and predictions may be computed by the brain. Finally, we briefly discuss some recent advances in machine learning, where spatiotemporally structured methods (akin to neuronal sequences) and hierarchical networks have independently been developed for a wide range of tasks. We conclude that the investigation of specific dynamical and structural principles of sequential brain activity not only helps us understand how the brain processes information and generates predictions, but also informs us about neuroscientific principles potentially useful for designing more efficient artificial neuronal networks for machine learning tasks.

Up to the magical number seven: An evolutionary perspective on the capacity of short term memory

Working memory and its components are among the most determinant factors in human cognition. However, in spite of their critical importance, many aspects of their evolution remain underinvestigated. The present study is devoted to reviewing the literature of memory studies from an evolutionary, comparative perspective, focusing particularly on short term memory capacity. The findings suggest the limited capacity to be the common attribute of different species of birds and mammals. Moreover, the results imply an increasing trend of capacity from our non-human ancestors to modern humans. The present evidence shows that non-human mammals and birds, regardless of their limitations, are capable of performing memory strategies, although there seem to be some differences between their ability and that of humans in terms of flexibility and efficiency. These findings have several implications relevant to the psychology of memory and cognition, and are likely to explain differences between higher cognitive abilities of humans and non-humans. The adaptive benefits of the limited capacity and the reasons for the growing trend found in the present study are broadly discussed.

The Implementation of Therapeutic Alliance in the Rehabilitation of an Elite Pediatric Athlete with Salter-Harris Fracture: A Case Report

BACKGROUND/PURPOSE

Although research on the value of therapeutic alliance is prominent in other areas of health care, physical therapy research is limited. The purpose is to describe the incorporation of therapeutic alliance concepts throughout the rehabilitation of an elite pediatric athlete with a complicated recovery following a fracture to the distal femoral epiphysis.

CASE DESCRIPTION

A 14-year-old male was referred to physical therapy following an open reduction and internal fixation to address a type IV Salter-Harris fracture of the right distal femoral epiphysis. Post-operative care included immobilization in a brace for six weeks and he initiated physical therapy for four weeks (post-op weeks 6-10). At 10-weeks post-injury his range of motion and strength were severely limited compared to expected post-operative milestones. Due to these deficits an arthroscopic debridement of the subject's right knee, hardware removal, and manipulation under anesthesia was performed. The subject then reported to the physical therapist on post-operative day three for evaluation and treatment without bracing or weight-bearing restrictions.

OUTCOMES

The episode of care spanned 17 weeks and included 25 physical therapy sessions. To facilitate therapeutic alliance with the subject, clear communication and easily measurable goals were established and connected to the subject's relevant needs as an athlete. The plan of care was divided into three phases using "chunking" techniques to establish the rehabilitation priorities. The subject demonstrated improved range of motion, strength and was able to return to hydroplane racing and won a national championship in his age group.

DISCUSSION

The unique aspect of this case was the incorporation of therapeutic alliance concepts and techniques into the rehabilitative management of a subject with a complicated fracture to the distal femoral epiphysis. The physical therapist built trust with the subject and facilitated a successful return to elite hydroplane boat racing.

LEVEL OF EVIDENCE

4.

STUDY DESIGN

Case Report.

Active inference on discrete state-spaces: A synthesis

Active inference is a normative principle underwriting perception, action, planning, decision-making and learning in biological or artificial agents. From its inception, its associated process theory has grown to incorporate complex generative models, enabling simulation of a wide range of complex behaviours. Due to successive developments in active inference, it is often difficult to see how its underlying principle relates to process theories and practical implementation. In this paper, we try to bridge this gap by providing a complete mathematical synthesis of active inference on discrete state-space models. This technical summary provides an overview of the theory, derives neuronal dynamics from first principles and relates this dynamics to biological processes. Furthermore, this paper provides a fundamental building block needed to understand active inference for mixed generative models; allowing continuous sensations to inform discrete representations. This paper may be used as follows: to guide research towards outstanding challenges, a practical guide on how to implement active inference to simulate experimental behaviour, or a pointer towards various in-silico neurophysiological responses that may be used to make empirical predictions.

Somatodendritic consistency check for temporal feature segmentation

The brain identifies potentially salient features within continuous information streams to process hierarchical temporal events. This requires the compression of information streams, for which effective computational principles are yet to be explored. Backpropagating action potentials can induce synaptic plasticity in the dendrites of cortical pyramidal neurons. By analogy with this effect, we model a self-supervising process that increases the similarity between dendritic and somatic activities where the somatic activity is normalized by a running average. We further show that a family of networks composed of the two-compartment neurons performs a surprisingly wide variety of complex unsupervised learning tasks, including chunking of temporal sequences and the source separation of mixed correlated signals. Common methods applicable to these temporal feature analyses were previously unknown. Our results suggest the powerful ability of neural networks with dendrites to analyze temporal features. This simple neuron model may also be potentially useful in neural engineering applications.

The authors propose a learning rule for a neuron model with dendrite. In their model, somatodendritic interaction implements self-supervised learning applicable to a wide range of sequence learning tasks, including spike pattern detection, chunking temporal input and blind source separation.

Chunking to improve verbal forward spans in Korsakoff's syndrome

Chunking is a mnemonic strategy that involves organizing information into appropriate units. Our article examined the use of this strategy on forward and backward span performance in Korsakoff's syndrome. Fifteen patients with Korsakoff's syndrome and 17 age-and-education matched healthy controls participated to the study. Digit span performance (both forward and backward) was tested before and after chunking training. Results demonstrated an increased performance on the forward spans after chunking training in the patients with Korsakoff's syndrome, but no beneficial effect was observed on the backward spans in these participants. Controls demonstrated a chunking effect on both forward and backward span performance. Our findings suggest that a simple training in chunking may be useful as part of a cognitive strategy training for improving working memory performance in patients with Korsakoff's syndrome.

The importance of different learning stages for motor sequence learning after stroke

The task of learning predefined sequences of interrelated motor actions is of everyday importance and has also strong clinical importance for regaining motor function after brain lesions. A solid understanding of sequence learning in stroke patients can help clinicians to optimize and individualize rehabilitation strategies. Moreover, to investigate the impact of a focal lesion on the ability to successfully perform motor sequence learning can enhance our comprehension of the underlying physiological principles of motor sequence learning. In this article, we will first provide an overview of current concepts related to motor sequence learning in healthy subjects with focus on the involved brain areas and their assumed functions according to the temporal stage model. Subsequently, we will consider the question of what we can learn from studies investigating motor sequence learning in stroke patients. We will first focus on the implications of lesion location. Then, we will analyze whether distinct lesion locations affect specific learning stages. Finally, we will discuss the implications for clinical rehabilitation and suggest directions for further research.

Skill in discrete keying sequences is execution rate specific

The present study tested the hypothesis that in motor sequences, the interval between successive movements is critical for the type of representation that develops. Participants practiced two 7-key sequences in the context of a discrete sequence production (DSP) task. The 0-RSI group practiced these sequences with response stimulus intervals (RSIs) of 0, which is typical for the DSP task, while the long-RSI group practiced the same sequences with unpredictable RSIs between 500 and 2000 ms. The ensuing test phase examined performance of these familiar and of unfamiliar sequences for both groups under both RSI regimes. The results support our hypothesis that the motor chunks that 0-RSI participants developed could not be used with long RSIs, whereas the long-RSI participants developed sequence representations that cannot be used with 0 RSIs. A new, computerized, sequence awareness task showed that long-RSI participants had limited sequence knowledge. The sequencing skill developed by long-RSI participants can, therefore, not have been based on explicit knowledge.

Backward Planning a Craniomaxillofacial Trauma Curriculum for the Surgical Workforce in Low-Resource Settings

BACKGROUND

Trauma is a significant contributor to global disease, and low-income countries disproportionately shoulder this burden. Education and training are critical components in the effort to address the surgical workforce shortage. Educators can tailor training to a diverse background of health professionals in low-resource settings using competency-based curricula. We present a process for the development of a competency-based curriculum for low-resource settings in the context of craniomaxillofacial (CMF) trauma education.

METHODS

CMF trauma surgeons representing 7 low-, middle-, and high-income countries conducted a standardized educational curriculum development program. Patient problems related to facial injuries were identified and ranked from highest to lowest morbidity. Higher morbidity problems were categorized into 4 modules with agreed upon competencies. Methods of delivery (lectures, case discussions, and practical exercises) were selected to optimize learning of each competency.

RESULTS

A facial injuries educational curriculum (1.5 days event) was tailored to health professionals with diverse training backgrounds who care for CMF trauma patients in low-resource settings. A backward planned, competency-based curriculum was organized into four modules titled: acute (emergent), eye (periorbital injuries and sight preserving measures), mouth (dental injuries and fracture care), and soft tissue injury treatments. Four courses have been completed with pre- and post-course assessments completed.

CONCLUSIONS

Surgeons and educators from a diverse geographic background found the backward planning curriculum development method effective in creating a competency-based facial injuries (trauma) course for health professionals in low-resource settings, where contextual aspects of shortages of surgical capacity, equipment, and emergency transportation must be considered.

Interactive reservoir computing for chunking information streams

Chunking is the process by which frequently repeated segments of temporal inputs are concatenated into single units that are easy to process. Such a process is fundamental to time-series analysis in biological and artificial information processing systems. The brain efficiently acquires chunks from various information streams in an unsupervised manner; however, the underlying mechanisms of this process remain elusive. A widely-adopted statistical method for chunking consists of predicting frequently repeated contiguous elements in an input sequence based on unequal transition probabilities over sequence elements. However, recent experimental findings suggest that the brain is unlikely to adopt this method, as human subjects can chunk sequences with uniform transition probabilities. In this study, we propose a novel conceptual framework to overcome this limitation. In this process, neural networks learn to predict dynamical response patterns to sequence input rather than to directly learn transition patterns. Using a mutually supervising pair of reservoir computing modules, we demonstrate how this mechanism works in chunking sequences of letters or visual images with variable regularity and complexity. In addition, we demonstrate that background noise plays a crucial role in correctly learning chunks in this model. In particular, the model can successfully chunk sequences that conventional statistical approaches fail to chunk due to uniform transition probabilities. In addition, the neural responses of the model exhibit an interesting similarity to those of the basal ganglia observed after motor habit formation.

Discrete Sequential Information Coding: Heteroclinic Cognitive Dynamics

Discrete sequential information coding is a key mechanism that transforms complex cognitive brain activity into a low-dimensional dynamical process based on the sequential switching among finite numbers of patterns. The storage size of the corresponding process is large because of the permutation capacity as a function of control signals in ensembles of these patterns. Extracting low-dimensional functional dynamics from multiple large-scale neural populations is a central problem both in neuro- and cognitive- sciences. Experimental results in the last decade represent a solid base for the creation of low-dimensional models of different cognitive functions and allow moving toward a dynamical theory of consciousness. We discuss here a methodology to build simple kinetic equations that can be the mathematical skeleton of this theory. Models of the corresponding discrete information processing can be designed using the following dynamical principles: (i) clusterization of the neural activity in space and time and formation of information patterns; (ii) robustness of the sequential dynamics based on heteroclinic chains of metastable clusters; and (iii) sensitivity of such sequential dynamics to intrinsic and external informational signals. We analyze sequential discrete coding based on winnerless competition low-frequency dynamics. Under such dynamics, entrainment, and heteroclinic coordination leads to a large variety of coding regimes that are invariant in time.

Planning and navigation as active inference

This paper introduces an active inference formulation of planning and navigation. It illustrates how the exploitation-exploration dilemma is dissolved by acting to minimise uncertainty (i.e. expected surprise or free energy). We use simulations of a maze problem to illustrate how agents can solve quite complicated problems using context sensitive prior preferences to form subgoals. Our focus is on how epistemic behaviour-driven by novelty and the imperative to reduce uncertainty about the world-contextualises pragmatic or goal-directed behaviour. Using simulations, we illustrate the underlying process theory with synthetic behavioural and electrophysiological responses during exploration of a maze and subsequent navigation to a target location. An interesting phenomenon that emerged from the simulations was a putative distinction between 'place cells'-that fire when a subgoal is reached-and 'path cells'-that fire until a subgoal is reached.

Hierarchical Chunking of Sequential Memory on Neuromorphic Architecture with Reduced Synaptic Plasticity

Chunking refers to a phenomenon whereby individuals group items together when performing a memory task to improve the performance of sequential memory. In this work, we build a bio-plausible hierarchical chunking of sequential memory (HCSM) model to explain why such improvement happens. We address this issue by linking hierarchical chunking with synaptic plasticity and neuromorphic engineering. We uncover that a chunking mechanism reduces the requirements of synaptic plasticity since it allows applying synapses with narrow dynamic range and low precision to perform a memory task. We validate a hardware version of the model through simulation, based on measured memristor behavior with narrow dynamic range in neuromorphic circuits, which reveals how chunking works and what role it plays in encoding sequential memory. Our work deepens the understanding of sequential memory and enables incorporating it for the investigation of the brain-inspired computing on neuromorphic architecture.

Itinerancy between attractor states in neural systems

Converging evidence from neural, perceptual and simulated data suggests that discrete attractor states form within neural circuits through learning and development. External stimuli may bias neural activity to one attractor state or cause activity to transition between several discrete states. Evidence for such transitions, whose timing can vary across trials, is best accrued through analyses that avoid any trial-averaging of data. One such method, hidden Markov modeling, has been effective in this context, revealing state transitions in many neural circuits during many tasks. Concurrently, modeling efforts have revealed computational benefits of stimulus processing via transitions between attractor states. This review describes the current state of the field, with comments on how its perceived limitations have been addressed.