At B1 you learned the subjunctive as a set of triggers: reported speech, conditionals, wishes, certain conjunctions. That gets you a long way, but it eventually fails, because it treats mood as a reflex of the conjunction — "see að, write a subjunctive." Real Icelandic does not work that way. The same að-clause can take the indicative or the subjunctive, and the choice changes the meaning. This page replaces the trigger list with a single organising principle that predicts mood even in sentences no list anticipated. (For the forms themselves see verbs/subjunctive-forms; for the basic triggers see verbs/subjunctive-overview. This page is about selection — which mood, and why.)
The one principle: assertion versus non-assertion
Strip away the rules and one idea remains: the indicative asserts; the subjunctive does not. The indicative presents a proposition as something the speaker commits to as fact — vouches for, takes responsibility for. The subjunctive presents a proposition as not asserted by the speaker: it is reported as someone else's, or hypothesised, or desired, or doubted, or merely entertained as one possibility among others. Everything else follows from this.
Notice what this is not. It is not "subjunctive = unreal" — plenty of true things take the subjunctive (when reported). It is not "subjunctive = after að" — að is mood-neutral. The variable is always the same: is the speaker putting their own weight behind this proposition, or holding it at arm's length?
The present subjunctive forms (komi, sé, hafi, fari, viti) handle open, still-live non-assertions; the past subjunctive (kæmi, væri, hefði, færi, vissi) handles the remoter, backshifted, or counterfactual ones. But the choice between subjunctive and indicative — the thing this page is about — is governed entirely by assertion versus non-assertion, independent of tense.
Factive verbs assert; non-factive verbs report
The clearest demonstration is the contrast between factive and non-factive predicates, because the same surface frame (verb + að + clause) flips mood depending on the verb's meaning.
A factive predicate presupposes that its complement is true — vita "know", muna "remember", gleyma "forget that", vera ljóst "be clear". You cannot know something false, so knowing it presupposes the fact, and the indicative is natural: you are committed to its truth.
Ég veit að hún kom í gær.
I know that she came yesterday. 'vita' is factive — her coming is presupposed true — so the indicative 'kom' is natural.
Ég man að við vorum þarna saman.
I remember that we were there together. 'muna' presupposes the truth of its complement → indicative 'vorum'.
A non-factive predicate of belief or saying — halda "think", telja "reckon", segja "say", fullyrða "assert" — makes no commitment to the truth of its complement. You can think something false. So the speaker is reporting a belief, not vouching for a fact, and the subjunctive is the norm.
Hann heldur að hún hafi komið í gær.
He thinks she came yesterday. 'halda' is non-factive — a reported belief — so the subjunctive 'hafi komið'.
Þau telja að þetta sé misskilningur.
They reckon this is a misunderstanding. 'telja' reports an opinion → subjunctive 'sé'.
Lay the minimal pair side by side and the principle is undeniable: ég veit að hún kom (I vouch for it — indicative) versus hann heldur að hún hafi komið (he merely believes it — subjunctive). The mood tracks commitment, and the verb's factivity sets the default commitment.
The genuine alternation: mood as a commitment marker
Here is where the trigger-list model collapses and the principle earns its keep. After many verbs the mood is not fixed — both moods occur, and the difference is the speaker's stance. This is the most sophisticated point about Icelandic mood, and it is real, not a textbook simplification.
Take segja "say". The default is the subjunctive — neutral reporting, no stance. But a speaker can switch to the indicative to signal that they personally endorse the reported content as fact:
Hann segir að hún sé veik.
He says she's ill. Subjunctive 'sé' — neutral relay; I take no stance on whether it's true.
Hann segir að hún er veik, og það er rétt — ég talaði við hana sjálfur.
He says she's ill, and it's true — I spoke to her myself. Indicative 'er' — I'm now vouching for it as fact.
The same machinery explains why vita "know" resists the subjunctive: knowing is commitment, so the indicative is its natural partner, and a subjunctive after vita would be self-contradictory (claiming to know something while declining to assert it). And it explains the scale in between: halda (mere belief, low commitment → subjunctive) … segja (reporting, neutral → subjunctive default, indicative if endorsed) … vita (knowledge, full commitment → indicative). Mood is functioning as a quiet evidential / commitment device — it tells the listener how much you stand behind the embedded claim.
Non-specific relative clauses: the subjunctive of "whoever"
The same non-assertion logic reaches into relative clauses, a context the basic trigger list never mentions. When a relative clause modifies a non-specific or hypothetical head — "whoever comes", "anyone who can", "a book that would help" (not yet identified) — the verb goes subjunctive, because you are not asserting that such a person/thing exists or did anything; you are describing a type, openly.
Sá sem komi fyrstur fær verðlaunin.
Whoever comes first gets the prize. Non-specific head 'sá sem' — no particular person is asserted to come → subjunctive 'komi'.
Við leitum að einhverjum sem kunni íslensku.
We're looking for someone who knows Icelandic. The someone is hypothetical, not yet identified → subjunctive 'kunni'.
Contrast a specific, real referent, where the indicative returns because you are asserting something about an actual individual:
Þetta er maðurinn sem kann íslensku.
This is the man who knows Icelandic. A specific, real man → indicative 'kann'.
The pair einhver sem kunni (hypothetical → subjunctive) versus maðurinn sem kann (actual → indicative) is the relative-clause echo of halda að … sé versus vita að … er. One principle, two constructions. (This deserves its own treatment — see complex/subjunctive-vs-indicative-relative.)
Why a trigger list eventually misleads
A mechanical rule like "after að use the subjunctive" is a useful crutch at B1 and a liability at B2, because it gets the factive cases wrong (it would wrongly subjunctivise ég veit að hún kom) and it cannot handle the alternations at all (it has no way to express the segir … sé / segir … er contrast). Worse, it teaches you to look at the conjunction when you should be looking at your own commitment. The conjunction að is mood-neutral; it is the predicate above it and your stance toward the content that select the mood.
This matters for English speakers specifically, because English gives you almost no help. English collapses "I know that she came", "he thinks that she came", and "whoever comes first" into a single indicative across the board. There is no audible mood difference to transfer. So you have to build the assertion/non-assertion sensor from scratch — but once built, it is a single sensor that handles reported speech, belief, knowledge, wishes, relatives, and conditionals alike.
Common Mistakes
❌ Ég veit að hún hafi komið.
Over-applied subjunctive — 'vita' is factive and asserts the fact; it takes the indicative: 'Ég veit að hún kom/er komin'.
✅ Ég veit að hún er komin.
I know she has arrived. Factive 'vita' → indicative.
This is the classic over-correction once a learner discovers the subjunctive: applying it after every að. Knowing presupposes truth, so the indicative is required.
❌ Hann heldur að hún kom í gær.
Under-applied subjunctive — 'halda' is non-factive (a reported belief), so the complement is the subjunctive 'hafi komið', not the indicative 'kom'.
✅ Hann heldur að hún hafi komið í gær.
He thinks she came yesterday. Non-factive belief → subjunctive.
❌ Sá sem kemur fyrstur fær verðlaunin (meaning 'whoever comes first', no one identified).
For the open, non-specific 'whoever', the relative verb is subjunctive: 'Sá sem komi fyrstur'. The indicative would assert a specific person comes.
✅ Sá sem komi fyrstur fær verðlaunin.
Whoever comes first gets the prize. Non-specific head → subjunctive 'komi'.
❌ Treating 'segir að … er' as always wrong.
Not an error — 'Hann segir að hún er veik' is grammatical and meaningful: the indicative signals the speaker endorses the claim. The subjunctive is merely the neutral default.
✅ Hann segir að hún sé veik. / Hann segir að hún er veik.
He says she's ill (neutral report / + I vouch for it). Both occur; the mood marks your stance.
The mistake here is believing the rule is absolute. After segja, both moods are available; choosing between them is choosing how much you commit.
Key Takeaways
- One principle replaces the trigger list: the indicative asserts (speaker commits); the subjunctive does not (reported, hypothetical, desired, doubted, non-specific). The conjunction að is mood-neutral.
- Factive predicates (vita, muna) presuppose truth → indicative; non-factive predicates of belief/saying (halda, telja, segja) report → subjunctive default.
- Many contexts genuinely alternate: after segja, the indicative endorses the content, the subjunctive merely relays it. Mood becomes an evidential / commitment marker — hence vita … er (knowledge) versus halda … sé (belief).
- Non-specific relative clauses take the subjunctive (einhver sem kunni "someone who knows"); specific, real referents take the indicative (maðurinn sem kann).
- A mechanical "after að → subjunctive" rule misleads at B2: it over-subjunctivises factives and cannot capture the alternations. Build the assertion sensor instead — one sensor for all of it.
Now practice Icelandic
Reading grammar gets you part of the way. The exercises are where it sticks — free, no signup needed.
Start learning Icelandic→Related Topics
- The Subjunctive (viðtengingarháttur): OverviewB1 — An orientation to the Icelandic subjunctive mood — a living, everyday part of the language, not a literary relic — covering its four big triggers (reported speech, conditionals, wishes/hopes, and certain conjunctions) and why English speakers, with only a vestigial subjunctive of their own, systematically and audibly leave it out.
- Subjunctive in Reported SpeechB1 — The single most frequent subjunctive trigger in Icelandic: indirect speech introduced by að (and hvort/wh-words) after verbs of saying, thinking, hoping, and asking. The reported clause goes into the subjunctive to mark that the content is REPORTED, not asserted — present subjunctive (sé, komi, fari) under a present matrix verb, past subjunctive (væri, kæmi, færi) under a past one (backshift). Indicative can creep in for facts the speaker personally vouches for, making the mood a subtle evidentiality device.
- Reported Speech and Sequence of MoodB2 — The full machinery of indirect speech in Icelandic: the shift into the subjunctive, the backshift of tense into the PAST subjunctive under a past matrix verb, the adjustment of pronouns and deictics (hér to þar, í dag to þann dag, núna to þá), and reported questions (hvort / wh + subjunctive) and commands (að + subjunctive or infinitive). The key insight: Icelandic backshifts to the past SUBJUNCTIVE, not merely a past indicative as in English, so a single form væri encodes both pastness and reportedness.
- Mood in Relative and Adverbial ClausesC1 — The subtle mood alternation inside relative and adverbial clauses, beyond the basic subjunctive triggers. A relative clause takes the subjunctive when its head is non-specific or hypothetical ('a man who knows Icelandic, any such man' → kunni) and the indicative when the referent is a specific, actual individual (kann). The same specificity logic reaches into temporal and purpose clauses. English marks this distinction only thinly, with 'any' versus 'the', so the mood must be built from scratch.
- Long-Distance Reflexives: the Famous sigC1 — The construction that made Icelandic central to syntactic theory: the reflexive sig / sér / sín can be bound NON-LOCALLY — by the subject of a higher clause, across one or more clause boundaries — provided the intervening clauses are SUBJUNCTIVE. An indicative complement blocks the long-distance link and leaves only the local reading. The subjunctive, Icelandic's other flagship feature, is the licenser; one generalisation ties the two together.
- Mood and Tense ErrorsB2 — A catalogue of the verb-mood and verb-tense slips English speakers make in Icelandic — indicative where the subjunctive is required (reported speech, counterfactuals, þótt/svo að), and the perfect where the preterite belongs (with past-time adverbs). Two distinct root causes: English's dead subjunctive feeds the mood errors; English's looser perfect feeds the tense errors.