Case Preservation and Quirky Case in Depth

There is one fact about Icelandic that appears in every serious syntax curriculum on Earth, and it is this: when an argument is marked for a case lexically — because the verb arbitrarily demands it — that case survives every syntactic operation you can put the argument through. Make the clause passive, raise the argument into a higher clause, embed it under a control verb, mark it accusatively from outside — the lexical case clings to the noun like a fingerprint. The most spectacular instance, the one that reshaped the theory of grammatical relations, is the passive of a dative-object verb: the object becomes the subject, occupies the subject position, behaves like a subject in every testable way — and is still dative. Honum var hjálpað "he was helped," with a dative "subject" honum. This page is about why that happens, what it proves, and how to wield it. We assume the introduction to quirky subjects (verbs/quirky-subjects-syntax) and the basics of raising (complex/raising-and-control); here we go to the bottom of preservation itself.

Two kinds of case: structural vs lexical

The whole phenomenon rests on a distinction English completely hides. Some case on a noun is structural — assigned by the position the noun lands in. A subject is nominative because it is a subject; a direct object is accusative because it is a direct object. Move it, and its case changes to match the new position. This is the only kind of case English (faintly) has: he saw him, he is nominative by virtue of being subject, him accusative by virtue of being object.

Other case is lexical (also called inherent or quirky) — assigned by the particular verb as a brute lexical fact, independent of position. hjálpa "help" simply demands that its object be dative; sakna "miss" demands genitive; vanta "lack/need" gives its subject accusative. There is no semantic logic you can fully derive these from — they are listed in the verb's lexical entry, the way English "depend on" lists its preposition. The defining test for which kind of case you have is exactly preservation:

The crown jewel: dative subjects of the passive

Now apply the passive. The passive promotes an object to subject. With an ordinary structural-accusative object, the promoted noun shows up nominative, exactly as English leads you to expect: Lögreglan handtók manninn "the police arrested the man" → Maðurinn var handtekinn "the man was arrested," with nominative maðurinn and an agreeing participle handtekinn. So far, no surprise. The object's accusative was structural, so when it moved to subject position it took on the case of that position: nominative.

But take a verb whose object is lexically dative, like hjálpa or stela "steal (from)." Passivise it. The object moves to subject position — and arrives still dative. The case does not update to nominative, because it was never structural to begin with; it was lexically stamped on the noun and travels with it.

Stare at Honum var hjálpað. The noun honum sits in the highest argument position of the clause; it triggers the subject behaviours; nothing else competes for subjecthood. Yet it is dative. The participle is correspondingly frozen in the neuter-singular default hjálpað — it does not agree with honum, precisely because honum is not nominative and only a nominative subject triggers participle agreement. (Compare the agreeing handtekinn above, where the subject was nominative.) The non-agreement is itself a second piece of evidence: the grammar is treating honum as a dative subject, and dative subjects don't control agreement.

Why this proves quirky nouns are real subjects

Why does this single fact matter so much that it appears in every textbook? Because it settles an argument about what a "subject" is. One could imagine that honum in Honum var hjálpað is not a subject at all but just a fronted dative object, with the clause being subjectless. Icelandic shows it is a genuine subject by a battery of independent tests — it raises like a subject, it can be the silent PRO of a control clause, it occupies the prefield, it is the antecedent for subject-oriented reflexives, it inverts with the finite verb in questions. A dative noun passing all the subject tests, while remaining dative, is impossible under the old assumption that "subject = nominative." It forced linguists to separate two things English had conflated: grammatical function (subject) and morphological case (nominative). A subject need not be nominative. Icelandic is the language where you can see the two come apart cleanly, and the wedge that pries them apart is case preservation.

Accusative is preserved too — and the accusative-subject verbs

Preservation is not a quirk of the dative; it is the behaviour of all lexical case. A few verbs lexically assign accusative to their subject — vanta "lack/need," langa "want," dreyma "dream." These accusative subjects survive operations exactly as datives do. The cleanest demonstration is raising: lift such a subject into virðast "seem," and the accusative comes up untouched. There is also a class where an accusative object's case is visible across operations.

Preservation under raising: the dative goes up unchanged

The other showcase operation is raising (treated structurally on complex/raising-and-control). A raising verb such as virðast "seem" has no subject of its own; it pulls up the lower clause's subject. If that lower subject is quirky, the quirky case rides up with it, because virðast assigns no case to overwrite it. The result is a dative — or accusative — noun sitting as the subject of virðist.

Note the deep consistency: whether the dative becomes a subject via the passive (Honum var hjálpað) or via raising (Honum virðist leiðast), the mechanism is the same — the case was lexical, no later operation assigns over it, so it persists into the derived subject position. Two unrelated-looking operations behave identically, which is exactly what a single principle ("lexical case is preserved") predicts and a pile of ad-hoc rules would not.

Preservation under control: the silent subject keeps its case

Even when the quirky subject is silent — the PRO of a control clause — its lexical case is still active, visible through agreement on predicates and floated quantifiers. A control verb's infinitival complement headed by a quirky-subject verb forces the understood subject to carry that quirky case, and a predicate adjective inside surfaces in it.

Why English speakers get this wrong

English has nothing like this, and the gap creates one very specific, very stubborn error. English passive always gives a nominative subject — "he was helped," subject "he," plainly nominative-looking. So the English-trained instinct, on meeting the Icelandic passive of hjálpa, is to reach for the nominative: *Hann var hjálpað. This is simply ungrammatical. hjálpa assigned dative to that argument, and the dative is preserved into the passive subject: it must be honum. The same trap hits stela (→ Bílnum var stolið, not *Bíllinn var stolið), bjóða (→ Mér var boðið), and every other dative-governing verb. A related sub-error is making the participle agree (*Honum var hjálpaður); since the subject is not nominative, the participle stays in the frozen neuter hjálpað. The fix is a habit: before passivising, ask what case the verb assigns to that argument in the active. If it was lexical dative or genitive, the passive subject keeps it, and the participle freezes.

Common Mistakes

The flagship mistake: nominativising the passive subject of a dative-object verb. The lexical dative survives the passive, so the subject is honum, and the participle does not agree.

Key Takeaways

• Icelandic distinguishes structural case (assigned by position; changes when the noun moves — e.g. an accusative object becoming a nominative subject under the passive) from lexical/quirky case (assigned by the particular verb; never changes).
• Lexical case is preserved across every operation: passive, raising, control, ECM. This is the defining property and the single most-cited fact in Icelandic syntax.
• The showcase is the passive of a dative-governing verb, which yields a dative subject with a frozen neuter participle: Honum var hjálpað "he was helped," Bílnum var stolið "the car was stolen," Mér var boðið "I was invited" — never the nominative *Hann var hjálpað, never an agreeing *hjálpaður.
• Accusative lexical case is preserved too (Hana virðist vanta peninga), and so is the dative under raising (Honum virðist leiðast) and under control (on a silent PRO).
• The theoretical payoff: a dative noun passing all subject tests while staying dative proves that grammatical function (subject) and morphological case (nominative) are separate — a subject need not be nominative. Case preservation is the wedge that splits them, and Icelandic is the language where you can watch it happen.
• English speakers' chief error is nominativising the dative passive subject (*Hann var hjálpað) and making the participle agree (*hjálpaður). The cure: check the verb's active case first; lexical dative/genitive survives the passive, and the participle freezes.