Quick Inquiry:
Are you curious about where our behavior comes from? Is it the firing of neurons deep within the brain, or some more complex causal network operating behind the scenes?
In neuroscience, the "cause of behavior" is often simply equated to "neural mechanisms." Scientists, using precise techniques like optogenetics, have achieved real-time control over behavior and subsequently claimed to have discovered certain neural mechanisms. However, does this reductionist explanation, focused on instantaneous neural activity, truly reveal the complete causes of behavior?
In this process, might we be overlooking the essence of cognitive processes, the coordinated action of the whole brain, and even erasing the unique history and meaning of the organism as a subject? This article will challenge this traditional perspective, introducing non-reductionist and diachronic causal concepts, exploring how to reconstruct a more comprehensive explanatory framework for behavior in neuroscience, and rethinking together: is the root of behavior truly mechanisms, or a life story that transcends mechanisms?
▷Potter, Henry D., and Kevin J. Mitchell. "Beyond Mechanism—Extending Our Concepts of Causation in Neuroscience." European Journal of Neuroscience 61.5 (2025): e70064.
In neuroscience research, the search for the "causes of behavior" is often simply equated with the search for "neural mechanisms." This research perspective typically involves a threefold simplification of causality:
First, reducing essential problems at the cognitive process level to the neural mechanism level; Second, simplifying overall brain activity to the activity of isolated brain regions; Finally, replacing the diachronic examination of processes evolving over time with observations of instantaneous states.
Although modern neuroscience has achieved remarkable success in identifying instantaneous neural mechanisms and enabling precise, real-time control over behavior, we argue that this does not mean we have fully understood the deep causality of behavior. This simplification is particularly likely to lead to two dangerous tendencies: it can exclude cognitive elements from the explanatory framework and may even erase the organism's subjectivity itself. To truly understand the causal relationships behind behavior, we need not only to know what results from neural activation but also to ask why these results occur.
This article introduces several established non-reductionist and diachronic causal concepts from philosophy (including criterial causation, trigger cause and structural cause, system constraints, macroscopic causality, historical specificity, and semantic causality, etc.), providing theoretical support for neuroscientists to build a more complete causal explanation for behavior. This cluster of concepts can explain mental and agent causation in a scientifically verifiable way and re-establish cognitive function and the organism itself as autonomous causal agents. Through this theoretical integration, we hope to rebuild the proper status of cognitive subjects and organisms within the neuroscience explanatory framework.
Table of Contents:
01 Origin of the Problem: The Dilemma of Causal Explanation in Neuroscience
02 Critique of Triple Reductionism: Limitations of the Driving View of Causation
03 Reconstruction of Causal Types
04 System Constraints and Macroscopic Causality
05 Diachronic Explanatory Framework: Structural and Final Causes
06 Meaning Construction and the Return of Subjectivity
07 Implications for Experimental Paradigms: An Expanded Framework for the Causes of Behavior
08 The Importance of Diachronicity and Non-Reductionism
Origin of the Problem:
The Dilemma of Causal Explanation
in Neuroscience
Where does our behavior truly come from? This question is not only the philosophical problem of free will but also a core puzzle that neuroscientists have been striving to answer. Modern neuroscience believes that to explain how the brain produces behavior, relevant research requires at least three basic assumptions:
First, all behavior and thought activities originate from neural activity;
Second, there is a causal relationship between neural activity and behavior;
Third, to truly explain a phenomenon (such as a certain behavior), its causal relationship must be found—this is known in the theory of scientific explanation as "Causal Explanationism".
Precisely because of this, identifying the causal relationships within the brain and between the brain and behavior has become a major task of neuroscience. As Ross and Bassett stated: "The core goal of neuroscience is to elucidate the causal structure of the brain—whether at the microscopic level of molecular and cellular interactions or the macroscopic scale of neural circuits, brain regions, and network-level activity." Similarly, Barack also proposed that neuroscientists consistently focus on two fundamental questions: "What brain activities lead to specific behaviors? What brain activities lead to other neural activities?" The research objective is: to pinpoint the key nodes of neural causality to explain the mechanisms by which target behaviors or neural events occur.
This research paradigm naturally leads scientists to investigate the neural mechanisms of behavior. As summarized by Ross and Bassett: "It is widely accepted in academia that a true neuroscience explanation must include the elucidation of brain mechanisms—where 'mechanism' specifically refers to the microscopic causal details capable of producing specific brain functional outcomes." Within this framework, understanding the cause of behavior is equivalent to revealing specific neural mechanisms: whether it is the activity of single neurons, the operation of neural circuits, or population coding, as long as its activity pattern reliably triggers a certain behavior, it is considered the fundamental cause of that behavior.
This research paradigm has profoundly shaped the methodology of neuroscience. In neuroimaging studies, scientists attempt to identify characteristics of brain activity associated with specific behaviors or mental states (in humans or other organisms), which are assumed to be potential causes contributing to the related behavior. Complementary to this, brain lesion studies have verified this mechanistic research paradigm through both localization and decomposition: demonstrating that a certain brain region is not only an active area for behavior but also a necessary condition for its production—whether it's episodic memory, face recognition, or language function, their normal operation relies on the integrity of specific neural structures.
The emergence of optogenetics technology in 2005 pushed neural causal research into a new phase. This technological breakthrough, along with pharmacological interventions, transcranial magnetic stimulation, and other techniques, built a powerful toolbox for neural manipulation. Researchers could precisely control the activity of specific neural units (whether single neurons, pathways, microcircuits, or whole-brain networks) and then verify whether changes in specific neural structures would inevitably lead to changes in specific behaviors or mental states (Figure 1).
▷Figure 1: Principles and Methods of Using Optogenetics in Neuroscience Research.
This "intervention-response" research paradigm, widely considered the gold standard for exploring causal relationships, has led to breakthroughs in analyzing behavioral mechanisms. Through this paradigm, scientists have successfully identified the neural coding characteristics behind specific behaviors (such as avoidance behavior). These neural states satisfy the dual criteria of being both "sufficient conditions" and "necessary conditions."
Sufficient Condition: When the target neural state is activated through optogenetics, even in contradictory situations, the animal can still produce specific behaviors or cognitive responses, indicating that the activity of these neurons is sufficient to "evoke" this behavior.
Necessary Condition: If we block the activity of target neurons through neural inhibition or lesioning, and the behavior is absent or impaired, we can confirm that this neural structure is an essential component for the behavior's production.
With this evidence, researchers have achieved precise control over animal behavior—by activating or deactivating specific neural mechanisms, they can control behavioral output as if flipping a switch.
Another important premise for this research is that the animal's regular behavior in its natural environment is also controlled by the specific firing patterns of these neurons. This means we can not only explain the mechanisms of specific behaviors under laboratory conditions but also extend this causal explanation to natural settings. As the Deisseroth team emphasized: "This integrated research path allows us to precisely identify the primary causal underpinnings of physiological functions and behavioral patterns at the cellular, circuit, and even whole-brain scales, across acute or chronic temporal dimensions—these underpinnings are both necessary and sufficient conditions."
So, the core question arises: how should these groundbreaking findings be interpreted, and what answers do they provide to the original question of "why behavior occurs"? When optogenetic manipulation can precisely control animal behavior like a puppet, it is easy to think that the manipulated neural variable is the ultimate answer to the behavior's generation—they seem to be the "responsible agent" or "control center" behind the behavior. Especially when we can control behavior in real-time through external intervention, it strengthens the idea that we have found the "origin of behavior." After all, if our understanding of how the brain generates behavior is deep enough to allow for arbitrary tuning through neural manipulation, what else is there left to explore?
Critique of "Triple Reductionism":
Limitations of the Driving View of Causation
This perspective, which explains internal brain and brain-behavior causal relationships, is called the "driving view of causation." This metaphorical expression is frequently seen in optogenetics research reports, for example:
"A subpopulation of lateral hypothalamic GABAergic neurons... specifically drives feeding behavior in mice";
In Caenorhabditis elegans research, scientists created a "neural signal propagation atlas." "By directly measuring signal propagation and building mathematical models, the atlas elucidates how upstream neural activity drives downstream neural responses"; more broadly, how "stimulation of one part of the network drives activity in other parts."
In Drosophila brain research, researchers built a causal model of the fly brain called the "effectome" by systematically optogenetically stimulating different brain regions and recording cascading reactions. This model can accurately predict how activating any neural node will drive the propagation trajectory of downstream effect waves.
"A long-term goal of neuroscience is to build causal models of nervous systems, which would allow us to explain animal behavior in terms of dynamic interactions between neurons."
——
This series of "driving metaphors" stems from the cognitive tradition deeply rooted in the classic research paradigm of simple reflex systems—they are both the starting point for human understanding of neural signal transmission and the cornerstone of many neuroscience textbooks. In the reflex system model, we typically imagine the process from sensory input to behavioral output as a linear transmission chain:
Each neural node, like a domino, drives the activity of subsequent nodes, ultimately triggering a preset behavior; when attempting to apply this cognitive framework to more complex brain systems, one might assume that simply scaling up and stacking the logic of these simple neural circuits can explain the operational mechanisms at the macroscopic nervous system and even whole-brain level.
This unidirectional feed-forward driving causal cognitive paradigm can be traced back to William James's famous assertion in 1890:
Though I am in favor of assisted suicide in cases of extreme severe and terminal physical illness, why do I find it unacceptable for patients with incurable mental illness?
▷Figure 2: The Cognitive Framework of Triple Causal Reduction. Viewing neural mechanisms as key explanatory factors for behavior involves three paths of simplification: 1) Vertical reduction at the ontological level—dimensionality reduction from the cognitive level to the neural level; 2) Horizontal reductionism—research through isolation and decomposition; 3) Temporal reduction—focusing only on the synchronous instantaneous states of the nervous system.
This idea of understanding neural activity as unidirectional "driving" can lead to three levels of "reductionism" (Figure 2):
Vertical Reduction Perspective
Although we can still describe behavioral control using mental states like beliefs, desires, or cognitive decision-making processes, these are not considered true causal explanations. What truly matters is the "neural substrate" corresponding to these mental states (i.e., neural mechanism activity), as they are the fundamental causes driving behavior. According to this view, mental states and cognitive processes are explained as "epiphenomena"—within the causal explanation framework, the subject's conscious deliberation or cognitive process itself has no substantive influence on behavior.
Horizontal Reduction Perspective
This perspective assumes that we can decompose the nervous system into different neural units and attribute the causality of specific behaviors solely to the independent activity of certain units, neglecting the broader neural environment. Consequently, the organism as a causal subject is gradually removed from the causal explanation framework of its own behavior, or even disappears entirely. Even when extended to neural circuits and macroscopic systems, the core logic remains: the organism's real-time behavior is controlled only by a subsystem of specific neural units.
Temporal Reduction
This perspective is the most subtle—it solely attributes the organism's behavior to the instantaneous activation of specific neural mechanisms, believing that examining the current neural activity pattern is sufficient to exhaust the behavioral drivers. This perspective depicts behavior as the product of a Markovian neural process. In this framework, the historical context that shapes neural activity, as well as the diachronic characteristics of the organism as a temporally extended entity, are excluded from causal explanatory considerations.
"Markov" is named after the Russian mathematician Andrey Andreyevich Markov, referring to a system whose future state depends only on its current state, not on its past history (memoryless). "Diachronic" refers to the study of how things change and develop over time, as opposed to "synchronic" (focusing on a static analysis at a specific point in time).
The current simplified perspective in neuroscience on synchronous neural mechanisms actually constructs a one-sided and misleading framework for explaining the causality of behavior. This limitation stems from an overly narrow concept of causality. When we understand causality only through the reductionist synchronous framework (such as the driving view of causation), the explanation for "the causes of behavior" is necessarily limited and excludes the organism itself from the causal picture. Crucially, this framework neglects a key fact: neural activity patterns have specific meaning for the organism, and the causal efficacy of the nervous system precisely depends on this meaning connection.
Therefore, we need to provide new explanatory frameworks for neuroscience, introducing established non-reductionist and cross-temporal causal theories from philosophy. These theories can both remold the organism as an autonomous causal subject and ensure the scientific rigor of causal explanations. It is particularly emphasized that a complete understanding of the organism's causality must adopt a diachronic perspective—highlighting the specific meaning of neural activity patterns for the organism and incorporating the temporal dimension into the analysis framework. This way, while retaining the independent explanatory value of cognitive processes, we can establish a model of their relationship with neural processes, avoiding the theoretical pitfall of reducing cognition to neural activity.
Reconstruction of Causal Types:
From Physical Forces to Causal Pluralism
(1) Productive Causes and Dependence Causes:
Have We Overlooked Environmental Factors?
Causality is often simply equated by the public with "physical force." This view holds that a "cause" is an event that triggers a result through energy transfer—as List and Menzies described as causal "oomph," similar to the force process of billiard ball collisions. Philosophers refer to this concept as "productive" causality, an idea also visible in the "driving" rhetoric of neuroscience discussed earlier (although synaptic transmission essentially does not involve the direct transfer of energy or physical force).
Another concept of causality widely discussed in philosophy is called "difference-making" or "dependence" causality. This theory views causes as "counterfactual difference-makers"—that is, any variable that could change the course of events, provided that variable differs from its actual state. This captures our everyday causal intuition: when we consider A to be the cause of B, it usually implicitly involves a counterfactual conditional judgment like "If A had not occurred, then B would not have occurred."
Clearly, the "difference-making" view of causality encompasses "productive (or driving) causes" that produce results through physical forces (such as the "push" of kinetic energy transfer), and its scope is broader—any necessary condition that enables a productive cause to produce a specific effect is considered a causal element.
Take the example of a baseball breaking a window:
Productive Cause: The trajectory of the baseball is naturally the direct productive cause—the process of kinetic energy transferring from the ball to the window, applying physical force to the glass's molecular bonds, causing them to break and shatter.
Dependence Causes: The event also depends on other "necessary conditions," such as the tensile strength of the glass and the material properties of the ball. If these conditions were to change specifically (e.g., the glass having bulletproof properties, or the baseball being replaced with a foam ball), the shattering event would not occur.
Most physical events result from the combined action of productive and dependence causes. However, in explaining events, we often selectively ignore dependence causes and focus instead on productive causes. This bias stems from "pragmatic considerations": researchers are usually more interested in difference-making factors that are directly related to the event and highly specific. In the example of the broken window, while necessary, dependence conditions like the glass's tensile strength are inherent properties of the material, possessing universality and stability, and thus lack sufficient "causal salience" to provide incremental explanatory value.
The mainstream "driving" view of causation in neuroscience holds that only "productive causes" like neuronal firing have explanatory power and causal salience for behavior. However, this premise fundamentally conflicts with the essential characteristics of neurobiology. In the neural mechanisms of behavior, dependence conditions are not constant like glass strength but evolve dynamically, exerting causal efficacy as "functional constraints" by modulating neuronal firing thresholds and other mechanisms. The existence of these conditions itself is an evolutionary product enabling the nervous system to achieve "meaning sensitivity." The following sections will systematically analyze: 1) The typology of diverse dependence conditions in the nervous system; 2) Their formation mechanisms; 3) How they collectively construct a causal response system based on the meaning of neural activity.
▷
(2) Criterial Causation;
Why the Brain Isn't Simply Unidirectionally "Driven"?
Conceptualizing brain operation solely through the driving metaphor that prioritizes "productive causes" creates a misleading impression: that the causal relationships between neurons and within neural circuits are essentially feedforward, sequential, and deterministic, with neurons passively driven by their presynaptic input. But as most neuroscientists know, this does not fully describe how neuronal communication actually works.
In reality, the response of a neuron to incoming activity depends heavily on the configuration of its synaptic connections and other biophysical parameters of the cell (such as its current membrane potential). For example, if neuron B receives input from neurons A and C, and the synapse from A is "stronger," then A's signal is more likely to be "heeded" by B, and vice versa. If cell B's membrane potential is already close to the firing threshold, then only a small stimulus from A and C might be needed for B to fire; but if B is in a more negative membrane potential state, perhaps more signals are needed to elicit an action potential. That is, the weight and nature of the synapse between neuron A and B, combined with the context of all other presynaptic inputs to B, and B's overall electrophysiological properties, collectively represent what Tse calls the "criteria" for neuronal firing—the necessary conditions for a neuron to "exert its effect."
These criteria specifically define the types of presynaptic input a neuron must receive (and the types of input that cannot activate it) in order to generate an action potential. For example, these criteria might include a firing threshold based on the number of action potentials arriving within a specific time window. More often, however, they specify complex spatiotemporal input patterns to which the neuron is causally sensitive.
For instance:
Due to its specific configuration of excitatory and inhibitory synapses, a neuron might require the input signal to exhibit a specific spatial pattern (e.g., implementing a logic "AND/OR" gate input combination) to "release its effect."
Another neuron might be sensitive to specific temporal patterns, such as a particular frequency or timing of inputs.
Thus, a neuron's firing criteria belong to a type of "dependence cause": by changing these criteria (e.g., changing the weights of its input synapses), one can control whether the neuron fires, even if the presynaptic input remains unchanged. Tse calls this causal relationship "criterial causation."
Crucially, in understanding neuronal communication (and thus how the brain generates behavior), these criterial dependence causes cannot be eliminated in explanation like dependence conditions in the non-biological world. The reasons are:
First, these criteria are not generic properties of neurons. They are contingent and highly specific characteristics possessed by individual neurons based on their particular synaptic configuration and intracellular state. Therefore, one cannot predict whether a postsynaptic neuron will fire based solely on the presynaptic action potentials.
Second, neuronal input conditions are dynamic. They are not fixed neuronal properties but continuously change due to synaptic remodeling and the cell's recent firing history. Thus, even with information about presynaptic action potentials and their prior criterial configuration, one cannot predict the firing state of the postsynaptic neuron.
Given this, some scholars have proposed: "The representation of synaptic states may be more explanatory than neuronal firing patterns in describing the state of a neural network. Indeed, the function of synapses is to regulate firing activity within neural circuits by setting the causal sensitivity criteria for neuronal firing."
As Tse comprehensively argues, the ability to change neuronal firing criteria through synaptic remodeling (sometimes in real-time) is central to the brain's mechanism for generating behavior. The configuration and weights of input synapses for any given neuron are shaped by a threefold influence: long-term evolutionary history, individual learning experiences, and the organism's current state (including its cognitive activity). It is these criteria that endow neurons with specific functionality and selective sensitivity, enabling them to serve the organism's needs.
However, some might worry that the concept of "criterial causation" merely refers to situations where multiple different upstream causes must act together to produce a single downstream effect—thus, ultimately, this situation is still fully compatible with the "driving" view of causation. From a certain perspective, this objection has merit.
But the core value of the criterial causation concept lies precisely in its focus on revealing the fundamental role of "hidden dependencies" in constructing such "many-to-one" scenarios. This concept prompts us to consider: why and how does the system form a particular configuration such that specific upstream causes trigger specific downstream effects? The answer to this question is not only key to understanding basic interneuronal communication but also the foundation for explaining the mechanisms of overall brain behavior generation.
The driving view of causation holds that upstream neurons simply drive downstream neuronal activity when sufficiently activated. The neurophysiology of neuronal communication requires us to overturn this traditional causal view and incorporate the concept of "criterial causation" into the theoretical framework: due to the sensitivity of downstream neurons to input types, their potential to interpret input signals should be emphasized (Figure 3). This forces us to delve deeper into: Why and how are neurons configured to respond specifically to input signals?
▷Figure 3: The Driving Metaphor Reversed. The top panel depicts the driving relationship between neurons A and B: B is essentially a "passive element"—A's activity drives B's activity. The bottom panel reverses this relationship, highlighting the active role of neuron B: based on the criteria embodied in its synaptic connections and cellular physiology, B performs an "interpretive processing" of its input signals.
It's particularly important to note: The existence of "criterial causation" in the brain forces us to rethink how optogenetics experiments are interpreted. Even if a group of neurons is found through optogenetic techniques to trigger a specific behavior, this does not mean that the activity of these neurons is a complete explanation for that behavior. In fact, these neurons are able to trigger the behavior because their activity occurs within a nervous system that has been configured in a particular way.
Therefore, to truly understand "what factors caused this behavior to occur?", the "overall system configuration" must be taken into account. Indeed, the concept of "criterial causation" provides a theoretical basis for understanding how system configuration imbues neural activity patterns with specific meaning (and this meaning supports the causal efficacy of the neural activity). By modulating criteria, organisms can exert "top-down causation," changing neural sensitivity in real-time to actively guide their own behavior. This requires us to expand our causal conceptual system and return to the "causal pluralism" proposed by Aristotle.
(3) Causal Pluralism:
Why Are "Purpose" and "Form" Also Important?
In neuroscience, embracing a pluralistic view of causality is essential for a comprehensive understanding of the causes of behavior. This viewpoint is not new; as early as ancient Greece, Aristotle's famous "Four Causes" proposed multiple types of causal relationships: the material cause, the efficient cause, the formal cause, and the final cause.
The "material cause" and "efficient cause" broadly correspond to the "mechanistic" explanations in modern neuroscience, which form the basis of the "synchronous productive causes" underlying the "driving metaphor." The "formal cause" is a relatively vague concept, generally referring to the set of essential properties that make something a specific type of entity (rather than another type)—that is, the "characteristic form" that material takes when composing the thing. This concept can be likened to the specific configuration of the nervous system (including synaptic connection patterns) and the "causal effect" of the information content represented/instantiated by that configuration. Finally, Aristotle, through the concept of the "final cause," asked: why does something happen? What is its ultimate purpose? Acknowledging "purposiveness" itself can be a driver of events. Formal and final causes essentially belong to the "diachronic causal category"—the former reflects how the system is specifically configured through historical events, while the latter points to the "future-oriented functional properties" that the system achieves.
Aristotle thus adopted a "causal pluralism" approach, viewing these different types of causal relationships as based on distinct but equally valid perspectives or causal types, together forming complementary explanatory paths for natural phenomena. This line of thought is continued in Niko Tinbergen's Four Questions for Animal Behavior, who proposed that a complete explanation of behavior requires answering four questions:
1. Function (or adaptation): What is the survival value of this behavior for the animal?
2. Evolution (or phylogeny): How did this behavior evolve?
3. Motivation (or mechanism): What are the immediate causes triggering this behavior?
4. Development (or ontogeny): How did this behavior develop during the individual's lifespan?
Unfortunately, the history of science has sometimes rejected formal and final causes, deeming only matter and direct forces as "true" scientific explanations. For example, Francis Bacon, a founder of 17th-century scientific methodology (whose thought profoundly influenced the establishment of scientific thinking paradigms), argued that "science should only concern itself with material and efficient causes, i.e., mechanistic explanations or matter in motion, which are productive causal relationships"; while relegating formal and final causes to the realm of metaphysics, or what he called "magic."