Telling Control from Raising

On the surface they are identical. Hann virðist sofa "he seems to sleep" and Hann vill sofa "he wants to sleep" share the shape subject + matrix verb + infinitive — same word order, same parts of speech, same length. Yet one is raising (the subject was lifted up from the lower clause) and the other is control (the subject belongs to the matrix verb, with a silent PRO below). The two pages that build these constructions — raising, ECM, and control and infinitival clauses and PRO — establish what each structure is. This page does the complementary job: it hands you the diagnostic tests that tell them apart in any sentence you meet. There are five, and the last one — quirky case — is a litmus test that exists in Icelandic and almost nowhere else, sharp enough to settle the question in a single glance. (We assume the basics from the two construction pages and link out rather than re-derive them.)

The core difference the tests detect

Every diagnostic below is a probe for one underlying fact: a raising verb (virðast, sýnast, reynast, vera líklegur til) gives its surface subject no thematic role — the noun is there only because it moved up, and the real meaning-giver is the lower verb. A control verb (vilja, reyna, lofa, vona, ætla, ákveða) gives its subject a genuine role — the wanter, the tryer, the promiser — and has a separate silent subject (PRO) below. So:

• Raising = one noun, two positions (it raised; matrix θ-role: none).
• Control = two subjects, one silent (they corefer; matrix θ-role: yes).

Each test exploits this. If the matrix subject has no role of its own, it can be a meaningless element (an expletive, an idiom chunk) and can carry the lower verb's case. If the matrix verb assigns a real role, it demands a meaningful, sentient-where-required, structurally-case-marked subject of its own. Hold that contrast; the five tests simply make it visible.

Test 1: the thematic-role test (the conceptual baseline)

Ask: does the matrix verb assign a role to its surface subject? With a raising verb, the subject is not the seemer/turner-out — nobody "seems" as an action; virðast comments on a whole situation. With a control verb, the subject genuinely does the matrix action — reyna requires a tryer, vilja a wanter.

This is the conceptual baseline, but as a test it can be slippery — intuitions about "who does what" blur. The next four tests turn the θ-role fact into something mechanical and checkable.

Test 2: the expletive test — það raises, but control blocks it

The expletive það ("it/there," the dummy subject of weather verbs, existentials, and impersonals) carries no meaning and no role. So it can only appear in a subject slot that demands no role — which is exactly the raising subject slot, and exactly not the control one. Raising verbs accept an expletive subject; control verbs reject it.

The contrast is decisive. Það virðist … is fine because virðast needs no one to do anything; \Það vill … (meaning "there wants to be") fails because *vilja needs a wanter, and an expletive cannot want. (Það vill svo til að … "it so happens that …" is a fixed idiom, a different vilja, and not a counterexample to this test.) So: substitute the expletive það. If the sentence survives, you are in raising; if it collapses, you are in control.

Test 3: the idiom-chunk test — idioms raise but cannot be controlled

An idiom like ljón liggja á veginum "lions lie in the road" (= there are obstacles) has a fixed subject, ljón "lions," that means nothing on its own — it only means "obstacles" inside the whole idiom. Such a frozen chunk can be displaced by raising (the idiom is reconstructed in the lower clause, where it keeps its idiomatic reading), but it cannot be the subject of a control verb, because a control verb assigns its subject a literal, independent role — and the moment you force ljón to be a literal "lions" that "wants" or "tries," the idiom shatters.

The logic: idioms keep their special meaning only when their parts stay in their original configuration (or move together by raising, which reconstructs them). Control puts a real role on the subject, prying the chunk loose from the idiom and destroying it. A second, well-travelled idiom makes the same point — teningnum er kastað "the die is cast":

Test 4: selectional restrictions — control demands a fitting subject

Because a control verb assigns its subject a real role, that role comes with selectional restrictions — reyna "try" and vilja "want" (in their core senses) want a sentient, volitional subject, something capable of trying or wanting. A raising verb imposes no such restriction, since it gives the subject no role; the only restrictions come from the lower verb. So an inanimate or non-volitional subject that the lower verb is happy with will pass under raising but jar under control.

The asymmetry is the diagnostic: swap in an inanimate, non-agentive subject. If the sentence stays natural, the verb imposed no role on it — raising. If it turns anomalous (the subject is being asked to do something it cannot), the verb assigned an agentive role — control.

Test 5 (the litmus): quirky-case preservation — Icelandic's decisive diagnostic

Here is the test that English cannot offer and that settles the matter instantly. Some Icelandic verbs assign a quirky (non-nominative) case to their subject: leiðast "be bored" takes a dative subject (honum leiðist "he is bored"), vanta "need/lack" takes an accusative subject (hann vantar "he needs"). Now run the diagnostic:

• A raising verb assigns no case of its own; it inherits whatever the lower verb assigned. So when a quirky-subject verb is embedded under virðast, the matrix subject surfaces in the quirky case — Honum virðist leiðast, dative honum, kept from leiðast.
• A control verb assigns its own structural nominative to its subject (and the lower quirky verb's case lands on the silent PRO, invisibly). So the matrix subject of a control verb is nominative, never quirky.

The single fact — does the matrix subject carry the lower verb's quirky case, or its own nominative? — flips raising and control apart with no ambiguity:

Look at the minimal pair Honum virðist leiðast versus Hann vill … leiðast. Same lower verb (leiðast), same embedded meaning ("be bored"), but the matrix subject is dative honum under virðast and nominative hann under vilja. The only explanation: virðast (raising) preserved leiðast's dative on the raised subject, while vilja (control) imposed its own nominative on its own subject. A quirky-subject verb is therefore a litmus test: embed it, and read the case on the matrix subject. Quirky case → raising; nominative → control.

That last pair is the sharpest demonstration of all: reynast ("turn out / prove," raising) and reyna ("try," control) are near-homographs, easily confused — yet the case settles them at once. Reynast can show a preserved quirky dative (mér reynist …); reyna always takes a plain nominative (ég reyni …). Where English would have no signal whatsoever, Icelandic writes the answer on the subject.

The tests, side by side

The same surface string — subject + verb + infinitive — yields opposite results on every probe:

Why English speakers conflate them

English has both constructions but gives no morphological signal, so the learner's intuition is built on surface word order alone — and the order is identical. "He seems to be bored" (raising) and "He wants to be bored" (control) look the same, and English "he" is the same word in both, carrying no informative case. The result is a persistent conflation of virðast (raising) with vilja / reyna (control), and two telling errors. First, learners nominativise the raised quirky subject — producing *hann virðist leiðast when leiðast forces the dative honum under raising. Second, they try to insert an expletive or idiom subject under a control verb (\það vill vera vandamál), not feeling the role-clash that an Icelandic ear registers immediately. The cure is to stop reasoning from the English surface and *run a test — and the fastest, because it is purely mechanical, is the quirky-case litmus: find a quirky-subject verb, embed it, and read the case.

Common Mistakes

Raising preserves the lower verb's quirky case; nominativising the raised subject both is ungrammatical and erases the very signal that proves it is raising.

The expletive test in action: if you want an expletive subject, you need a raising verb, not a control verb.

Don't reach for raising virðast when you mean a flat control assertion; they are different structures with different meanings.

Under control, the matrix subject is nominative (the verb's own case); the lower verb's quirky case stays below on PRO — the mirror image of raising's case preservation.

The idiom-chunk test: idioms can only ride a raising verb; a control verb's real role on the subject destroys the frozen reading.

Key Takeaways

• All five tests probe one fact: raising gives the matrix subject no thematic role (it merely moved up); control gives it a real role and has a separate silent PRO below.
• Expletive test: raising accepts dummy það (Það virðist vera vandamál); control rejects it (\Það vill vera vandamál*).
• Idiom-chunk test: idioms survive under raising (ljón virðast liggja á veginum = obstacles) but die under control (the subject is forced into a literal role).
• Selectional test: inanimate/non-volitional subjects are fine under raising (steinninn virðist …) but anomalous under control (#steinninn reynir …).
• Quirky-case litmus (the decisive, Icelandic-specific test): a raising verb preserves the lower verb's quirky case on the matrix subject (Honum virðist leiðast, dat.; Hana virðist vanta, acc.); a control verb imposes its own nominative (Hann vill … leiðast, nom.), with the quirky case landing on PRO. Quirky → raising; nominative → control. The near-homographs reynast (raising) and reyna (control) are told apart instantly this way.
• English shows no case here, so learners conflate virðast (raising) with vilja/reyna (control); the fix is to run a test, fastest of all the quirky-case litmus.