Loading field map
The philosophy of causation is the branch of metaphysics that investigates what it means for one thing to cause another. It asks whether causation is a fundamental feature of reality, a pattern in our experience, or a conceptual tool we impose on the world. The subfield examines the nature of the causal relation, the kinds of things that can be causes and effects, and the principles that govern causal reasoning. Its central questions include: What distinguishes a causal sequence from a mere correlation? Do causes necessitate their effects, or is causation a matter of probability? Is causation reducible to more basic facts, or is it an irreducible feature of the universe? The stakes are high: causation is woven into scientific explanation, moral responsibility, legal attribution, and everyday action.
The philosophy of causation has ancient roots, but its modern form crystallized in the early modern period. Aristotle distinguished four types of cause—material, formal, efficient, and final—but the subfield now focuses almost exclusively on efficient causation: the relation between a prior event or state and the change it produces. Medieval thinkers debated divine causation and human freedom within a theological framework. The decisive shift came with David Hume in the 18th century, who argued that we never perceive a necessary connection between cause and effect, only constant conjunction and a felt expectation. Hume’s skeptical challenge set the agenda for subsequent philosophy: either causation is a projection of the mind, or it must be grounded in something more robust than mere regularity.
After Hume, Immanuel Kant attempted to rescue causation as a necessary category of the understanding, imposed by the mind on experience. In the 19th and early 20th centuries, John Stuart Mill developed a regularity theory of causation and linked it to inductive method. The logical positivists of the early 20th century, influenced by Hume, sought to eliminate metaphysical notions like necessary connection, treating causal statements as summaries of observable regularities. Since the mid-20th century, the field has diversified into several competing approaches, each responding to perceived inadequacies in the others.
Regularity theories, rooted in Hume, hold that causation is nothing more than constant conjunction: event A causes event B if and only if events of type A are always followed by events of type B. This approach avoids appeal to unobservable necessary connections and fits with a broadly empiricist epistemology. Its central problem is distinguishing genuine causation from accidental regularities. For example, night regularly follows day, but day does not cause night. Mill attempted to refine the theory by requiring that the regularity hold under all relevant circumstances, but this introduces the need to specify what counts as relevant. A further difficulty is that many causal relations are probabilistic rather than deterministic: smoking causes lung cancer, but not every smoker develops it. Regularity theories struggle to accommodate probabilistic causation without collapsing into mere correlation. Despite these limitations, the regularity approach remains influential as a minimal, ontologically parsimious starting point, and its core insight—that causation is tied to patterns of events—is preserved in more sophisticated successors.
Counterfactual theories analyze causation in terms of what would have happened if the cause had not occurred. The basic idea, developed systematically by David Lewis in the 1970s, is that an event A causes an event B if and only if, had A not occurred, B would not have occurred. This approach captures the intuitive notion that causes make a difference: they are necessary for their effects in the circumstances. Lewis grounded counterfactuals in possible worlds semantics: a counterfactual is true if the nearest possible world where the antecedent holds is one where the consequent also holds. The theory handles probabilistic causation by comparing probabilities across possible worlds.
Counterfactual theories face several challenges. They must specify which possible worlds count as "near enough" to the actual world, a problem that becomes acute when causes are redundant or overdetermined. For instance, if two assassins both shoot the victim, each shot is sufficient for death, but neither is necessary; the counterfactual test would wrongly deny that either shot caused the death. Lewis addressed this with a more complex account involving causal chains and dependence, but the problem remains disputed. Counterfactual theories have been highly influential in philosophy of science, law, and history, where counterfactual reasoning is common. They also connect naturally to theories of explanation and intervention.
Probabilistic theories treat causation as a matter of probability raising: a cause increases the probability of its effect. This approach, developed by Patrick Suppes, Wesley Salmon, and others in the 1970s and 1980s, is designed to handle indeterministic causation, where causes do not guarantee their effects. The core idea is that C causes E if and only if P(E|C) > P(E|not-C), where the probabilities are understood as objective chances or frequencies. This captures the sense in which smoking causes lung cancer even though most smokers do not develop it.
The probabilistic approach faces the problem of spurious correlations: a common cause can make two effects probabilistically dependent even though neither causes the other. For example, a falling barometer reading is correlated with rain, but does not cause it; both are effects of low atmospheric pressure. To handle this, probabilistic theories require that the probability-raising relation hold when all other relevant factors are held fixed—a condition that is difficult to specify without circularity. Salmon and others developed the concept of "screening off" to distinguish genuine causes from spurious correlates. Probabilistic theories have been influential in epidemiology, social science, and quantum mechanics, where indeterminism is fundamental. They are often combined with counterfactual or mechanistic approaches.
Mechanistic theories hold that causation consists in the operation of a physical mechanism that connects cause to effect. This approach, revived in the late 20th century by philosophers such as Wesley Salmon (in his later work) and Peter Machamer, Lindley Darden, and Carl Craver, emphasizes the productive, continuous process that transmits energy, force, or information from cause to effect. A mechanism is typically understood as a structured set of entities and activities that produce a regular change from a starting condition to a termination condition.
Mechanistic theories address a weakness of regularity and counterfactual approaches: they explain why a cause produces its effect, rather than merely describing a pattern. They also handle cases of causal preemption and overdetermination more naturally, since the mechanism provides a concrete connection. However, the mechanistic approach faces difficulties in defining what counts as a mechanism without circularity, and in applying to domains like quantum mechanics, where continuous processes may not exist. It also struggles with causation in the social sciences, where mechanisms are often unobservable or poorly understood. Despite these challenges, mechanistic theories are widely used in philosophy of biology, neuroscience, and medicine, where researchers routinely seek underlying mechanisms.
Interventionist theories, developed by James Woodward and others, analyze causation in terms of what would happen under ideal interventions. The central idea is that C causes E if and only if an ideal intervention that changes C would change E, while holding other factors fixed. This approach is closely related to counterfactual theories but emphasizes the practical, experimental character of causal reasoning. It is explicitly designed to connect causation to scientific practice, where experiments and controlled interventions are the gold standard for establishing causal claims.
Interventionist theories avoid some metaphysical puzzles by not requiring a deep account of what causation is in itself; they focus on the conditions under which causal claims are warranted. They handle probabilistic causation naturally, since interventions can change probabilities. The approach is widely used in philosophy of science, statistics, and machine learning, where causal graphical models and do-calculus provide formal tools. A limitation is that the notion of an "ideal intervention" is itself a theoretical construct that may not be realizable in practice. Critics also argue that interventionist theories presuppose a causal understanding of intervention, making them circular as a reductive account of causation. Proponents respond that the theory is not intended as a reductive analysis but as a characterization of the role of causation in empirical inquiry.
These approaches are not mutually exclusive, and many philosophers combine elements from multiple theories. For example, a probabilistic theory may be supplemented with a mechanistic account to explain why probabilities are raised. Counterfactual and interventionist theories are closely related: an intervention can be understood as a special kind of counterfactual. Regularity theories provide the empirical basis for probabilistic and counterfactual accounts, while mechanistic theories offer a deeper explanation of why regularities hold. The field is characterized by productive overlap rather than strict division.
Disagreement persists over which approach is fundamental. Some philosophers argue that causation is a primitive, irreducible feature of reality, and that all theories are attempts to capture its manifestations. Others hold that causation can be reduced to more basic facts, such as patterns of counterfactual dependence or physical processes. The debate often turns on whether one prioritizes metaphysical parsimony, fit with scientific practice, or intuitive adequacy.
Contemporary philosophy of causation is a vibrant, technically sophisticated field. Formal tools from probability theory, graph theory, and logic are widely used. The development of causal graphical models and Bayesian networks by computer scientists and statisticians has created a rich interdisciplinary dialogue. Philosophers now engage with causation in quantum mechanics, where Bell's theorem and nonlocality challenge classical notions of causal structure; in biology, where downward causation and emergence are debated; and in the social sciences, where causal inference from observational data is a central methodological problem.
A major ongoing debate concerns the relationship between causation and laws of nature. Some philosophers hold that causation is derivative from laws; others argue that laws are generalizations about causal relations. The problem of causal exclusion—whether mental causation is possible given physical determinism—remains active in philosophy of mind. The metaphysics of causal powers and dispositions has been revived, with some philosophers arguing that objects have intrinsic causal capacities that are not reducible to patterns of events.
The field has also seen increased attention to causation in non-Western philosophical traditions, including Buddhist and Indian philosophical accounts of causality (such as the doctrine of dependent origination) and Chinese correlative cosmology. These traditions offer alternative frameworks that do not assume the event-based, linear model dominant in Western philosophy, and they are increasingly recognized as contributing substantive insights rather than mere historical curiosities.
No single approach commands universal acceptance. The field is characterized by a plurality of methods and a recognition that causation may be a multifaceted phenomenon that resists simple reduction. The most influential contemporary work often combines formal rigor with attention to actual scientific practice, and the boundaries between philosophy of causation and related fields—philosophy of science, metaphysics, epistemology, and statistics—are increasingly porous.