Causation refers to a relationship in which an event, condition or intervention produces or contributes to a change in another event, condition or outcome. A causal claim is stronger than a claim that two variables are associated because it asserts a mechanism or process through which differences in one factor affect another under specified conditions. In scientific research causal reasoning concerns questions such as whether an exposure changes an outcome, whether an intervention produces an effect or whether modifying a factor would alter a result. The precise meaning of a causal claim depends on the scientific framework and the population, intervention and circumstances being considered.
Establishing Causation requires more than observing that two variables occur together. Researchers commonly consider temporal ordering, alternative explanations, confounding, consistency of evidence, plausible mechanisms, experimental manipulation and the relationship between exposure and outcome. Randomised experiments can provide strong evidence because random assignment can reduce systematic differences between groups under appropriate conditions. Observational studies can also provide causal evidence when supported by suitable designs, assumptions, natural experiments, longitudinal information and careful adjustment for confounding. No single statistical technique automatically transforms an association into a causal relationship.
Confounding is a central problem in causal inference. A confounder is a variable associated with both an exposure and an outcome that can create or distort an observed association if not appropriately addressed. For example, an apparent relationship between two variables may partly or entirely reflect a third variable that influences both. Researchers can address confounding through randomisation, restriction, matching, stratification, statistical adjustment or other causal inference methods depending on the design and assumptions. Adjustment is not universally beneficial because controlling for certain variables can itself introduce bias when those variables lie on a causal pathway or are affected by the exposure.
Causal reasoning distinguishes several concepts that are often confused. Necessary causes are factors that must be present for an outcome to occur under specified conditions, while sufficient causes are conditions that can produce the outcome. A contributory cause increases the probability of an outcome without being the only possible cause. Causal effects can also be heterogeneous, meaning an intervention may produce different effects for different populations or circumstances. Therefore a statement that one factor causes another does not necessarily mean that the cause is the sole explanation or that every exposed individual will experience the same outcome.
Randomised experiments are an important tool for investigating Causation because researchers assign interventions rather than merely observing naturally occurring differences. Randomisation helps balance measured and unmeasured characteristics on average, while control groups provide a comparison for estimating the effect of the intervention. Blinding can reduce certain expectation and measurement effects. Experimental evidence still requires appropriate implementation, sufficient adherence, valid outcome measurement and suitable analysis. Generalising an experimentally estimated causal effect to a different population or setting also requires consideration of differences in context and effect modifiers.
Causal inference from observational data often relies on explicit assumptions about how variables relate to one another. Directed acyclic graphs can represent proposed causal structures and help identify potential confounders, mediators and selection variables. Counterfactual approaches define a causal effect by comparing an outcome under one exposure condition with the outcome that would have occurred for the same unit under an alternative condition. Because both potential outcomes cannot generally be observed simultaneously for the same unit, researchers use study designs and assumptions to estimate the relevant contrast. These frameworks make causal assumptions more explicit rather than removing the need for them.
Causation is often communicated using terms such as causal effect, intervention effect, mechanism, determinant, risk factor and causal pathway, but these expressions can have different technical meanings. A risk factor is not necessarily a direct cause because it may be correlated with other causal variables. A mediator can lie between an exposure and an outcome, meaning that adjusting for it may answer a different causal question. Reverse causation occurs when the apparent direction of influence is opposite to the one initially proposed. Recognising these distinctions prevents causal language from being inferred automatically from statistical association.
Causal conclusions are necessarily tied to a defined question and context. A study may provide evidence that changing one variable affects an outcome under specific conditions without demonstrating that the same effect occurs everywhere or for everyone. Replication, convergence of evidence from different methods and examination of alternative explanations strengthen causal understanding. Scientific Causation is therefore not simply a stronger form of Correlation but a conclusion supported by research design, temporal information, causal assumptions, empirical evidence and substantive understanding of the mechanism connecting variables.
