What Is a Scientific Explanation?
Published:
Science does not merely ask what happens.
It asks why.
Why do planets orbit?
Why does ice melt?
Why do species change?
Why does a disease spread?
Why did this bridge fail?
A scientific explanation connects a phenomenon to a broader structure that makes it understandable.
But not all explanations have the same form.
Some appeal to laws.
Some to causes.
Some to mechanisms.
Some to statistical patterns.
Some to unification.
The philosophy of science has tried to understand what these forms share.
Description Is Not Explanation
Suppose we observe:
the pressure of a gas increases when its volume decreases.
That is a pattern.
An explanation asks why.
One answer invokes molecular motion and collisions.
The descriptive regularity becomes intelligible through a deeper mechanism.
Science often moves from:
what
to
how
to
why.
Prediction Is Not Explanation
A model may predict perfectly without explaining.
A machine-learning system can forecast which patients are at risk while revealing little about causal mechanism.
Conversely, an explanation may be good even when exact prediction is impossible.
Evolutionary theory explains adaptation.
It cannot predict every future mutation.
Prediction and explanation overlap.
They are not identical.
Hempel’s Covering-Law Model
Carl Hempel proposed an influential account called the deductive-nomological model.
A phenomenon is explained by showing that it follows deductively from:
- general laws,
- initial conditions.
For example:
Law: gases expand when heated under specified conditions.
Initial conditions: this gas was heated.
Therefore: it expanded.
The event is “covered” by a law.
Deductive-Nomological Explanation
In the ideal form:
- state relevant laws,
- state initial conditions,
- derive the event.
If the derivation is valid, the explanandum follows logically.
This captures an important feature of physics.
But it does not fit every scientific explanation.
Statistical Explanation
Many phenomena are probabilistic.
Smoking increases risk of lung cancer.
A radioactive nucleus has a decay probability.
A treatment improves average outcomes.
Hempel extended his approach with statistical models.
The explanation may show that an event was highly probable under specified conditions.
But probability raises new problems.
A low-probability event can still occur.
How should it be explained?
The Flagpole Problem
A famous objection to simple covering-law accounts uses a flagpole.
Given:
- the height of the pole,
- the angle of the sun,
- laws of optics,
we can derive the length of the shadow.
But mathematically, given the shadow length and sun angle, we can derive the pole height.
Does the shadow explain the pole?
Usually not.
The asymmetry shows that derivability alone is insufficient.
Explanation seems to require causal direction.
Causal Explanation
A causal explanation identifies factors that produced the event.
The window broke because the ball struck it.
The patient improved because the antibiotic eliminated the bacterial infection.
The eclipse occurred because orbital geometry aligned the bodies.
Causation gives direction to explanation.
Causes explain effects more naturally than effects explain causes.
Mechanistic Explanation
A mechanistic explanation describes the entities, activities, and organization producing a phenomenon.
For example:
How does a neuron generate an action potential?
Answer through:
- ion channels,
- membrane potential,
- ion gradients,
- voltage-dependent dynamics.
The mechanism explains how the process works step by step.
This form is especially important in biology and neuroscience.
Mechanism vs Law
Mechanistic explanations need not reduce to one universal law.
Biological systems often depend on organized structures.
A heart pumps because chambers, valves, electrical conduction, and muscle contraction are arranged in a particular way.
The explanation depends on organization.
Knowing fundamental physics alone would not be an efficient explanation.
Functional Explanation
Biology often uses functional explanations.
The heart pumps blood.
Eyes enable vision.
Roots absorb water.
Such statements can mean:
this trait contributes to a system-level capacity, or this trait was historically selected because of its effect.
Function is not automatically purpose in a conscious sense.
Scientific functional explanation needs careful causal grounding.
Historical Explanation
Some phenomena are explained through history.
Why are marsupials concentrated in Australia?
Why does a language have a particular irregular form?
Why does a species possess a strange anatomical structure?
The answer may depend on historical sequence rather than universal law alone.
Evolutionary and geological explanations are often path-dependent.
Path Dependence
In a path-dependent system, present structure depends on the route taken.
Two systems governed by the same general laws can end differently because their histories differ.
This limits explanations that appeal only to timeless laws.
History itself becomes explanatory.
Unification
Another idea is that explanation comes from unification.
A powerful theory explains many apparently different phenomena using one framework.
Newton unified:
- falling bodies,
- planetary motion,
- tides.
Maxwell unified electricity, magnetism, and light.
Unification reduces the number of independent assumptions needed to understand the world.
Kitcher’s Unification View
Philosopher Philip Kitcher developed an account in which explanation consists partly in deriving many phenomena from a small set of argument patterns.
The more the theory unifies, the more explanatory power it has.
This captures why scientists value broad theories.
But unification alone can also miss causal detail.
Different kinds of explanation may complement each other.
Mathematical Explanation
Sometimes mathematics explains directly.
Why do soap films form certain shapes?
Energy minimization and geometry can explain.
Why do some population patterns emerge?
Network topology may matter.
Why are particular symmetries unavoidable?
Group structure may constrain possibilities.
Not every scientific explanation is a story about one event causing another.
Constraint-Based Explanation
Some explanations show that alternatives are impossible or highly restricted.
Conservation laws constrain processes.
Symmetry restricts allowed interactions.
Topology can prevent certain transformations.
In these cases, explanation comes from structural constraint.
The event occurs because the space of possibilities is narrow.
Equilibrium Explanation
Thermodynamics often explains macroscopic behavior through equilibrium principles.
Why does heat flow from hot to cold?
Why does gas spread through a container?
The explanation may appeal to overwhelmingly larger numbers of high-entropy microstates.
This is statistical and structural rather than one simple causal chain.
Explanatory Depth
An explanation can answer one “why” and still invite another.
Why does the apple fall?
Because of gravity.
Why gravity?
Because spacetime is curved.
Why does mass-energy curve spacetime?
Because Einstein’s equations relate them.
Why those equations?
The regress continues.
Scientific explanation usually stops at the level relevant to the question.
Appropriate Level
A good explanation matches the scale of inquiry.
Why did the car stop?
“Because electromagnetic interactions between atoms…” is physically true but explanatorily poor.
“Because the driver pressed the brake and friction slowed the wheels” is better.
Fundamental detail is not always explanatory depth.
Relevance matters.
Reduction
Some explanations reduce a phenomenon to lower-level processes.
Temperature can be explained statistically through microscopic motion.
Chemical bonding through quantum mechanics.
Inheritance partly through molecular biology.
Reduction can be powerful.
But not every useful explanation must be fully reduced.
Higher-level regularities may remain indispensable.
Emergent Explanation
Some phenomena are best explained at collective levels.
Traffic jams.
Phase transitions.
Evolutionary dynamics.
Economic bubbles.
Conscious states, perhaps.
Lower-level laws constrain them.
But large-scale patterns may require new variables and concepts.
This tension between reduction and emergence will receive an entire section later.
Explanation and Causation
Does every explanation require causation?
No.
Mathematical identities can explain patterns without temporal causes.
Symmetry can explain conservation.
Geometry can explain constraints.
But causal explanation remains central when the question concerns why one event happened rather than another.
Different “why” questions demand different answers.
Counterfactuals
A causal explanation often supports counterfactuals.
If the cause had been absent, would the effect have changed?
The bridge collapsed because a support failed.
Had the support remained intact, collapse would probably not have occurred.
This counterfactual dependence is a powerful way of distinguishing causes from mere correlations.
Manipulation
Some philosophers connect causation to intervention.
If changing X while holding relevant factors fixed changes Y, then X is causally relevant to Y.
This idea links explanation directly to experiment.
Causal explanation becomes tied to possible manipulation.
Explanatory Relevance
An explanation should identify information that matters to the phenomenon.
Suppose a patient recovers.
“The patient recovered because they were wearing blue socks” may be true in the sense that both facts occurred.
But the socks are irrelevant.
Explanation requires more than association.
It requires relevance under the mechanism or causal structure.
Too Much Explanation
An explanation can contain too much detail.
Listing every molecular collision in a hurricane would not improve meteorological understanding.
Good explanations compress.
They reveal the variables that matter.
This is one reason explanatory science depends on abstraction.
Explanation and Understanding
A scientific explanation is successful partly when it produces understanding.
We can answer:
What would happen if conditions changed?
Which factors matter?
How does the phenomenon connect to other knowledge?
What mechanism produces it?
Understanding is more than memorizing a prediction.
It is the ability to navigate the structure.
Can Everything Be Explained?
Earlier we asked whether everything can be explained.
Scientific explanation shows why the question is hard.
Every explanation uses:
- laws,
- models,
- mechanisms,
- boundary conditions,
- background assumptions.
Those may themselves need explanation.
At some point, explanatory chains reach deeper theory, brute facts, necessity, or open questions.
Science does not promise an infinite regress ending in certainty.
A Working Definition
A useful working definition is:
A scientific explanation shows how a phenomenon follows from, is produced by, or is constrained by empirically supported laws, mechanisms, structures, or historical conditions.
This definition is intentionally plural.
Science explains in more than one way.
The Next Problem
Many explanations are causal.
But causation is easy to confuse with correlation.
Two variables can move together because one causes the other.
Or because both share a common cause.
Or because the relationship is accidental.
So the next question is unavoidable:
What is causality, and why is correlation not enough?