1. Introduction
The similarities and differences of noun phrases and clauses are an old puzzle. Work on nominalizations has examined the extent to which the two may share internal structure (Chomsky Reference Chomsky, Jacobs and Rosenbaum1970), and the external distribution of the CP resembles that of the DP, as Thráinsson (Reference Thráinsson1979) argued for at length, focusing on Icelandic (using the terminology of the time).Footnote 2 Yet the CP and the DP are not identical, and one difference is that CP is not obviously associated with morphological case.
This squib shows that the evidence that Sigurðsson (Reference Sigurðsson1991) used to reveal the case of PRO can be extended to the Icelandic CP. Just as certain elements overtly agree with the case of PRO, the same types of elements overtly agree with the case of CP. This is shown in (1), where the dative case of a CP subject is realized morphologically on the element báðu ‘both’.Footnote 3
(1) [cp Að hann hefði logið og svikið loforð] var báðu haldið
[cp that he had lied and betrayed promise].dat was both.dat held
fram.
forth
‘It was claimed both that he had lied and broken a promise.’
The case agreement with the CP is interesting because it is exactly this type of evidence that provides the strongest support for the case-marking of PRO. Such evidence can thus be extended to support the view that CP can be case-marked and that the case of a CP can be realized morphologically. The present study does not contradict the view that a CP may not (always) need case (cf. Pesetsky Reference Pesetsky1982); rather, it provides evidence that when a CP is case-marked, it is not only in terms of some abstract licensing mechanism but also in terms of a morphologically realizable property.
Here, I adopt the view that CP can be case-marked directly. However, the main contribution of the squib is not to explore the structural details of configurations in which a CP appears to be in a case position, but rather to show that case is indeed assigned in environments where the argument is a CP rather than a DP. These findings may also support theories wherein the CP itself would not be case-marked, but rather linked to some silent element that has a case value, for example, a CP trace or a silent pronoun. Such theories will be discussed below.
The paper is organized as follows. Section 2 presents some background on the syntax of case in Icelandic and the types of diagnostics that have been applied in the literature to detect elusive case-marking as in the case of PRO. Section 3 presents the core evidence for overt case agreement with CP. Section 4 discusses some theoretical implications of the findings. Section 5 concludes.
2. Background on case and its empirical diagnostics
Icelandic grammatical subjects are not always nominative; they can appear in any morphological case (Andrews Reference Andrews, Maling and Zaenen1990 [1976]; Thráinsson Reference Thráinsson1979; Zaenen et al. Reference Zaenen, Maling and Thráinsson1985; Sigurðsson Reference Sigurðsson1989). While subjects in the language are nominative by default and direct objects accusative, some verbs assign other cases to their subjects as demonstrated by the verb leiðast ‘to be bored’ in (2) which takes a dative subject.
(2) Henni/*Hún leiddist.
her.dat/*she.nom bored.
‘She was bored.’ (Sigurðsson Reference Sigurðsson1991: 328)
Analyzing a dative as a grammatical subject does not carry much weight unless we agree on some diagnostics for subjecthood. The literature takes the compatibility of an argument with the position of PRO in control infinitives to be a reliable sign of subject status. In Icelandic, PRO can correspond to a nominative (3) as well as a dative (4) element.
(3) Strákana langar ekki til [að PRO segja sögu].
boys.the.acc wants not for [to PRO.nom tell story]
‘The boys do not want to tell a story.’
(4) Hana langar ekki til [að PRO leiðast].
her.acc wants not for [to PRO.dat bore]
‘She does not want to be bored.’ (Sigurðsson Reference Sigurðsson1991: 328)
We know from (2) that ‘to be bored’ takes a dative subject and that PRO in (4) corresponds to the dative, but an unpronounced element does not show any dative morphology. The crucial evidence, for the literature on non-nominative subjects as well as the present study on the CP, comes from elements like floating quantifiers which agree overtly with the case of the subject. The following examples from (Sigurðsson Reference Sigurðsson1991: 331) demonstrate this point.
(5) Strákunum leiddist öllum í skóla.
boys.the.dat bored all.dat in school
‘The boys were all bored in school.’
(6) Strákarnir vonast til [að PRO leiðast ekki öllum í skóla].
boys.the.nom hope for [to PRO.dat bore not all.dat in school]
‘The boys hope that they will not all be bored in school.’
Example (5) shows that the floating quantifier ‘all’ agrees in case with an overt dative subject. Example (6) furthermore shows that the quantifier agrees in the same way with a dative PRO. As argued by Sigurðsson, such examples are evidence that non-nominative subjects exist and also that PRO is case-marked. Note that the controller in the main clause is in the nominative case, and thus cannot be the source of the dative. The evidence is quite convincing because the floating quantifier manifests overt morphological case. Having reviewed these background facts about Icelandic case and how it can be detected, we are well equipped to turn to the case of CP.
3. Overt case agreement with CP
In this section I focus on Icelandic predicates which can take either a DP complement or a CP complement and show evidence that a CP is case-marked the same way as a DP in the same position. Consider first the verb segja ‘to tell/say’.
(7) María sagði [dp sögu].
Mary.nom told [dp story.acc]
‘Mary told a story.’
(8) María sagði [cp að hann hefði komið, séð og sigrað].
Mary.nom told [cp that he had come seen and conquered]
‘Mary said that he had come, seen, and conquered.’
The main verb in these examples is a canonical nom-acc verb so the direct object in (7) is in the accusative case. The DP and the CP have a similar distribution cross-linguistically, and these similarities are especially clear in Icelandic as shown by Thráinsson (Reference Thráinsson1979); thus it is natural to ask whether the CP participates in the case system by virtue of being a similar syntactic object. Consider the passivized variants below with a floating quantifier ‘all’.
(9) Sagan var öll sögð (af Maríu).
story.nom was all.nom told (by Mary)
‘The whole story was told (by Mary).’
(10) [cp Að hann hefði komið, séð og sigrað] var allt sagt.
[cp that he had come seen and conquered].nom was all.nom told
‘All of it was said, that he had come, seen and conquered.’
As in English, the theme of a nom-acc verb is realized in the nominative case in an Icelandic passive as shown in (9).Footnote 4 Here, the quantifier agrees with the subject in case, as well as gender (feminine) and number (singular). A parallel example with a CP theme is shown in (10). Here, the quantifier is nominative, neuter, singular. The values are consistent with CP being case-marked in the same way as DP, but they are also compatible with these being default values for the relevant features, with no actual agreement taking place. Consider, in contrast, the Icelandic verb halda fram ‘claim’, literally ‘hold forth’, which is shown in the active voice below.
(11) María hélt [dp þessu] fram.
Mary.nom held [dp this.dat] forth
‘Mary claimed this.’
(12) María hélt fram [cp að hann hefði sigrað].
Mary held forth [cp that he had won]
‘Mary claimed that he had won.’
Example (11) shows that this verb lexically assigns dative case to its object, and thus one can ask whether the CP in (12) receives lexical dative case as well. Note that the relative position of the object and the particle fram is sensitive to the heaviness of the object; a DP object is pronounced after the particle, just as the CP is, if the DP is heavy as in (13). Note that dative morphology is realized on multiple elements in this example due to noun-modifier concord.
(13) María hélt fram [dp einhverju fáránlegu rugli sem enginn
Mary held forth [dp some.dat ridiculous.dat nonsense.dat that nobody
trúði].
believed]
‘Mary claimed some ridiculous nonsense that nobody believed.’
Lexical case is preserved under passivization. The examples in (14) show this; the theme is realized as a dative subject. These examples also contain a floating quantifier that agrees with the dative subject case. This is shown for floating einn ‘one’ (14a), báðir ‘both’ (14b), and allur ‘all’ (14c).
- (14)
a. [dp Þessu] var einu haldið fram.
[dp this.dat] was one.dat held forth
‘Only this was claimed.’
b. [dp Þessu og hinu] var báðu haldið fram.
[dp this.dat and the.other.dat] was both.dat held forth
‘This and that other thing were both claimed.’
c. [dp Þessu, hinu og fleiru] var öllu haldið fram.
[dp this.dat the.other.dat and more.dat] was all.dat held forth
‘This, and that other thing, and some more things, were all claimed.’
Crucially, a CP in this position also triggers dative agreement on a floating quantifier. The examples in (15) show this for the same quantifiers as above.
- (15)
a. [cp Að hann hefði sigrað] var einu haldið fram.
[cp that he had won].dat was one.dat held forth
‘It was only claimed that the he had won.’
b. [cp Að hann hefði logið og svikið loforð] var báðu
[cp that he had lied and betrayed promise].dat was both.dat
haldið fram.
held forth
‘It was claimed both that he had lied and broken a promise.’
c. [cp Að hann hefði komið, séð og sigrað] var öllu haldið
[cp that he had come, seen and conquered].dat was all.dat held fram.
forth
‘That he had come, seen, and conquered were all claimed.’
This is interesting because the type of evidence that uncovered the case of PRO (Sigurðsson Reference Sigurðsson1991) is also found when a finite CP is the locus of the case-marking. The evidence also carries over to infinitival CP. The verb sleppa ‘skip’ takes a dative argument as the following passives show.
- (16)
a. [dp Þessu] var einu sleppt.
[dp this.dat] was one.dat skipped
‘Only this was skipped.’
b. [dp Þessu og hinu] var báðu sleppt.
[dp this.dat and the.other.dat] was both.dat skipped
‘This and that other thing were both skipped.’
c. [dp Þessu, hinu og fleiru] var öllu sleppt.
[dp this.dat the.other.dat and more.dat] was all.dat skipped
‘This, and that other thing, and some more things, were all skipped.’
The examples in (16) are parallel to (14), but the difference is that in the case of sleppa ‘skip’, the dative DP position alternates with an infinitival clause rather than with a finite CP. The examples in (17), parallel to (15), show that this CP also triggers case agreement on a floating quantifier.
- (17)
a. [cp Að yfirheyra Jón] var einu sleppt.
[cp to interrogate John].dat was one.dat skipped
‘Interrogating John was the only thing skipped.’
b. [cp Að yfirheyra og dæma Jón] var báðu sleppt.
[cp to interrogate and sentence John].dat was both.dat skipped
‘Interrogating and sentencing John were both skipped.’
c. [cp Að yfirheyra, dæma og hengja Jón] var öllu sleppt.
[cp to interrogate, sentence and hang John].dat was all.dat skipped
‘Interrogating, sentencing, and hanging John were all skipped.’
For agreement with an accusative element, we can consider impersonal vanta ‘to be lacking/missing’, shown with an accusative DP in (18). It will be assumed here that the CP in (19) is also in the accusative case but the case diagnostics are ambiguous here because unlike the case of a DP, a floating quantifier with a CP is always in the neuter singular (plausibly default values because CP lacks ϕ-features) and this yields nom/acc syncretism in the quantifier, for example, allt ‘all’ in (19).
(18) Hún mætti en [dp strákana] vantaði alla.
she attended but [dp boys.acc.m.pl] lacked all.acc.m.pl
‘She attended but all the boys were missing.’
(19) Margt tókst en [cp að hún, hann og þau mættu] vantaði
Many succeeded but [cp that she, he and they attended] lacked
allt.
all.acc.n.sg
‘Many things worked out but not that she, he and they would attend.’
The examples are consistent with the idea that a floating quantifier can realize an accusative case of a CP and although the form allt is homophonous with the nominative, the dative examples above suggest that this is the realization of an accusative case value.
Some verbs in Icelandic take genitive objects in the active voice like sakna ‘to miss’ in (20). As before, it is possible to float an agreeing genitive quantifier in the passive in (21).
(20) Jón saknaði [dp stelpnanna].
John missed [dp girls.the.gen]
‘John missed the girls.’
(21) [dp Stelpnanna] var allra saknað.
[dp girls.the.gen] were all.gen missed
‘The girls were all missed.’
As with the dative CPs seen above, a floating quantifier can agree with the genitive case of a CP; see (22).
(22) [cp Að hann hefði komið, séð og sigrað] var alls saknað.
[cp that he had come seen and conquered].gen was all.gen missed
‘That he had come, seen and conquered were all missed.’
The case agreement facts show that a CP in a case position in Icelandic can trigger overt morphological case agreement on floating quantifiers. Let us consider the implications of this.
4. Theoretical implications
The previous section speaks against any analysis where a CP argument blocks case-assignment. The case mechanism is clearly active in environments where the argument is a CP rather than a DP. A related question is which element gets the case value, that is, whether (i) CP is directly case-marked or (ii) case is assigned to some covert element that CP is linked to. I adopt the former analysis because it is, in my opinion, the most straightforward approach for Icelandic. I elaborate on this view in this section. Even if future investigations conclude that the case values are really associated with some covert CP-associated element and not the CP itself, the empirical point remains that case is indeed assigned in environments where the argument is a CP rather than a DP. I will now consider analyses according to which the case-marked element is either a CP trace or a silent pronoun and show why I prefer an analysis where CP is directly case-marked.
First, let us consider an analysis where the element with the case value is a CP trace. Stowell (Reference Stowell1981) developed a Case Resistance Principle (CRP) according to which a CP cannot be case-marked directly whereas a trace of its movement can be. According to the CRP, a CP of the type under discussion cannot stay in a case-position or be case-marked directly. When CP appears to occur in subject position, as in the crucial examples above, Stowell adopts a variant of Koster's (Reference Koster and Keyser1978) analysis, proposing that an apparent subject CP is actually in a higher Topic position, associated with a silent element in the true subject position. To adhere to the CRP, a CP must undergo string-vacuous topicalization out of the subject position, or alternatively, extrapose to the right. Accordingly, only a trace of a CP can be case-marked, not the CP itself. Although a floating quantifier can be understood as always arising from movement that strands Q (see Sportiche Reference Sportiche1988), and thus leaves a QP-internal DP trace behind, I follow Stowell (Reference Stowell1981: 153) in assuming that the CRP concretely requires a CP in the subject position to “move to a non-A position”.
This line of reasoning depends on Koster's motivation for placing apparent subject CPs in Topic position. Koster argued that the unavailability of topicalization in subordinate clauses correlates with the unavailability of apparent (clause-initial) CP subjects in the same clauses. The correlation would be explained if the Topic position were unavailable in general in the relevant subordinate clauses. However, as shown by Thráinsson (Reference Thráinsson1979: 102), Icelandic allows topicalization in subordinate clauses where English would not allow it.
(23) Jón veit [cp að Harald elskar María].
John knows [cp that Harold.acc loves Mary.nom]
‘John knows that it is Harold that Mary loves.’
The fact that topicalization is grammatical in this environment, where a CP subject would still need to extrapose (as in English), suggests that whatever demands the extraposition of a CP subject in an embedded clause is not related to the availability of topicalization. This undermines the motivation for adopting the Koster account for Icelandic. It seems more likely that a clause-initial CP really is in the subject position in Icelandic, and thus CP itself is case-marked rather than its trace, although there might be cross-linguistic variation in the structural position of subject-like CPs (see Lohndal Reference Lohndal2014). Thus, it remains a plausible analysis that Icelandic CP can be case-marked directly.
Another approach to maintaining a caseless CP is to say that a CP which appears to be in a case position is in fact an appositive/parenthetical element on a case-bearing pronoun, the only difference being that this pronoun is sometimes covert. Under this analysis, the dative CP examples would have the following structure, dative ‘it’ being covert (as in Li Reference Li, Chappell, Djamouri and Wiebusch2013).
(24) [dp Því/
$\emptyset $ [cp að hann hefði sigrað]] var einu haldið fram.[dp it.dat [cp that he had won]] was one.dat held forth
‘The only thing that was claimed was that he had won.’
Here, the silent structure needs to be motivated. Any analysis along such lines would suggest that a configuration with an overt pronoun and a CP should have the same syntactic properties as an apparent bare CP. It would therefore have to explain why the structure with the overt pronoun (25) is a robust island for the purpose of extraction whereas the bare CP is not (26) (see Thráinsson Reference Thráinsson1979: 195–197; Wood Reference Wood2017).
(25) * [dp Þessu viðhorfi]i talaði Jón um [dp það [cp að hann hefði
[dp this opinion].dat spoke John about [dp it [cp that he had
haldið fram ti]]
held forth tDAT]]
Intended: ‘John spoke about having claimed this opinion (to be right).’
(26) ? [dp Þessu viðhorfi]i talaði Jón um [cp að hann hefði haldið
[dp this opinion].dat spoke John about [cp that he had held
fram ti]
forth tdat]
‘John spoke about having claimed this opinion (to be right).’
These examples show that there is no silent pronoun on top of the CP in (26). The most straightforward analysis of these facts is that the CP itself is the locus of case valuation.
Having considered two types of analysis where an (apparent) case-marked CP would be caseless but associated with a silent case-bearing element, I conclude that a theory with a directly case-marked CP is more straightforwardly appropriate for the Icelandic data under discussion. Of course, such brief comments are not the final word on whether CP is case-marked directly or needs some additional structure to appear in a case position. Settling such questions in a definitive manner is a task for the future. However, any version of case theory in which the case mechanism is effectively turned off when the argument is a CP is incompatible with the evidence.
These findings also resonate with certain other observations about CP case in Icelandic. Consider, for example, constructions with a CP subject and an accusative DP object as shown below for a finite (27) and an infinitival (28) CP (see Sigurðsson Reference Sigurðsson, Brandner and Zinsmeister2003; Thráinsson Reference Thráinsson2007: 195–196).
(27) [cp Að Gunnar skyldi ekki hafa mætt] drap alveg stemninguna.
[cp that Gunnar should not have attended] killed totally mood.the.acc
‘It killed the mood that Gunnar did not attend.’
(28) [cp Að syngja of hátt] gæti angrað hana.
[cp to sing too loudly] could irritate her.acc
‘Singing too loudly could irritate her.’
In a case theory like the dependent case approach of Marantz (Reference Marantz and Reuland2000), see also Yip et al. (Reference Yip, Maling and Jackendoff1987), an accusative argument should not surface in the absence of a nominative argument. Here, the most obvious analysis of (27–28) is that the CP is the nominative element in question. This type of evidence fits well with the view that CP can be case-marked but the evidence is indirect because there is no nominative morphology in these examples.
If the discussion in this paper is on the right track and the CP and the DP are quite similar with respect to case valuation, there are a couple of ways in which the theory can accomodate such a situation. One way is to posit a case feature on the C head. Another is to say to that the CP and the DP share some edge structure. For example, there could be a KP (case phrase) on top of each of them. No attempt will be made here to settle such questions as they are beyond the scope of a squib. CP can be case-marked but the formal details of what that means remain open for discussion.
5. Conclusion
This paper has shown that the Icelandic evidence which has been used so fruitfully to analyze the case of PRO is also relevant for the issue of CP case-marking. A chain of evidence was reviewed which emerged out of the literature on non-nominative subjects and the case of PRO and this line of research was extended to the case of CP by studying floating quantifiers which agree overtly in case-marking with a CP. This was taken to be evidence that CP can be case-marked. Yet, it was acknowledged that it may also be the case that CP is not case-marked directly in such cases but can rather be associated with a silent intermediate element that can host the relevant case values. At a general level, the paper offers an avenue of future inquiry by revealing a parallelism between the diagnostics available in Icelandic for detecting the case of PRO and the case of CP.
