Evidence Pros
Down the Rabbit Hole

Knowledge Library

Scientific Laws, Theories, Models, and Assumptions

Every claim on Evidence Pros depends on some combination of observations, measurements, laws, theories, models, equations, and assumptions. The categories are not interchangeable, and none of them counts as evidence on its own. Use this library to see what each category can and cannot establish — and to trace which assumptions any given argument inherits.

Law vs. Theory — before you read further

  • • A theory does not 'graduate' into a law.
  • • A law describes a repeated relationship under stated conditions.
  • • A theory attempts to explain why observed relationships occur.
  • • A model represents a system and generates predictions.
  • • Mathematics expresses relationships. It does not establish that its starting assumptions describe physical reality.

Knowledge classification chart

Twenty categories. Each states what it can and cannot establish, the evidence it requires, an example, its common misuse, and whether it is independently testable.

  • Observation

    Sometimes testable

    Something directly witnessed or detected by a person or instrument.

    Can establish
    That an event or state was perceived at a specific time, place, and condition.
    Cannot establish
    Why the event occurred, or that the perception was accurate without corroboration.
    Required evidence
    Observer identity, time, location, conditions, and instrument if any.
    Common misuse
    Treating an inferred cause ('gravity pulled it down') as if it were the observation itself.
    Example
    At 06:42 UTC on 2024-04-08 from Mazatlán, the sun was fully occluded by the moon.

  • Measurement

    Independently testable

    A quantified observation produced with an instrument or defined method.

    Can establish
    A numeric value with units, within stated uncertainty.
    Cannot establish
    That the measured quantity is caused by the mechanism the observer assumes.
    Required evidence
    Instrument, calibration, method, raw reading, uncertainty.
    Common misuse
    Reporting a measurement without uncertainty or calibration provenance.
    Example
    Sea-level atmospheric pressure = 101.3 kPa ± 0.1 kPa (calibrated barometer, 15 °C).

  • Raw data

    Independently testable

    Measurements before interpretation, filtering, or model-fitting.

    Can establish
    What the instruments recorded.
    Cannot establish
    The physical meaning of the numbers by itself.
    Required evidence
    Original files or logs, timestamped, unmodified.
    Common misuse
    Presenting model-corrected or averaged values as raw.
    Example
    The unaltered CSV of a barometric logger's per-second readings.

  • Fact

    Independently testable

    A repeatedly confirmed observation within stated conditions.

    Can establish
    That the observation holds under the conditions where it has been checked.
    Cannot establish
    That it holds outside those conditions, or that its cause is understood.
    Required evidence
    Multiple independent observations agreeing within uncertainty.
    Common misuse
    Extrapolating a fact beyond the conditions where it was confirmed.
    Example
    Water freezes at 0 °C at 101.3 kPa in a still, pure sample.

  • Hypothesis

    Independently testable

    A proposed, testable explanation for an observation.

    Can establish
    A candidate mechanism worth testing.
    Cannot establish
    That the mechanism is true, merely because the hypothesis is plausible.
    Required evidence
    A statement precise enough to make a falsifiable prediction.
    Common misuse
    Calling an untestable narrative a hypothesis.
    Example
    Hypothesis: atmospheric refraction lifts distant objects above the geometric horizon.

  • Prediction

    Independently testable

    What should occur if a hypothesis or model is correct.

    Can establish
    A commitment made before results are collected.
    Cannot establish
    The truth of the underlying model; predictions can succeed for wrong reasons.
    Required evidence
    Preregistered value, direction, and uncertainty, dated before the test.
    Common misuse
    Adjusting the prediction after seeing the result.
    Example
    If the Earth is a sphere of radius 6371 km, an object 20 km away sits ~31 m below the geometric horizon.

  • Experiment

    Independently testable

    A controlled procedure designed to test a prediction.

    Can establish
    Whether the predicted result occurs under stated conditions.
    Cannot establish
    A general truth from a single trial, or a result if controls are absent.
    Required evidence
    Documented setup, controls, procedure, raw measurements, uncertainty.
    Common misuse
    Calling a demonstration an experiment when no control or falsifier exists.
    Example
    Bedford Level canal sighting with theodolite calibrated to a plumb line at fixed height.

  • Scientific law

    Independently testable

    A mathematical or descriptive relationship repeatedly observed under defined conditions.

    Can establish
    That the relationship holds within its stated domain.
    Cannot establish
    Why the relationship holds; a law describes, it does not explain.
    Required evidence
    Broad empirical support, defined conditions, known limits.
    Common misuse
    Treating 'law' as absolute or as an explanation.
    Example
    PV = nRT for an ideal gas, within temperature and pressure ranges where the ideal assumption applies.

  • Scientific theory

    Sometimes testable

    A broad explanatory framework supported by multiple lines of evidence.

    Can establish
    A coherent explanation that generates further testable predictions.
    Cannot establish
    Certainty. A theory does not 'graduate' into a law.
    Required evidence
    Multiple independent lines of confirming evidence; stated falsification conditions.
    Common misuse
    Dismissing a theory as 'just a theory', or treating it as proven because widely accepted.
    Example
    Plate tectonics: continents drift on convecting mantle plates.

  • Model

    Sometimes testable

    A simplified representation used to calculate or predict behavior.

    Can establish
    Predictions within the model's assumed regime.
    Cannot establish
    That the physical world matches the model, only that its outputs align within tolerance.
    Required evidence
    Stated inputs, assumptions, equations, and validation range.
    Common misuse
    Presenting model output as a direct measurement of nature.
    Example
    The ideal gas model; the standard atmospheric refraction coefficient k ≈ 0.13.

  • Principle

    Independently testable

    A foundational relationship or rule used within a field.

    Can establish
    A working rule with wide empirical support in its domain.
    Cannot establish
    A universal truth; principles have scope limits.
    Required evidence
    Historical experiments plus continued replication.
    Common misuse
    Applying a principle outside the fluid, geometry, or scale where it was validated.
    Example
    Archimedes' principle: buoyant force equals the weight of displaced fluid.

  • Assumption

    Sometimes testable

    Something accepted for purposes of analysis but not established by that analysis.

    Can establish
    Nothing on its own; it is a starting condition.
    Cannot establish
    Its own truth. Conclusions depending on it inherit its uncertainty.
    Required evidence
    Explicit statement, and separate evidence if the assumption is to be trusted.
    Common misuse
    Treating an assumption as an observed fact once buried inside a calculation.
    Example
    Assume the Earth is a sphere of radius 6371 km when computing horizon drop.

  • Axiom

    Not independently testable

    A starting proposition used within a logical or mathematical system.

    Can establish
    The logical structure that follows from it.
    Cannot establish
    That the physical world obeys the system.
    Required evidence
    Consistency within the system; nothing more.
    Common misuse
    Treating a mathematical axiom as a physical claim.
    Example
    Euclid's parallel postulate.

  • Equation

    Not independently testable

    A mathematical relationship between variables.

    Can establish
    How named quantities relate, given the equation's assumptions.
    Cannot establish
    That the equation describes a physical system, without measurement.
    Required evidence
    Formula, defined variables, units, and validated conditions.
    Common misuse
    Presenting an equation as proof, or 'the math shows it', without measured inputs.
    Example
    F = m·a.

  • Simulation

    Sometimes testable

    A model executed computationally.

    Can establish
    What the model produces under given inputs.
    Cannot establish
    A physical result. A simulation is not an experiment.
    Required evidence
    Model source, inputs, versions, and comparison to physical measurement.
    Common misuse
    Presenting a rendering or animation as photographic evidence.
    Example
    A ray-traced globe image at fixed camera altitude and standard atmosphere.

  • Interpretation

    Sometimes testable

    An explanation applied to observations or data.

    Can establish
    A candidate meaning consistent with the data.
    Cannot establish
    That the interpretation is unique; competing interpretations may fit equally.
    Required evidence
    Explicit interpretive framework and enumeration of alternatives.
    Common misuse
    Presenting the interpretation as the observation.
    Example
    Interpreting a red-shifted spectrum as recession velocity.

  • Testimony

    Sometimes testable

    A person's report of an experience or event.

    Can establish
    What the person says they experienced.
    Cannot establish
    That the described event physically occurred as described, without corroboration.
    Required evidence
    Identity, date, conditions, and any recording of the report.
    Common misuse
    Weighing testimony as if it were instrumented measurement.
    Example
    A pilot's statement about visible horizon curvature at cruise altitude.

  • Consensus

    Not independently testable

    Agreement among people or institutions.

    Can establish
    That agreement exists.
    Cannot establish
    The physical truth of what is agreed upon. Consensus is not evidence.
    Required evidence
    None beyond the fact of agreement itself.
    Common misuse
    Using consensus as a substitute for the underlying observations and experiments.
    Example
    Most textbooks state Earth is an oblate spheroid.

  • Falsification

    Independently testable

    A result that contradicts a testable prediction.

    Can establish
    That the prediction is wrong under the tested conditions.
    Cannot establish
    That every prediction of the underlying model is wrong.
    Required evidence
    Preregistered prediction, controlled test, matching uncertainty.
    Common misuse
    Declaring a model falsified from a single anomalous, unreplicated result.
    Example
    A Foucault pendulum failing to rotate as predicted at a given latitude.

  • Replication

    Independently testable

    Independent repetition of a test to determine whether the result persists.

    Can establish
    Whether the original result is reproducible.
    Cannot establish
    The cause of the result; only its persistence.
    Required evidence
    Independent team, equipment, and location applying the same procedure.
    Common misuse
    Counting reruns by the same team as independent replications.
    Example
    Multiple laboratories independently reproducing a pendulum-precession measurement.

Verification rules

  1. Never label something a law merely because a source calls it one.
  2. Never treat an equation as physical proof.
  3. Never treat a theory as proven solely because it is widely accepted.
  4. Never dismiss a theory solely because it remains called a theory.
  5. Require original sources, experiments, measurements, and replication records.
  6. Separate verified performance from philosophical interpretation.
  7. State clearly when competing models predict the same observation.
  8. Never invent a quantitative prediction for a model that has not supplied one.
  9. Connect every law, theory, model, equation, and assumption to the claims that depend on it.