eduKateSG · Mathematical Logic · Worked learning guide
Find the reasoning step you need
Separate no solutions, one solution and several solutions. Learn how identity expresses exact counts and why proving uniqueness does not prove existence.
1. Separate at least one, at most one and exactly one
3. Read identity as sameness, not resemblance
5. Express exactly two without counting names twice
7. Distinguish one result per input from a shared result
9. Practise with domain and identity checks
Full chapter index · Practice and checking · Mathematical Logic subject guide
A problem may have no solution, one solution or several solutions. “There is at most one” does not decide between the first two possibilities. “There is at least one” does not decide between the last two. The gold standard of mathematical logic keeps these promises separate.
This guide develops existence, uniqueness and identity through original equations, finite interpretations and proof repairs. It is written for learners who can read simple quantifiers and want to understand what an answer such as “the solution” commits them to.
The setting is classical first-order logic with identity and a nonempty domain. Names denote objects, two names may denote the same object, and identity means being the very same object. Start with Mathematical Logic for the broader map.
Chapter index
Chapters 1–4
Chapters 5–8
Let P(x) mean that x solves the stated problem. Existence is ∃x P(x): at least one object in the domain has the property. It allows two, ten or infinitely many such objects.
“At most one” can be expressed as ∀x∀y((P(x) ∧ P(y)) → x = y). Whenever two objects both have the property, they must be identical. This holds when exactly one object qualifies, but it also holds when none qualifies.
Exactly one combines the two demands. One convenient formula is ∃x(P(x) ∧ ∀y(P(y) → y = x)). A witness exists, and every object with the property is that witness.
The notation ∃!x P(x) is often used as an abbreviation for unique existence. The exclamation mark adds a specific logical claim; it is not emphasis or enthusiasm.
The three-way diagnostic
First ask whether any object qualifies. If none does, existence is false and “at most one” is true. If one qualifies, both are true. If two distinct objects qualify, existence is true and “at most one” is false.
A student can therefore prove the uniqueness condition correctly without having solved the original existence question. The proof has completed one task, and the remaining task must still be addressed.
This matters whenever a phrase such as “the answer” appears before a witness has been supplied. Ordinary language often assumes the relevant object exists. Formal reasoning should either establish that assumption or state it explicitly.
A good final result says which claim has been proved: existence, at-most-one, or exactly-one. Avoid compressing all three into the vague statement that the answer is unique.
Consider x² = 4. Over the real numbers, x = 2 and x = −2 both solve the equation. There is at least one solution, but there is not at most one.
Over the nonnegative real numbers, only x = 2 qualifies. Exactly one solution exists. Over the positive integers, the same unique solution remains. Over {0,1}, no solution exists.
The written equation stayed the same. The domain changed which objects were available. Existence and uniqueness belong to the equation together with its domain, rather than to the symbols alone.
Now consider x² = −1. There are no real solutions, so its at-most-one condition is true over the reals. Over the complex numbers there are two distinct solutions, i and −i, so at-most-one is false.
Why an empty solution set passes at-most-one
To violate at-most-one, find two distinct solutions. With no solutions, no such pair exists. The conditional that defines at-most-one has no true antecedent pair capable of making it false.
This is not a claim that an absent solution is secretly a unique existing object. It is the difference between a prohibition on multiple solutions and a promise of one solution.
A practical correction is to write two lines: “There cannot be two distinct solutions” and “Here is a solution.” The first controls multiplicity; the second supplies existence. Together they justify exactly one.
For an equation-solving task, also substitute the proposed value into the original equation and check membership in the domain. A value obtained after squaring, cancelling or extending the domain may fail one of these checks.
Identity x = y means that x and y denote the same object. Sharing a colour, size, label, mark or numerical measurement does not automatically establish identity of the objects being measured.
Two library copies may have the same title while remaining distinct physical books. Two records may display the same name while referring to different people. Conversely, two different names may refer to the same person.
In standard first-order semantics, different constant symbols are not automatically assigned different objects. If a and b occur in a formula, an interpretation may assign both to one object unless a ≠ b is required.
This affects arguments about counting. From P(a) and P(b), it does not follow that at least two objects have P. Add a ≠ b to establish distinct witnesses.
Substitution has a defined setting
In ordinary first-order logic with identity, identical objects can be substituted in extensional predicates without changing the truth of the predication. If a = b and P(a), then P(b).
But “has the same examination score as” is generally an equivalence relation on people, not identity of people. Replacing one person with another inside every predicate would be invalid. The relation preserves equality of the selected measurement, not all properties.
Likewise, statements about what someone believes can depend on how an object is described. Such contexts are not automatically represented by a simple extensional predicate with unrestricted substitution of co-referring descriptions. This guide uses ordinary extensional first-order examples.
The teaching check is to finish the phrase “same ___.” Same object, same value and same property are different claims. Logic becomes clearer when that blank is filled before an equality symbol is introduced.
Suppose P(x) means x is a winning token in a finite game model. Write ∃x(P(x) ∧ ∀y(P(y) → y = x)). The first conjunct says the chosen x qualifies. The second says every qualifying y is that very x.
Test the formula in a domain {a,b,c}. If only b is winning, choose x = b. For y = a and y = c, P(y) is false, so the conditional holds. For y = b, identity holds. The formula is true.
If a and b are both winning and distinct, no choice works. Choosing x = a fails at y = b; choosing x = b fails at y = a; choosing x = c fails the requirement P(x). The formula is false.
If no token is winning, every proposed x fails the existence conjunct. The universal uniqueness condition cannot rescue the missing witness.
Avoid an overly strong formula
The expression ∃x∀y(P(y) ∧ y = x) says every object both qualifies and equals x. In a nonempty domain it requires the entire domain to contain only one object, which qualifies. That is far stronger than saying one object in a larger domain has P.
Similarly, ∃x∀y(P(y) ↔ y = x) is a correct alternative for exactly one P-object. Here the biconditional makes P true precisely at x. Its equivalence with the earlier formula can be checked by using y = x and the reflexivity of identity.
A student should be able to explain the role of each connective. The conditional excludes additional qualifying objects while allowing nonqualifying objects to exist. Replacing it with a conjunction accidentally changes the size and content of the whole domain.
“At least two objects have P” is ∃x∃y(P(x) ∧ P(y) ∧ x ≠ y). The inequality makes the witnesses distinct. Without it, the same object could satisfy both existential choices.
“At most two” can be stated as ∀x∀y∀z((P(x) ∧ P(y) ∧ P(z)) → (x = y ∨ x = z ∨ y = z)). Any three qualifying objects include a repeated object, so three distinct qualifying objects are impossible.
Exactly two combines at-least-two with at-most-two. Another compact form is ∃x∃y(x ≠ y ∧ ∀z(P(z) ↔ (z = x ∨ z = y))). It says the P-objects are precisely the two distinct witnesses.
A finite interpretation
Let the domain be {a,b,c}, with P true of a and c only. Choose x = a and y = c. For z = a or c, both sides of the biconditional are true. For z = b, both are false. Exactly two is true.
Now suppose the constants m and n both denote a. The statements P(m) and P(n) are still true, but they identify only one object. Counting written names would overcount.
This distinction matters in data work. Two rows are not automatically two people; two identifiers are not automatically two entities unless the identification rules establish that. Conversely, merging rows because their visible labels match may incorrectly combine distinct entities.
The formal lesson is about witness identity. An exact-count claim needs both enough distinct qualifying objects and a condition excluding further ones. Providing only a list can establish a lower bound, while providing only an exclusion can leave existence unresolved.
Consider 3x + 2 = 11 over the real numbers. For existence, propose x = 3 and substitute: 3×3 + 2 = 11. Thus at least one real solution exists.
For uniqueness, suppose u and v are both solutions. Then 3u + 2 = 11 and 3v + 2 = 11. Subtracting gives 3(u − v) = 0, hence u = v. There cannot be two distinct solutions.
The proof uses properties of real arithmetic, including cancellation by the nonzero coefficient 3. It is not a purely logical consequence of the equation’s visual shape in every possible algebraic structure.
A parameter reveals the condition
For ax = b over the reals, if a ≠ 0, the unique solution is b/a. If a = 0 and b ≠ 0, there are no solutions. If a = 0 and b = 0, every real x is a solution.
An argument that divides by a without a ≠ 0 can turn these three cases into one incorrect answer. The missing condition affects both existence and multiplicity.
A complete classification keeps the cases visible. For a ≠ 0, demonstrate the witness and prove uniqueness. For a = 0, inspect the original equation before performing any cancellation.
Preserve the original problem
If an equation is transformed through an implication rather than an equivalence, a root of the transformed equation may not solve the original. Squaring both sides is a familiar example. Candidate generation and solution verification are separate stages.
The gold standard is therefore more than an elegant rearrangement. The final candidate belongs to the stated domain, satisfies the original condition and is the only such candidate under the stated assumptions.
A relation F(x,y) can describe an output y for input x. The claim “every input has exactly one output” is ∀x∃y(F(x,y) ∧ ∀z(F(x,z) → z = y)).
This permits the output to depend on x. It does not say every input has the same output. Over the integers, y = x + 1 gives exactly one output for each input, and the outputs vary.
A shared-output claim would require one y to serve every x, with an appropriate uniqueness condition. It is a different specification.
Per-input uniqueness does not imply injectivity
A function can send two different inputs to the same output while still assigning exactly one output to each input. For f(x) = x² over the integers, inputs 2 and −2 both have output 4.
Injectivity requires that equal outputs come only from equal inputs. In relational notation, one formulation is ∀x∀x′∀y((F(x,y) ∧ F(x′,y)) → x = x′), alongside the relevant function conditions.
Totality asks whether every permitted input has an output. Single-valuedness asks whether each input has at most one. A partial function can be single-valued without being total.
These distinctions are useful when reading a specification. “Each registered application has one reference number” does not alone establish that different applications receive different numbers, that every possible application is registered, or that every number corresponds to an application.
The repair is to state the domain, relation and separate promises. Then test each promise with a small table. One output per row, no repeated output across rows and no missing row are distinct table properties. A single informal phrase should not be asked to carry all three.
| Argument gap | What has actually been shown | What remains |
|---|---|---|
| Any two solutions are equal | At most one solution | Supply an existing solution |
| A candidate satisfies the equation | At least one solution | Exclude distinct alternatives |
| Two names satisfy P | Possibly one or two P-objects | Establish the names denote distinct objects |
| Every input has one output | Per-input unique existence | Check injectivity separately |
| A rearranged equation has one root | A transformed problem is classified | Verify equivalence and the original condition |
A good repair preserves the valid part. If the at-most-one proof is sound, do not discard it because existence was omitted. Add the missing witness argument.
If existence is easy but uniqueness is false, correct the conclusion. For x² = 4 over the reals, listing both solutions is the right repair. Restricting the domain merely to force a unique answer would change the problem unless that restriction is justified.
If two candidate witnesses turn out to be identical, the appropriate count may decrease. The written appearance of two constants does not override the interpretation.
A short explanation test
Ask: “Can a problem with no solutions have at most one solution?” Yes. Then ask: “Can it have exactly one solution?” No. The learner should explain the existence requirement rather than rely on the sound of the word one.
Next ask: “If every input has exactly one output, must every output belong to exactly one input?” No. The direction of the relation has changed. A function’s definition and injectivity answer different questions.
These checks reveal whether the learner can move between formula, model and ordinary language. They are more informative than recognising the symbol ∃! in isolation.
Exercise 1: classify the solution set
For x² = 9, classify existence and uniqueness over the reals, nonnegative reals and the domain {0,1,2}.
Over the reals, existence is true and at-most-one is false because ±3 qualify. Over nonnegative reals, exactly one solution exists: 3. Over {0,1,2}, existence is false and at-most-one is true.
Exercise 2: two names
Assume P(a), P(b) and a = b. Does at least two P-objects follow? No. A one-object interpretation with both names denoting the same qualifying object satisfies the premises.
Exercise 3: at most one without existence
Let P(x) mean x² + 1 = 0 over the reals. Show at-most-one holds and exactly-one fails. There are no qualifying real objects, so no pair of distinct qualifying objects exists; no existential witness exists either.
Exercise 4: a relational table
Inputs are {1,2,3}. The relation contains (1,A), (2,A) and (3,B), with no other pairs. Every input has exactly one output. Injectivity fails because 1 and 2 share A. The range contains A and B.
Transfer check
Write a claim involving “the,” “only one” or “unique” from a mathematical problem. Split it into existence and at-most-one. Identify whether names or labels could refer to the same object, and state the domain.
Then test a no-object case and a two-object case. If the proposed formula accepts both while claiming exactly one, it is missing a condition. If it rejects a larger domain containing one qualifying object and several nonqualifying ones, it may have accidentally restricted the entire domain.
Frequently asked questions
Does unique mean at most one or exactly one? Usage can vary in informal discussion. State the intended claim explicitly. A unique existing solution requires both existence and at-most-one.
Does a ≠ b follow because the letters differ? No. Standard first-order interpretations allow different names to denote the same object.
Is equality the same as having the same properties in a selected list? No. Agreement on selected properties can hold between distinct objects. Identity is sameness of object.
Can unique existence change when the domain changes? Yes. A solution excluded from one domain may be included in another, and additional objects can supply additional solutions.
Connected reading
Return to Mathematical Logic for interpretations and proof rules. Use Logic for necessary and sufficient conditions, and Discrete Mathematics for relations and finite structures.
Sources and scope
Identity and counting notation were checked against forall x: Calgary, Identity. The domain and interpretation conventions are supported by the Open Logic Project syntax and semantics reference. Accessed 22 October 2026. Equations, models and repairs here are original instructional examples. Applications involving records illustrate logical distinctions and do not establish a particular institution’s identification rules.
Contents · Previous chapter · Continue to Mathematical Logic
Connected Mathematical Logic guides
Quantifier order and dependent choices