Causality

• 1. Introduction

  ›

  Causality is the invisible thread that binds every event to what came before it. Strike a match, and a flame appears. Drop a glass, and it shatters. Intuitively, the connection seems obvious. But ask a philosopher, a physicist, or a psychologist to define exactly what that connection is, and the answers diverge in startling ways. Does the cause actually produce the effect, or do we simply observe one thing following another and assume a link? Can causality even be perceived, or only inferred? And does the structure of cause and effect actually give rise to time itself, rather than the other way around? These questions have occupied thinkers from Aristotle's lecture halls to twentieth-century physics laboratories, and they remain alive in courtrooms, epidemiology labs, and economic models today.

• 2. Aristotle's Four Answers

  ›

  Aristotle identified four distinct kinds of answer to any "why?" question, and a mistranslation from ancient Greek has muddied the waters ever since. The Greek word he used, often rendered in English as "cause," would be better translated as "explanation." Each of his four modes offered a different type of answer: the material cause (what something is made of), the formal cause (its structure and shape), the final cause (its purpose or goal), and the efficient cause (what set it in motion). The efficient cause is closest to what most people mean today when they say something caused something else. To illustrate, Aristotle offered the bronze of a statue as its material cause, the craftsman's shaping as its efficient cause, and health as the final cause of a man's evening walk. In the philosophical tradition from Thomas Aquinas through the late Middle Ages, scholars ranked these four modes in a hierarchy, placing the final cause highest. Aquinas sought to trace the chain of efficient causes back to a single unmoved first mover, which he identified with God. Francis Bacon and Niccolo Machiavelli, writing after the Middle Ages, narrowed the word "cause" to mean almost exclusively what Aristotle had called the efficient cause, a shift that shaped how the concept has been understood ever since.

• 3. Hume's Challenge

  ›

  David Hume took aim at the assumption that we can ever directly perceive a causal connection. His argument, rooted in his opposition to rationalism, was blunt: pure reason cannot prove the reality of efficient causality. What we actually experience, he said, is one event following another, again and again, until habit and custom lead the mind to expect the second whenever it sees the first. In Part III, section XV of his book A Treatise of Human Nature, Hume laid out eight criteria for judging whether two things might stand in a cause-and-effect relationship. The first three required that cause and effect be contiguous in space and time, that the cause precede the effect, and that there be a constant union between them. The influence of this framework was enormous. Max Born, writing in 1949, built on Hume's framework by distinguishing two kinds of causation: nomic or generic causation, where cause and effect are linked by general probabilistic laws, and singular causation, where a particular complex of events is linked by antecedence and contiguity. The psychologist Patricia Cheng later attempted to reconcile Hume's view with Immanuel Kant's competing claim that people carry innate assumptions about causes. Her power PC theory proposed that people filter observations through an intuition that causes have the power to generate or prevent their effects.

• 4. Counterfactual and Probabilistic Theories

  ›

  David Lewis, in his 1973 paper "Causation," proposed a formal definition built on counterfactual conditionals: event E causally depends on event C if, and only if, C's occurrence would bring about E and C's absence would prevent E. This approach traces back to Hume's own phrasing, that the cause is that "where, if the first object had not been, the second never had existed." The full development of counterfactual causation required twentieth-century work on possible-world semantics. Lewis extended the framework in a second paper, "Counterfactual Dependence and Time's Arrow," to account for why causation runs forward in time and not backward. One persistent challenge is overdetermination, where an effect has more than one sufficient cause. If Alice and Bob both throw bricks at a window and it breaks, removing either throw alone still leaves the window broken. The Halpern-Pearl definitions of causality were developed specifically to handle such cases. Separately, probabilistic theories address the intuition that determinism is too strict a requirement. Smoking does not guarantee cancer in any individual case, yet it plainly increases the likelihood. Informally, A probabilistically causes B if knowing that A occurred raises the probability that B will occur. Distinguishing genuine probabilistic cause from spurious correlation, however, requires adding conditions about temporal order and plausible mechanisms, a task that different researchers have resolved in different ways.

• 5. Causality, Time, and Physics

  ›

  Alfred Robb showed that the properties of antecedence and contiguity, two features of causality, are topological in nature and can serve as the foundation from which notions of time and space are derived. Physicist Max Jammer described how Albert Einstein's postulate opens the way to constructing the causal topology of Minkowski space. This means that, in the framework of special relativity, causality is metaphysically prior to time and space rather than the reverse. A concrete consequence of this priority is the speed-of-light limit on causal efficacy. If causal influence could travel faster than light, then using the Lorentz transform of special relativity, one could construct a reference frame in which an observer would see an effect precede its cause, violating the entire structure of causality. Wave packets illustrate the distinction neatly: energy, which has causal efficacy, travels at a wave packet's group velocity, and that velocity cannot exceed the speed of light. Phase, which is not causal, travels at the phase velocity, which can exceed light speed without contradiction. In general relativity, causal considerations require that the universe's semi-Riemannian manifold be orientable, so that "future" and "past" are globally definable. In engineering, a causal system is one whose output depends only on current and past input values; any system that depends on future input values is termed acausal. Cellier, Elmqvist, and Otter have argued that physics is essentially acausal, noting that Ohm's law V = IR does not specify whether voltage causes current or current causes voltage.

• 6. Causal Inference in Science and Statistics

  ›

  Judea Pearl's causal calculus, also known as do-calculus, provides a formal system for inferring interventional probabilities from conditional probabilities when some variables remain unmeasured. One practical result of this theory is the backdoor criterion, a mathematical definition of confounding that helps researchers identify which variables to measure and adjust for in order to isolate a causal effect. The foundational idea of learning causal structure from data goes back to Sewall Wright's 1921 work on path analysis. George Rebane and Pearl later developed a recovery algorithm that rests on Wright's distinction between three types of causal substructures possible in a directed acyclic graph. In epidemiology, Austin Bradford Hill built on the work of Hume and Karl Popper and suggested, in his paper "The Environment and Disease: Association or Causation?," that researchers consider strength, consistency, specificity, and temporality when distinguishing causal from noncausal associations. In management, Kaoru Ishikawa developed a cause-and-effect diagram in the 1960s, now called an Ishikawa diagram or fishbone diagram, to categorize causes systematically for quality control in manufacturing. Critics have since noted that Ishikawa's method does not distinguish between necessary and sufficient conditions, a gap Ishikawa himself apparently did not recognize. In economics, regression analysis is used to infer contributory causality from observational data, but the technique requires ruling out reverse causation and omitted variable bias, neither of which can be guaranteed with certainty.

• 7. Causality Across Disciplines

  ›

  Psychologists at the University of California, Berkeley, David Sobel and Alison Gopnik, designed a device called the blicket detector to probe how even young children learn causal powers. Their research showed that children quickly learn about a new causal power of an object and spontaneously use that information in naming and classifying it. Anjan Chatterjee at the University of Pennsylvania and Jonathan Fugelsang at the University of Waterloo have used neuroscience techniques to study the neural underpinnings of causal launching events, in which one object sets another in motion. In law, legal cause must be demonstrated to hold a defendant liable for a crime or civil wrong, and causation is also an essential element for claiming remedies under international trade law. In history, John Gaddis has distinguished between exceptional and general causes, and between immediate, intermediate, and distant causes. His example is telling: in accounting for what happened at Hiroshima on the 6th of August 1945, he argues, greater importance attaches to President Truman's order to drop an atomic bomb than to the Army Air Force's decision to carry out that order. Hindu philosophical texts from the Vedic period, dating roughly to 1750-500 BCE, contain early discussions of karma as a principle of cause and effect across lifetimes. The Bhagavad-gita 18.14 identifies five causes for any action: the body, the individual soul, the senses, the efforts, and the supersoul. Buddhist philosophy formulated the related concept of pratityasamutpada, or dependent origination, holding that everything arises in dependence upon multiple causes and conditions and that nothing exists as a singular, independent entity.

• 8. Philosophical Weight of a Single Concept

  ›

  J. L. Mackie's analysis of ordinary causal talk introduced the concept of the INUS condition: a cause is an insufficient but non-redundant part of a condition which is itself unnecessary but sufficient for the effect. A short circuit in a house is his canonical example. Taken alone, the short circuit would not have ignited a fire; without it, given everything else present, the fire would not have occurred. Mackie's account runs into difficulty, however, when applied to joint effects of a common cause: it mistakenly identifies one effect as an INUS condition for a second effect of the same cause, even when the two effects bear no causal relation to each other. Modern regularity theories aim to correct this by appealing to non-redundant regularities. The broader debate between necessary and sufficient causation has practical stakes. If x is necessary for y, then y cannot occur without x; if x is sufficient for y, then x alone guarantees y. A contributory cause, the third category, is neither necessary nor sufficient but genuinely adds to the effect in conjunction with other co-occurring factors. Whether philosophy can ultimately reduce causality to something more fundamental, such as empirical regularities, counterfactual dependencies, probabilistic relations, or mechanisms, remains open. Nobel laureate Herbert A. Simon and philosopher Nicholas Rescher argued that the asymmetry of causation is better captured as a functional relation between variables in a system of equations than as any form of logical implication. Whether pursued through Aristotle's four explanatory modes, Hume's eight criteria, Lewis's chains of counterfactual dependence, or Pearl's causal graphs, the concept continues to resist any single, universal formulation.