In many situations, one outcome happening tells us nothing about another. A system works, a coin lands heads, a sensor triggers, a value exceeds a threshold — and these situations unfold without affecting each other. Probability treats this kind of "no influence" as a distinct idea.
Independence captures the situations where events stand on their own. Nothing about one event changes how we think about the other, and no new information is gained from seeing one occur. This idea appears everywhere: repeated experiments, separate components, unrelated conditions, or processes that evolve without interaction.
The rest of the page develops what independence means, how it is expressed formally, and how it connects to other probability concepts.
Draw a card, and someone tells you one fact about it. Does that change the chance that it is an ace?
Before introducing any formal definition, it helps to understand the basic idea. Two events do not influence each other when the occurrence of one provides no information about the other. Learning that one situation happened does not change how we think about the likelihood of the second.
Learning the suit or the colour leaves the chance at 1/13; learning that it is a face card drops it to 0.
what you learn about the card
P(ace) after learning it
nothing yet
4/52 = 1/13
it is a heart
1/13 — unchanged
it is red
2/26 = 1/13 — unchanged
it is a face card
0 — changed
A heart is as likely to be an ace as any card: 1 ace among 13 hearts. Face cards contain no aces at all.
Independence is the absence of an update.
This is an information-based view: independence is about the absence of update. If seeing one event occur leaves our expectations about the other exactly as they were before, then the two events behave independently.
The definition
Two events are independent when learning that one of them occurred leaves the probability of the other unchanged.
So "ace" and "heart" are independent events, while "ace" and "face card" are not.
Independence is a statement about information, not about physical separation.
The suit and the rank in the second row belong to the same card, yet neither tells anything about the other.
Independent is not the same as mutually exclusive.
"Ace" and "face card" can never happen together, and that is exactly why they are dependent: learning one rules the other out.
This perspective captures the core intuition and prepares the ground for the formal definition that follows.
The rest of the page covers the formal definition, notation, the independence formula, visual representations, examples, how independence fails, conditional independence, problem solving, common mistakes and connections to other concepts.
Formal Definition of Independence (In Words)
Fact: Independent of itself only at 0 or 1
Take B=A: knowing A occurred makes A certain, so A is independent of itself only if it was already certain or impossible. In symbols, P(A∩A)=P(A) must equal P(A)2, which forces P(A)=0 or 1.
Two events are considered independent when knowing that one has occurred does not alter the chance of the other. In other words, the likelihood of event A remains exactly the same whether event B happens or not, and vice versa.
This definition focuses on the idea of unchanged information. If the occurrence of one event never forces us to revise our expectation about the other, the two events meet the formal standard of independence, even before introducing any symbolic expressions.
Independence Notation
Notation
Independence Notation
The defining equations of independence for a single pair are written — and owned — at conditional probability notation. This page's marks begin where two events end: what independence looks like across a whole collection, inside a condition, between random variables, and compressed into statistics' favorite abbreviation. All of them are catalogued among the probability symbols.
For a collection, the product must hold for every sub-collection: three events are mutually independent only when the triple equation and all three pairwise equations hold — four statements, not one. For n events the count is 2n−n−1.
CasesPairwise independence — all pairs checked, the triple not — is strictly weaker: with two fair coins, "first is heads", "second is heads", "both show the same" are independent in every pair, yet the triple product fails. When a text calls a collection "independent" it means mutual unless it says otherwise.
Do not confuseOne equation standing in for all of them. Checking only the triple product, or only the pairs, each accepts collections that are not mutually independent — the shortfall runs in both directions.
A⊥B, X⊥⊥Y
A is independent of B
The whole defining equation in a single mark, borrowed from the perpendicularity symbol — the mathematical keyboard carries it. Between random variables, X⊥Y asserts that the joint distribution factors: every event about X is independent of every event about Y, infinitely many statements at once.
Also writtenWith the stroke doubled, X⊥⊥Y — Dawid's notation, standard in the graphical-models literature precisely so that single ⊥ keeps its geometric job; elementary texts avoid the symbol entirely and write the equation out.
Do not confuseOrthogonality proper. For zero-mean random variables E[XY]=0 means uncorrelated — a strictly weaker property than independence, so reading probability's ⊥ with geometry's meaning quietly downgrades the claim.
A⊥B∣C
A is independent of B, given C
The bar and the perp compose: inside the world where C holds, the product factors — P(A∩B∣C)=P(A∣C)P(B∣C). Conditional Independence below tells this story in prose; this line is its formula.
CasesThe slots take events or whole variables, X⊥Y∣Z — the workhorse mark of graphical models and the reasoning behind Bayes' theorem's sequential use; in those texts it usually wears the doubled stroke.
Do not confuseThe unconditional kind. Neither implies the other: dependence can vanish once a shared cause is fixed — the How Independence Fails patterns — and independent events can turn dependent inside a condition; both directions genuinely fail.
X1,…,Xn∼iidF
independent and identically distributed, drawn from F
Statistics' compression: one tag making two separate claims — mutual independence of the whole collection, and one shared distribution declared by the tilde.
CasesSpelled iid, i.i.d., or IID by house style; the tag rides above the tilde or trails the sentence. It is the standing assumption of sample-based statistics — sums, averages and the law of large numbers all open with "let X1,…,Xn be i.i.d.".
Do not confuseA single claim. Independent-but-differently-distributed and identically-distributed-but-dependent each break exactly one half of the abbreviation — and only the two halves together license the standard limit theorems.
Events and Probability Rules: Notation at a Glance
The marks from the notation cards across the pages on events and probability rules: how each is read and what it means. Every group links to the full card on its page.
The intuitive idea of independence becomes precise when expressed in terms of probabilities. Two events are independent exactly when their joint occurrence behaves like the product of their separate chances:
This statement captures the idea that combining the events does not introduce any new influence between them. It is the compact formal expression of "no change in information."
An equivalent way to view the same idea is through conditional probabilities:
P(A∣B)=P(A)
P(B∣A)=P(B)
Pitfall: Calling disjoint events independent
Treating "roll a 1" and "roll a 2" on one die as independent because they share no outcomes fails the test: P(A∩B)=0, but P(A)P(B)=361. Disjoint events with positive probability are dependent: P(A∣B)=0, not 61.
Each form highlights a different aspect, but they all represent the same underlying condition: the occurrence of one event leaves the probability of the other untouched.
In practice the formula works as a test. Given two events with known probabilities, compute P(A)P(B) and compare it with the probability that both occur; the events are independent only when the two numbers agree. The joint probability P(A∩B) has to be found separately, from the outcomes the two events share.
The product rule as an area
The unit square is all of the probability. Column A has width P(A) = 0.4. Event B takes the bottom 0.3 of the height in both columns, so P(B | A) = P(B | not A) = P(B) = 0.3. The blue rectangle A ∩ B is then 0.4 wide and 0.3 high, and its area is exactly P(A) · P(B) = 0.12. Equal heights in both columns and the product formula are the same fact.
Change either column's height and the blue area stops being P(A) · P(B).
Independence is a numerical test: two events are independent exactly when the chance of both equals the product of their chances.
Visual Representations
Independence can be understood more clearly by comparing it to situations where events do influence one another.
Venn-style view: Although real probabilities cannot be read from the areas of a standard Venn diagram, the picture helps convey the idea: the region representing A contains no "information distortion" from B, and vice versa. The overlap simply reflects the product structure implied by independence.
Tree diagram view: A probability tree makes independence especially clear. When events are independent, the branches for one event look the same regardless of whether the other event occurred. The structure of the tree does not change from one branch to the other, visually showing that no event alters the chances of the other.
These representations help highlight the contrast with dependent situations, where the shapes or branch weights change once one event is known to have occurred.
A waffle chart with equal columns: independence
Each column is a value of one variable and the shaded share in every column is the same: knowing the column tells you nothing about the shading, which is what independence looks like. Unequal shares across columns would mean dependence. Drag the shares apart and watch the events become dependent on the conditional probability waffle chart.
Examples show where such equal shares actually arise.
Examples
Independence shows up in many simple and practical situations:
1. Repeated Trials Consider flipping a fair coin twice. The result of the first flip does not affect the result of the second. If A is "first flip is heads" and B is "second flip is heads," then P(A∩B)=P(A)P(B), reflecting the independence of the trials.
2. Separate Components Imagine two unrelated sensors operating in different parts of a system. If their detections come from unrelated mechanisms, the event "sensor 1 triggers" and the event "sensor 2 triggers" behave independently. Observing one does not update our belief about the other.
3. Contrast With Dependence Suppose A is "it rains today" and B is "the ground is wet." These events are not independent: knowing B changes how we evaluate A. This contrast helps clarify what true independence looks like.
4. Table-Based Illustration A simple table of outcomes where every combination is equally likely (such as rolling two dice) often provides an easy demonstration of independent structure: each coordinate behaves as if the other were irrelevant.
Pitfall: Multiplying dependent events
Taking P(rain∩wet)=P(rain)P(wet), with rain on 30% of days and wet ground on 41%, gives 0.3⋅0.41≈0.12. Rain makes wet ground likely: if it is wet on 90% of rainy days, P(rain∩wet)=0.3⋅0.9=0.27.
In the first, second and fourth examples the two events come from separate mechanisms: two flips, two sensors, two dice. In the third, the wet ground is partly produced by the rain, so the two events are tied together by how the situation works. Problems usually state the separate probabilities of such events, and the chance that both happen is the quantity left to find.
These examples show both the appearance of independence and how it differs from scenarios where events influence one another.
The four scenarios above can be lined up with their independence verdict and the reason behind it.
Scenario
Verdict
Why
Two flips of a fair coin (A: 1st is H; B: 2nd is H)
independent
the second flip's mechanism is unaffected by the first
Two separate sensors firing (A: sensor 1; B: sensor 2)
independent
detections come from unrelated mechanisms
Rolling two dice (A: first die value; B: second die value)
independent
every (a, b) cell is equally likely; each coordinate is unaffected by the other
Rain today (A) and wet ground (B)
dependent
B is a downstream effect of A; learning B substantially updates P(A)
Rain and wet ground: dependent
Example 3 with numbers chosen for illustration: rain on 30% of days, the ground wet on 90% of rainy days and on 20% of dry ones. The wet part of the rain column is much taller than in the no-rain column, so learning the ground is wet changes the chance of rain. P(rain ∩ wet) = 0.27, far from P(rain) · P(wet) = 0.3 · 0.41 ≈ 0.12: the product rule fails.
Given wet ground, the chance of rain rises from 0.3 to 0.27/0.41 ≈ 0.66.
How Independence Fails (Dependence Patterns)
Many situations look independent at first glance but are not. Dependence appears whenever the occurrence of one event changes how we evaluate another.
A common failure pattern is shared causes. Two events may seem unrelated, but both are influenced by the same underlying factor. Observing one then provides information about the other.
Another pattern is structural restriction. When events draw from a limited set of possibilities, the occurrence of one may remove options for the other, creating dependence.
Fact: One die, two independent events
On one fair die, "even" has probability 21 and "at most 2" has 31; both happen only on a 2, with probability 61=21⋅31, so they are independent. Change the second event to "at most 3" and 61=41: now they are dependent.
A shared mechanism is a reason to check, not a verdict. Two events about a single roll of one fair die, such as "the roll is even" and "the roll is at most 2", come from the same roll and even share an outcome, the 2. Whether learning one changes the chance of the other is still settled only by the numbers: the probability that both occur against the product of the two separate probabilities.
Dependence also arises through conditioning. Events that are independent in isolation may become dependent once additional information is known, or dependent events may appear independent only within a restricted context.
Recognizing these patterns is essential, because assuming independence where it does not exist is one of the most common sources of error in probability reasoning.
Conditional Independence
In some situations, two events may influence each other in general, but become unrelated once additional information is known. This phenomenon is called conditional independence.
Here, the relationship between events depends on a third event or condition. Knowing this extra information can block the flow of influence between them, so that learning about one event no longer changes how we think about the other.
In symbols, for events A and B given a third event C:
Pitfall: Assuming independence survives a condition
With two fair coins, "first is heads" and "second is heads" are independent. Keeping them independent given "exactly one head" fails: each then has probability 21, but both together have probability 0, not 41.
This idea appears frequently in real systems: hidden variables, background conditions, or common causes can create apparent dependence that disappears once the underlying factor is taken into account. Conditional independence plays a central role in probabilistic modeling, graphical models, and Bayesian reasoning.
Whether two events are independent is a statement about a particular setting. The same pair can be checked once with no extra information and again inside a condition C, and the two checks use different numbers: P(A), P(B) and P(A∩B) in the first, and the same probabilities given C in the second.
Independence belongs to a setting: a new condition can create it or destroy it.
Independence in Problem Solving
Recognizing independence can dramatically simplify probability problems. When events are independent, complex joint situations break into simpler pieces that can be handled separately.
Independence allows probability trees to collapse into repeated patterns, makes joint probabilities easier to compute, and reduces the number of cases that must be considered. Many models in practice rely on independence assumptions precisely because they make reasoning tractable.
At the same time, independence should never be assumed blindly. In problem solving, the key skill is not using independence, but justifying it — understanding why one event truly does not influence another in the given context.
Common Mistakes
Independence is often misused or misunderstood, leading to incorrect conclusions.
A frequent mistake is confusing disjoint events with independent ones. Disjoint events cannot occur together, while independent events can — and usually do.
Another error is assuming independence simply because events look unrelated. Shared causes, hidden constraints, or limited resources often introduce dependence even when it is not obvious.
Independence is also mistakenly treated as permanent. Events that are independent in one context may become dependent once additional information is introduced, and vice versa.
Carefully checking assumptions is essential, because incorrect independence assumptions can invalidate an entire probability argument.
The pitfalls above can be packaged with the underlying confusion and the correct view.
Mistake
Why it's wrong
Correct view
Confusing disjoint with independent
disjoint events cannot co-occur; independent events can and usually do
disjoint events with positive probability are always dependent
Assuming independence because events "look unrelated"
shared causes and hidden constraints often introduce hidden dependence
verify P(A ∩ B) = P(A) P(B), or justify causally
Treating independence as permanent
conditioning can both create and destroy independence
independence is context-dependent; always specify the conditioning set
Confusing P(A ∩ B) with P(A) + P(B)
addition is for disjoint unions, not co-occurrence
use the product P(A) · P(B), and only under independence
A 2 by 2 contingency table
Independence is a check on the table: the cell for A and B must equal the product of the row total and the column total, divided by the grand total. Disjoint events fail that check badly, because their shared cell is zero while the product is not, so disjoint and independent are opposite situations. Edit the cells and test the product rule on the contingency tables explorer.
The connections below place independence among the other core ideas.
Connections to Other Probability Concepts
Independence does not stand alone. It interacts directly with many of the central ideas in probability.
Conditional probability explains how probabilities change when information is known; independence describes when they do not change.
Total probability combines contributions from different cases and often relies on independence assumptions to simplify models.
Bayes' reasoning depends critically on understanding when events are independent or conditionally independent.
Random variables extend independence from events to numerical quantities.
Joint distributions reflect independence through their factorization structure.
Seeing these connections makes independence easier to recognize and prevents it from being treated as an isolated rule rather than a structural idea running through probability.
Independence at a Glance
The table below condenses independence into a single reference — its plain meaning, the joint and conditional formulas that express it, its distinction from disjointness, the patterns by which it fails, the role of conditional independence, why it matters in practice, and the pitfall most often confused with it.
Aspect
Statement
Note / example
Plain meaning
knowing one event occurred does not change the probability of the other
events can become independent (or dependent) once a third event is known
central to graphical models and Bayesian reasoning
Why it matters
factorizes joint distributions, collapses tree branches, reduces cases
justifying independence is the real work; using it is the reward
Most common pitfall
confusing disjoint with independent
disjoint events with positive probability are never independent
Independence FAQ
What's the difference between independent and disjoint events?
+
They are near-opposites. Disjoint events cannot happen together: P(A ∩ B) = 0. Independent events satisfy P(A ∩ B) = P(A)P(B), which is nonzero whenever both are possible. So two disjoint events with positive probability are strongly dependent — learning one occurred rules the other out completely. The similar-looking equations are what make the pair easy to confuse.Read more →
Does pairwise independence imply mutual independence?
+
No. Mutual independence demands the product rule for every sub-collection, not just the pairs — for three events that is four equations, and for n events it is 2ⁿ − n − 1. Toss two fair coins and take “first is heads”, “second is heads”, “both match”: every pair is independent, yet the triple product fails.Read more →
Does conditional independence imply independence?
+
No, and neither direction holds. Two events can be dependent in general yet independent once a shared cause is fixed — conditioning removes the link. They can also be independent outright yet become dependent inside a condition. Independence and conditional independence are separate claims, and establishing one says nothing about the other.Read more →
What does the symbol A ⊥ B mean?
+
A is independent of B — the whole defining equation compressed into one mark, borrowed from the perpendicularity symbol. Graphical-models texts double the stroke so that a single ⊥ keeps its geometric job, while elementary texts skip the symbol and write the equation out. Between random variables it asserts that the joint distribution factors.Read more →
What does i.i.d. mean?
+
Independent and identically distributed — one tag making two separate claims. Independent: the whole collection is mutually independent, not merely pairwise. Identically distributed: every variable follows the same distribution. Breaking either half breaks the abbreviation, and both together are what license results such as the law of large numbers.Read more →
Can an event be independent of itself?
+
Only if its probability is 0 or 1. Taking B=A, the product rule asks for P(A∩A)=P(A)2, and since A∩A=A this reads P(A)=P(A)2, whose only solutions are 0 and 1. In words, learning that A occurred makes A certain, so the information changes nothing only when A was already certain or impossible.Read more →
Can two events from the same roll of a die be independent?
+
Yes. Coming from one roll is a reason to check, not a verdict. On a fair die, “even” has probability 21 and “at most 2” has 31; both happen only on a 2, with probability 61=21⋅31, so they are independent. Change the second event to “at most 3” and the product becomes 41, while both still happen only on a 2, so the events are dependent.Read more →
