RESEARCH DOSSIER
Neural Cell Types and Functional Circuit Organization
Abstract. Neural cell types and functional circuit organization describe the relationships between cellular identity, spatial anatomy, synaptic connectivity, physiological dynamics, and experimentally established behavioral function. This fundamental research monograph synthesizes the historical development of cellular neuroscience and circuit physiology, molecular and multimodal cell taxonomies, excitatory and inhibitory projection systems, astrocytic and microglial contributions, local and interareal circuit motifs, and the methods used to connect structure to function. Original experiments and contemporary atlases are analyzed through histology, transcriptomics, Patch-seq, paired electrophysiology, calcium imaging, electron microscopy, connectome reconstruction, and selective perturbation. Comparative evidence from retina, neocortex, insect nervous systems, and human–mouse cell atlases reveals both recurring organizational principles and important limits to cross-species inference. The central synthesis distinguishes anatomical wiring, observed activity, and causal contributions to psychological capacities while examining sampling, reconstruction errors, state dependence, glial regulation, and animal-level replication. The work is a critical narrative synthesis of published scientific sources and reports no new empirical data or formal meta-analysis.
Keywords: neural cell types; neuronal taxonomy; glia; astrocytes; microglia; inhibitory interneurons; projection neurons; neural circuits; synaptic connectivity; functional connectomics; electron microscopy; single-cell transcriptomics; Patch-seq; circuit motifs; neural causality; behavioral neuroscience
Scientific Scope and Levels of Explanation
Neural cell types and functional circuit organization concern the relations among cellular identity, spatial position, synaptic connection, physiological operation, and the generation of organized behavior. The research object spans the molecular features of neurons and glia, projection patterns and synaptic arrangements, and the operation of interacting populations during defined tasks. An account of circuit organization must specify its unit of observation: a genetically profiled cell, an anatomical neuron class, a directed edge, a recurrent population, a brain region, or an organism performing a task. These units produce complementary descriptions, but their definitions do not automatically coincide. Work synthesizing the neocortical circuit emphasizes recurring motifs alongside substantial anatomical and functional variation Harris and Shepherd (2015).
This monograph uses a critical narrative synthesis of published original experiments, anatomical atlases, physiological measurements, comparative studies, and contemporary connectomic data. It neither reports newly collected measurements nor estimates pooled intervention effects. Evidence is assessed according to what a method actually identifies: transcriptional clustering provides molecular similarity, tracers expose paths, paired recordings reveal physiologically detectable connections, and behavior under selectively altered activity can support causal claims. A structural graph documents potential routes of influence; the consequences of traversing those routes depend on synapse strength, cell state, context, and ongoing activity. The distinction is central to the interpretation of connectomic research Bargmann and Marder (2013).
The analysis follows a practical sequence from historical definitions to cell taxonomy, circuit motifs, measurement, causal inference, and psychological explanation. The main questions are whether cell categories retain functional meaning across brain regions and species; which circuit motifs are sufficiently reliable to support explanatory generalization; how nonneuronal cells alter those motifs; and which experimental combinations establish a connection between a reconstructed pathway and behavior. Questions about channel gating, detailed reward learning, and long-term synaptic plasticity are addressed only where their evidence changes interpretation of circuit architecture. They belong principally to neighboring research dossiers, whereas this work evaluates the organization and evidential status of the network itself Tremblay et al. (2016).
From Neuron Doctrine to Circuit Physiology
The late nineteenth-century differentiation of individual nerve cells provided the anatomical basis for later circuit descriptions. Light microscopy, selective stains, and observations of neurites made it possible to describe discrete cells and directional arrangements. The decisive methodological transition occurred as investigators began to connect morphology with measurable electrical events. In the squid giant axon, Hodgkin and Huxley derived a quantitative account of membrane currents and action-potential generation, providing a physiological account of excitability that could be related to neuronal communication Hodgkin and Huxley (1952). Its mathematical success concerned membrane dynamics, not the complete organization of a behavioral network, and the distinction remains important when a single-cell mechanism is generalized.
Hubel and Wiesel's physiological studies of cat primary visual cortex demonstrated systematic response selectivity for oriented contours and spatial stimulus configurations Hubel and Wiesel (1962). Their findings made receptive-field measurement a central way to identify functional differences among neurons. The experiments established that local populations participate in differentiated sensory computations, yet a receptive field alone does not reveal every synaptic source that generates it. Subsequent connectomic and imaging methods therefore complemented rather than displaced classical extracellular recordings. The history of circuit neuroscience is best understood as the gradual accumulation of mutually constraining measurements: morphology, electrical response, projection, and task-dependent intervention.
Modern cell and circuit classifications inherit these historical methodological divisions. A neuron may exhibit a stable firing phenotype in one preparation while changing its response with behavioral state; a projection-defined class may contain multiple transcriptomic subclusters; a cell with a similar molecular signature may participate in substantially different local networks. The synthesis of local cortical connectivity emphasizes that classification by location, morphology, physiology, and molecular signature captures related but distinct aspects of organization Jiang et al. (2015). A historical theory retains scientific value when it generates specific tests. Its descriptive vocabulary, however, should not determine which distinctions count as functional without independent experimental assessment.
Defining Neurons, Glia, and Cell Types
Neurons are polarized excitable cells whose electrical and chemical processes support signaling through specialized compartments and contacts. Glial populations include astrocytes, oligodendrocyte-lineage cells, and microglia, with heterogeneous developmental origins and functions. The term cell type requires an explicit classification rule. Molecular taxonomies partition cells according to expressed genes and other features; morphological taxonomies use dendritic and axonal structures; physiological classes use response patterns and intrinsic properties; projection classes are identified by source and target. These criteria can agree in some circumstances and diverge in others. Transcriptomic work across mouse neocortical areas establishes both shared and area-dependent signatures Tasic et al. (2018).
The scientific meaning of a category is determined in part by its reproducibility across animals, regions, developmental periods, measurement platforms, and analytic pipelines. A cluster defined by expression thresholds is an empirical grouping within a dataset and resolution, while a behaviorally meaningful cell type requires a stronger set of convergent claims about function and causal participation. Human–mouse comparative transcriptional analysis identifies conserved broad organizational features together with substantial species-dependent cellular characteristics Hodge et al. (2019). Thus a cell label transferred across species must be accompanied by evidence concerning which measured properties persist, and at what level of resolution.
Cell identity is also spatial. A transcriptional population may occur in particular layers or nuclei, and its contribution depends on its inputs, outputs, and anatomical neighbors. Spatial registration and molecular profiling have transformed the scale at which such distributions can be studied, but the choice of clustering algorithm, sampling coverage, and gene selection affects boundaries between proposed types. The whole-mouse-brain atlas developed by Yao and colleagues illustrates the scope of contemporary molecular classification and the challenge of interpreting extensive hierarchies of cell populations Yao et al. (2023). For this monograph, type is a testable multi-criterion category whose role must ultimately be evaluated in a specified circuit and task.
Historical Methods of Functional Localization
Localization emerged through a sequence of comparisons among anatomy, neurological deficits, stimulation effects, sensory mapping, and pathway analysis. A lesion can show that a region is necessary for the measured capacity under particular conditions, although effects may reflect interruption of passing fibers, network diaschisis, adaptation, or changes in motor expression. Electrical recording can reveal task-associated activity but leaves open whether that activity is a cause, consequence, or correlate of the observed operation. Linking the two methods creates stronger constraints on explanation because correspondence between a response map and a behavioral deficit is more informative than either observation alone. Classical sensory physiology demonstrated why receptive fields required an anatomical and physiological interpretation Hubel and Wiesel (1962).
Anatomical pathways are organized at multiple scales. Long-range projections connect distributed cortical and subcortical territories, while local circuits transform incoming activity through recurrent and feedforward interactions. In their comparative architectural analysis of primate cortical areas, Felleman and Van Essen described distributed hierarchical processing with multiple interareal connections Felleman and Van Essen (1991). The contribution was an anatomically grounded network account rather than an assertion that information always travels in one direction or in a single serial chain. Contemporary studies accordingly distinguish ascending and descending projections, laminar origins and terminations, and dynamic patterns of integration across tasks.
Functional localization is therefore a conditional causal assignment rather than a permanent one-to-one mapping from a brain coordinate to a mental faculty. The same anatomical territory can participate in multiple tasks, and the same behavioral output can arise through alternative circuits or compensatory routes. Histological maps help establish the spatial reference needed for cross-study comparison. The ultrahigh-resolution postmortem human brain model known as BigBrain illustrates how three-dimensional cytoarchitectonic registration can enrich structural analysis while remaining separate from measurements of live network dynamics Amunts et al. (2013). A defensible circuit claim identifies anatomical extent, experimental preparation, operational outcome, and the limits of generalization.
Anatomical Organization Across Scales
A neural circuit can be represented as a multiscale system of cells, directed synaptic edges, pathways, and interacting regions. At the microscopic scale, dendritic arborization, axonal branches, synaptic location, and local cellular neighborhoods shape the opportunities for signal exchange. At the mesoscopic scale, projections aggregate into identifiable layers, nuclei, pathways, and recurrent modules. At the macroscopic scale, regional networks coordinate activity over anatomical distances. These scales are linked through physical tissue but they measure different properties, and an average connection strength at one level cannot be interpreted as the full state of individual synapses at another Harris and Shepherd (2015).
Directed anatomical connectivity indicates which signals might travel through a system, but functional coupling must be inferred from activity measured within a defined state. A pair of strongly correlated neurons need not be directly connected: shared inputs, feedback paths, and common changes in arousal can create correlation without a monosynaptic edge. Conversely, anatomically connected cells may show weak contemporaneous correlations because their connection is gated by timing, inhibition, or synaptic efficacy. Bargmann and Marder describe this gap between a structural connectome and its active physiological realization, emphasizing why the same wiring diagram can support multiple operating modes Bargmann and Marder (2013).
Modern reconstruction increases structural precision while introducing new questions of sampling and representation. Each connectomic volume covers a finite tissue extent, and neurons whose axons leave the imaged region can have missing distant targets. Automated segmentation must identify processes across dense electron microscopy images, and synapse detection introduces uncertainty that propagates into graph measures. The informative network object is therefore a graph accompanied by its acquisition conditions, annotation reliability, and behavioral context. The analysis of local cortical circuits by Song and colleagues shows how nonrandom connection patterns can reveal candidate motifs, while interpretation requires attention to the recording selection process and anatomical population sampled Song et al. (2005).
Excitatory Projection Neurons and Long-Range Outputs
Excitatory projection neurons occupy diverse cortical and subcortical positions and create many of the long-range paths by which distinct regions exchange activity. Their classification often combines transmitter phenotype, projection target, cortical layer, dendritic morphology, and molecular identity. In neocortex, corticocortical, corticothalamic, and subcerebral projection classes differ in principal outputs and laminar distributions, but a broad projection label can still include subtypes with different intrinsic conductances, local inputs, and task responses. This variation is central to contemporary descriptions of the neocortical circuit Harris and Shepherd (2015). The behavioral meaning of a projection depends on the larger pathway it joins and the state of downstream targets.
Connectivity must be studied directionally. Tracing a neuron from a region of origin to one or more destinations identifies a pathway, while postsynaptic recordings test whether that pathway transmits a detectable effect in a particular preparation. Long-range signaling may recruit direct excitation, local inhibitory interneurons, or broader recurrent networks. A cell's axonal destination is thus a useful part of its identity without being a complete specification of the computation performed. The extensive connectivity study by Jiang and colleagues demonstrated that morphologically differentiated cells participate in connectivity patterns that depend on class and synaptic partner Jiang et al. (2015). Such findings motivate projection-resolved experiments rather than treating all glutamatergic neurons in one region as interchangeable.
Single-cell atlases permit a further refinement of projection categories through coordinated measurements of gene expression and anatomical location. Tasic and collaborators found molecularly distinct and partially shared cell populations across cortical areas, illustrating why area should be part of the classification model Tasic et al. (2018). Yet transcriptomic similarity does not directly estimate transmitter release, synaptic weight, or behavioral necessity. A responsible circuit interpretation triangulates molecular class, projection pattern, recorded response, and perturbation. The resulting explanation can identify when changing an output-defined cell population alters the propagation of a signal, while keeping separate whether the manipulation specifically affects perception, motor execution, attention, or another downstream variable.
Inhibitory Interneurons and Cell-Type-Specific Control
Inhibitory interneurons are often organized by complementary molecular, physiological, and anatomical features. Broad populations expressing parvalbumin, somatostatin, or vasoactive intestinal peptide provide useful entry points into cortical organization, while each broad population contains substantial diversity. Many parvalbumin-associated interneurons strongly influence perisomatic excitability; several somatostatin-associated groups target dendritic compartments; vasoactive intestinal peptide-associated pathways frequently participate in disinhibitory motifs. These are common patterns rather than universal identities across every cortical area and developmental stage. Tremblay, Lee, and Rudy synthesize the relationship between cortical interneuron morphology, targeting patterns, and network operation Tremblay et al. (2016).
The functional impact of inhibition depends on synaptic position, membrane properties, network timing, and the level of ongoing excitation. Perisomatic inhibition can modulate the transfer of a neuron's integrated input to its spike output; dendritic inhibition can regulate particular incoming branches; interneuron-to-interneuron inhibition can release downstream populations from local restraint. A circuit can therefore use multiple inhibitory routes to implement context-dependent changes in responsiveness. Jiang and colleagues mapped physiologically measurable connections among morphologically differentiated adult cortical neurons, helping expose the specificity and asymmetry of local inhibitory networks Jiang et al. (2015). Their observations concern sampled connection probabilities and synaptic dynamics rather than exhaustive wiring across the brain.
Experiments that selectively silence or excite genetically defined interneurons can test their contribution to sensory gain, temporal coordination, or behavioral choice. Interpretation nevertheless requires knowledge of which subtypes were actually targeted, whether network feedback compensated for the manipulation, and whether the behavioral effect could arise from altered arousal or movement. The presence of a molecular marker alone is insufficient to identify a unique psychological operation. Comparing inhibition across regions is particularly informative when the same molecular labels correspond to similar but not identical physiological roles. Circuit accounts should state the class definition, measured downstream effect, and context in which a proposed inhibitory motif operates Tremblay et al. (2016).
Astrocytes and Synaptic Microenvironments
Astrocytes organize a substantial component of the cellular environment in which neuronal circuits operate. Their processes contact synapses and vascular structures, participate in extracellular ion regulation, and help influence transmitter clearance and metabolic support. A neuronal connection therefore has a physiological setting comprising neighboring cells, extracellular chemistry, and state-dependent regulatory processes. The review by Allen and Eroglu integrates evidence that astrocytes contribute to synapse formation, maturation, maintenance, and elimination through multiple molecular interactions Allen and Eroglu (2017). These results support treating local circuit organization as a multicellular system, although individual molecular mechanisms vary across developmental stage, brain region, and experimental model.
Demonstrating an astrocytic influence on a circuit requires care about intervention specificity and timing. Manipulations that change extracellular potassium handling, glutamate uptake, intracellular signaling, or metabolic supply can alter neuronal excitability through several routes simultaneously. An observed change in network activity may therefore indicate a causal glial contribution without uniquely identifying the intended molecular pathway. Physiological and imaging experiments can establish how astrocytic processes respond to local activity, but measured calcium signals should not automatically be interpreted as direct records of information transmission equivalent to neuronal action potentials. The cell biology of astrocyte–synapse interactions provides an explanatory framework for these distinct types of evidence Allen and Eroglu (2017).
Astrocytes also illustrate the importance of scale. A single cell can interact with many local contacts, yet each contact may experience different molecular and temporal conditions. A synaptic map reconstructed from electron microscopy captures spatial proximity and ultrastructure while omitting much of the ongoing transport and regulation that alters signal transfer. Consequently, a complete model of functional circuit organization must allow glial variables to change the effective behavior of a fixed neuronal graph. It should also distinguish demonstrated cellular interactions from broader claims about consciousness, cognition, or behavioral regulation that require separate task-based testing. The gap between structural connectivity and network function is a general problem for connectomics Bargmann and Marder (2013).
Microglia, Synaptic Remodeling, and Development
Microglia are resident immune-related cells of the central nervous system that contribute to tissue surveillance, development, and the remodeling of neural environments. During development, neural circuits undergo changes in synapse number, partner selection, and functional refinement that are related to experience and spontaneous activity. Stevens and colleagues reported a role for the classical complement cascade in developmental central nervous system synapse elimination, establishing a molecular mechanism through which synaptic elements can be marked for removal Stevens et al. (2007). Their work helped connect developmental pruning to cellular processes that could be experimentally disrupted and assessed.
Schafer and colleagues subsequently described activity- and complement-dependent microglial contributions to the refinement of postnatal neural circuits Schafer et al. (2012). The causal inference arises from combinations of anatomical observations, molecular interventions, and developmental outcomes. It is important to distinguish identification of an engulfment-related process from proof that all pruning across brain regions uses one mechanism. Synapse elimination can involve multiple cell types and signaling systems, and the effect of manipulating a complement pathway may depend on the age, region, and condition of the animal. Developmental circuit organization is therefore better explained through defined processes and contexts than a universal model of microglia as simple synapse removers.
In psychological research, microglial circuit effects matter because early changes in connectivity can influence the maturation of sensory, motor, and social behavior. A developmental behavioral difference after an immune or complement manipulation is evidence of organism-level consequences, but it does not by itself identify which synapses or cell types mediate the effect. Multi-level studies would link target engagement, structural remodeling, physiological operation, and behavior while evaluating competing changes in general health and activity. Contemporary glial research also motivates broader definitions of neural circuit membership, provided that interpretations retain the distinction between a molecular mechanism demonstrated in a preparation and a psychological function inferred from a behavioral task Schafer et al. (2012).
Molecular Atlases and Multimodal Cell Taxonomy
Single-cell and single-nucleus transcriptomics make it possible to classify large populations according to patterns of gene expression. Repeated sampling across anatomical locations reveals major neuronal and glial families as well as nested subtype distinctions, producing hierarchical molecular atlases. The whole-mouse-brain analysis of Yao and colleagues shows how extensively these cell populations can be mapped across diverse regions Yao et al. (2023). Related work by Zhang and collaborators spatially locates molecularly defined populations within intact tissue, adding positional information needed to interpret cell classes in circuit context Zhang et al. (2023). Together, these strategies increase taxonomic coverage while making the criteria used to define a stable type more explicit.
Molecular classification is a measurement-based inference. Tissue dissociation or nuclear isolation can alter which populations are recovered; expression profiles depend on detection depth, cell state, sample age, tissue preservation, and analytic choices. Clustering resolution determines whether a population appears as one group or several related subgroups. Spatial methods preserve location while often measuring a different set of features or at different sensitivity. The appropriate interpretation of a transcriptomic cluster therefore includes its sampling frame, discriminating markers, and relationship to other classification modalities. Tasic and colleagues provide a foundational comparison of transcriptomic identities across cortical areas, revealing both recurring and region-specific features Tasic et al. (2018).
Multimodal taxonomy seeks to establish when molecular identity aligns with morphology, projection, electrophysiology, and spatial niche. Patch-seq integrates electrical recording, cellular morphology, and transcriptome profiling from individual neurons, offering a direct means of testing relations among these descriptive axes Cadwell et al. (2016). Even under ideal measurement, functional identity may be conditional: the same cell can contribute differently to behavior as inputs, internal state, and network partners change. A practical taxonomy for circuit explanation should therefore report both stable defining properties and context-dependent functions. Cross-species comparison becomes especially important because molecularly corresponding cell populations may differ in proportion, morphology, and gene expression Hodge et al. (2019).
Synapses, Connectivity, and Effective Influence
Synapses are anatomical and physiological sites where communication between cells can occur, yet a count of synaptic contacts is not a complete measure of effective influence. A connection may be excitatory, inhibitory, modulatory, or mixed in its network consequences, depending on transmitter, receptors, postsynaptic compartment, and surrounding circuit. Functional connectivity describes statistical dependence among recorded signals; effective connectivity adds a model of directed influence; structural connectivity describes anatomical adjacency or pathways. These terms should be operationally distinguished because an indirect route or common input can generate strong statistical coupling between cells without any direct synaptic edge. Bargmann and Marder explain why connectome data must be placed within dynamic models Bargmann and Marder (2013).
Paired intracellular recording can reveal whether one cell produces a detectable postsynaptic response in another and estimate features such as latency, short-term dynamics, and connection reliability. It nevertheless samples selected cells accessible within a slice, and a missing measured response can reflect severed axons, weak contacts, or the recording conditions. Song and colleagues identified nonrandom connection patterns among local cortical neurons, illustrating that a circuit is not simply a homogeneous random network Song et al. (2005). Such results motivate hypotheses about recurrent subgraphs, but they do not make a unique behavioral computation emerge from the connectivity statistics alone. Independent physiological and behavioral tests remain necessary.
At the population level, the effect of a synapse depends on network state and convergent input. Presynaptic firing that is strongly effective during one behavioral condition may have little detectable consequence during another, because inhibitory conductance, membrane potential, neuromodulation, and competing activity change the transfer function. A complete functional model should therefore join directed graphs with temporal dynamics and specify the counterfactual intervention that would test each proposed influence. Measurements can support different inferential strengths: an anatomical contact supports a possible route, a recorded monosynaptic response supports a functioning connection, and a selective perturbation with behavioral controls can help establish causal participation Jiang et al. (2015).
Feedforward, Feedback, and Recurrent Motifs
Circuit motifs describe recurring arrangements of connections that can support characteristic information transformations. A feedforward path sends activity from an upstream population to a downstream population; a feedback path returns information toward an earlier processing stage; recurrent motifs contain loops whose states depend on prior activity. Local inhibitory arrangements can implement gain control, competition, disinhibition, or temporal coordination, while excitatory recurrence can amplify or sustain activity. The existence of a particular motif in a reconstructed circuit is a structural statement. Whether it performs a proposed operation depends on synaptic properties, external drive, and the broader network in which it is embedded Harris and Shepherd (2015).
Nonrandom structural motifs have been documented in cortical slice preparations. Song and collaborators reported patterns of strong connections and connectivity motifs that differed from simple independent-connection expectations Song et al. (2005). The results helped motivate circuit theories based on structured recurrence and assemblies, but slice selection and the limited sampling of neuron pairs constrain generalization. A motif's function cannot be read directly from a static wiring diagram: a recurrent loop can stabilize activity, amplify noise, or create oscillatory dynamics depending on parameter values and input. Simulation and direct perturbation should therefore be treated as tests of circuit hypotheses, not inevitable interpretations of motif counts.
Feedback also complicates causal localization because a perturbation at one stage can alter signals far from its anatomical origin. In distributed hierarchies, ascending and descending pathways connect multiple areas, and responses are shaped by expectations, ongoing motor plans, and changes in behavioral state Felleman and Van Essen (1991). A local electrode may register consequences of this larger network rather than only local stimulus computation. Multi-site measurements, targeted disruption, and temporally resolved analysis can distinguish candidate directions of influence. The most useful motif descriptions include the identity of participating cell classes, connection direction, signal timing, and experimental circumstances in which a proposed functional operation is observed.
Cortical Layers and Interareal Hierarchies
Cortical lamination supplies an anatomical coordinate system for describing recurrent local interactions and long-range routes. Laminar position is associated with typical but nonexclusive patterns of input, output, and local connectivity. Interareal projection systems differ in the layers from which axons arise and the layers in which they terminate, providing one basis for hierarchical descriptions. Felleman and Van Essen's comparative analysis of primate visual pathways made such distributed hierarchy a foundational network concept Felleman and Van Essen (1991). It demonstrated extensive reciprocal connectivity and parallel pathways, which limits simple representations of a fixed serial processing ladder.
Laminar identification must be combined with cell class and region. A transcriptomic cell population sampled from one cortical layer cannot be assumed to perform the same computation in a different area, even when related marker genes are expressed. The neocortical synthesis of Harris and Shepherd highlights how recurrent circuit themes coexist with area-specific details Harris and Shepherd (2015). Tasic and collaborators further show that cortical areas contain transcriptionally related yet distinguishable populations Tasic et al. (2018). This convergence supports the use of laminar information as a powerful anatomical descriptor while resisting a single uniform cortical algorithm imposed independently of evidence.
The functional importance of hierarchy depends on empirical task conditions. Sensory evoked responses can propagate through several pathways, and feedback can alter the responses of neurons at earlier stations. Activity correlations within a region can reflect global arousal, eye movements, or task-dependent input as readily as a local computation. Testing a hierarchical mechanism therefore requires appropriately timed interventions, anatomical knowledge of projection direction, and measurements of relevant behavioral variables. Models should state whether hierarchy refers to anatomical order, temporal processing, representational abstraction, or causal influence. These properties may correlate in a task, but no single measurement establishes all of them Felleman and Van Essen (1991).
Retinal Circuits and Direction Selectivity
The retina offers a particularly tractable system for linking cell types, precise anatomical contacts, and sensory computation. Photoreceptors, bipolar cells, amacrine cells, and retinal ganglion cells participate in layered microcircuits whose response properties can be measured under controlled visual stimulation. Different ganglion-cell populations encode features of visual scenes, and circuit arrangements contribute to temporal sensitivity, contrast processing, and direction selection. The strength of the retinal example lies in the possibility of measuring identifiable stimulus transformations while reconstructing local partner relationships. Hubel and Wiesel's cortical receptive-field research established a parallel physiological tradition at a different anatomical stage of vision Hubel and Wiesel (1962).
Briggman, Helmstaedter, and Denk combined physiological characterization and high-resolution circuit reconstruction to examine wiring specificity in retinal direction-selectivity circuits Briggman et al. (2011). The study demonstrates how a functional phenotype and an identified pattern of connections can constrain each other. Such convergence provides more than a correlation between an anatomical region and sensory behavior: it connects candidate cell populations and synaptic partner selection to a well-specified computational property. Nevertheless, the results concern the investigated retinal circuit and experimental preparation. General claims about cortical computation or organism-level visual decisions require additional evidence.
Retinal circuitry also illustrates that cellular function can depend on relationships among several classes rather than one uniquely decisive neuron type. An identified ganglion-cell response emerges within a network of excitatory and inhibitory input, temporal filtering, and patterned connectivity. Altering a cell class may disrupt selectivity by changing a necessary connection or by indirectly shifting the balance of the network. Researchers can strengthen conclusions by combining anatomical identification, receptive-field measurement, paired physiology, and selective perturbations. The retinal model thus supplies a clear methodological principle for the wider monograph: a convincing functional circuit explanation links named cells, directed interactions, quantitative signals, and an outcome that is defined independently of the intervention Briggman et al. (2011).
Network States, Neuromodulation, and Degeneracy
A nervous system's connection graph persists across many short-term changes in behavior, while the operating state of that graph can change rapidly. Neuromodulators alter intrinsic membrane properties, synaptic release, and interactions among cells, changing which pathways are functionally influential. Marder reviews how identified neural circuits can produce different outputs under distinct modulatory conditions Marder (2012). This evidence establishes an essential limit of purely structural explanations: the same set of anatomical edges can support more than one effective dynamical organization. Circuit identity must therefore include information about state variables when predictions concern timing, gain, persistence, or behavioral expression.
Degeneracy describes circumstances in which different cellular or parameter configurations produce comparable functional outcomes. This is an empirical problem for causal inference because an intervention can be compensated by other pathways, while similar behavioral outputs may mask different neural implementations. In a highly recurrent network, acute inhibition of one node can have broad downstream effects and longer-term changes may recruit alternative activity patterns. Bargmann and Marder emphasize the importance of bridging connectome descriptions with physiological dynamics Bargmann and Marder (2013). Their argument complements local measurements: a reliable anatomical map is indispensable, yet the model must also specify how the system operates under the conditions being explained.
For psychological explanation, state dependence means that circuit findings should be accompanied by behavioral context and measurement of arousal, movement, sensory engagement, and relevant internal variables when feasible. A population classified through stable transcriptional markers can take part in different causal chains during exploration, attention, or rest. Broad statements that a cell type produces a single mental function obscure this variability. A stronger formulation specifies the intervention, the predicted network-level response, the measured behavioral operation, and the domain of conditions under which the effect appears. Such descriptions preserve mechanistic precision while allowing stable anatomical structure and variable functional regimes to coexist Marder (2012).
Histology, Cytoarchitecture, and Spatial Registration
Histology provides the spatial foundation for cell and circuit research. Tissue staining, immunolabeling, serial sections, and high-resolution reconstruction reveal cell density, morphology, laminar boundaries, fiber organization, and the distributions of molecular markers. Cytoarchitectonic boundaries can be related to nuclei and cortical fields, allowing the anatomy studied in one preparation to be compared with that in another. The BigBrain project introduced an ultrahigh-resolution three-dimensional model of a human postmortem brain based on histological reconstruction Amunts et al. (2013). Such reference models enable anatomical registration and quantitative tissue descriptions, but they do not directly record signaling, plasticity, or behavior in living subjects.
Spatial correspondence is itself an empirical inference. Boundaries defined by cellular density may differ from boundaries drawn by gene expression, connectivity, functional imaging, or macroscopic landmarks. Tissue shrinkage, section deformation, individual variation, and algorithmic registration can introduce error. Researchers should therefore state the coordinate system, anatomical landmark, segmentation criterion, and uncertainty that support each location claim. Whole-brain spatial cell atlases provide a complementary strategy by linking molecular populations to their positions within tissue Zhang et al. (2023). Their spatial resolution and transcript detection vary with methodology, making convergence between independent registration methods particularly valuable.
Anatomical maps become explanatory when linked to a specified physiological or behavioral effect. Identifying a cell-rich layer is descriptive; showing that neurons originating there communicate with a target is connective; demonstrating that the pathway influences a defined task is functional. These conclusions can be assembled across studies when the underlying anatomical labels align reliably. Incomplete spatial registration can otherwise create the appearance of a dispute where studies sampled different subdivisions or species. The defensible goal is an atlas whose boundaries remain traceable to observational procedures and whose correspondence with circuit function is established through separate evidence, not assumed from anatomical elegance Amunts et al. (2013).
Patch Clamping, Patch-seq, and Functional Phenotyping
Intracellular and patch-clamp methods characterize neuronal membrane properties and synaptic responses through controlled current or voltage manipulation. They can measure intrinsic excitability, synaptic currents, and responses to defined inputs, making them indispensable for identifying mechanisms that cannot be inferred from transcriptional data alone. Hodgkin and Huxley's work illustrates the historical importance of quantitative electrophysiology for causal models of membrane behavior Hodgkin and Huxley (1952). Paired recordings extend the logic to local circuits by examining whether activation of a presynaptic cell generates a measurable postsynaptic response. Their sensitivity, however, depends on slice integrity, recording conditions, and the class of connections under investigation.
Patch-seq links an individual recorded neuron's electrical properties to morphology and molecular identity. Cadwell and colleagues developed a profiling approach combining electrophysiological, transcriptomic, and morphological observations from single neurons Cadwell et al. (2016). This cross-modal integration can test whether a transcriptomic cluster corresponds to a reproducible pattern of electrical behavior or dendritic architecture. Disagreement among features may indicate genuine biological diversity or measurement limitations. Consequently, a cluster's validity should be evaluated by independent predictive performance rather than simply by whether several descriptive methods produce visually similar groupings.
An experimental circuit model requires additional sampling and causal tests after phenotyping. A neuronal response elicited by current injection describes cellular capacity within a preparation; it does not establish the cell's contribution during natural movement or perception. Experiments that relate Patch-seq profiles to projection targets, in vivo responses, and behavioral perturbations provide increasingly informative triangulation. Cell dissociation and transcript recovery may introduce selection effects, while morphological reconstruction may be truncated in slices. Explicit reporting of these limitations permits comparison across laboratories and clarifies which conclusions concern cellular identity, synaptic connectivity, or behavioral relevance Cadwell et al. (2016).
Electron Microscopy and Connectome Reconstruction
Serial electron microscopy reveals synaptic ultrastructure and allows researchers to reconstruct neurites and their candidate synaptic partners within dense tissue. By converting images into segmented neurons and annotated contacts, connectomics generates directed graphs with spatially explicit evidence. Bock and collaborators integrated in vivo visual physiology with electron microscopy of cortical neurons, demonstrating an early strategy for jointly studying responses and their local anatomical context Bock et al. (2011). Reconstructing an entire local volume provides remarkable coverage of structures that are difficult to infer from paired recordings, but the measured graph is still limited to the tissue captured and the reliability of its annotations.
Modern reconstruction requires image alignment, automated segmentation, proofreading, synapse identification, and error-aware graph analysis. Fragmented processes, erroneously joined neurites, ambiguous contacts, and edges crossing a volume boundary can affect inferred motifs, path lengths, and cell classes. An observed contact is strong evidence of an anatomical relationship when validated, but contact morphology alone is not sufficient to estimate physiological gain or determine behavioral necessity. The MICrONS Consortium combined large-scale structural reconstruction with functional recordings in mouse visual cortex, making these distinctions unusually visible at population scale MICrONS Consortium (2025). Its contribution is a rich joint dataset whose causal interpretation still requires experimental triangulation.
The graph abstraction is both powerful and selective. It can represent neurons as nodes and synapses as edges while retaining annotations for location, contact number, cell class, and morphological features. Yet it compresses continuous geometry, synaptic molecular composition, local extracellular conditions, and activity history into simplified variables. Researchers should therefore preserve the underlying reconstruction metadata and state the assumptions used to transform tissue into a network model. Adult fly-brain work demonstrates the feasibility of dense whole-brain wiring descriptions in a defined species, while also showing why an anatomical connectome is a starting point for behavioral modeling rather than a complete dynamic account Dorkenwald et al. (2024).
Calcium Imaging, Population Measurement, and State
Calcium imaging permits simultaneous measurement of activity-associated fluorescence across many neuronal elements, often during sensory stimulation or behavior. It has greatly expanded the number of cells that can be related to local network dynamics and anatomical structures. Fluorescence signals, however, depend on indicator kinetics, expression, imaging depth, motion correction, and the process that converts intracellular calcium changes into an estimate of spiking. A calcium transient can be informative about activity without matching the temporal precision of an extracellular action-potential recording. Studies joining visual function with reconstruction illustrate the value of measuring responses before assigning anatomical connectivity Bock et al. (2011).
Population recordings also face inferential limitations because correlated activity can arise from common inputs, slow behavioral state changes, and recurrent interactions. Selectivity measured in a visual task may vary with locomotion, gaze, arousal, or preceding stimulation. Functional connectomics addresses some of these limitations by registering recorded neurons to anatomical reconstructions, allowing researchers to ask whether cells with particular response properties exhibit distinctive connection patterns. The large-scale mouse visual-cortex study conducted by the MICrONS Consortium joins functional measurements and synaptic reconstruction within a defined tissue volume MICrONS Consortium (2025). The dataset supports analyses of relations between tuning and wiring, while its sampling remains restricted to its animal and preparation.
Statistical analysis should distinguish decoding accuracy, response reliability, and neural causality. A decoder can predict a stimulus or behavior from population activity while exploiting shared covariance or task-related signals that are incidental to the proposed computation. Cross-validation can assess predictive stability but does not establish whether the decoded cells are necessary for a behavioral capability. A stronger experimental program combines activity measurement with interventions and independent outcome variables, retaining temporal and spatial controls. The interpretation of simultaneous neural data should also make explicit whether the outcome is a sensory feature, a behavioral response, an internal state, or an abstract model parameter Bargmann and Marder (2013).
Perturbation, Lesions, and Behavioral Causal Inference
Causal claims about circuit function require interventions whose targets and off-target consequences can be evaluated. Lesions, pharmacological manipulations, genetic disruptions, electrical stimulation, and optogenetics have different spatial and temporal profiles. Boyden and colleagues demonstrated millisecond-timescale optical control of genetically targeted neurons, providing a major tool for selectively altering activity Boyden et al. (2005). Modern applications can test whether changing a defined cell population or projection affects a measured outcome. The inference still depends on actual targeting specificity, expression pattern, light delivery, physiological confirmation, and whether the task distinguishes sensory from motor or motivational changes.
Necessity and sufficiency should be understood within specified experimental conditions. Silencing a pathway during a task can support a necessity claim for normal performance in that task, but the deficit may arise from general changes in movement, engagement, or attention. Activating a population and observing an outcome supports intervention-dependent efficacy while leaving open whether natural behavior uses the same mechanism. Compensation and network feedback can also make a necessary circuit appear dispensable under chronic manipulation. Comparisons among acute and chronic interventions, dose or intensity controls, temporal controls, and task variants help constrain these alternatives Marder (2012).
The strongest functional circuit studies triangulate anatomical targeting, physiological target engagement, and independent behavioral measurement. Reversibility, blinded scoring, cross-animal replication, and direct measurement of the manipulated activity improve confidence. A single lesion or optical experiment can identify a causal dependency but rarely establishes a complete information-processing chain. Investigators should distinguish where an effect occurs from how a circuit implements it, then test predicted intermediate steps. Retinal functional-connectomic studies exemplify the value of aligning activity-defined computations with connection patterns Briggman et al. (2011). Each causal claim should therefore include a cell-class definition, manipulation window, network response, behavioral outcome, controls, and the model of alternative routes.
Drosophila Connectomics and Complete Networks
Small nervous systems have become important comparative laboratories for testing how anatomical network structure relates to coordinated activity and behavior. The larval fruit-fly connectome reported by Winding and collaborators mapped an extensive synaptic network in a juvenile insect brain, providing a basis for identifying major paths, recurrent loops, and interacting cell classes Winding et al. (2023). The value of such a system lies in its relatively comprehensive coverage, which permits graph-theoretic predictions that would be harder to evaluate in a partially imaged mammalian circuit. Comprehensiveness of wiring, however, does not establish all physiological parameters or determine an organism's behavior without appropriate task data.
Dorkenwald and colleagues reported a neuronal wiring diagram for an adult fruit-fly brain, extending dense reconstruction to a different developmental stage and a large anatomically differentiated network Dorkenwald et al. (2024). The reported graph contains approximately 139,000 neurons and tens of millions of synaptic connections, but these descriptive counts should be interpreted with the study's tracing and synapse-identification procedures in view. Completeness at one anatomical resolution leaves questions concerning neuromodulatory signaling, gap junctions, synaptic weights, developmental variability, and dynamic state. A whole-brain anatomical map is uniquely useful for evaluating candidate communication paths, but a graph-based simulation remains a model whose parameters require independent calibration.
Comparative insect work raises a general theoretical question: how much of neural functional organization is conserved across different nervous-system architectures, and which computations are realized through different cellular configurations? Broad principles such as recurrence, divergence, and targeted inhibition can be investigated across species, but homologous genetic labels and comparable behavioral tasks need separate validation. Distinguishing conserved organizational principles from species-specific solutions is scientifically productive precisely because neural evolution does not force a one-to-one correspondence among cell types or gross anatomical divisions. Circuit-based comparative psychology should connect anatomical organization to task behavior rather than infer equal function from superficial similarity Winding et al. (2023).
Functional Connectomics in Mouse Visual Cortex
Functional connectomics attempts to register neurons measured during activity to the same neurons and synapses identified in ultrastructural reconstruction. The MICrONS Consortium reported a large dataset spanning multiple areas of mouse visual cortex, integrating in vivo visual-response measurements and dense electron-microscopic anatomy MICrONS Consortium (2025). This combination enables questions that neither method can answer alone: whether cells sharing response properties have particular connection probabilities, how local and long-range projections align with functional tuning, and whether the paths predicted by a structural network are compatible with observed population responses. The resulting evidence supports increasingly constrained circuit hypotheses while preserving the distinction between statistical and causal inference.
The large number of cells and synapses in such a reconstruction should not be confused with a large number of independently sampled animals. The core MICrONS resource derives from a defined experimental preparation in one mouse, so the anatomy can be analyzed extensively while estimates of population-to-population generalizability require independent replication MICrONS Consortium (2025). Functional measurements are also conditional on the visual stimuli, behavioral state, indicator kinetics, and registration precision. Treating each segmented cell as an independent replicate of a species-level circuit organization would inflate evidential confidence. The appropriate units of generalization include the mouse, cortical area, cell class, sampled tissue volume, and recording session.
The research advance is methodological as well as descriptive. It brings together a connectivity graph, physiological features, and spatial localization within one registration framework, allowing explicit tests of hypotheses concerning cell-type-dependent wiring. Results should be interpreted alongside older functional-anatomical studies, including the combined imaging and microscopy approach of Bock and colleagues Bock et al. (2011). A direct path from wiring to a named psychological function still requires perturbation of selected circuits during a task, with controls for alternative behavioral and sensory variables. Large-scale data therefore strengthens the design of causal experiments by identifying more precise candidate cells and edges rather than replacing behavioral evidence.
Human–Mouse Comparisons and Limits of Transfer
Mouse circuit studies permit detailed recordings, tissue access, and experimental interventions that are usually unavailable in living human research. This methodological advantage creates a translation problem. Hodge and colleagues compared human and mouse cortical cell transcriptomes, reporting conserved broad classes together with divergent molecular features across species Hodge et al. (2019). A label such as inhibitory interneuron or corticocortical projection neuron may be useful across both systems at one level of description while failing to preserve particular genes, proportions, shapes, or network roles at a finer level. Cross-species inference should therefore specify which features support correspondence and which remain untested.
Human anatomical resources can complement molecular comparisons. BigBrain provides a high-resolution histological reference of human tissue, enabling spatial analysis of cytoarchitecture and laminar organization Amunts et al. (2013). Such resources are especially valuable when anatomical regions or layer definitions need to be reconciled with transcriptomic samples and imaging studies. They remain static anatomical descriptions. Human functional evidence often relies on neuroimaging, intracranial recordings in specific clinical contexts, neuropsychological lesions, or noninvasive perturbation, each of which samples different scales and populations. Combining these methods can establish convergent constraints without pretending that a mouse synapse map has already been directly replicated in human tissue.
Generalization must also consider developmental stage, sex, disease status, and the distinction between laboratory behavior and naturally occurring psychological performance. A mouse visual-cortex intervention may reveal a reliable sensory mechanism without directly defining how human language, reflective judgment, or social behavior is implemented. Comparative research should compare operational tasks and intermediate neural mechanisms, identify the level of homology claimed, and seek independent human evidence before importing circuit-level explanations into broad psychological theory. Cross-species similarities are scientifically powerful when their resolution and boundaries are made explicit Hodge et al. (2019).
Functional Circuit Organization and Psychological Explanation
Psychological explanation seeks to connect measured neural processes to capacities such as sensation, action selection, memory, attention, and adaptation. Circuit models contribute when they identify a plausible mechanism linking cell activity to information transformation and behavior. A successful explanation specifies the task, stimulus or internal event, relevant outputs, targeted cellular populations, directed interactions, and predicted behavioral changes under intervention. Receptive-field mapping and retinal wiring studies illustrate this progression from functional selectivity to more specific candidate circuits Hubel and Wiesel (1962) Briggman et al. (2011). Their strength is the alignment of independent anatomical and physiological measures with well-defined perceptual operations.
A cell type may participate in many psychological processes, because neural functions are distributed across task-dependent connections. Conversely, a named process such as attention can involve multiple specialized circuits and cannot be localized through one marker or a single graph statistic. Theoretical analysis should resist treating descriptive labels as causal explanations. For example, the presence of a recurrent motif is consistent with several dynamical outcomes until the parameters and experimental operating conditions are specified. The cortical network perspective developed by Harris and Shepherd and the structure-function analysis by Bargmann and Marder provide complementary constraints on such interpretations Harris and Shepherd (2015) Bargmann and Marder (2013).
The implications extend to computational and artificial systems, but the evidential bridge must be explicit. Artificial neural networks may use graph motifs or population representations that admit formal comparison with biological circuits, while their implemented operations, optimization procedures, substrates, and learning rules differ substantially. Structural resemblance can motivate a testable hypothesis about robustness or information routing; it does not establish biological homology, equivalent mental states, or shared psychological mechanisms. Within the Artificial Era, a productive cross-system research program defines common task measurements and controlled perturbations before asserting a functional equivalence. Neural circuit evidence thereby contributes a rigorous comparative methodology rather than a metaphor for artificial cognition.
Unresolved Problems and Research Priorities
The field faces a continuing problem of cell-type granularity. Transcriptomic technologies can distinguish progressively smaller clusters, yet circuit models require categories that support reproducible predictions about connectivity and physiology. One research priority is to test whether new subtypes improve out-of-sample prediction across animals, developmental periods, and brain areas. Atlas studies by Yao and Zhang provide extensive molecular and spatial foundations for such work Yao et al. (2023) Zhang et al. (2023). Taxonomic uncertainty should be documented as an empirical property rather than erased by the choice of one clustering threshold. Cross-modal datasets offer a way to test when molecular, morphological, and functional descriptions truly converge.
A second priority is the causal interpretation of massive connectomes. Dense graphs now capture synaptic pathways with unprecedented coverage, but the mapping from anatomy to signal flow is state-dependent, and errors in segmentation, registration, and selected tissue boundaries remain material. Replication across animals and tasks is necessary even when a dataset contains millions of reliably traced structures. The MICrONS and adult-fly resources define important methodological opportunities while illustrating the need for biological replication and dynamical measurement MICrONS Consortium (2025) Dorkenwald et al. (2024). Circuit simulations should report which parameters were fitted, which outcomes were independently predicted, and whether alternative graph structures could explain the observations.
A third priority is integrating glial, immune, and developmental factors with neural network dynamics. Evidence concerning astrocyte–synapse interactions and microglial contributions to circuit refinement shows that function depends on the biological environment as well as neuronal edges Allen and Eroglu (2017) Schafer et al. (2012). Longitudinal experiments could connect these mechanisms to changing circuit physiology and behavior across development, experience, and recovery. Finally, human translation requires cross-species models that preserve uncertainty rather than importing local animal findings into generalized claims about mental life. More precise experimental causal designs, transparent atlas correspondence, and openly described inference boundaries will improve cumulative integration across cellular neuroscience and psychology Hodge et al. (2019).
Conclusion
Neural cell types and functional circuit organization constitute an integrated scientific object in which classification, anatomical location, connectivity, physiological state, and behavior must be examined together. Molecular atlases have substantially expanded the resolution of cell taxonomy, while histology, paired recording, Patch-seq, and electron-microscopic reconstruction identify complementary aspects of biological organization. Dense connectomes expose structural pathways and motifs; functional recording and intervention establish how some of those pathways operate under defined conditions. The converging evidence shows that anatomically related cell populations can be distinguished by their projection patterns, physiological properties, and participation in local or distributed networks Tasic et al. (2018) Cadwell et al. (2016) MICrONS Consortium (2025).
The central scientific conclusion is that a wiring diagram is an indispensable structural constraint on neural function and an incomplete explanation of circuit operation. Synaptic efficacy, inhibition, neuromodulation, network recurrence, glial regulation, and developmental remodeling alter what anatomically stable connections do over time. A molecule-level grouping is informative when it improves predictions of circuit properties, and a proposed neural mechanism is strengthened when independent anatomical, physiological, and behavioral measures converge. Experiments on retinal direction selectivity, cortical circuit specificity, complement-dependent pruning, and activity-dependent perturbation exemplify how research can move from descriptive classification to carefully bounded causal inference Briggman et al. (2011) Jiang et al. (2015) Stevens et al. (2007) Boyden et al. (2005).
For psychology, the value of circuit research lies in specifying the implemented mechanisms of perception, movement, attention, and other capacities at a resolution appropriate to each question. Claims should be scaled to the evidence, with organism-level replication distinguished from the number of observed cells and with cross-species correspondence treated as an empirical hypothesis. The next phase of research will connect multimodal cell atlases, dynamic physiology, causal experimentation, and independently validated behavior across species and developmental contexts. Such integration will convert increasingly detailed neural maps into progressively stronger explanations of how living systems organize and regulate psychological function Bargmann and Marder (2013) Hodge et al. (2019).
References
- Allen, N. J., & Eroglu, C. (2017). Cell biology of astrocyte-synapse interactions. Neuron, 96(3), 697–708. https://doi.org/10.1016/j.neuron.2017.09.056
- Amunts, K., Lepage, C., Borgeat, L., Mohlberg, H., Dickscheid, T., Rousseau, M.-É., Bludau, S., Bazin, P.-L., Lewis, L. B., Oros-Peusquens, A.-M., Shah, N. J., Lippert, T., Zilles, K., & Evans, A. C. (2013). BigBrain: An ultrahigh-resolution 3D human brain model. Science, 340(6139), 1472–1475. https://doi.org/10.1126/science.1235381
- Bargmann, C. I., & Marder, E. (2013). From the connectome to brain function. Nature Methods, 10, 483–490. https://doi.org/10.1038/nmeth.2451
- Bock, D. D., Lee, W.-C. A., Kerlin, A. M., Andermann, M. L., Hood, G., Wetzel, A. W., Yurgenson, S., Soucy, E. R., Kim, H. S., & Reid, R. C. (2011). Network anatomy and in vivo physiology of visual cortical neurons. Nature, 471, 177–182. https://doi.org/10.1038/nature09802
- Boyden, E. S., Zhang, F., Bamberg, E., Nagel, G., & Deisseroth, K. (2005). Millisecond-timescale, genetically targeted optical control of neural activity. Nature Neuroscience, 8, 1263–1268. https://doi.org/10.1038/nn1525
- Briggman, K. L., Helmstaedter, M., & Denk, W. (2011). Wiring specificity in the direction-selectivity circuit of the retina. Nature, 471, 183–188. https://doi.org/10.1038/nature09818
- Cadwell, C. R., Palasantza, A., Jiang, X., Berens, P., Deng, Q., Yilmaz, M., Reimer, J., Shen, S., Bethge, M., Tolias, K. F., Sandberg, R., & Tolias, A. S. (2016). Electrophysiological, transcriptomic and morphologic profiling of single neurons using Patch-seq. Nature Biotechnology, 34(2), 199–203. https://doi.org/10.1038/nbt.3445
- Dorkenwald, S., Matsliah, A., Sterling, A. R., Schlegel, P., Yu, S.-C., McKellar, C. E., Lin, A., Costa, M., Eichler, K., Yin, Y., Silversmith, W., Schneider-Mizell, C., Jordan, C. S., Brittain, D., Halageri, A., Kuehner, K., Ogedengbe, O., Morey, R., Gager, J., . . . Murthy, M. (2024). Neuronal wiring diagram of an adult brain. Nature, 634, 124–138. https://doi.org/10.1038/s41586-024-07558-y
- Felleman, D. J., & Van Essen, D. C. (1991). Distributed hierarchical processing in the primate cerebral cortex. Cerebral Cortex, 1(1), 1–47. https://doi.org/10.1093/cercor/1.1.1-a
- Harris, K. D., & Shepherd, G. M. G. (2015). The neocortical circuit: Themes and variations. Nature Neuroscience, 18, 170–181. https://doi.org/10.1038/nn.3917
- Hodge, R. D., Bakken, T. E., Miller, J. A., Smith, K. A., Barkan, E. R., Graybuck, L. T., Close, J. L., Long, B., Johansen, N., Penn, O., Yao, Z., Eggermont, J., Höllt, T., Levi, B. P., Shehata, S. I., Aevermann, B., Beller, A., Bertagnolli, D., Brouner, K., . . . Lein, E. S. (2019). Conserved cell types with divergent features in human versus mouse cortex. Nature, 573, 61–68. https://doi.org/10.1038/s41586-019-1506-7
- Hodgkin, A. L., & Huxley, A. F. (1952). A quantitative description of membrane current and its application to conduction and excitation in nerve. The Journal of Physiology, 117(4), 500–544. https://doi.org/10.1113/jphysiol.1952.sp004764
- Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. The Journal of Physiology, 160(1), 106–154. https://doi.org/10.1113/jphysiol.1962.sp006837
- Jiang, X., Shen, S., Cadwell, C. R., Berens, P., Sinz, F., Ecker, A. S., Patel, S., & Tolias, A. S. (2015). Principles of connectivity among morphologically defined cell types in adult neocortex. Science, 350(6264), aac9462. https://doi.org/10.1126/science.aac9462
- Marder, E. (2012). Neuromodulation of neuronal circuits: Back to the future. Neuron, 76(1), 1–11. https://doi.org/10.1016/j.neuron.2012.09.010
- MICrONS Consortium. (2025). Functional connectomics spanning multiple areas of mouse visual cortex. Nature, 640, 435–447. https://doi.org/10.1038/s41586-025-08790-w
- Schafer, D. P., Lehrman, E. K., Kautzman, A. G., Koyama, R., Mardinly, A. R., Yamasaki, R., Ransohoff, R. M., Greenberg, M. E., Barres, B. A., & Stevens, B. (2012). Microglia sculpt postnatal neural circuits in an activity and complement-dependent manner. Neuron, 74(4), 691–705. https://doi.org/10.1016/j.neuron.2012.03.026
- Song, S., Sjöström, P. J., Reigl, M., Nelson, S., & Chklovskii, D. B. (2005). Highly nonrandom features of synaptic connectivity in local cortical circuits. PLoS Biology, 3(3), e68. https://doi.org/10.1371/journal.pbio.0030068
- Stevens, B., Allen, N. J., Vazquez, L. E., Howell, G. R., Christopherson, K. S., Nouri, N., Micheva, K. D., Mehalow, A. K., Huberman, A. D., Stafford, B., Sher, A., Litke, A. M., Lambris, J. D., Smith, S. J., John, S. W. M., & Barres, B. A. (2007). The classical complement cascade mediates CNS synapse elimination. Cell, 131(6), 1164–1178. https://doi.org/10.1016/j.cell.2007.10.036
- Tasic, B., Yao, Z., Graybuck, L. T., Smith, K. A., Nguyen, T. N., Bertagnolli, D., Goldy, J., Garren, E., Economo, M. N., Viswanathan, S., Penn, O., Bakken, T., Menon, V., Miller, J., Fong, O., Hirokawa, K. E., Lathia, K., Rimorin, C., Tieu, M., . . . Zeng, H. (2018). Shared and distinct transcriptomic cell types across neocortical areas. Nature, 563, 72–78. https://doi.org/10.1038/s41586-018-0654-5
- Tremblay, R., Lee, S., & Rudy, B. (2016). GABAergic interneurons in the neocortex: From cellular properties to circuits. Neuron, 91(2), 260–292. https://doi.org/10.1016/j.neuron.2016.06.033
- Winding, M., Pedigo, B. D., Barnes, C. L., Patsolic, H. G., Park, Y., Kazimiers, T., Fushiki, A., Andrade, I. V., Khandelwal, A., Valdes-Aleman, J., Li, F., Randel, N., Barsotti, E., Correia, A., Fetter, R. D., Hartenstein, V., Priebe, C. E., Vogelstein, J. T., Cardona, A., & Zlatic, M. (2023). The connectome of an insect brain. Science, 379(6636), eadd9330. https://doi.org/10.1126/science.add9330
- Yao, Z., van Velthoven, C. T. J., Kunst, M., et al. (2023). A high-resolution transcriptomic and spatial atlas of cell types in the whole mouse brain. Nature, 624, 317–332. https://doi.org/10.1038/s41586-023-06812-z
- Zhang, M., Pan, X., Jung, W., Halpern, A. R., Eichhorn, S. W., Lei, Z., Cohen, L., Smith, K. A., Tasic, B., Yao, Z., Zeng, H., & Zhuang, X. (2023). Molecularly defined and spatially resolved cell atlas of the whole mouse brain. Nature, 624, 343–354. https://doi.org/10.1038/s41586-023-06808-9