What Is Evidence?

Evidence is information that changes what we should believe

Evidence is information that counts for or against a claim. It does not have to make a claim certain. Its basic role is to change how reasonable the claim is compared with the alternatives. A photograph, measurement, document, experiment, witness statement or statistical pattern can all be evidence when they bear on whether something is true.

This means evidence should always be considered in relation to a particular claim. The same fact may strongly support one proposition, weakly support another and be irrelevant to a third. Philosophers of evidence therefore connect evidence closely with justification: evidence matters because it can make a belief more reasonable to hold. This is also why asking simply “Is there evidence?” is often insufficient. The more useful questions are: evidence for what, how strongly, produced how, and compared with which alternatives?

Relevance is the first test, not the last

An item is relevant when it would make the claim more or less likely than it would otherwise be. This basic idea appears in many fields. The United States Federal Rules of Evidence, for example, define relevant evidence in terms of whether it has a tendency to make a consequential fact more or less probable. Philosophy uses different terminology, but the underlying relationship is similar: evidence must bear on the proposition under consideration.

Relevance alone does not make evidence strong. A rumour might be slightly relevant to an allegation while remaining unreliable. A laboratory measurement may be highly relevant but weakened by a faulty instrument. A witness may have been present but unable to see clearly. Evidence must therefore be evaluated for both relevance and quality.

Different questions require different kinds of evidence

There is no single form of evidence that is best for every question. Claims about physical processes may be tested through measurement, observation and experiment. Historical questions often depend upon documents, artefacts, contemporary records and independent accounts. Legal proceedings may use testimony, physical exhibits, expert evidence and documentary records. Questions about populations commonly require statistical evidence rather than individual anecdotes.

Personal experience can also be evidence, particularly for what a person experienced. But an experience and its explanation are not identical. Feeling a presence is evidence that a person had a particular experience; it does not by itself establish what caused that experience. The same distinction matters when evaluating eyewitness testimony, memory and personal experience.

Logical and conceptual reasoning play a different role again. A contradiction can show that a set of propositions cannot all be true together. A valid deduction can establish what follows from stated premises. But valid reasoning cannot make an unsupported premise factually true. Evidence and reasoning work together: evidence informs the premises we should accept, while reasoning determines what those premises support.

Reliability depends on how the evidence was produced

Evidence is stronger when the process that produced it is capable of distinguishing truth from error. A calibrated instrument is generally more useful than an unknown device. A record created at the time of an event may be more dependable than a recollection written decades later. A study designed to control confounding variables may tell us more about causation than an uncontrolled observation.

Source reliability also matters. We should ask whether a source had direct access to the relevant information, whether it has expertise in the subject, whether its methods are transparent, whether errors can be corrected and whether important conflicts of interest exist. None of these factors creates an infallible authority. They help estimate how much weight a source deserves.

Bias does not automatically make evidence false. Governments, companies, campaign groups, religious institutions, critics and researchers may all have perspectives or interests. The appropriate response is not to discard every interested source, but to distinguish what the source is in a position to establish from what it merely asserts or interprets.

Independent evidence is stronger than repetition

Ten websites repeating one original report do not provide ten independent pieces of evidence. They may represent a single claim copied nine times. Independence matters because genuinely separate observations are less likely to share exactly the same error, fabrication or misunderstanding.

Corroboration becomes more persuasive when different lines of evidence converge. A historical claim supported independently by contemporary documents, physical evidence and records from opposing parties is generally stronger than the same claim supported only by later retellings. In science, similar reasoning helps explain why replication and converging results can increase confidence.

This distinction is particularly important online. Search results can create an illusion of consensus when many pages trace back to one press release, social-media post or news agency. A good investigation follows information back to its origin before counting sources.

Scientific evidence becomes stronger through testing and cumulative inquiry

Science does not rely upon one universal ritual called “the scientific method”. Different disciplines use different methods, but systematic observation, experimentation, modelling, hypothesis testing, statistical reasoning and critical scrutiny are recurring features. The Stanford Encyclopedia of Philosophy notes both this methodological diversity and the importance of testing how well hypotheses fit observations.

A successful test can support a hypothesis without proving it permanently. Scientific conclusions remain open to revision if better data, improved methods or competing explanations emerge. This fallibility is not a weakness of evidence-based reasoning; it is a mechanism for correcting error.

Replication and reproducibility can also contribute to confidence. The National Academies describes them as important parts of science's self-correcting process while warning that replication alone is not the sole measure of reliable science. What matters is the cumulative body of evidence, the quality of the methods and whether findings survive serious attempts at checking them.

In evidence-based medicine, the Cochrane framework illustrates another important principle: not all bodies of evidence deserve the same confidence. Its GRADE approach considers factors including risk of bias, inconsistency, indirectness, imprecision and publication bias. The broader lesson applies well beyond medicine: evidence should be weighted according to how it was obtained and what uncertainties remain.

Evidence is not the same as proof or certainty

In mathematics and formal logic, proof can establish a conclusion from axioms or premises through valid deduction. Most claims about the world do not have that kind of certainty. Historical, scientific, medical and everyday conclusions are normally probabilistic: the evidence can make one explanation substantially more reasonable than another without eliminating every conceivable possibility.

Demanding absolute certainty before accepting any empirical claim would make ordinary knowledge almost impossible. At the opposite extreme, treating any supportive observation as sufficient evidence makes us vulnerable to coincidence, selective reporting and wishful thinking. Rational confidence should therefore be proportional to both the amount and the quality of the evidence.

This is especially important for extraordinary claims. A claim that conflicts with a large body of well-tested knowledge must explain not only the new observation but also why the existing evidence appeared to support the established account. Strong claims need evidence capable of discriminating between the proposed explanation and more ordinary alternatives.

Absence of evidence can matter when evidence should exist

The phrase “absence of evidence is not evidence of absence” is only partly correct. A failure to find something tells us little when we had little chance of detecting it in the first place. Searching one small patch of ocean and failing to see a particular fish provides little evidence that the species does not exist anywhere.

But absence becomes informative when a claim predicts evidence that should be readily detectable. If a proposed event should have produced extensive contemporary records, measurable effects or repeatable observations, a careful failure to find those expected traces can count against the claim. The strength of negative evidence therefore depends on the probability that the evidence would have appeared if the claim were true.

How to judge evidence in practice

A useful evidence assessment does not ask whether a source feels persuasive. It breaks the problem into questions that can be examined consistently:

• Claim: What exactly is being asserted?
• Relevance: Does this information actually bear on that claim?
• Source: Who produced the evidence, and how could they know?
• Method: How was the information observed, measured, recorded or inferred?
• Independence: Are apparently separate sources genuinely independent?
• Corroboration: Do different forms of evidence converge on the same conclusion?
• Alternatives: What competing explanations could produce the same observation?
• Bias and error: What incentives, methodological weaknesses or cognitive biases could distort the result?
• Uncertainty: Which parts of the conclusion are well established and which remain doubtful?
• Falsifiability: What evidence would cause confidence in the claim to decrease?

The final question is particularly valuable. If no possible evidence could count against a claim, the claim is protected from correction rather than supported by evidence. Good reasoning makes room for revision.

Evidence should be judged by consistent standards

One of the hardest parts of evaluating evidence is applying the same standard to conclusions we favour and conclusions we dislike. Confirmation bias can lead people to accept weak supporting evidence while demanding near-perfect proof from opposing claims. Political, religious and personal commitments can intensify that tendency.

A disciplined approach therefore asks what level of evidence the claim deserves before knowing whether the evidence supports our preferred conclusion. The appropriate threshold depends partly on the claim's probability, specificity and consequences, an issue explored further in How Much Evidence Should a Claim Require?

Evidence does not remove every disagreement, because people can dispute methods, background assumptions and the weight given to different facts. But transparent standards make disagreements more productive. They allow us to identify whether the dispute concerns the evidence itself, its reliability, an alternative explanation or the values used to make a decision.