Consider the sentence Vulcan does not exist. It is true. The planet that nineteenth-century astronomers posited inside Mercury's orbit was never there. Yet classical first-order logic cannot straightforwardly say so. Standard semantics requires every name to be assigned an object in the domain, and from that requirement it follows, as a theorem, that every named thing exists.
This is not a minor bookkeeping issue. The inference from Vulcan is a planet to something is a planet is classically valid. So is the inference from everything is self-identical to Vulcan is self-identical, and from there to something is Vulcan. Logic is supposed to be topic-neutral, yet it acquires existential commitments the moment we introduce a singular term.
Free logic is the family of systems that removes this commitment. It is logic 'free' of existential presuppositions about its terms. What follows examines why such systems are needed, and how two principal variants, negative and positive, distribute truth across sentences containing empty terms. Each choice, as we will see, encodes a substantive metaphysical position.
Why Classical Logic Overcommits
In classical predicate logic, existential generalisation (from Fa infer ∃xFx) and universal instantiation (from ∀xFx infer Fa) hold without restriction. Combined with the reflexivity of identity, they yield ∃x(x = a) for any term a whatever. Read as a claim, this says that whatever we can name exists. The theorem is not discovered about the world. It is an artefact of the requirement that every constant be assigned a referent.
Three sources of pressure show the cost. First, empty names: 'Vulcan', 'Sherlock Holmes' and 'Zeus' are grammatically names, and we reason with them competently. Second, definite descriptions: 'the largest prime' and 'the present king of France' look like singular terms, but nothing satisfies them. Russell's response was to deny that descriptions are terms at all and to paraphrase them into quantified sentences. That is a costly revision of surface form, and free logic lets us avoid it.
Third, and most consequential for metaphysics, are possible objects. In Kripke's varying-domain semantics for quantified modal logic, an object may exist at one world and not at another. If a name rigidly designates an individual across worlds, then at worlds where that individual is absent the name is empty relative to that world's domain. Applying classical rules there validates the converse Barcan formula and the necessary existence of whatever is named. Contingentists reject precisely these results.
Free logic's core modification is modest. It introduces an existence predicate, E!, typically definable as ∃x(x = a), and restricts the quantifier rules. From ∀xFx and E!a, infer Fa. From Fa and E!a, infer ∃xFx. The quantifiers still range over existents. What changes is that terms are no longer guaranteed to land within that range. The rest of the system depends on how we evaluate atomic sentences whose terms land nowhere.
TakeawayClassical logic's existential commitments are not facts about the world but artefacts of its semantics. Free logic turns the question of what exists into something logic leaves open rather than settles by fiat.
Negative Free Logic: No Subject, No Truth
Negative free logic, developed by Tyler Burge and Rolf Schock among others, adopts a simple principle. An atomic sentence is true only if all its terms denote and their denotations stand in the relevant relation. If any term is empty, the atomic sentence is false. 'Vulcan is a planet', 'Vulcan orbits the Sun' and even 'Vulcan = Vulcan' all come out false.
The view has an attractive rationale. Atomic predication naturally reports that some object instantiates a property or relation. Where there is no object, there is nothing to do the instantiating. One notable structural consequence is that E!a becomes equivalent to a = a, so self-identity amounts to existence. The reflexivity of identity holds only for terms that denote. Sentential negations of atomic sentences with empty terms come out true: 'It is not the case that Vulcan is a planet.'
The puzzles begin with negative existentials. 'Vulcan does not exist' is true when parsed as ¬E!a, the negation of an atomic sentence. But suppose we use lambda abstraction to form the complex predicate λx¬E!x, 'is non-existent', and apply it to Vulcan. That is now an atomic predication with an empty term, so negative semantics makes it false. Lambda conversion, the equivalence between applying an abstracted predicate and asserting the corresponding open sentence, therefore fails.
The upshot is a sharp scope distinction. Sentence negation and predicate negation diverge: 'Vulcan is not a planet' is true, but 'Vulcan is a non-planet' is false. Negative free logic thereby enforces a Russellian thesis that non-existence cannot be a property anything has, since there is nothing to have it. Whether this is a discovery or a stipulation is open to dispute. Intuitive truths such as 'Holmes is a detective' also come out false, so fictional discourse must be handled by operators like 'according to the story'.
TakeawayNegative free logic treats predication as requiring a subject. Non-existence can then be asserted of a name but never instantiated by anything, which is a logical expression of the view that being is prior to being-so.
Positive Free Logic and the Meinongian Temptation
Positive free logic allows some atomic sentences with empty terms to be true. Its minimal commitment, central to Karel Lambert's work, is that a = a holds for every term, whether or not it denotes. Identity is reflexive without restriction. What fails is the inference from self-identity to existence. 'Pegasus is Pegasus' is true, while 'something is Pegasus' is false.
The most perspicuous semantics uses two domains. An interpretation supplies an inner domain, over which the quantifiers range, and an outer domain that contains it. Every term denotes something in the outer domain, and E! picks out the inner domain. Predicates receive extensions over the outer domain. 'Pegasus is winged' can therefore be true, because the outer-domain object assigned to 'Pegasus' falls within the extension of 'winged'.
Read realistically, this is close to Meinong's theory of objects. On that theory there are items that lack being yet genuinely have properties, so the golden mountain is golden and the round square is round. Russell objected that unrestricted characterisation yields an existent golden mountain that does not exist. Neo-Meinongians respond by restricting characterisation. Terence Parsons distinguishes nuclear from extranuclear properties, and Edward Zalta distinguishes encoding a property from exemplifying it.
Yet the formalism does not compel the metaphysics. Lambert himself treated the outer domain as a semantic device rather than an ontological inventory. Positive free logic can also be given semantics that never mention non-existent objects. The dual-domain model shows that Meinongian talk is consistent; it does not show that it is true. The decisive question is whether 'there are things that do not exist' is a coherent use of 'there are'. Quine famously answered that it is not.
TakeawayA semantics that assigns objects to empty names shows that talk of non-existents can be made consistent. Whether those objects belong to reality or are mere bookkeeping is a further question that no model theory can settle.
Free logic isolates a question that classical logic suppresses: whether reference and existence can come apart. Its minimal reform, restricting the quantifier rules by means of an existence predicate, is common ground. The variants diverge over how atomic truth behaves when reference fails.
Negative free logic ties predication to existence and pays for it with failures of lambda conversion and a strict separation between sentential and predicate negation. Positive free logic preserves unrestricted identity and permits truths about the non-existent, inviting a Meinongian reading without strictly requiring one. Neutral free logics, which leave such sentences without a truth value, offer a third route.
The choice among them is not purely technical. To decide which sentences with empty terms are true is to decide what predication requires, and that is a question about the structure of reality itself.