Parasitic Gaps and Across-the-Board Extraction

This page sits at the edge of what is known about Icelandic extraction, and it earns its C2 label honestly: the data are subtle, some judgements are gradient, and a few points are debated even among specialists. We will be precise about where the firm ground ends. The starting fact is a hard constraint you already half-know from complex/wh-movement-and-islands: you cannot, in general, pull a phrase out of one half of a coordinate structure. I met John and Mary does not let you ask *Who did you meet John and ?. The Coordinate Structure Constraint (CSC) bans it. But there are two famous escape hatches. The first is across-the-board (ATB) extraction: you can extract if you extract the same element from both conjuncts at once, leaving a gap in each. The second is the parasitic gap (PG): a "free-rider" gap, normally illicit on its own, that becomes acceptable when there is a genuine extraction gap elsewhere to license it. Icelandic adds the twist that makes the whole topic worth a dedicated page: because every noun wears its case overtly, Icelandic lets you see a constraint that is invisible in English — both gaps must want the same case.

The baseline: you cannot extract from one conjunct

First, fix the constraint we are escaping. Coordinate two verb phrases or two objects, and Icelandic — like English — forbids reaching into just one of them to extract a wh-phrase. The conjunction og "and" (and en "but," eða "or") closes the coordinate structure off; pulling material out of a single conjunct violates the CSC and produces sharp ungrammaticality.

The ban is robust; native judgements here are not gradient. With that fixed, the interesting question is: what can legally leave a coordination?

ATB extraction: extract from both conjuncts at once

The answer is across-the-board extraction. If a single wh-phrase corresponds to a gap in every conjunct — it is "missing" from both halves in the same role — extraction is licit. You are not reaching into one conjunct and ignoring the other; you are extracting across the board, with one wh-phrase binding a gap in each. Coordinate two clauses that both have a missing object, and one hvern "who(m)" can question both gaps at once.

The structural intuition: coordination demands parallelism, and ATB is extraction that respects the parallelism — it removes the same thing from both sides, so the conjuncts stay balanced. Extracting from one conjunct only would leave a lopsided structure, and that is what the CSC rules out. ATB is the symmetrical, balanced exception.

The case twist: ATB requires matching case in both gaps

Here is where Icelandic earns the page. In English, Who does María love and Jón admire? works and you never notice anything about case, because the wh-word who is invariant and the two verbs both take a plain object anyway. But Icelandic verbs assign different cases. elska "love" takes an accusative object; dá "admire" also accusative — so Hvern elskar María og dáir Jón is fine, with hvern (accusative) satisfying both gaps. Now coordinate a verb that takes accusative with one that takes dative, and ATB breaks: the single extracted wh-phrase would have to be accusative for one gap and dative for the other at the same time, which is impossible. One form, two incompatible case demands.

Compare elska (acc) with hjálpa "help" (dat). Hvern elskar María needs accusative hvern; Hverjum hjálpar María needs dative hverjum. Try to ATB them and there is no single wh-form that fits both gaps:

So the deep finding, visible only in a case language: ATB extraction requires that both gaps assign the extracted phrase the same case. The Coordinate Structure Constraint is not just about whether you extract from both conjuncts — it interacts with case identity across the gaps. English satisfies the requirement vacuously (its wh-word has no informative case), so the constraint is invisible there; Icelandic spells it out. This is the kind of fact that put Icelandic at the centre of extraction theory: the case morphology makes an abstract identity condition observable.

When the two conjuncts genuinely share a case, ATB is robust even across quirky case — coordinate two dative-object verbs and dative hverjum serves both:

Parasitic gaps: a second gap that rides on a real one

The second phenomenon is rarer, more delicate, and where the honest acknowledgement of gradience matters most. A parasitic gap is a gap that would be illegal on its own — typically inside an adjunct (an án þess að "without," áður en "before," or other adverbial clause), or inside a subject, both of which are normally islands you cannot extract from — but which becomes acceptable because there is a real, independently licensed extraction gap elsewhere in the sentence to "license" it. The real gap is the host; the parasitic gap is the free-rider that survives only in its company.

The schema, on the English model, is "Which paper did you file [real gap] without reading [parasitic gap]?" — the gap inside without reading would be impossible alone (*Which paper did you sleep without reading? is bad), but it is rescued by the real object gap in file.

Two honest caveats belong here, prominently. First, parasitic gaps are marginal and stylistically heavy in Icelandic; many speakers strongly prefer to fill the second position with a resumptive pronoun (…án þess að lesa hana "without reading it") rather than leave a gap, and for some speakers the true gap version ranges from awkward to barely acceptable. Acceptability is gradient, not a clean yes/no, and it improves when the gap is deeper, the adjunct is non-finite (án þess að + infinitive), and the sentence is short. Treat these as a real but fragile construction, not an everyday tool.

The case twist again: the parasitic gap must match case

The second caveat is the page's payoff, and it parallels ATB. Because Icelandic spells case, a parasitic gap is best — for the speakers who accept it at all — when the case wanted at the parasitic gap matches the case of the real gap. In Hvaða skýrslu lagðir þú frá þér án þess að lesa , both leggja (frá sér) and lesa take an accusative object, and hvaða skýrslu is accusative throughout: the real gap and the parasitic gap agree in case. When the two positions would demand different cases, acceptability drops sharply — the single fronted phrase cannot simultaneously be the accusative the host wants and, say, the dative the adjunct verb wants.

So both escape hatches — ATB and parasitic gaps — turn out to obey a case-identity condition that English cannot show you. This is the unifying insight: the coordinate-structure / island machinery does not operate over bare syntactic positions alone; it operates over positions with their case demands, and a single extracted phrase can satisfy two gaps only if those gaps agree on case. Overtly case-marked Icelandic is the laboratory where this becomes visible.

Why English speakers get this wrong

English gives almost no useful intuition here, and what it gives is misleading. First, the CSC itself is sometimes flouted by learners who, hearing that Icelandic word order is "free," try to extract from a single conjunct (*Hvern hitti ég Jón og ); the freedom is in main-clause fronting, not in cracking open coordinate structures. Second, and more insidious, English's caseless wh-word who lets Who does she love and help? pass without any case thought, so learners assume the Icelandic equivalent is equally free and produce case-mismatched ATB (*Hvern elskar María og hjálpar Jón) — the very thing the case-identity condition forbids. Third, learners over-reach with parasitic gaps, treating the English construction as freely available when in Icelandic the natural choice is usually a resumptive pronoun, and a true gap is marginal and case-sensitive. The corrective discipline: when coordinating or building an adjunct gap, check that every gap the fronted phrase binds assigns it the same case; if they don't, the extraction is out, and you should reach for a resumptive pronoun or restructure.

Common Mistakes

You may only extract from a coordinate structure across-the-board (from every conjunct), never from a single conjunct.

ATB extraction is only legal when every conjunct assigns the extracted phrase the same case. Coordinate verbs with the same case government.

When a parasitic gap would require a different case from the real gap, drop the gap and use a resumptive pronoun.

Key Takeaways

• The Coordinate Structure Constraint bans extracting from a single conjunct (*Hvern hitti ég Jón og ). This ban is firm, not gradient.
• ATB extraction is the legal exception: one wh-phrase may extract if it leaves a gap in every conjunct simultaneously (Hvern elskar María og dáir Jón ).
• The Icelandic-specific finding: ATB requires matching case in all gaps. Coordinating an accusative-object verb with a dative-object verb breaks ATB, because no single wh-form satisfies both gaps. English hides this; case-marked Icelandic reveals it.
• Parasitic gaps — a second gap inside an island (adjunct/subject) licensed by a real wh-gap — exist in Icelandic but are marginal, gradient, and stylistically heavy; most speakers prefer a resumptive pronoun, and acceptability is best when the parasitic gap matches the real gap's case.
• The unifying insight: both escape hatches obey a case-identity condition across gaps — the extraction machinery operates over positions with their case demands, not bare positions. Overtly case-marked Icelandic is where this becomes observable.
• English speakers err by cracking open single conjuncts, by producing case-mismatched ATB (their caseless who gives no warning), and by over-using parasitic gaps where a resumptive pronoun is the natural choice. The discipline: every gap a fronted phrase binds must assign it the same case.