Definiteness Effects and Specificity

Two facts about Icelandic that learners normally meet in completely different chapters turn out to be the same fact. The first: you can say Það er köttur í garðinum "there is a cat in the garden" but not \Það er kötturinn í garðinum "there is the cat in the garden" — the existential *resists definite subjects. The second: a relative clause can sit in the subjunctive (maður sem kunni íslensku "a man who would know Icelandic") or the indicative (maðurinn sem kann íslensku "the man who knows Icelandic"), and the choice tracks something about the head noun. The hidden link is a single semantic dimension — specificity: does the noun phrase pick out a particular, already-real referent, or merely describe a type that may or may not be instantiated? That one question drives both the existential restriction and the relative-clause mood. This page unifies them. (For the mechanics of existentials see syntax/existential-sentences; for the relative-clause mood itself, complex/subjunctive-vs-indicative-relative; for the article, nouns/definite-vs-indefinite. Here we tie the threads together.)

The existential definiteness restriction

An existential sentence asserts that something exists — that the world contains an instance of some description. Its job is to introduce a new referent. That job is incompatible with a definite noun phrase, because a definite phrase signals that the referent is already known, already on the table. You cannot introduce as new something you have flagged as old. So Icelandic existentials, like English ones, admit indefinite subjects and reject definite ones — the definiteness restriction.

Now try a definite. kötturinn "the cat," Jón, kennarinn "the teacher," þessi bók "this book" — all are barred from the existential subject slot, for the same reason English bars "*there is the cat": you cannot existentially introduce something already presupposed to exist.

The pattern is identical to the one governing the transitive expletive construction: there too the low, post-verbal subject must be indefinite. That is no coincidence — the existential restriction and the TEC restriction are the same restriction, because both involve a referent occupying a "presented/introduced" position, and only indefinites can be presented.

Definite early, indefinite late: how definiteness orders the clause

The restriction is one face of a broader principle that organises Icelandic word order: definite (given, known) material gravitates early, indefinite (new) material gravitates late. This is the discourse logic of information structure — you anchor on what is already shared, then add what is new. A definite subject sits up front as the ordinary subject; a brand-new indefinite is held back, often pushed into the post-verbal "presented" position with the help of expletive það. So the choice between Kötturinn er í garðinum and Það er köttur í garðinum is not random style — it reflects whether the cat is old news (definite, early) or fresh news (indefinite, late, presented by the existential).

This early/late split is why the definiteness restriction feels so natural once you see it: the existential is the construction for late, new referents, and definites are early, old referents, so they simply never meet. The construction and the restriction are two views of one ordering principle.

Specific vs non-specific: the second face of the same notion

Definiteness is about whether the referent is known; specificity is the closely related question of whether the speaker has a particular referent in mind at all. The two usually align (definites are specific, existential indefinites are non-specific), but indefinites can go either way, and Icelandic makes the difference visible — most sharply in the mood of a relative clause.

Consider Ég leita að manni sem talar fimm tungumál "I'm looking for a man who speaks five languages." This is ambiguous in English but Icelandic can disambiguate. On the specific reading there is a particular such man — you know who he is, you are looking for him. On the non-specific reading you want any man satisfying the description, and no particular individual is intended (perhaps none exists yet). The specific reading takes the indicative in the relative clause (the man and his language-skill are real); the non-specific reading licenses the subjunctive (the man is hypothetical, a description not yet anchored to a real individual).

This is the crux: a non-specific antecedent licenses the subjunctive in its relative clause; a specific antecedent takes the indicative. The subjunctive is doing exactly what it does everywhere — marking something as not-yet-real, hypothetical, sought rather than possessed. A non-specific "a man who speaks five languages" is a wanted description, not a real individual, so the relative clause that describes him is subjunctive. (The full treatment is on complex/subjunctive-vs-indicative-relative; the point here is why the mood tracks the noun.)

The unification: one notion, two phenomena

Now stand back and see the single mechanism. Both phenomena answer the same question — does the referent (already) exist as a particular thing?

• In the existential, the answer must be "not yet / newly" for the construction to apply at all: existentials introduce non-presupposed referents, so they demand indefinite (non-presupposed) subjects. A definite — a presupposed, already-existing referent — is excluded.
• In the relative clause, the same question selects the mood: a particular, existing referent makes the relative clause a statement of fact (indicative); a non-particular, merely-described, possibly-not-yet-existing referent makes it hypothetical (subjunctive).

So one semantic dimension — call it specificity/existence of the referent — surfaces twice: as a restriction on a subject position (existentials/TEC) and as a choice of verbal mood (relatives). They look like unrelated rules in two different chapters, but they are governed by the same underlying fact about the noun phrase. This is the unifying insight, and it has real predictive power: when you see a subjunctive relative clause, you can infer the antecedent is non-specific, and when you see an existential, you can infer its subject is non-presupposed. Two syntactic symptoms, one semantic cause.

The last pair is the clincher: the only difference between them is whether the medicine's existence is asserted (statement) or questioned (interrogative), and that single difference flips the relative-clause mood from indicative to subjunctive — because it flips the antecedent from specific to non-specific. Specificity is doing the work.

Generic bare plurals: a third symptom of the same logic

The same machinery explains why Icelandic uses bare (article-less) plurals for generic statements — claims about a kind, not about particular individuals. Hundar gelta "dogs bark" is about dogs in general; there is no particular pack in mind, so no definite article and no existential introduction — just a bare plural making a kind-level statement. This is, again, the non-specific end of the scale: a generic refers to a type, not to specific, presupposed individuals, so it takes neither the definite article (which would specify) nor the existential frame (which would introduce particulars).

Notice the contrast in the last pair: Hundar gelta (bare, generic, kind-level) versus Hundarnir gelta (definite, particular, episodic). The same noun moves along the specificity scale as you add or omit the article, exactly as it moved along it in the existential and relative-clause cases. Three constructions — existential subjects, relative-clause mood, bare-plural generics — all read the same semantic scale.

Why this is hard for English speakers

English shares the existential definiteness restriction (you also can't say "*there is the cat"), so that part transfers. Two things do not transfer. First, English does not mark specificity with mood: "a secretary who knows German" is the same string whether you mean a particular person or anyone fitting the bill, and the verb is "knows" either way. So English speakers systematically use the indicative in Icelandic non-specific relatives where the subjunctive is needed — *ritari sem kann for a wanted (non-specific) secretary, instead of ritari sem kunni. Second, English speakers, having no mood cue, fail to hear the specificity distinction at all, and so miss the unifying logic — they learn "definite can't go in existentials" and "some relatives are subjunctive" as two disconnected rules, never seeing that both answer the question does this referent particularly exist?

Common Mistakes

A definite referent is already known and cannot be existentially introduced. Indefinite for the existential, ordinary subject slot for the definite.

Non-specific antecedents license the subjunctive. English's invariant "knows" tempts the indicative, but a sought referent is hypothetical.

Specificity cuts both ways: a real, particular referent must take the indicative. Don't over-apply the subjunctive.

Generics are non-specific (about a kind), so they take a bare plural, not the definite article.

Key Takeaways

• One semantic notion — does the referent (already) exist as a particular thing? — drives several Icelandic phenomena that look unrelated.
• Existential definiteness restriction: Það er/eru admits only indefinite subjects (Það er köttur), because existentials introduce new referents; a definite is already known and takes the ordinary subject slot (Kötturinn er...). The same restriction governs the post-verbal subject of the transitive-expletive construction.
• Definite early, indefinite late: given/known material anchors the front of the clause; new/indefinite material is held back, often presented post-verbally with expletive það.
• Specificity → relative mood: a specific (real, particular) antecedent takes the indicative (mann sem talar); a non-specific (sought, merely-described, possibly-unreal) antecedent licenses the subjunctive (mann sem tali). Asserting vs questioning a referent's existence flips the mood (lyf sem læknar vs lyf sem lækni).
• Generic bare plurals (Hundar gelta "dogs bark") sit at the non-specific, kind-level end of the same scale — no article (which would specify), no existential frame (which would introduce particulars). Adding the article makes it particular: Hundarnir gelta.
• The unification: the existential restriction, the relative-clause mood, and bare-plural generics are three symptoms of one underlying distinction (specific/existing vs non-specific/non-presupposed). English shares the existential restriction but does not mark specificity with mood, so English speakers wrongly use the indicative in non-specific relatives and miss that the phenomena are one.