An experiment is a systematic investigation in which researchers deliberately arrange or manipulate conditions to observe the effects on a phenomenon of interest. The defining feature is intervention: an investigator changes at least one aspect of a system or assigns different conditions and measures resulting outcomes. Experiments are used to test hypotheses, investigate causal relationships, estimate effects, compare treatments and examine mechanisms. They can occur in laboratories, field environments, clinical settings, engineering facilities or other contexts.
A simple experiment may manipulate one factor and measure one outcome, while complex experiments can involve many variables, treatment groups, repeated measurements and sophisticated statistical designs. Experiments differ from purely observational studies because the researcher deliberately controls or assigns conditions rather than simply recording naturally occurring variation.
Experimental variables are central to experimental design. An independent variable is a factor manipulated or assigned by the researcher, while a dependent variable is an outcome measured to determine whether it changes in response. Controlled variables are conditions maintained consistently where possible. A treatment refers to an experimental condition applied to subjects, objects or systems, while a control condition provides a comparison under specified circumstances. Variables must be operationally defined so that other researchers can determine exactly what was manipulated and measured. For example, an investigation into temperature and reaction rate must specify temperature conditions, reaction composition, measurement procedure and time interval. Precise definitions reduce ambiguity and improve reproducibility.
Experiments can test causation because intervention changes the explanatory factor while other relevant conditions are controlled or balanced. If changing the independent variable consistently produces a corresponding change in the dependent variable and alternative explanations have been adequately addressed, causal inference becomes stronger. Randomisation is frequently used to distribute unknown or uncontrolled influences across experimental groups. Replication helps distinguish systematic effects from random variability. Blinding can reduce the influence of expectations when subjective measurement or human behaviour is involved. These principles are particularly important in medical, psychological and biological experiments but also apply more broadly. The strength of causal inference depends on the quality of the entire design rather than simply the existence of a treatment and an outcome.
Laboratory experiments provide substantial control over environmental conditions and can make precise measurements possible. Researchers can isolate specific variables, repeat trials and use specialised instruments. However, laboratory conditions can differ from real world environments, potentially limiting external validity. Field experiments occur in natural or realistic settings and can provide greater ecological relevance, but they generally involve less control over environmental variability.
Some experiments use model organisms, simulated systems or physical models to investigate processes that would be difficult or unethical to study directly. The choice of experimental setting depends on the research question. A highly controlled experiment may establish a mechanism clearly, while a field experiment may determine whether the mechanism operates under realistic environmental conditions.
Experimental design includes considerations such as sample size, replication, randomisation, blocking, treatment allocation, measurement precision and statistical power. Replication can refer to repeated experimental units within a study and should be distinguished from repeated measurements of the same unit. Pseudoreplication occurs when nonindependent observations are incorrectly treated as independent replicates, potentially producing misleadingly precise results. Blocking can reduce the influence of known sources of variation by grouping similar experimental units before treatment assignment. Factorial designs can investigate multiple independent variables simultaneously and test interactions between them. Appropriate design can therefore increase the information obtained from a given number of observations while reducing confounding and improving statistical efficiency.
Measurement is an essential part of experimentation because conclusions depend on the quality of recorded outcomes. Instruments should be calibrated where appropriate, and researchers should assess accuracy, precision, sensitivity and detection limits. Measurement error can be random or systematic. Random error contributes to variability, whereas systematic error can consistently bias results in one direction. Repeated measurements can help estimate random variation but do not automatically correct systematic bias. Researchers may use validated protocols, reference materials, instrument calibration and independent measurements to improve reliability. Data recording procedures should be specified in advance where possible to reduce selective reporting and analytical flexibility.
Experiments have ethical and practical constraints. Human and animal research may require ethical review, informed consent or adherence to welfare standards depending on the jurisdiction and research type. Some variables cannot ethically be manipulated because doing so would create unacceptable risks. Physical systems may also be inaccessible or too large, small, slow or dangerous for direct experimentation. Researchers may therefore combine experiments with observational studies, simulations, historical evidence and theoretical modelling.
A limitation of experimentation is that artificial control can sometimes alter the phenomenon being studied. Another is that an experiment designed to isolate one mechanism may not capture complex interactions occurring in real systems. These limitations do not invalidate experimental evidence but determine how broadly its conclusions should be applied.
A good experiment produces evidence that can be evaluated and, ideally, independently repeated. Researchers normally specify the question, hypothesis, design, materials, procedures, variables, treatment allocation, measurements, statistical analysis and limitations. Results may support the hypothesis, contradict it or reveal a more complicated relationship than expected. An experiment does not automatically prove a hypothesis because alternative explanations, sampling limitations and measurement uncertainty can remain. Strong experimental conclusions arise when the design directly addresses the hypothesis, relevant confounders are controlled, measurements are reliable and results are consistent across appropriate replications. Experiments are therefore powerful tools for causal investigation, but their value depends fundamentally on rigorous design, execution, analysis and interpretation.
