About this Research Topic
The term takes on different meanings depending on disciplinary contexts and traditions (e.g. cognitive science / psycholinguistics, neuroscience or computer science), but it generally encompasses one or more aspects of dealing with temporally structured information, including learning, recognition, transduction, prediction and production. Thus, at its core, the problem presents many different facets and degrees of complexity and expresses itself either explicitly in observable behavioral phenotypes or implicitly in abstract cognitive domains. Importantly, the acquisition of sequential structure and regularities in time appears to be a predominantly implicit and unsupervised process, with sensitivity to order and timing being a pervasive characteristic of the mammalian neocortex. Understanding the phenomenon in its many facets and, particularly, grounding its expression in neurobiological mechanisms is fundamental to establish causal links between brain, cognition and behavior.
Regardless of how the process is expressed and the peculiarities of the task at hand, the computation is biophysically realized in neural hardware. To a first approximation, neural circuits are recurrently coupled networks and, as such, are naturally suited to discover structure in temporal information.
Contingent on the architectural details, the power of recurrent circuits spans the entire hierarchy of computational complexity, from recognizing simple sequences to Turing-completeness. In contrast, biophysical models aiming to elucidate cortical operations, such as spiking neural networks in conjunction with non-supervised learning rules, are typically restricted to sequences of significantly lower complexity.
Recent studies unveiled promising candidate mechanisms, ranging from cellular (dendritic computation) and synaptic level (reward-modulated plasticity) to architectures involving distinct cortical and sub-cortical regions. These could serve as bases for new biologically-inspired computational models that go beyond investigating individual task features, validating existing and generating new hypotheses that in turn can drive experimental research. Such computational models can also be instrumental in bridging the conceptual and discursive gaps between neuroscience and the cognitive and linguistic domains, which study the topic through more formal and rigorous frameworks but often rely on highly abstract models to support their findings and/or are agnostic to neural mechanisms.
This Research Topic calls on theoretical, computational and experimental works that explore and advance our understanding of the mechanisms that endow neural circuits with general capabilities for learning and processing sequential structure. Relevant problems may range from pattern perception, examining the timing and temporal properties of pattern representations, to rule learning, segmentation and chunking, or compositionality. These and other aspects may be explored in the context of any area of cognition and behavior (e.g., navigation, reasoning, language, etc).
We aim to bring together contributions that investigate biophysical processes and thus require a minimal degree of biological verisimilitude. However, artificial neural networks incorporating clear biological properties, such as excitation and inhibition, will also be considered. Authors are invited to focus on (but not limited to) novel mechanisms, architectures and computational principles supporting sequence processing, including neuronal morphology, plausible (local or reward-based) learning rules and circuit motifs. Additionally, we welcome modeling studies rooted in the cognitive sciences, psychology or psycholinguistics, which involve established paradigms and more complex tasks, and attempt the transition from abstract to biologically more detailed models. The scope further extends to neurophysiological experiments on both human and non-human subjects, that scrutinize the neural activity/biophysical processes during the aforementioned or similar behavioral tasks.
Even though we primarily welcome Original Research submissions, we also accept other types of articles. We particularly encourage Perspective and Opinion pieces that provide a synergistic view and strategies for integrating cross-domain knowledge towards a deeper, mechanistic understanding of how the brain exploits temporal regularities in the environment.
Keywords: Sequence processing, Temporal sequences, Symbolic sequences, Serial order, Sequential behavior, Biophysical models, Spiking networks, Biological learning, Functional circuits, Rule learning, Synaptic plasticity, Neurophysiological experiments
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.