Embedded V2 and Bridge Verbs

You learned the V2 rule as a fact about main clauses: the finite verb sits second, and fronting anything but the subject forces inversion. You may also have learned, from subordinate clause order, that subordinate clauses are different — the subject stays first, ekki sits before the verb, and inversion is suppressed. Both are true as baselines. But Icelandic has a property that sets it apart from almost every other Germanic language: V2 can hold inside a subordinate clause too. After the right kind of main verb, you can topicalize inside an að-clause and trigger inversion, exactly as in a main clause: Ég veit að á morgun fer Jón "I know that tomorrow Jón is leaving," with the subordinate verb fer sitting second, before its subject Jón. This is embedded V2, and Icelandic's freedom with it is why linguists call it a symmetric V2 language — V2 in both main and embedded clauses. English and Mainland Scandinavian (Swedish, Danish, Norwegian) cannot do this freely. This page is about when embedded V2 is allowed, the verbs that license it, and the deep link between word order, mood, and assertion.

The baseline: subordinate clauses are usually subject-first

Begin with the ordinary subordinate clause, the one most textbooks describe. After að "that," the neutral order is subject – finite verb, and the sentence adverbs (ekki, oft, alltaf) sit before the finite verb, not after it. There is no inversion; fronting is suppressed. This is the "asymmetric" behaviour shared by all Germanic languages — embedded clauses look syntactically quieter than main clauses.

Against this baseline, embedded V2 is the marked, attention-drawing option — and the striking thing is that Icelandic allows it at all.

Embedded V2: topicalize and invert inside the að-clause

Now the phenomenon. After certain main verbs, you can take a subordinate að-clause and front a non-subject inside it — a time adverb, an object, a place — and when you do, the embedded finite verb jumps to second position within the clause, inverting with the embedded subject. The að stays where it is; everything after it behaves like a miniature main clause.

Look at Ég veit að á morgun fer Jón. Everything after að is a perfect main-clause V2 string: fronted adverb á morgun, finite verb fer, then subject Jón. The subordinate clause has, in effect, its own prefield that something can move into, forcing the finite verb to second position behind it. That is exactly what does not happen in English or in Mainland Scandinavian, where fronting inside a that/att-clause is impossible — the embedded clause has no available prefield.

Bridge verbs: who licenses embedded V2

Embedded V2 is not allowed under every main verb. It is licensed by a class called bridge verbs — verbs of saying and thinking whose complement clause is an assertion: something the speaker (or the subject) puts forward as true. The core members are:

These verbs all introduce a complement that is asserted — a piece of information offered as holding. Their að-clauses are "windows" through which a main-clause structure is visible; hence "bridge." By contrast, non-bridge contexts — factive verbs like sjá eftir "regret," vera ánægður með "be glad that," and especially clauses that are not asserted (questions, complements of efast um "doubt," relative clauses, adverbial clauses with þegar, af því að) — generally resist embedded V2. You cannot freely topicalize inside them.

The three-way correlation: bridge verb + V2 + indicative

Here is the insight that ties the page together and that competitors never connect. Embedded V2 does not travel alone. It correlates with the indicative, and the indicative in turn signals assertion — the speaker's commitment to the embedded proposition. The three line up:

• A bridge verb introduces an asserted complement.
• An asserted complement is indicative (the embedded proposition is presented as fact).
• An indicative, asserted complement is exactly the environment that licenses embedded V2.

So when you see embedded V2, you are looking at an asserted, indicative complement under an assertive bridge verb — three properties moving together. The flip side is just as revealing: a subjunctive complement marks the proposition as not asserted (reported, doubted, hypothetical), and subjunctive complements resist embedded V2. Mood and word order are reading the same thing — whether the embedded clause is asserted — off two different signals.

The practical reading of this correlation: embedded V2 is a commitment marker. By using it, the speaker signals that the embedded clause is something they are asserting, not merely relaying or entertaining. A reported but non-committal complement (subjunctive, under segja in pure reportative use) keeps the quiet subordinate order; an asserted complement can light up with V2. So the choice of subordinate word order interacts with mood and with how much the speaker stands behind the embedded proposition — a genuinely three-way correlation among bridge verb, V2, and indicative.

Why English and Mainland Scandinavian can't do this

The contrast is sharp and worth stating plainly. English has no embedded V2 at all: you can say "I know that Jón is leaving tomorrow," but never "*I know that tomorrow is Jón leaving" or "*He said that this book had María read." English fronting ("This book, María has read") works only in main clauses; inside that, the subject must stay first. Mainland Scandinavian (Swedish, Danish, Norwegian) is the same — these languages have main-clause V2 but asymmetric V2: they lost embedded V2, and inside att/at the order is subject-first with adverbs before the verb. Icelandic (with Yiddish) is the standout that kept symmetric V2, allowing the embedded clause to behave like a main clause under bridge verbs. So an English or Swedish speaker's instinct — "you can't reorder inside that" — is correct for their language and exactly wrong for Icelandic. The fix is to recognise that an Icelandic að-clause under a bridge verb has its own prefield and can be topicalized and inverted just like a main clause.

Common Mistakes

If you topicalize inside the að-clause, V2 applies inside it too — the finite verb inverts with the embedded subject, exactly as in a main clause.

Don't import the English/Mainland-Scandinavian ban on embedded reordering. Bridge verbs license embedded V2 in Icelandic.

Embedded V2 needs an assertive (bridge) context. Under doubt, factives, and questions, the proposition isn't asserted, so V2 is blocked.

Inversion changes order, not agreement. The inverted finite verb still agrees with the embedded subject.

Key Takeaways

• Icelandic is a symmetric-V2 language: the verb-second rule holds inside subordinate að-clauses too, so you can topicalize and invert within them — Ég veit að á morgun fer Jón, Hann sagði að þessa bók hefði María lesið.
• The baseline subordinate clause is subject-first with adverbs before the verb; embedded V2 is the marked option, where the að-clause gains its own prefield and behaves like a main clause.
• Embedded V2 is licensed by assertive bridge verbs of saying/thinking (segja, halda, telja, vita, vona, fullyrða) and blocked by factives (sjá eftir) and non-asserted contexts (doubt, questions, adverbial clauses).
• The three-way correlation: bridge verb + embedded V2 + indicative all move together, because all three track assertion — the speaker's commitment to the embedded proposition. Subjunctive (non-asserted) complements resist embedded V2.
• English and Mainland Scandinavian have asymmetric V2 — no embedded V2 at all — so the instinct "you can't reorder inside that" is correct for them and wrong for Icelandic.
• Inversion under embedded V2 changes order, not agreement: the finite verb still agrees with its own (embedded) subject.