Hostname: page-component-74656f9884-j7cmm Total loading time: 0 Render date: 2026-11-09T21:45:34.233Z Has data issue: false hasContentIssue false
  • English
  • Français

The movement analysis of temporal adverbial clauses1

Published online by Cambridge University Press:  19 October 2009

LILIANE HAEGEMAN
LILIANE HAEGEMAN*
Affiliation:
University of Ghent, Department of English, Rozier 44, B-9000 Ghent, Belgiumliliane.haegeman@UGent.be
Rights & Permissions [Opens in a new window]

Abstract

In the literature it has been proposed that temporal adverbial clauses can be derived by wh-movement of an operator (e.g. when) to the left periphery (Geis 1970, 1975; Enç 1987: 655; Larson 1987, 1990; Dubinsky & Williams 1995; Declerck 1997; Demirdache & Uribe-Etxebarria 2004: 165–70). After reviewing the arguments that have been proposed in favour of such a movement analysis, the article provides additional empirical evidence in support of the analysis. The data concern so-called Main Clause Phenomena (MCP) or Root phenomena, that is, syntactic phenomena such as argument fronting, Locative Inversion, preposing around be, VP preposing and Negative Inversion, which in English are by and large restricted to main clauses. The unavailability of these MCP in temporal adverbial clauses follows directly from the movement account. The movement analysis will be extended to conditional clauses and factive clauses.

Information

Type
Research Article
Information
English Language & Linguistics , Volume 13 , Issue 3 , November 2009 , pp. 385 - 408
Copyright
Copyright © Cambridge University Press 2009

1 Introduction: aim and scope of the article

The central focus of this article is the internal syntax of adverbial clauses such as the when clause in (1).

  1. (1) When I heard this song, I remembered my first love.

I will first focus on such temporal when clauses; at a second stage, I will widen the discussion to include other temporal adverbial clauses and conditional clauses.

In the literature it has at various points been proposed that the adverbial clause in (1) is a headless relative and is derived by wh-movement of an operator (when) to the left periphery (Geis Reference Geis1970, Reference Geis1975; Enç Reference Enç1987: 655; Larson Reference Larson1987, Reference Larson1990; Dubinsky & Williams Reference Dubinsky and Williams1995; Declerck Reference Declerck1997; Demirdache & Uribe-Etxebarria Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004: 165–70) and some evidence has been put forward. I will first present the arguments proposed in favour of the movement analysis of adverbial when clauses. I will then show that additional empirical evidence is available confirming the movement analysis. The data discussed concern so-called Main Clause Phenomena (MCP) or Root phenomena, that is, syntactic phenomena which in English are by and large restricted to main clauses. The relevant MCP are excluded from when clauses. In particular it will be shown that while adjuncts may appear to the left of the canonical subject position in temporal clauses introduced by when, fronted arguments may not appear to the left of the subject. This argument/adjunct asymmetry follows directly from the movement account. In later parts of the article, I broaden the analysis to include temporal clauses introduced by before, after, since and while, conditional clauses and factive clauses.

The theoretical framework used is that of generative grammar, but throughout the article, the discussion will be kept at a fairly nontechnical level. There are a number of conceivable implementations of the analysis and the choice depends on theoretical concerns which, as far as I can tell, are not directly relevant for the empirical issue discussed here.

The article is organized as follows. Section 2 presents the analysis of the temporal clause in (1) in terms of a movement derivation as it has been discussed in the literature and reviews the main arguments proposed so far. Section 3 shows that the movement account allows us to make the correct predictions concerning the distribution of fronted constituents in English when clauses. To the best of my knowledge, the data discussed in this section, some of which are well known, have not been brought to bear on the movement account of when clauses. This section thus offers new empirical support for the movement analysis of the temporal clause in (1). The final part of the section extends the analysis to temporal clauses introduced by before, after, until and (temporal) since. Section 4 considers adverbial clauses which are compatible with MCP. Section 5 extends the movement derivation to conditional clauses. Section 6 shows how the incompatibility of MCP with English factive clauses can also be accounted for by a movement analysis. Section 7 is a summary of the article.

2 The syntax of temporal adverbial clauses

2.1 The proposal: operator movement in temporal clauses

It has been proposed by, among others, Geis (Reference Geis1975), Enç (Reference Enç1987: 655), Larson (Reference Larson1987, Reference Larson1990), Dubinsky & Williams (Reference Dubinsky and Williams1995), Declerck (Reference Declerck1997), Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004), that temporal adverbial clauses such as (1) are derived by the movement of a clause-internal operator to the left periphery, that is, the domain to the left of the canonical subject position. This is schematically represented in (2), in which the relevant operator is when.Footnote 2

  1. (2) [CP wheni [TP I [VP heard this song] ti]]

2.2 Arguments in favour of the movement account

Intuitively, the movement derivation of adverbial clauses is certainly appealing. The very fact that when, the ‘conjunction’ introducing temporal clauses, is isomorphic to an interrogative operator as well as to a relative pronoun (see below, (8a)) supports this idea. In other languages too, temporal clauses are introduced by a ‘conjunction’ which doubles up as an interrogative element (French: quand, Italian: quando, Spanish: cuando, Catalan: quan, Dutch: wanneer) or as a relative element (Norwegian: når; Stephens Reference Stephens, Butt and King2006). There is also diachronic support for the movement analysis of when clauses. Declerck (Reference Declerck1997: 45) writes:

When is originally a question word, which at the end of the Old English period came to be used as a relative. As a relative it could be used either with a temporal noun as antecedent or as a free (headless) relative. In the latter use it corresponded exactly to the present-day conjunction when.

Declerck (Reference Declerck1997: 51–6) provides a detailed discussion of the development of when from an interrogative element to its use as a conjunction. I do not repeat his discussion here but refer the reader to Declerck's own work.

In other cases, the ‘conjunction’ introducing the temporal clause may correspond to a clause-internal temporal adverb. For instance, in Dutch the temporal adverb toen (‘then’) in (3a) can also be used to introduce an adverbial clause (3b).Footnote 3 If we generalize the movement analysis of temporal adverbial clauses, it could be postulated that the conjunction toen originates in a lower adverbial position and moves to the clause typing position (SpecCP).

  1. (3)

Returning to English and to when clauses, let us assume that clauses are layered structures in which the Verb Phrase (VP) domain encodes thematic relations, the Tense Phrase (TP) encodes modal, temporal and aspectual information, and the Complementizer Phrase (CP) contains discursive information such as illocutionary force and topic or focus related concepts (see Rizzi Reference Rizzi and Haegeman1997 for a detailed discussion of the internal structure of CP). In (4a) the DPFootnote 4this song is a topic: it is moved from its VP-internal object position to the left periphery of the clause. (4b) is a tree diagram. The movement in (4a) is an instance of so-called ‘short’ movement: the moved object DP does not leave the CP-domain in which it starts out.

  1. (4)

    1. (a) This song, he had heard.

    2. (b)

In (4c) the object this song leaves the complement CP, and moves to the CP domain of the matrix clause. This illustrates so-called ‘long’ movement:

  1. (4)

    1. (c) [CP [This song]i, [TP he [VP said [CP ti [TP he had [VP heard ti already]]]]]].

According to a movement approach, when in (1)/(2) undergoes short movement: it moves from the TP domain of the clause for which it establishes a temporal relation to the CP of that same clause.

The argument that is standardly advanced in favour of the movement account of temporal adverbial clauses (cf. Geis Reference Geis1970, Reference Geis1975; Larson Reference Larson1987, Reference Larson1990) concerns the observation that in addition to cases of short movement, illustrated in (1)/(2), one also finds instances in which the operator has undergone long movement. The example used in the literature to illustrate this point is (5). (5) is ambiguous: the time expressed by when might be either that of the higher clause within the adverbial clause, giving the so called ‘high construal’ in (i), with the corresponding derivation, or it may be that of the lower clause in the adverbial clause, giving what is referred to as the ‘low construal’ in (ii), with the corresponding derivation. (i) illustrates short movement of the operator, (ii) illustrates long movement, with an operator originating in the embedded clauses and targeting the higher clause of the adverbial adjunct.

  1. (5) I saw Mary in New York when she claimed that she would leave.

    1. (i) high construal: ‘I saw her at the time that she made that claim.’

      I saw Mary in New York [CP wheni [TP she claimed [CP that [TP she would leave]] ti]]

    2. (ii) low construal: ‘I saw her at the time of her presumed departure.’

      I saw Mary in New York [CP wheni [TP she claimed [CP ti that [TP she would leave ti]]]]

      (Larson Reference Larson1987: 169–70)

If the low-construal reading is derived by the movement of an operator from a lower clause to a higher clause, then we expect that it will not be available in those contexts in which movement from a lower clause to a higher clause is independently known to be excluded. This prediction is correct. Consider (6a), which retains only the high-construal reading (discussed in Larson Reference Larson1990: 170; Demirdache & Uribe-Etxebarria Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004: 165–76).

  1. (6)

    1. (a) I saw Mary in New York when she made the claim that she would leave.

    2. (i) ‘I saw her at the time that she made that claim.’

    3. (ii) *‘I saw her at the time of her presumed departure.’

Differently from (5), the temporal when clause in (6a) contains a complex noun phrase, headed by the noun claim which takes a that-clause as its complement. In order to attain the low-construal reading (ii) in (6a), the operator when would have had to be extracted from inside the clause embedded under the noun claim. (6b) represents the (unavailable) low-construal reading of (6a):

  1. (6)

    1. (b) *I saw Mary in New York [CP wheni [TP she made [DP the claim [CP ti that [TP she would leave ti]]]]]

Given the movement account, the difference in availability of the low-construal reading between (5) and (6) can be related to the difference in grammaticality between (7a) and (7b): in (7a) the relative operator (which) can be extracted from the clausal complement (‘CP’) of the verb claim, but a similar extraction is not possible from the clausal complement of the noun claim because the DP boundary blocks extraction.

  1. (7)

    1. (a) This is the text [CP whichi [TP she claimed [CP ti that [TP she wrote ti]]]]

    2. (b) *This is the text [CP whichi [TP she made

      [DP the claim [CP ti that [TP she wrote ti]]]]]

For similar arguments see also Rudanko (Reference Rudanko1981), Stump (Reference Stump1985) and Declerck (Reference Declerck1997), among many others.

2.3 An implementation of the movement account: temporal adverbial clauses as free relatives

Given that there is no overt antecedent for the moved operator when, the movement analysis of adverbial clauses implies that they are represented as headless or free relatives (Bresnan & Grimshaw Reference Bresnan and Grimshaw1978: 345–67; Demirdache & Uribe-Etxebarria Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004; Declerck Reference Declerck1997; Citko Reference Citko2000; Liptàk Reference Lipták2005; Stephens Reference Stephens, Butt and King2006; Tomaszewicz to appear; Ürögdi Reference Ürögdi2008). As also signalled by Declerck, when clauses may actually also function as relative clauses with antecedents, as shown in (8a):Footnote 5

  1. (8)

    1. (a) The day when I heard this song . . . .

A further argument advanced in support of the relative analysis of temporal clauses is the alternation between when in (1) and whenever in (8b), an alternation also found in free relatives (9) (see the discussion in Declerck Reference Declerck1997).Footnote 6

  1. (8)

    1. (b) Whenever I hear this song, I remember my first love.

  1. (9)

    1. (a) I will eat what you cook.

    2. (b) I will eat whatever you cook.

The precise formal implementation of the movement analysis of adverbial clauses ultimately depends on the analysis of relative clauses in general and that of free relatives in particular. This is not the place to go into these issues in detail (see Bresnan & Grimshaw Reference Bresnan and Grimshaw1978 for an early discussion, and Van Riemsdijk Reference Riemsdijk2006 for a recent survey of the generative debate on free relatives). The analysis also depends on the theory of Tense representation that one adopts. For the present article I adopt the proposal on the derivation of temporal adverbial clauses as elaborated by Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004), which is couched in their theory of Tense representation.

Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004) adopt a Reichenbachian (Reference Reichenbach1947) approach to tense interpretation and propose that temporal relations in the clause are expressed through the projections T(ense)P and Asp(ect)P, where TP selects AspP. Simplifying somewhat, T takes as its external argument the Utterance time, and as its internal argument the Reference time, the latter encoded in the specifier of AspP. Asp takes as its external argument the Reference time, and as its internal argument the Event time, which is encoded in VP.

The authors further propose that temporal adjuncts are prepositional phrases, possibly with null P (e.g. last week), in which P takes as its complement a ‘temporal argument’, a DP with temporal interpretation, the so-called ZeitP. Thus in before Christmas, the temporal P is before, and the ZeitP complement is the DP Christmas (see Demirdache & Uribe-Etxebarria Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004: 265). Adverbial clauses are also temporal PPs in which P selects a ZeitP. Temporal clauses introduced by until/after/before, for instance, are headed by an overt preposition. I briefly return to these below. Temporal when clauses are ‘concealed’ PPs (Demirdache & Uribe-Etxebarria Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004: 169): the temporal P is null, as is the head of its complement, the ZeitP. Thus (10a) has the same underlying structure as (10b) (2004: 177, n.12):

  1. (10)

    1. (a) [PP at [ZeitP the time [CP when Zooey arrived]]]

    2. (b) [PP Ø [ZeitP Ø [CP when Zooey arrived]]]

The when clauses in (10a) and (10b) are both analysed as relative clauses, derived by operator movement. (10c) is Demirdache & Uribe-Etxebarria's representation for (10b). I will assume here that the landing site of when is the same as that of relative pronouns (but see note 6).

  1. (10)

    1. (c)

It is proposed that the moved temporal operator originates in SpecAspP. As discussed by Demirdache & Uribe-Etxebarria, this will capture the observation made by Hornstein (Reference Hornstein1990: 63) that a time adjunct in a temporal clause cannot modify the Reference time. Thus, while the adverbial modifier at 3 p.m. may modify either the Reference time or the Event time in (11a), it can only modify the Event time in (11b):

  1. (11)

    1. (a) John had left at 3 p.m.

      (Demirdache & Uribe-Etxebarria Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004: 172, their (46b))

    2. (b) Nobody had been surprised when John had left at 3 p.m.

In (11b) when is associated with the Reference time, encoded in SpecAsp; at 3 p.m. is associated with the Event time, encoded in the VP. I refer to Demirdache & Uribe-Etxebarria's own discussion (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004: 172) for more details.

The analysis allows one to account for the contrast between the ‘high’ and ‘low’ construal readings of temporal clauses illustrated in (5): in (5i), with the ‘high’ construal, the temporal operator originates in SpecAspP of the higher clause and thus modifies claimed, in (5ii) the adverbial operator originates in SpecAspP of the lower clause, i.e. the complement of claimed, and modifies the time of leave.

The derivation of when clauses as (free) relatives finds crosslinguistic support. To mention but two examples here: Lipták (Reference Lipták2005) shows that Hungarian temporal adjunct clauses lend themselves to an account in terms of relativization. In particular, as illustrated in (12a), a subset of temporal clauses contain the relative marker a. As shown in (12b), a is ‘a morpheme that adorns relative wh-phrases in Hungarian’ (Lipták Reference Lipták2005: 139).

  1. (12)

The same conclusions are drawn by Zribi-Hertz & Diagne (Reference Zribi-Hertz, Diagne, Sauzet and Zribi-Hertz2003: 23) with respect to the derivation of temporal clauses in Wolof:

L'ensemble de données présenté ci-dessus conduit à conclure que les syntagmes étiquetés plus haut ‘circonstancielles de temps’ ne sont autres que ce qu'on a pu appeler ailleurs des ‘relatives indépendantes adverbiales’, c'est-à-dire des DP incluant un circonstant relativisé dépourvu de contenu lexical.

(The data presented here lead to the conclusion that the constituents labelled ‘temporal adverbial clauses’ are none other than what are also labelled ‘independent adverbial relatives’, that is to say, DPs with a nonlexicalized relativized adverbial adjunct. (Tr. LH))

The next section provides empirical support for the assumption that when clauses are derived by operator movement.

3 Empirical support for the movement analysis

3.1 Argument fronting in adverbial clauses

The movement account of adverbial clauses finds empirical support from, and thus also sheds new light on, the long-standing observation (see Emonds Reference Emonds1970, Reference Emonds1976, Reference Emonds, Adger, Cat and Tsoulas2004; Hooper & Thompson Reference Hooper and Thompson1973) that argument fronting is not possible in English temporal clauses. To the best of my knowledge, this type of evidence has not been advanced in the literature in support of the movement analysis. Consider the data in (13):

  1. (13)

    1. (a) She started to write this column last year.

    2. (b) This column she started to write last year.

    3. (c) When she started to write this column last year, I thought she would be fine.

    4. (d) *When this column she started to write last year, I thought she would be fine.

In order to achieve particular discourse effects (topicalization/focalization), the direct object of write in (13a) can be fronted to the left of the subject in (13b). Argument fronting like that illustrated in (13b) is classified as a Main Clause Phenomenon (MCP) or a root transformation (Emonds Reference Emonds1970, Reference Emonds1976, Reference Emonds, Adger, Cat and Tsoulas2004; Hooper & Thompson Reference Hooper and Thompson1973), because this type of fronting is restricted to main clauses and a subset of embedded clauses (see Heycock Reference Heycock2006 for a recent survey of the relevant subset). Argument fronting is not licit in the temporal when clause in (13d). Various proposals have been made to account for the distribution of MCP. Many of them rely on a distinctive property of the root clause which licenses the MCP. In the absence of this property, MCPs are excluded. The relevant property may be pragmatic/semantic (Hooper & Thompson Reference Hooper and Thompson1973; Green Reference Green1976, Reference Green1996; Sawada & Larson Reference Sawada and Larson2004) or structural (Haegeman Reference Haegeman2003a; Emonds Reference Emonds, Adger, Cat and Tsoulas2004; De Cat Reference De Cat2004, Reference De Cat2007). In this section I will show that the ungrammaticality of (13d) follows straightforwardly from the movement account of when clauses.

Observe first of all that it would not be correct to say that when clauses do not allow any kind of fronting, and that they lack the positions in the CP domain which are required for fronting.Footnote 7When clauses display an adjunct/argument asymmetry with respect to the material that precedes the canonical subject position: while argument fronting is excluded (13d), circumstantial adjuncts can precede the subject (13e):

  1. (13)

    1. (e) When last year she started to write this column, I thought she would be fine.

Such asymmetries are typically found in extraction contexts. In (14a) subject extraction across the adjunct last year is grammatical; in (14b) subject extraction across the argument this new column is ungrammatical; in (15a) the indirect object to whom crosses an adjunct; in (15b) an intervening direct object gives rise to a severe degradation in grammaticality. (16) shows that the adjunct/argument asymmetry is relevant for long relativization.

  1. (14)

    1. (a) This is the author who last year started to write a new column.

    2. (b) *This is the author who this new column started to write last year.

  2. (15)

    1. (a) These are the patients to whom, in the present circumstances, we should give the cooked vegetables.

    2. (b) *??These are the patients to whom, the cooked vegetables, we should give in the present circumstances.

  3. (16)

    1. (a) These are the patients to whom Mary suggested that, in the present circumstances, we should give the cooked vegetables.

    2. (b) *??These are the patients to whom Mary suggested that, the cooked vegetables, we should give in the present circumstances.

These data show that while fronted arguments act as interveners for (long) relativization, initial adjuncts do not intervene in relativization. For reasons of space I cannot go into a general discussion of intervention here; see Haegeman (to appear).

On an account according to which adverbial when clauses are free relatives, the ungrammaticality of (13d) above is accounted for in the same way as that of the (b) examples of (14–16). For instance, in the same way that the relative operator who cannot cross the argument this new column in (14b), the temporal relative operator when cannot cross the argument this column in (13d). (13f) represents the structure:

  1. (13)

    1. (f) * [CP whenj her columni [TP she [VP began to write ti] tj]]

Conversely, in the same way that the relative operator who can cross the adjunct last year in (14a), the temporal relative operator when can cross the adjunct last year in (13e).Footnote 8

3.2 Preposing in when clauses

Not only argument fronting is incompatible with the left periphery of adverbial clauses. Other fronting operations are also ungrammatical: (17) illustrates Locative Inversion (Culicover & Levine Reference Culicover and Levine2001 and references cited there), (18) illustrates preposing around be (Hooper & Thompson Reference Hooper and Thompson1973: 467; Emonds Reference Emonds1976), (19) illustrates VP preposing (Hooper & Thompson Reference Hooper and Thompson1973: 466; Emonds Reference Emonds, Adger, Cat and Tsoulas2004: 78), (20) illustrates Negative Inversion (Hooper & Thompson Reference Hooper and Thompson1973: 467; Emonds Reference Emonds1976; Culicover Reference Culicover and Delfitto1991; Haegeman Reference Haegeman, Horn and Kato2000; Sobin Reference Sobin2003, and references cited there). All of these are ungrammatical in adverbial clauses.

  1. (17)

    1. (a) We were all much happier when the Browns lived upstairs.

    2. (b) *We were all much happier when upstairs lived the Browns.

      (Hooper & Thompson Reference Hooper and Thompson1973: 496 (their (253))

  2. (18)

    1. (a) When the company directors were present at the meeting, nothing of substance was ever said.

    2. (b) *When present at the meeting were the company directors, nothing of substance was ever said.

  3. (19)

    1. (a) When you've passed these exams, you'll get the degree.

    2. (b) *When passed these exams you have, you'll get the degree.

  4. (20)

    1. (a) When I did not agree with a single proposal, I left the meeting.

    2. (b) *??When not a single proposal did I agree with, I left the meeting.Footnote 9

If one takes when clauses to be temporal free relatives, the ungrammaticality of the (b) examples in (17–20) can again be ascribed to an intervention effect: the temporal operator when illicitly crosses another fronted constituent. Locative Inversion in (17) preposes a locative constituent to the left periphery of the clause, i.e. to the left of the canonical subject position (cf. Den Dikken & Naess Reference Dikken and Naess1993; Rizzi & Shlonsky Reference Rizzi, Shlonsky and Frascarelli2006). As seen in (17c) below, fronting of the relevant locative constituent gives rise to an intervention effect for interrogative when. A similar intervention effect is found in (18b): preposing around be can be analysed along the same lines as Locative Inversion, with the fronting of a predicate to the left periphery (Carre Reference Carre2006). As shown in (18c), interrogative when cannot cross the fronted predicate. VP preposing fronts a contextually salient VP to the left periphery of the clause. Again the ungrammaticality of (19b) and of (19c) below is the result of an intervention effect. Finally, Negative Inversion too implicates movement to the left periphery of the clause and hence we predict that this should be incompatible with additional movement. Thus (20b) and (20c) are also the result of an intervention effect.Footnote 10

  1. (17)

    1. (c) *I was wondering when upstairs lived the Browns.

  2. (18)

    1. (c) *I am trying to remember when present at the meeting were the company directors.

  3. (19)

    1. (c) *I was wondering when passed these exams he had.

  4. (20)

    1. (c) *I am trying to remember when not a single proposal did they agree with.

Argument fronting, Locative Inversion, preposing around be, VP preposing and Negative Inversion all block leftward movement of an operator. If when clauses are derived by leftward movement of the temporal operator when, then such clauses are expected to be incompatible with the fronting operations listed here.

3.3 Extending the proposal: temporal clauses

3.3.1 A generalization of the account

One of the initial arguments in favour of a movement derivation which one finds cited in the literature was the observation that adverbial clauses allow low construal of the temporal operator, as illustrated in (5ii), which is taken to reflect long movement. In English, ambiguity due to the high/low construal is also found with temporal adverbial clauses introduced by until, before, after and (temporal) since (Geis Reference Geis1970; Larson Reference Larson1987, Reference Larson1990). This is shown in (21):

  1. (21)

    1. (a) I can't leave until John says I can leave. (Based on Larson Reference Larson1990: 170 (2b))

    2. (b) I saw Mary in New York before/after John said that she left.

      (Larson Reference Larson1987: 261 (45a))

    3. (c) I haven't been there since I told you I was there. (Larson Reference Larson1990: 170 (2d))

Recall from section 2.3 that according to Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004), temporal adjuncts such as after Christmas are PPs in which a temporal P (after) selects Christmas, a temporal DP, their ZeitP, as its complement. When clauses are relative clauses in which the temporal operator when is moved from SpecAspP to SpecCP, and which modify a ZeitP (which may be overt as in (10a) or null as in (10b)), which is itself the complement of a (possibly null) temporal P. I refer to section 2.3 for more details. Temporal adverbial clauses introduced by a preposition (until, after, before, since) are PPs in which the (overt) temporal P selects a null ZeitP. ZeitP is again modified by a relative clause, here derived by the leftward movement of a null temporal operator. A sentence such as (22a) would have the representation in (22b), though see also Dubinsky & Williams (Reference Dubinsky and Williams1995) for an alternative which remains compatible with a movement analysis.

  1. (22)

    1. (a) Until Zooey arrived, . . .

    2. (b)

Once again the analysis leads to the prediction that additional fronting operations will give rise to intervention effects. (22c–e) show that in until clauses, adjunct fronting is possible while argument fronting and VP preposing are ungrammatical:

  1. (22)

    1. (c) I used to work on an old desktop until one day that was stolen and . . .

    2. (d) *I won't take time off until this handout I have finished.

    3. (e) *You won't be allowed in until passed these exams you have.

3.3.2 Two problems

The operator movement account excludes argument fronting in temporal clauses because the fronted argument would block the movement of the operator from within TP to the clause typing position (CP). Thus temporal adverbial clauses are predicted to be incompatible with argument fronting and with MCP. Note that temporal clauses introduced by before, after and since differ from when clauses and from until clauses in that fronting is more severely restricted: not only argument fronting (23), but also adjunct fronting (24), is excluded:

  1. (23)

    1. (a) *Before this baker's shop I discovered last year, I used to bake my own bread.

    2. (b) *Since this baker's shop she discovered, she has never bought her bread anywhere else.

  2. (24)

    1. (a) *Before last year I discovered this baker's shop, I used to bake my own bread.

    2. (b) *Since last year she discovered this baker's shop, she has never bought her bread anywhere else.

These data suggest that such adverbial clauses have a reduced left periphery. It is not clear which property of the clauses is at the basis of this further restriction. One option is that the restriction is due to the fact that before, after and since are themselves prepositions. But since until is also a preposition and until clauses do allow adjunct fronting (22a), it cannot be the prepositional nature of before, after and since that is at issue. At this point I do not have an account for this and I hope to return to it in future work. See McCloskey (Reference McCloskey, Zanuttini, Campos, Herburger and Portner2006) for discussion.

While clauses pose another problem (already noted by Geis Reference Geis1975). Recall that one of the prime reasons for adopting the movement account in adverbial clauses was the fact that low construal is available (cf. the discussion of (5,6) in section 2.2). In this respect, it has to be noted that low construal is unavailable with temporal while (Geis Reference Geis1970, Reference Geis1975; Stump Reference Stump1985; Larson Reference Larson1990: 174, his (11a)):

  1. (25) I didn't see Mary in New York while she said she was there.

At the same time, like when clauses, while clauses are incompatible with argument fronting (26a), while remaining compatible with adjunct preposing (26b):

  1. (26)

    1. (a) *While this paper I was revising last night, I thought of another analysis.

    2. (b) While last night I was revising my paper, I thought of another analysis.

The argument/adjunct asymmetry in (26) suggests that there is an intervention effect, and hence that while clauses are derived by operator movement to the left periphery. Such an analysis would account for the fact that temporal while clauses are generally incompatible with other MCP. (27a) shows that temporal while clauses are incompatible with Locative Inversion, (27b) shows that they resist preposing around be, and (27c) shows they are incompatible with VP-preposing.Footnote 11

  1. (27)

    1. (a) *While upstairs were three plain clothes policemen, the thieves emptied the house.

    2. (b) *While present at the party were some under-age children, they showed X-rated films.

    3. (c) *While watching them a plain clothes policeman was, they emptied the house.

The argument/adjunct asymmetry observed for temporal while clauses supports the movement account. The absence of low construal can therefore not be taken as evidence against such an account, and will have to find an independent explanation. At this point I have nothing further to offer by way of an explanation for the absence of low construal with while but I will return to this in future work (see also section 5). For discussions of the availability of high/low-construal readings in temporal adverbial clauses, see also Stephens (Reference Stephens, Butt and King2006) on Norwegian når (high/low construal) vs da (no low construal) and Lipták (Reference Lipták2005) on Hungarian (see below) and on Serbian, among others.

4 Peripheral adverbial clauses

So far I have only discussed temporal adverbial clauses which modify the event expressed in the main clause. In Haegeman (Reference Haegeman2003a) I argue extensively that such clauses are truly embedded and integrated to the main clause and I label such integrated adverbial clauses ‘central’ adverbial clauses. However, not all clauses labelled as ‘adverbial clauses’ are structurally integrated. Often what appears to be the same conjunction may introduce a fully integrated temporal clause but it may also introduce a clause that is less closely integrated to the associated clause. In Haegeman (Reference Haegeman2003a) I label adverbial clauses that are not syntactically integrated ‘peripheral’ adverbial clauses.Footnote 12

Concessive/contrastive while clauses are a case in point. Unlike temporal while clauses, which are temporally subordinated to the main clause, concessive/contrastive while clauses are temporally independent from the main clause. In the attested example (28a) the first while clause is contrastive/concessive and temporally independent, but the second one is temporal, and temporally subordinated, witness the use of the present tense decides with future reference in the latter. Concessive clauses introduced by (al)though and whereas, as well as since clauses expressing reason − with since roughly corresponding to ‘given that’ − also belong to the class of peripheral adverbial clauses.Footnote 13

  1. (28)

    1. (a) Whileconc the lawsuit [challenging the legitimacy of lethal injection; LH] probably won't stop the use of lethal injection altogether, it will certainly delay its use whiletemp the supreme court decides what to do. (The Guardian G2, 12.12.2003, p. 4, col. 4)

Peripheral adverbial clauses serve to bring to the fore a contextually salient proposition which constitutes the privileged discourse background for the processing of the proposition expressed in the associated clause. The juxtaposition of the associated clause and the peripheral clause gives rise to contextual effects in the sense of Sperber & Wilson (Reference Sperber and Wilson1986). For instance, in (28a) the peripheral while clause provides complementary information to the assertion that the ongoing lawsuit will delay the use of lethal injection.

The peripheral while clause is not interpreted as a temporal modifier of the associated clause. As shown by the use of will in both the peripheral clause and the associated clause in (28a), there is no temporal dependency between the event expressed in the peripheral while clause and that in the main clause. The temporal independence of the peripheral while clause and the associated clause is also clear in (28b), in which both the while clause and the associated clause have their own temporal adjunct, on Sunday and yesterday respectively:

  1. (28)

    1. (b) While on Sunday Dr Abdullah said the alliance had not discussed with Washington the possibility of US ground forces being deployed on alliance-held territory, yesterday he said that both sides were discussing possible US air cover for a mojahedin ground offensive, probably on Kabul. (The Guardian, 9.10.2001, p. 3, col. 3)

Since they are not temporal modifiers, but are used to represent contextually salient propositions, one option is to postulate that peripheral adverbial clauses such as, for instance, while clauses are not introduced by a temporal operator and hence that they are not derived by movement of such an operator. Either peripheral adverbial clauses lack an operator in their left periphery or, alternatively, if there is an operator, this is not a temporal operator; it has been inserted directly in the CP domain, and hence, crucially, it has not moved from TP to CP.

If there is no operator movement in peripheral adverbial clauses, then there should be no intervention effect with respect to other fronting operations, and we predict that they will be compatible with Main Clause Phenomena. This is confirmed by the examples in (29).Footnote 14 In (29a), from Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985), an argument is fronted in a concessive while clause; in the attested (29b) the object of contain is fronted in a concessive though clause. The attested (29c) illustrates the application of Locative Inversion in a concessive while clause:

  1. (29)

    1. (a) His face not many admired, while his character still fewer felt they could praise. (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1378)Footnote 15

    2. (b) We don't look to his paintings for common place truths, though truths they contain none the less. (The Guardian, G2, 18.2.2003, p. 8, col. 1)

    3. (c) At one end of the tranquil valley stood the towering form of Tirch Mir, the highest mountain in the Hindu Kush, while at the other swept a row of icy peaks. (Travel, The Guardian, 13.1. 2007, p. 2, col. 2)

  1. (30) is an attested nontemporal since clause, which displays argument fronting:

    1. (30) It is amazing how this view could have spread about someone who changed the image of causes like Aids and landmines, and in doing so showed a possible new role for the royals. It is particularly ironic since so much of what Diana did for her fellow humans she did with no concern for publicity whatsoever. (The Guardian, G2, 31.8.2004, p. 9, col. 2)

Observe that an alternative account is conceivable which will allow for Main Clause Phenomena in peripheral adverbial clauses, while maintaining that the latter are derived by operator movement. In temporal adverbial clauses, main clause phenomena such as argument fronting are excluded because the fronted constituent will intervene on the movement path of the temporal operator, which originates within the TP domain of the adverbial clause (SpecAspP according to Demirdache & Uribe-Etxebarria Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004), and moves to the clause typing projection (CP). The intervening fronted arguments occupy a position higher than the starting point of the moved temporal operator and lower than its landing site.

Peripheral while clauses could be paraphrased as ‘while it is true that’, ‘while we can admit that’. It has been argued that such peripheral while clauses are associated with speaker anchoring and with illocutionary force (for more discussion see Haegeman Reference Haegeman2003a, Reference Haegeman, Zanuttini, Campos, Herburger and Portner2006a, Reference Haegeman2006b, Reference Haegeman2007, and also Komagata Reference Komagata2003). If the encoding of speech acts is associated with a specific projection, then peripheral while clauses might be argued to contain a temporal operator, but this operator will be associated with the speech act. There is a lot of discussion on the syntactic representation of speech acts, but there is a convergence that speech act is encoded by a functional projection high in the left periphery (see Ernst Reference Ernst2002: 70ff; Speas & Tenny Reference Speas, Tenny and Sciullo2003; Meinunger Reference Meinunger2004; Hill Reference Hill2007; Abraham Reference Abraham2008). One might then propose that in peripheral adverbial clauses a temporal operator associated with a high Speech Act phrase in the left periphery moves to a yet higher clause-typing position in the left periphery. If the relevant operator movement takes place in a syntactic domain which does not overlap with the domain in which topicalization and focalization apply, there will not be any intervention effects.

5 Conditional clauses as free relatives (Bhatt & Pancheva Reference Bhatt and Pancheva2002, Reference Bhatt and Pancheva2006)

5.1 Argument/adjunct asymmetries and a movement analysis

We have seen that temporal when clauses display an asymmetry between arguments and adjuncts with respect to fronting to the left of the canonical subject position: (13d) and (13e) are repeated for convenience in (31):

  1. (31)

    1. (a) *When this column she started to write last year, I thought she would be fine.

    2. (b) When last year she started to write this column, I thought she would be fine.

A similar asymmetry is found in conditional if clauses: (32a) is attested:

  1. (32)

    1. (a) If on Monday the share price is still at the current level then clearly their defence doesn't hold much water. (The Observer, 11.7.2004, business, p. 22, col. 5)

    2. (b) *If these exams you don't pass, you won't get the degree.

Other fronting operations too are illicit in conditional clauses: (33a) illustrates Locative Inversion, (33b) illustrates preposing around be, (33c) illustrates VP preposing and (33d) illustrates Negative Inversion.

  1. (33)

    1. (a) *If upstairs live his parents things will be much simpler.

    2. (b) *If present at the party are under-age children, they won't be able to show the X-rated films.

    3. (c) *If passed these exams you had, you would have had the degree.

    4. (d) *If not a single book did you read, you'll never pass the exam.

The adjunct/argument asymmetry in (32) and the fact that MCP are ungrammatical (33) suggest that conditional clauses too are derived by operator movement. If a movement analysis is intuitively appealing for temporal adverbial clauses, it may at first sight appear less plausible for conditional clauses. Yet, Bhatt & Pancheva (Reference Bhatt and Pancheva2002, Reference Bhatt and Pancheva2006) argue for the derivation of conditional clauses by movement of an operator to SpecCP. They say: ‘Our proposal that [conditional clauses] are interpreted as free relatives amounts to the claim that they are definite descriptions of possible worlds’ (Bhatt & Pancheva Reference Bhatt and Pancheva2002: 15; Reference Bhatt and Pancheva2006: 655). For a similar proposal see also Arsenijević (Reference Arsenijević2006). A conditional clause such as (34a) would be derived by the leftward movement of a World operator (34b):

  1. (34)

    1. (a) If John arrives late

    2. (b) [CP OPw C° [John arrives late in tw]]

      (Bhatt & Pancheva Reference Bhatt and Pancheva2002: 15 (60); Reference Bhatt and Pancheva2006: 655 (46))

As was the case with temporal clauses, the intervention effects discussed above offer empirical support for Bhatt & Pancheva's analysis.

5.2 Absence of low construal

Bhatt and Pancheva observe that, unlike temporal clauses, conditional clauses do not allow the low construal found with temporal adverbial clauses (Geis Reference Geis and Geis1985; Bhatt & Pancheva Reference Bhatt and Pancheva2002, Reference Bhatt and Pancheva2006): while when in (35a) allows for high or low construal, the latter reading is unavailable for the conditionals in (35b) or (35c):

  1. (35)

To account for the absence of low construal in the conditional clause, Bhatt & Pancheva (Reference Bhatt and Pancheva2002, Reference Bhatt and Pancheva2006) propose that conditionals involve the creation of a World variable (‘w’), which needs to be locally bound. I refer to their work for more discussion.Footnote 16

The absence of low construal in conditional clauses is illustrated in an interesting way in West Flemish (WF). WF oa allows for both a temporal (‘when’) and a conditional (‘if’) reading. In the former reading (36a, b) low construal is available, while low construal is unavailable if oa has the conditional interpretation (36c):

  1. (36)

These data suggest that it is not the conjunction (oa) as such which is at stake but the nature of the moved operator. See also Bhatt & Pancheva (Reference Bhatt and Pancheva2002, Reference Bhatt and Pancheva2006) for a similar observation in relation to German.

5.3 Support for the movement account

The ban on MCP in conditional clauses discussed in section 5.1 can be interpreted as empirical evidence in favour of the movement analysis of conditional clauses. Some support for the movement derivation of conditional clauses may be drawn from their syntactic similarity to yes/no questions. Conditional clauses and yes/no questions are parallel in some respects. Embedded yes/no questions are introduced by the conjunction if, the same conjunction that is used for conditionals. As shown by (35c) above, subject−auxiliary inversion (SAI) may be used to derive a conditional clause, and SAI is also used to derive a (main) yes/no question, as in (37a) below.

The Germanic Verb Second (V2) languages offer some support for postulating a null interrogative operator in direct yes/no questions. The Dutch analogue of (37a), (37b), shows that in V2 languages, which have the finite verb in second position in root clauses, direct yes/no questions constitute an apparent exception to the V2 constraint in that here the fronted verb is the first constituent. This ‘exception’ can be eliminated if one adopts the view that yes/no questions contain an initial abstract interrogative operator in their left periphery (37c), which functions as the first constituent. Hence the finite verb is in second position. If we also assume that the relevant operator originates in a lower position, then yes/no questions are derived by movement.Footnote 17 The conditional clause in (37d), which is formally identical to the yes/no question in (37b), could then also be said to be derived by movement:

  1. (37)

    1. (a) Had he said he would leave?

If direct yes/no questions are derived by the movement of an interrogative operator to the left periphery, we may assume that the same interrogative operator is present in indirect yes/no questions and the movement analysis can hence be extended to the derivation of indirect yes/no questions. Accordingly, the embedded interrogative in (38a) would be derived by moving an operator to the left periphery (38b). Once again, the same derivation could be appealed to for the conditional analogue of this, with leftward movement of a World operator.

  1. (38)

    1. (a) I wonder if he said he would leave

    2. (b) [CP OPw if [he said he would leave in tw]]Footnote 18

Like conditionals, yes/no questions do not allow for low construal of the operator.Footnote 19 In (38a), for instance, the question bears on the polarity of the matrix clause (‘he said’) and not on the embedded clause (‘he would leave’).

With an operator movement account for yes/no questions and for conditional if-clauses, the ungrammaticality of argument fronting in indirect yes/no questions will also follow. Once again a fronted argument (such books in (39a), this book in (39b)) will block the movement of the yes/no operator to the left periphery.

  1. (39)

    1. (a) *Bill asked if such books John only reads at home. (Schachter Reference Schachter, Stowell and Wehrli1992: 108 (16a))

    2. (b) ??/*John knows whether this book Mary read. (Maki et al. Reference Maki, Kaiser and Ochi1999: 9, note 8, their (i))Footnote 20

On the basis of the data discussed above we can adopt Arsenijević's (2006) proposal that conditional clauses are the relative variant of yes/no questions (see also Haegeman in preparation):

I analyze conditionals as yes−no relative clauses: a [s] restrictive relative clauses in which the truth value of a proposition is restricted. The proposition represented by the conditional clause restricts the set of worlds compatible with the proposition represented by the head clause. Syntactically, the locus of modification is a functional projection called WorldP, the projection that specifies the truth value of clauses by containing the feature world with a value, [actual] or [possible]. (Arsenijević Reference Arsenijević2006)

6 Factive clauses

As discussed by Emonds (Reference Emonds1970, Reference Emonds1976, Reference Emonds, Adger, Cat and Tsoulas2004), Hooper & Thompson (Reference Hooper and Thompson1973), Maki et al. (Reference Maki, Kaiser and Ochi1999) and Heycock (Reference Heycock2006), among others, nonassertive complement clauses (see Hooper & Thompson Reference Hooper and Thompson1973; Haegeman Reference Haegeman2006b), and in particular complements of factive predicates, resist MCP, at least for many speakers.Footnote 21 In (40a) the complement clause of regret resists argument fronting; in (40b) VP preposing is illicit in the clause associated with bother; in (40c) the complement of forget is incompatible with preposing around be, and in (40d) the complement clause of surprise resists negative inversion. For more data see Hooper & Thompson (Reference Hooper and Thompson1973: 479ff.).Footnote 22

  1. (40)

    1. (a) *John regrets that this book Mary read.Footnote 23 (Maki et al. Reference Maki, Kaiser and Ochi1999: 3, their (2c))

    2. (b) *Sally plans for Gary to marry her, and it bothers me that marry her he will. (Hooper & Thompson Reference Hooper and Thompson1973: 479)

    3. (c) *I forgot that playing in the concert was Arthur Rubinstein.

    4. (d) *He was surprised that never in my life had I seen a hippopotamus.

The question arises whether the incompatibility of such clauses with argument fronting can also be related to an intervention effect. Two elements in the literature point to a tentatively positive answer to this question. On the one hand, various authors have postulated that factive clauses contain a propositional operator in their left periphery (see Melvold Reference Melvold, Cheng and Demirdache1991; Bianchi Reference Bianchi2000: 95). Secondly, comparative evidence may suggest that the relevant operator is moved from a lower position (see Collins Reference Collins1994; Aboh Reference Aboh, Brugè, Giusti, Nicolà Munaro, Schweikert and Turano2005). I refer the reader to the relevant discussions.

7 Summary

This article explored the movement analysis of adverbial clauses which has repeatedly been put forward in the literature, according to which temporal adverbial clauses are derived by the movement of a temporal operator from within the TP domain to a clause-typing position (SpecCP). It is shown that this accounts for the absence of a number of fronting phenomena in adverbial clauses in English. Structures which display argument fronting are predicted to be ungrammatical due to the fact that the fronted argument would occupy a position in the left periphery which would block an IP-internal operator moving to the same position. The article shows that the movement account can be extended to conditional clauses and to other domains which resist fronting operations in English.

Footnotes

2 I use the symbol ‘t’ for ‘trace’ to indicate the extraction site; the trace is co-indexed with the extracted constituent. I assume that a trace is a copy of the moved constituent. An alternative representation is that in (ia):

  1. (i) (a) [CP when [TP I. . . . [VP heard this song] when]]

Alternatively one might assume that when is a genuine complementizer: it spells out a C-head, and a null operator moves from the TP domain into the specifier of C, as shown in (ib):

  1. (i)

    1. (b) [CP OPi [C when [TP I. . . . [VP heard this song] ti]]]

For the relevance of the null operator analysis see also section 3.3.1.

3 For a similar case in Norwegian see Stephens (Reference Stephens, Butt and King2006), and for a similar use of þonne (‘then’) in Old English see Declerck (Reference Declerck1997: 51).

4 In the current terminology the projection of a noun is referred to as a Determiner Phrase (DP), the underlying assumption being that the head of the projection is not the lexical noun but the determiner.

5 Dutch toen (‘then’) may also function as a relativizer with a head noun:

  1. (i)

6 The anonymous reviewer asks what distinguishes when and whenever: in (ia/b) they seem to alternate, but they cannot always alternate, as shown in (ic/d):

  1. (i)

    1. (a) When I am depressed I eat chocolate.

    2. (b) Whenever I am depressed I eat chocolate.

    3. (c) The days when I am depressed I eat chocolate.

    4. (d) *The days whenever I am depressed I eat chocolate.

Put differently: why can when introduce both an antecedentless (‘free’) relative and one with an antecedent, while whenever can only introduce an antecedentless relative? The answer to this question depends on the particular analysis adopted for free relatives.

In terms of Demirdache & Uribe-Etxebarria's (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004) analysis (i) is indeed a puzzle. In their account, when occupies the same position in the ‘free relative’ in (ia) as in the ‘ordinary’ relative (ic), namely a specifier position in the C domain (see the discussion in the text). If, by analogy, whenever in (id) occupies the same specifier of the C domain inside the relative, the incompatibility with a nominal antecedent would have to be semantic: one might propose that the quantificational element ever is incompatible with the determiner contained in the antecedent.

If, contrary to Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004), one assumes that the wh-element introducing the antecedentless relative actually moves from a specifier position in the C-domain of the relative into the domain of the nominal antecedent, say to D, then in antecedentless/free relatives there will be no position available for a nominal antecedent with its own determiner. For (id) one could propose that ever occupies the D-position of the antecedent domain (cf. every) and when moves to it. Thus (id) would be ruled out syntactically because whenever would itself be in the position occupied by the antecedent. The latter proposal entails that, though superficially identical, the free relative when clause in (ia) and the headed relative when clause in (ic) have a different structure, with when occupying the specifier of the CP in (ia) and moving to the antecedent domain in (ic).

For full discussion of the two types of analyses of free relatives I refer to Van Riemsdijk (Reference Riemsdijk2006) and the references cited.

7 The fact that clitic left dislocation is possible in French (i), also shows that adverbial clauses must potentially have a left periphery.

  1. (1)

See Haegeman (Reference Haegeman2007). Not all French speakers accept (i). Thanks to Michal Starke and Amélie Rocquet for judgements.

8 The question obviously arises why circumstantial adjuncts do not give rise to intervention effects. This issue is beyond the scope of the present article. One possible account is that circumstantial adjuncts are base-generated in the left peripheral position. For some discussion see Haegeman (Reference Haegeman2003b, to appear) and the references cited there. For the semantic implications of the base generation analysis see also Cinque (Reference Cinque1990: 92–4); Bianchi, Squartini & Bertinetto (Reference Bianchi, Squartini, Bertinetto, Bertinetti, Bianchi, Higginbotham and Squartini1995); Bianchi (Reference Bianchi2000).

9 As pointed out by the anonymous reviewer for English Language and Linguistics, (20b) is not as bad as (19b). This needs further study. One point that differentiates the two patterns is that while negative inversion implicates a focalized negative constituent (Haegeman Reference Haegeman, Horn and Kato2000), VP preposing implicates the fronting of a VP that is contextually given (cf. Ward Reference Ward1988). In this respect VP preposing is more like topicalization and could be argued to be ‘D-linked’. It has been shown independently that the intervention effects of topicalized constituents are stronger than those of focused constituents in the left periphery. So typically, fronted topics in embedded domains create islands for extraction while negative inversion in an embedded clause leads to a weak island:

  1. (i)

    1. (a) *??These are the patients to whom Mary suggested that the cooked vegetables we should give in the present circumstances.

    2. (b) ?These are the patients to whom Mary said that on no account should we give the cooked vegetables.

The examples raise questions about the status of when in terms of intervention. I will return to this issue in later work. See also Haegeman (to appear) for an outline of the theory of intervention to be assumed.

Anticipating the discussion: with respect to text examples (40) it will again be observed that negative inversion leads to a lesser degradation. See also notes 11 and 22 below.

10 There is a substantive literature on the semantic and pragmatic effects of the various operations illustrated here. See Ward (Reference Ward1988) and much later work. I do not go into this here and restrict myself to the discussion of the syntax of the relevant constructions.

11 An anonymous reviewer points out that there is a difference between (27a) and (27b) on the one hand and (27c) on the other, in that the latter is far more degraded than the former two. I do not have a precise account to offer for this contrast. It shows clearly that VP fronting is like argument topicalization in creating a strong intervention effect (see note 9), and that the intervention effect due to Locative Inversion is weaker. A full account of this contrast will depend on a better understanding of Locative Inversion.

12 See Haegeman (Reference Haegeman2003a, Reference Haegeman, Chiba, Ogawa, Fuiwara, Yamada, Koma and Yagi1991/2009), Komagata (Reference Komagata2003) and references cited for a survey and for discussion.

A fully fledged discussion of the syntactic status of peripheral adverbial clauses is beyond the scope of this article. With respect to their syntactic relation to the associated clause various options are available. Haegeman (Reference Haegeman, Chiba, Ogawa, Fuiwara, Yamada, Koma and Yagi1991/2009) argues for an ‘orphanage’ approach according to which such adverbial clauses are not syntactically integrated. See also Haegeman, Shaer & Frey (to appear). An alternative option would be to assume with Gaertner (Reference Gaertner2001: 105–8) that such peripheral adverbial clauses have a paratactic relation to the associated clauses which is mediated by a specialized functional projection πP. In line with Gaertner's (Reference Gaertner2001: 107) proposal for German weil (‘because’), which introduces adverbial clauses with root properties, one might then assume that the ‘conjunction’ which introduces the peripheral adverbial clauses is merged in π.

13 In this respect Hornstein (Reference Hornstein1990) writes: ‘There is a secondary conjunctive interpretation that all these connectives (as, while, when) shade into. They get an interpretation similar to and in these contexts. And is not a temporal connective, and these conjunctive interpretations do not tell against the theory [of temporal subordination and complex tense structures]’ (Hornstein (Reference Hornstein1990: 206: note 19).

14 See also Komagata (Reference Komagata2003: 302) for additional examples.

15 The anonymous reviewer points out that (29a) is stylistically marked. I have nothing to say about this at this point.

16 Bhatt and Pancheva observe that low construal is available with conditionals formed by relativization:

  1. (i)

17 For postulating an operator in the CP domain of yes/no questions, see Barbiers (Reference Barbiers and Wurff2007: 102–3), for arguments from Dutch, and Den Dikken (Reference Dikken2006: 729). But see Roberts & Roussou (Reference Roberts, Roussou and Svenonius2002: 41) for a different viewpoint.

18 For the movement analysis, see among others Larson (Reference Larson1985), Den Dikken (Reference Dikken2006: 729), with evidence from the distribution of either in indirect questions introduced by whether and if.

19 In a different context, this point was also made in Ingham (Reference Ingham2008).

20 Maki et al. (Reference Maki, Kaiser and Ochi1999: 9, note 8), point out that (39b) is ‘marginal in American English and almost grammatical in British English’. The (British) speakers I consulted considered it ungrammatical.

21 See Maki et al. (Reference Maki, Kaiser and Ochi1999) and Haegeman (Reference Haegeman2006b) for discussion of speaker variation.

22 As pointed out by the anonymous reviewer, negative inversion leads to a lesser degradation. We have seen similar effects above. As suggested in notes 9 and 11, I propose that negative inversion is focalization (see Haegeman Reference Haegeman, Horn and Kato2000) and the other fronting operations in (40) can be assimilated to topicalization, which is known to lead to a stronger degradation in English.

23 On factive verbs and semifactives see Hooper & Thompson (Reference Hooper and Thompson1973: 480ff.) and also the discussion in Maki et al. (Reference Maki, Kaiser and Ochi1999: 9, esp. note 8).

References

Aboh, Enoch. 2005. Deriving relative and factive constructions in Kwa. In Brugè, Laura, Giusti, Giuliana, Nicolà Munaro, , Schweikert, Walter & Turano, Gina (eds.), Contributions to the thirtieth Incontro di Grammatica Generativa, 265–85. Venice: Libreria Editrice Cafoscarina.Google Scholar
Abraham, Werner. 2008. Illocutive Force is speaker and information source concern. Ms. University of Vienna/Munich.Google Scholar
Arsenijević, Boban. 2006. The correlative construction as a type of conditional clause. GLIF seminar, University Pompeu Fabra. 19 May 2006.Google Scholar
Barbiers, Sjef. 2007. On the periphery of imperative and declarative clauses in Dutch and German. In Wurff, Wim van der (ed.), Imperative clauses in generative grammar, 95–112. Amsterdam and Philadelphia: John Benjamins.CrossRefGoogle Scholar
Bhatt, Rajesh & Pancheva, Roumanya. 2002. A cross-constructional analysis of if clauses. Paper presented at the Syntax Seminar: interface in the CP domain, Zentrum fur Allgemeine Sprachwissenschaft, Berlin, 9 March 2002.Google Scholar
Bhatt, Rajesh & Pancheva, Roumanya. 2006. Conditionals. In Everaert & Van Riemsdijk (eds.), vol. 1: 638–87.Google Scholar
Bianchi, Valentina. 2000. On time adverbials. Rivista di Linguistica 12, 77–106.Google Scholar
Bianchi, Valentina, Squartini, Mario & Bertinetto, Pier Marco. 1995. Perspective point and textual dynamics. In Bertinetti, Pier Marco, Bianchi, Valentina, Higginbotham, James & Squartini, Mario (eds.), Temporal reference, aspect and actionality, vol. 1: Semantic and syntactic perspectives, 309–24. Turin: Rosenberg and Sellier.Google Scholar
Bresnan, Joan & Grimshaw, Jane. 1978. The syntax of free relatives in English. Linguistic Inquiry 9, 331–91.Google Scholar
Carre, Jonathan. 2006. Predicate around be constructions in English (A look at PABs and locative PABs). Student's paper: Master 2: linguistique anglaise. Université Charles de Gaulle, Lille III.Google Scholar
Cinque, Guglielmo. 1990. Types of A′ dependencies. Cambridge, MA: MIT Press.Google Scholar
Citko, Barbara. 2000. On the syntax and semantics of Polish adjunct clauses. Journal of Slavic Linguistics 8, 1–38Google Scholar
Collins, Chris. 1994. The factive construction in Kwa. Travaux de recherche sur le créeole haitien 23, 31–65. Université du Quebec à Montreal.Google Scholar
Culicover, Peter. 1991. Topicalization, inversion and complementizers in English. In Delfitto, Dennis et al. (eds.), Going Romance and beyond (OTS Working Papers), 1–44. Utrecht: University of Utrecht.Google Scholar
Culicover, Peter W. & Levine, Robert D.. 2001. Stylistic inversion in English: A reconsideration. Natural Language and Linguistic Theory 19, 283–310.CrossRefGoogle Scholar
De Cat, Cécile. 2004. Dislocation without movement. ZAS papers in Linguistics 35, 1, Proceedings of the dislocated elements workshop, November 2003, 77–109.CrossRefGoogle Scholar
De Cat, Cécile. 2007. French dislocation without movement. Natural Language and Linguistic Theory 25, 485–534.CrossRefGoogle Scholar
Declerck, Renaat. 1997. When-clauses and temporal structure. London: Routledge.Google Scholar
Demirdache, Hamida & Uribe-Etxebarria, Myriam. 2004. The syntax of time adverbs. In Guéron, Jacqueline & Lecarme, Jacqueline (eds.), The syntax of time, 143–80. Cambridge, MA: MIT Press.CrossRefGoogle Scholar
Dikken, Marcel den & Naess, Alma 1993. Case dependencies: The case of predicate inversion. Linguistic Review 10, 303–36.Google Scholar
Dikken, Marcel den. 2006. Either float and the syntax of co-or-dination. Natural Language and Linguistic Theory 24, 689–749.CrossRefGoogle Scholar
Dubinsky, Stanley & Williams, Kemp. 1995. Recategorization of prepositions as complementizers: The case of temporal prepositions in English. Linguistic Inquiry 26, 125–37.Google Scholar
Emonds, Joseph. 1970. Root and structure-preserving transformations. PhD dissertation, MIT.Google Scholar
Emonds, Joseph. 1976. A transformational approach to English syntax. New York: Academic Press.Google Scholar
Emonds, Joseph. 2004. Unspecified categories as the key to root constructions. In Adger, David, Cat, Cécile De & Tsoulas, Georges (eds.), Peripheries, 75–121. Dordrecht: Kluwer.CrossRefGoogle Scholar
Enç, Murvet. 1987. Anchoring conditions for tense. Linguistic Inquiry 18, 633–57.Google Scholar
Ernst, Thomas. 2002. The syntax of adjuncts. Cambridge: Cambridge University Press.Google Scholar
Everaert, Martin & Riemsdijk, Henk Van (eds.). 2006. The Blackwell companion to syntax. Oxford and Boston: Blackwell.CrossRefGoogle Scholar
Gaertner, Hans-Martin. 2001. Are there V2 relative clauses in German? Journal of Comparative Germanic Linguistics 3, 97–141.CrossRefGoogle Scholar
Geis, Michael. 1970. Adverbial subordinate clauses in English. PhD dissertation, MIT.Google Scholar
Geis, Michael. 1975. English time and place adverbials. Working Papers in Linguistics 18, Ohio State University. 1–11.Google Scholar
Geis, Michael. 1985. The syntax of conditional sentences. In Geis, Michael (ed.). Studies in Generalized Phrase Structure Grammar, 130–59. Columbus, OH: Department of Linguistics, OSU.Google Scholar
Green, Georgia. 1976. Main Clause Phenomena in subordinate clauses. Language 52, 382–97.CrossRefGoogle Scholar
Green, Georgia. 1996. Distinguishing main and subordinate clauses. MS, University of Illinois.Google Scholar
Haegeman, Liliane. 2000. Negative inversion, the Neg criterion and the structure of CP. In Horn, Laurence & Kato, Yasuhiko (eds.), Negation and polarity, 29–69. Oxford and New York: Oxford University Press.Google Scholar
Haegeman, Liliane. 2003a. Conditional clauses: external and internal syntax. Mind and Language 18, 317–39.CrossRefGoogle Scholar
Haegeman, Liliane. 2003b. Notes on long adverbial fronting in English and the left periphery. Linguistic Inquiry 34, 640–9.CrossRefGoogle Scholar
Haegeman, Liliane. 2006a. Argument fronting in English, Romance CLLD and the left periphery. In Zanuttini, Raffaella, Campos, Hector, Herburger, Elena & Portner, Paul (eds.), Negation, tense and clausal architecture: Cross-linguistic investigations, 27–52. Georgetown: Georgetown University Press.Google Scholar
Haegeman, Liliane. 2006b. Conditionals, factives and the left periphery. Lingua 116, 1651–69.CrossRefGoogle Scholar
Haegeman, Liliane. 2007. Operator movement and topicalization in adverbial clauses. Folia Linguistica 41, 279–325.CrossRefGoogle Scholar
Haegeman, Liliane. 1991/2009. Parenthetical adverbials: The radical orphanage approach. In Chiba, Shuki, Ogawa, Akira, Fuiwara, Yasuaki, Yamada, Norio, Koma, Osamu & Yagi, Takao (eds.), Aspects of modern English linguistics: Papers presented to Masatomo Ukaji on his 60th birthday, 232–54. Tokyo: Kaitakushi. Reprinted in Philippa Cook, Werner Frey & Claudia Maienborn (eds.), Dislocated elements in discourse: Syntactic, semantic, and pragmatic perspectives, 331–47. New York: Routledge.Google Scholar
Haegeman, Liliane. to appear. Speculations on the syntax of adverbial clauses. In Grohmann, Kleanthes & Tsimpli, Ianthi (eds.), Exploring the left periphery. Lingua thematic issue.Google Scholar
Haegeman, Liliane (in preparation) The movement derivation of conditionals. MS, University of Ghent, Dept of English.Google Scholar
Haegeman, Liliane, Shaer, Benjamin & Frey, Werner (2009). Postscript: Problems and solutions for orphan analyses. In Cook, Philippa, Frey, Werner & Maienborn, Claudia (eds.), Dislocated elements in discourse: Syntactic, semantic, and pragmatic perspectives, 348–65. New York: Routledge.Google Scholar
Heycock, Caroline. 2006. Embedded root phenomena. In Everaert & Van Riemsdijk (eds.), vol. II: 174–209.Google Scholar
Hill, Virginia. 2007. Romanian adverbs and the pragmatic field. Linguistic Review 24, 61–86.CrossRefGoogle Scholar
Hooper, John & Thompson, Sandra. 1973. On the applicability of Root Transformations. Linguistic Inquiry 4, 465–97.Google Scholar
Hornstein, Norbert. 1990. As time goes by. Tense and Universal Grammar. Cambridge, MA: MIT Press.Google Scholar
Ingham, Richard. 2008. Negative co-ordination in Middle English. Paper presented at conference on Continuity and Change in Grammar. University of Cambridge; 18–20 March 2008.Google Scholar
Kayne, Richard. 2005. Movement and silence. New York and Oxford: Oxford University Press.CrossRefGoogle Scholar
Komagata, Nobo. 2003. Information structure in subordinate and subordinate-like clauses. Journal of Logic, Language and Information 12, 301–18.CrossRefGoogle Scholar
Larson, Richard. 1985. On the syntax of disjunction scope. Natural Language and Linguistic Theory 3, 217–164.CrossRefGoogle Scholar
Larson, Richard. 1987. Missing prepositions and the analysis of English free relative clauses. Linguistic Inquiry 18, 239–66.Google Scholar
Larson, Richard. 1990. Extraction and multiple selection in PP. The Linguistic Review 7, 169–82.CrossRefGoogle Scholar
Lipták, Aniko. 2005. Relativization strategies in temporal adjunct clauses. LIVY Yearbook 5, 133–85. Philadelphia and Amsterdam: John Benjamins.Google Scholar
Maki, Hideki, Kaiser, Lizanne, & Ochi, Masao. 1999. Embedded topicalization in English and Japanese. Lingua 109, 1–14.CrossRefGoogle Scholar
McCloskey, James. 2006. Questions and questioning in a local English. In Zanuttini, Raffaella, Campos, Hector, Herburger, Elena & Portner, Paul (eds.), Negation, tense and clausal architecture: Cross-linguistic investigations, 86–126. Georgetown: Georgetown University Press.Google Scholar
Meinunger, André. 2004. Verb Second in German(ic) and mood selection in Romance. Paper presented at the Workshop on Clause Typing and the Left Periphery. Georgetown University Round Table.Google Scholar
Melvold, J. 1991. Factivity and definiteness. In Cheng, Lisa & Demirdache, Hamida (eds.), More papers on Wh-Movement (MIT Working Papers in Linguistics, no. 15), 97–117. Cambridge, MA: MIT.Google Scholar
Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1985. A comprehensive grammar of the English language. London: Longman.Google Scholar
Reichenbach, Hans. 1947. Elements of symbolic logic. Reprinted 1966. New York: Free Press.Google Scholar
Riemsdijk, Henk van 2006. Free relatives. In Everaert & Van Riemsdijk (eds.), vol. II: 338–82.CrossRefGoogle Scholar
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Haegeman, Liliane (ed.), Elements of grammar, 281–337. Kluwer: Dordrecht.CrossRefGoogle Scholar
Rizzi, Luigi & Shlonsky, Ur. 2006. Satisfying the subject criterion by a nonsubject: English locative inversion and heavy NP shift. In Frascarelli, Mara (ed.), Phases of interpretation, 341–62. Berlin: Mouton de Gruyter.CrossRefGoogle Scholar
Roberts, Ian & Roussou, Anna. 2002. The Extended Projection Principle as a condition on the Tense Dependency. In Svenonius, Peter (ed.), Subjects, expletives and the IPP, 125–55. Oxford: Oxford University Press.CrossRefGoogle Scholar
Rudanko, Juhani 1981. On a structural ambiguity in English subordinate clauses of time. English Studies 62, 53–5.CrossRefGoogle Scholar
Sawada, Miyuki & Larson, Richard. 2004. Presupposition and root transforms in adjunct clauses. NELS 34.Google Scholar
Schachter, Paul. 1992. Comments on Bresnan and Kanerva's ‘Locative inversion in Chichewa: a case study of factorization in grammar’. In Stowell, Tim & Wehrli, Eric (eds.), Syntax and semantics 26, 103–10. New York: Academic Press.Google Scholar
Sobin, Nicolas. 2003. Negative inversion as nonmovement. Syntax 6: 183–222.CrossRefGoogle Scholar
Speas, Margaret & Tenny, Carol. 2003. Configurational properties of point of view roles. In Sciullo, Anne-Marie di (ed.), Asymmetry in grammar, vol. I: Syntax and semantics, 315–44. Amsterdam and Philadelphia: John Benjamins.CrossRefGoogle Scholar
Sperber, Dan & Wilson, Deirdre. 1986. Relevance. Oxford: Blackwell.Google Scholar
Starke, Michal. 2001. Move dissolves into Merge: a theory of locality. PhD dissertation, University of Geneva. http://theoling.auf.net/papers/starke_michal/Google Scholar
Stephens, Nola M. 2006. Norwegian when-clauses. In Butt, Miriam & King, Tracy Holloway (eds.), Proceedings of the LFG06 conference. Universität Konstanz. Stanford, CA: CSLI Publications. http://clsli-publications.stanford.edu/Google Scholar
Stump, Gregory. 1985. The semantic variability of absolute constructions. Dordrecht: Reidel.CrossRefGoogle Scholar
Tomaszewicz, Barbara. To appear. Subjunctive mood in Polish. Proceedings of the FDSL7-conference. Frankfurt am Main: Peter Lang.Google Scholar
Ürögdi, Barbara. 2008. Temporal adverbial clauses with or without operator movement. MS, Research Institute for Linguistics, Hungarian Academy of Sciences.CrossRefGoogle Scholar
Ward, Gregory L. 1988. The semantics and pragmatics of preposing. New York: Garland.Google Scholar
Zribi-Hertz, Anne & Diagne, Lamine. 2003. Déficience flexionnelle et temps topical en Wolof. In Sauzet, Patrick & Zribi-Hertz, Anne (eds.), Typologie des langues d'Afrique et universaux de la grammaire, vol. 2: Benue-Kwa et Wolof, 205–31. Paris: L'Harmattan.Google Scholar