The Syntax of Adverbial Clauses and Its Consequences For Topicalisation - Compress

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

The syntax of adverbial clauses and its consequences for

topicalisation *
Liliane Haegeman, Université Charles de Gaulle – Lille III,
UMR 8258 Silex du CNRS, l.haegeman@skynet.be

1. Introduction

1.1. Aim of the paper

Based on data from English, the paper first argues that adverbial clauses are
not a homogenous group and that at least two types must be distinguished:
central adverbial clauses modify the proposition expressed by clause with
which they are related, and peripheral adverbial clauses provide background
propositions that are to be processed as the privileged discourse context for
the proposition expressed in the associated clause. These two types of clauses
differ in both external and internal syntax, and my claim will be that the left
periphery (in the sense of Rizzi 1997) of central adverbial clauses lacks the
projection of the functional head Force, responsible for speaker anchoring. The
second part of the paper takes a comparative perspective and explores the
consequences of this proposal for argument fronting. The analysis correctly
predicts that argument topicalisation in English is excluded in central adverbial
clauses. However, clitic left dislocation in Romance is possible in central
adverbial clauses in spite of the assumed absence of ForceP in their left
periphery. I will offer an account for the contrast exploring the role of Fin in
licensing topics.

1.2. Two types of adverbial clauses

The first part of this paper is concerned with the syntax of adverbial clauses. In
the recent syntax literature, adverbial clauses have often tended to be
discussed as one undifferentiated group with respect to syntactic properties
such as extraction, parasitic gap licensing etc1. Closer examination reveals that
they do not behave homogeneously and that, though adverbial clauses may
share some properties, sub-types can be distinguished. Basing the account on
English data, I distinguish between adverbial clauses whose semantic function
is to structure the event expressed in the associated clause and adverbial
clauses that structure the discourse. The latter type expresses propositions
that are to be processed as part of the discourse background for the
proposition expressed in the associated clause. For instance, adverbial
clauses introduced by the conjunction while either provide a temporal
specification of the event, as illustrated in (1a), or they provide a background
proposition which, combined with the proposition expressed by the associated
clause, will yield contextual implications and thus enhance the relevance of the
associated clause (in the sense of Sperber and Wilson (1986)), as illustrated in
(1b):

61
(1) a According to Smith, a group of Arkansas state troopers who
worked for Clinton while he was governor wanted to go public with
tales of Clinton’s womanising. (Guardian, G2, 12.3.2, page 3, col
2-3)(event time: 'during the time that')

b While [Dr Williams’] support for women priests and gay


partnerships might label him as liberal, this would be a misleading
way of depicting his uncompromisingly orthodox espousal of
Christian belief. (Guardian, 2.3.2, page 9, col 1-2) (background
assumption: 'whereas',)2

Similarly, clauses introduced by the conditional conjunction if either express a


condition for the realisation of the event expressed in the main clause (2a), or
they provide a proposition that serves as a background assumption which,
combined with the assertion of the associated clause, yields additional
inferences (2b).

(2) a. If your back-supporting muscles tire, you will be at increased risk


of lower-back pain (Independent on Sunday, Sports, 14.10.1,
page 29, col 3) (event-condition)

b If we are so short of teachers (‘Jobs crisis grows as new term


looms’, August 30), why don’t we send our children to Germany to
be educated? (Letters to the editor, Eddie Catlin, Norwich,
Guardian, 31.8.1, page 9, col 5) (‘given that', background
assumption)

Table 1 classifies a number of English conjunctions with respect to the clause


type they introduce. Some conjunctions can embed clauses either of the two
readings; other conjunctions introduce adverbial clauses that have only one
reading: either they always modify the event (before, after, until) or they always
contribute to discourse structuring (although, whereas, given that). 3

Conjunction Central adverbial clause Peripheral adverbial clause


Event structure Discourse structure
Before Event time
After Event time
Until Event time
As Event time Rationale
Since Event time Rationale
While Event time Background assumption (Contrast)
When Event time Background assumption (Contrast)
If Event-condition Background assumption/premise
Because/ ‘cos ‘Event-reason’ Argument/ reason
So that Purpose Result
Whereas Background assumption (contrast)
Although Concession/contrast
Given that Background assumption/premise

62
Table 1: Typology of adverbial clauses in English
1.3. Organisation of the paper

The paper is organised as follows. In section 2 I provide further evidence for


the distinction between the two types of adverbial clauses, focusing on their
external syntax. To account for the observed contrasts, I propose that
peripheral adverbial clauses are syntactically less integrated with the
associated clause than central event-related adverbial clauses. Section 3
shows that there are also distinctions with respect to the internal syntax of the
two types, with peripheral adverbial clauses tolerating what have come to be
known as root or main clause phenomena (MCP). MCP are excluded from
central adverbial clauses. Section 4 accounts for this contrast by proposing that
the CP-domain of central adverbial clauses is structurally deficient and lacks
the functional heads encoding speaker-related functions (speech time,
epistemic modality, illocutionary force). MCP depend on the presence of Force.
Sections 5 and 6 move to comparative data. In English central adverbial
clauses, topicalisation of arguments is predicted to be banned because it
depends on the encoding of speaker. However, while argument topicalisation is
disallowed in English central adverbial clauses, it is possible in Romance. In
section 7 I speculate that the fronted arguments in the Romance languages
examined can be licensed by the head Fin, the lowest inflectional head of the
CP domain. Section 7 concludes the paper.

2. External syntax: degrees of integration

In this section I provide evidence for the distinction between adverbial clauses
which is introduced above and illustrated in examples (1) - (2). I relate the
contrast to the degree of syntactic integration of the adverbial clause with the
associated clause.

2.1. Co-ordination of likes

Typically, only adverbial clauses with parallel interpretations can be co-


ordinated: in (3a), both conditional clauses will be read as event-related central
conditionals and in (3b) they will both be read as discourse-related peripheral
conditionals. There is no way in which the conditional clauses in (3) could be
given distinct interpretations with the first being event-related and the second
discourse-related (or vice versa):

(3) a The party is also in danger of alienating older people above the
poverty line, Mr Cable argues. ‘Both these groups will swing to the
Conservatives if the Tories are smart enough and if we have
nothing much to offer them.’ (Guardian, 11.2.2., page 6, col 5)
b But if Sir Richard has been tainted by the affair, and if Mr
Sixsmith’s role may not have been as entirely well-intentioned as
he claims, the individual most damaged by the row remains
Stephen Byers. (Guardian, 25.2.2, page 4, col 3)

63
2.2. Scope phenomena

Scopal properties distinguish the two adverbial clause types. Essentially,


central adverbial clauses can be within the scope of operators in the
associated clause4, while peripheral adverbial clauses are outside the scope of
such operators. This can be illustrated in a number of different ways, some of
which discussed below.

2.2.1. Temporal subordination

Central adverbial clauses are within the scope of temporal operators in the
associated clause, while peripheral adverbial clauses are not. This has been
pointed out in the literature. Hornstein (1993), for instance, discusses temporal
dependency as a property of what I have labelled central adverbial clauses.

temporal adjuncts headed by temporal connectives such as when, while,


after, before, as, until, and since interact with the tense of the matrix
clause. … There are rather specific tense-concord restrictions that
obtain between the tense of the matrix clause and the tense of the
modifying clause. These restrictions can be largely accounted for
structurally in terms of the C[onstraint] on D[erived] T[ense] S[tructures]
and the rule that combines these clauses into complex tense structures.
(Hornstein 1993: 43)

Concerning what I have labelled peripheral adverbial clauses Hornstein says:


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 (206:
note 19)

Hornstein's 'secondary conjunctive interpretation' corresponds to my discourse-


related interpretation. Note though that it would not be correct to treat
discourse-related adverbial clauses simply in terms of co-ordination. I will
illustrate this for contrastive while clauses. It is true that contrastive while is
semantically close to a co-ordinating conjunction, as shown by (4), in which
while can be replaced by but or by and:

(4) a John does a Ph.D in Oxford while he did his first degree in
Cambridge.
a' John does a Ph.D in Oxford and/but he did his first degree in
Cambridge.
b John reads the Guardian while Mary reads the Times.
b' John reads the Guardian and/ but Mary reads the Times.

64
However, contrastive while clauses do not share all the properties of co-
ordinated clauses. Ellipsis of the subject of the second co-ordinated clause is
possible in (5a) but in an adverbial clause introduced by contrastive while the
subject cannot be ellipted (5b): 5

(5) a John does a Ph.D. in Oxford but did his first degree in Cambridge.
b *John does a Ph.D. in Oxford while did his first degree in
Cambridge.

Gapping is possible with co-ordinated clauses but not in constructions with a


contrastive while clause:

(6) a John reads the Guardian and Mary the Times.


b *?John reads the Guardian while Mary the Times.

With respect to conditional clauses, Declerck and Reed (2001) distinguish two
types. The first, labelled the present perspective type, is said to be 'temporally
subordinated' to the matrix clause, the second which they label the future
perspective type is not temporally subordinated. A clause is temporally
subordinated when the interpretation of its tense forms depends on that of the
associated clause. Interestingly, Declerck and Reed relate the difference in
temporal relations to a difference in illocutionary force, a point to which I return.
When the Present Perspective System is used in the [conditional] sub-
clause [i.e. with temporal subordination, lh (2a)], the speaker makes a
single (but complex) prediction: she presents the contents of the two
clauses as forming a unit. […] when the Future Perspective System
[without temporal subordination lh] is used in both clauses [conditional
and associated clause, lh, (2b,7a,b below)], the speaker makes two
independent predictions: there are, as it were, two illocutionary speech
acts. (Declerck & Reed 2001: 131):

Temporally subordinated conditional clauses such as those in (2a) are central


conditionals. In such adverbial clauses a present tense is interpreted with
future reading when it is within the scope of a matrix future time expression. In
peripheral conditional clauses tense interpretation is independent: the tense of
the associated clause does not affect the interpretation of that of the adverbial
clause. For conditionals, the non-subordinated pattern is illustrated in (2b)
above, in which the present tense in the subordinate clause refers to a present
time, and it is also illustrated also in (7a) in which future time is expressed
independently in the adverbial clause. (7b) illustrates independent temporal
expression in discourse-related while clauses.

(7) a If I’m no longer going to be arrested for possessing cannabis for


my own consumption ('Cannabis laws eased in drugs policy
shake-up', October 24), shouldn’t I be able to grow my own?
(Jason Cundy, Letter to the editor Guardian, 25.11.1, page 9, col.
8)

65
b The French president, Jacques Chirac will tell George Bush in
Washington today that while France will continue to back the
military campaign in Afghanistan, the search for a political
framework for the country’s future must be intensified. (Guardian,
19.11.1, page 4, col 3)

Observe, though, that it cannot be argued that peripheral adverbial clauses


must never be temporally subordinated and that they are somehow
syntactically completely unattached (a claim I had made in my earlier work
(Haegeman 1991)). Peripheral adverbial clauses may be temporally
subordinated when together with the associated clause, they are embedded
under a higher verb:

(8) The EOC -commissioned research found that while girls were
aware of discrimination, less than 40 per cent of 15 and 16 year
olds thought girls and boys were treated the same in the family.
(Observer, 28.10.1, page 9, col 2)

The past tense of were in the contrastive while clause in (8) is not triggered by
the impact of the past tense of thought in the associated clause, but rather both
the past tense of were and that of thought are the result of the embedding the
two containing clauses under the higher past tense verb found.

2.2.2. Adjunct scope

As expected, in addition to tense, other adverbial operators may also have


scope over central adverbial clauses but they do not scope over peripheral
adverbial clauses. In (9a,b), the frequency adverbs always and usually scope
over the central adverbial clauses. In (9c) on the other hand, the frequency
adjunct often in the matrix clause does not scope over the peripheral adverbial
while clause, with its own independent adverb of frequency (always):

(9) a I always get home before the programme starts.


b ‘When a woman says she wants to go out and get a job, ‘ [Oliver
Reed] once remarked, ‘it usually means she’s hopelessly behind
with the ironing.’(Observer, 10.3.2, p 32, col 1)
c While Mary always drives to school, John often goes by bike.

2.2.3. Negation

Main clause negation may scope over central adverbial clauses, but peripheral
adverbial clauses cannot fall within the scope of a negative operator in an
associated clause, as illustrated by the contrast between (10a,b) and (10c):

(10) a He doesn’t drink while he is driving.


b He never drinks while he is driving.
c My husband doesn't smoke cigarettes, while he does occasionally
smoke a cigar.

66
In (10a) and (10b) the negation can be said to range over a complex event: 'he
does not drink-drive'. In (10c) two propositions are interpreted in parallel, only
one of these is negated.

67
2.2.4. Focus (cleft/only)

As illustrated by the clefting in (11), a focus operator in the matrix clause may
select the central adverbial clause (11a,b); a focus operator in the associated
clause cannot range over the peripheral adverbial clause (11c).

(11) a It is after I left that I realised he was my former teacher.


b It’s only while [time] you’re alive that human selfishness, or
whatever, is held against you (Independent on Sunday, Review
14.10.1, page 9, col 1)
c *It is while my mother was a housewife that my father used to
work in a brickyard.

2.2.5. Interrogative scope

As is to be expected by now, an interrogative operator in the matrix clause may


target a central adverbial clause (12a,b), though not the peripheral adverbial
clause (12c,d):

(12) a Do you ever read Belgian newspapers while you are abroad?
b While you are abroad, do you ever read Belgian newspapers?
c While Bush is clearly delighted to have Blair as an extra
ambassador for his policies at the moment, somebody to get on
those dangerous aeroplanes and rush around the Middle East
chatting up guys with difficult names in order to strengthen
America’s position, what kind of influence do we really imagine
Blair has on Bushes foreign policy? (Independent, Comment
1.11.1, page 5, col 3)

2.3. Prosody & orthography

Peripheral adverbial clauses are typically prosodically set off from the
associated clause by comma intonation, usually signalled by a comma in
writing. Sometimes, however, the peripheral adverbial clause is typographically
set off as if it were an independent clause. Some attested examples of this
practice are given in (13):

(13) a Today the party’s Treasury spokesman, Matthew Taylor, will


commit the Liberal Democrats to entering the next election with a
promise to pump more money into the NHS, possibly with a
special health tax. While an emergency debate this weekend is
likely to reaffirm Liberal Democrat scepticism about British
involvement in a war with Iraq. (Independent, 9.3.2, page 8, col 8)
b Maybe Tyson should redirect his energies towards a sport less
likely to bring out his violent side. Figure skating for example, or
synchronised swimming. Because this week’s ugly scenes
probably won’t be the last and every time the moral commentators
become even more outraged: these boxers – they are behaving in

68
a violent and aggressive manner! (Guardian, 26.1.2, page 8, col
3)

Based on the scope facts discussed in section 2.2, as well as the


prosodic/orthographic separation of the peripheral adverbial clauses, I have
elaborated an account (Haegeman 1984a,b,c,d, 1991) according to which
central adverbial clauses and peripheral adverbial clauses differ in their
external syntax, more precisely in the level and degree of integration with the
associated clause. Central adverbial clauses are adjoined to VP or to IP and
are part of the main clause; peripheral adverbial clauses are adjoined to CP,
hence their connection with the associated clause is much looser. My proposal
thus differs from the conclusions drawn by Declerck and Reed, who deny there
is a syntactic contrast:
a subordinate clause is a syntactically dependent clause. Such
questions as the scope of negation, focusing, modality, etc. in the head
clause are immaterial to this, as they pertain, not to syntactic, but to
semantic integration (Declerck and Reed 2001: 37-8)

Further evidence for the proposed syntactic distinction in terms of degrees of


embedding/syntactic integration is given in the next sections.

2.4. VP ellipsis and strict/sloppy identity

The data in (14) show that central adverbial clauses may be affected by VP-
ellipsis and that when such adverbial clauses contain a pronoun, VP-ellipsis
may lead to a so-called sloppy identity reading. Thus in (14a) so will Bill may be
interpreted as in (14b) or as in (14c), the latter illustrating the sloppy identity
reading:

(14) a Johni will leave the meeting before hisi paper is discussed and so
will Bill.
b (i) Bill will also leave the meeting before Johni’s paper is
discussed.
c (ii) Billj will also leave the meeting before hisj paper is discussed.

A temporal while clause may also be affected by VP-ellipsis and a sloppy


identity reading may arise:

(15) a Johni works most efficiently while hisi children are at school and
so does Bill.
b Bill also works most efficiently while John’s children are at school.
c Billj also works most efficiently while hisj children are at school.

On the other hand, contrastive while clauses, which I take to be peripheral, are
not affected by VP-ellipsis and VP-ellipsis does not lead to sloppy readings.
(16a) has the interpretation paraphrased in (16b) and there is no way in which
one will assign a sloppy identity interpretation to the pronoun his and assume
that James's wife is also unemployed. My analysis predicts this difference. In

69
(16a) the peripheral adjunct clause is attached outside the VP of the
associated clause, hence VP-ellipsis cannot affect the adverbial clause.

(16) a While hisi wife is unemployed, Johni has a high-powered job in the
city and so does Jamesj.
b Jamesj also has a high powered job in the city.

2.5. Parasitic gaps

Central adverbial clauses allow for parasitic gaps bound by operators in the
associated clause (17); parasitic gaps in peripheral adverbial clauses are
somewhat degraded:6

(17) a John is the guy they said they’ll hire [∅] - if I publicly criticise [∅] -
in order to get me to praise [∅] (Nissenbaum 2000)
b He is a man who if you know [∅] you will love [∅] (Jespersen
1931: 202)

(18) a This is the paper which I memorised [∅] while I was copying [∅].
b ?This is the paper which I myself enjoyed [∅] very much,
while/whereas you will probably dislike [∅].(cf. Postal 2001: 299
ff))

The contrast follows from an account that assumes that peripheral adverbial
clauses are not fully integrated with the associated clause. If the parasitic gap
phenomenon depends on a kind of semantic composition between the
adverbial clause containing the PG and the matrix clause containing the
operator and the 'real gap' (Nissenbaum 2000, chapter 3), it is reasonable to
propose that this complex predicate formation is subject to constraints of
locality. The syntactic independence of peripheral adverbial clauses entails
that they lack the required local relation with the associated clause, rendering
the formation of a complex predicate (and hence the existence of PG)
impossible. On the other hand, being near-coordinate, ATB extractions may be
expected to be marginally possible in peripheral adverbial clauses.
The contrast between the type of adverbial clauses can also be detected to
some extent in simple extraction. While extraction out of adjunct clauses is
generally somewhat degraded (see Sabel 2002 for a recent discussion and for
references), argument extraction out of a central adverbial clause of the event-
related type is marginally acceptable, with the effect of a weak subjacency
violation. Data such as (19) are occasionally found (cf. Haegeman 1987)

(19) a the details and the whole, which an artist cannot be great unless
he reconciles (Ru, Sel. 1.175, Jespersen, 1931: 202)
b a stranger, from that remote and barbarian Isle which the Imperial
Roman shivered when he named, paused.(Lytton, Pomp, v, ch
xi,153b, Poutsma 1926: 645)

70
Similar examples in which extraction is launched from a peripheral adverbial
clause are unattested; they would lead to full ungrammaticality:

(19) c ??This is the paper which I enjoyed the conference very much,
whereas I disliked [∅].(cf. Postal 2001: 299 ff.))7
d ??This is the paper which, whereas I disliked [∅], I enjoyed the
conference very much.(cf. Postal 2001: 299 ff.))

2.6. External syntax

Updating my earlier analysis I propose that central adverbial clauses are


merged with the matrix clause at an earlier point in the derivation than
peripheral adverbial clauses. While central adverbial clauses are adjoined to
vP or to an Inflectional projection (cf. Thompson 1994, Nissenbaum 2000).8
Peripheral adverbial clauses are merged with the associated clause after the
latter is fully projected, i.e. they are merged with a CP: the resulting structure
will be as in (20), a pattern close to co-ordination.9

(20) Peripheral adverbial clauses:


[CP1 adverbial clause [CP1 associated clause]]

3. Internal syntax: Main clause phenomena in peripheral adverbial


clauses

My earlier work (Haegeman 1984a,b,c,d, 1991) discussed the differences in


external syntax of the adverbial clauses. This section shows that the difference
in external syntax is paired with a difference in internal syntax. Summarising
the discussion below, I will show that, typically, peripheral adverbial clauses
give rise to so-called root phenomena (Emonds 1970, 2000) or Main clause
phenomena (MCP) (Hooper and Thompson 1973): they display a number of
syntactic patterns which are usually restricted to root clauses. MCP are not
available in central adverbial clauses.

3.1. Speaker oriented epistemic modals and adverbial clauses

One difference between central adverbial clauses and peripheral adverbial


clauses is that the latter may, and the former may not, contain expressions of
epistemic modality:

(21) a *Mary accepted the invitation without hesitation after John may
have accepted it (based on Verstraete 2001: 149)
b ??John works best while his children are probably/might be
asleep.
c The ferry will be fairly cheap, while/whereas the plane may/ will
probably be too expensive.

Epistemic modality is by definition speaker-related: it expresses the speaker's


evaluation concerning the likelihood of the state of affairs/event. This

71
evaluation is anchored to the speech time. Crucially, even if 'epistemic modals
can be morphologically associated with a past tense, … this morphological
marking does not express the speaker's past judgement. Either it is used for
tentativeness,… or it occurs in a context of indirect or free indirect speech'
(Verstraete 2001: 152, italics mine).

72
3.2. Illocutionary force

The availability of epistemic modality in peripheral adverbial clauses and its


absence in central adverbial clauses suggest that peripheral clauses encode
speaker-anchoring in a way that central adverbial clauses do not. This
distinction is confirmed by the fact that discourse-related adverbial clauses
have illocutionary force, a point signalled by Declerck and Reed (2001) for
peripheral conditional clauses (cf. section 2.2.1). Central adverbial clauses do
not have this illocutionary potential. There are a number of empirical facts
confirming the opposition.

3.2.1. Echoic effect in conditional clauses

Declerck and Reed signal that peripheral conditional clauses are echoic:
closed P-clauses [≈ peripheral conditional clauses lh]are always echoic
in one sense or another. They can echo straightforward statements
about the actual world, or they can echo Q-propositions about a
nonfactual world. However, the claim that closed P-propositions are
echoic need not mean that they have to be echoes of actual utterances.
They may also be echoes of an internal or mental proposition (thought)
such as the interpretation of an experience, perception etc. (Declerck
and Reed 2001:83)

Being echoic, peripheral adverbial clauses must 'echo' a speaker, hence they
encode encoding a speech act.

3.2.2. Tags: ( H&T 1973: 471.)

Further evidence for the encoding of illocutionary force in peripheral adverbial


clauses and its absence in central adverbial clauses comes from the
observation that peripheral adverbial clauses may and that central adverbial
clauses may not have their own question tags associated with them. In (22a)
the tag didn't she is related to the matrix clause; a tag hadn't they, which would
have to be related to the adverbial clause, is not possible, as seen in (22b).
Temporal while clauses show the same restrictions (23).

(22) a Mary went back to college after/before her children had finished
school, didn’t she?
b *Mary went back to college after/before her children had finished
school, hadn't they?

(23) a Bill took a degree at Oxford while his children were still very
young, didn’t he?
b *Bill took a degree at Oxford while his children were still very
young, weren’t they?

The situation is different with peripheral adverbial clauses. Sentence-final


contrastive while clauses cannot be followed by a tag relating to the associated

73
clause. Such a tag would have to precede the contrastive while clause (24a,b).
On the other hand, a contrastive while clause may have its own tag (24c):

(24) a *Bill took a degree at Oxford, while his daughter is studying at


UCL, didn’t he?
b Bill took a degree at Oxford, didn’t he, while his daughter is
studying at UCL.
c Bill took a degree at Oxford, while his daughter is studying at
UCL, isn’t she?

(25) is an attested example in which a question tag is associated with a


peripheral because clause:

(25) Henry III, or example, ruled for 56 years but his golden jubilee
was flop. ‘Henry III?’ they said, ‘Erm, now which one’s that then?
Cos Henry V is Agincourt, isn’t he…(Guardian, 2.2.2., page 8, col
2)

3.2.3. (Rhetorical) questions

Among peripheral adverbial clauses, because clauses and although clauses


may typically also be associated with interrogative force. In this case, the
interpretation of the adverbial clause will be that of a rhetorical question:

(26) a No one would have been too upset about her bad behaviour,
because wasn't that what writers were put on earth to do?
(Observer, 20.8.2000 page 27, col 8)
b News about the anti-American demonstrations which had begun
to appear in Berlin and other parts of Germany in the fortnight
since the summit hadn’t exactly helped sell what was supposed to
be Michelle's greatest success. Although what did the mid-west
care about Berlin? (BNC, Verstraete 2002: 147)

3.2.4. Imperatives (Verstraete 2002: 146)

Verstraete (2002: 146) signals that some peripheral adverbial clauses may also
have imperative force [see note 5]. I refer to his work for discussion.

3.3. Argument fronting in adverbial clauses.


In the literature on English, it has often been assumed that ‘embedded
topicalization is consistently impossible in an adjunct clause’ (Maki, Kaiser and
Ochi 1999: 4). This restriction is illustrated in (27) (see also Authier 1992):

(27) a *Before this book, Mary read, John had already read it (Maki,
Kaiser & Ochi 1999:4)
b *Before MY book, Mary bought, John had already bought YOURS
c *If some of these precautions you take, you will pass the exam.10

74
Observe, though, that the ban on topicalisation is considerably weakened in
peripheral adverbial clauses in English, as shown by the data in (28):

(28) a His face not many admired, while his character still fewer felt they
could praise. (Quirk et al 1378)
b [He had brought a number of records.] Although some of them I
really enjoyed, others were note very inspiring.

A similar contrast is to be found in Japanese11 (cf. Heycock 2002), where wa


topicalisation is not possible in central adverbial clauses but it is licit in
peripheral adverbial clauses. (29) illustrates the contrast in conditional clauses:

(29) a *Mosi sono yoona zassi-wa, (anata-ga) yome-ba,


if that like magazine-top (you-nom) read (conditional)-if
(anata-wa) yasai-ga sukini narimasu.
(you-top) vegetable-nom like become
'If these magazines, you read, you will come to like vegetables'
b Mosi sono yoona zassi–wa (anata-ga) sukide-nai (conclusive)-
naraba,
if that like magazine-top (you-nom) like-not-if
naze (anata-wa) (sorera-o) kai-tuzukerunodesu ka?
why you-top them-acc buy-continue, Q
'If such magazines, you don't like, why do you keep buying them?'

In Japanese, the choice of conjunction distinguishes the two types of


conditionals. Central conditional clauses are introduced by the conjunction ba,
peripheral conditional clauses are introduced by the complex conjunction
naraba. Interestingly, nara itself can be used as a subordinating conjunction.
Concerning the conditional sentence pattern [S1 nara] S2, Kuno (1973: 168)
says: ‘It is usually said that this pattern has a strong degree of assertion about
the statement represented by S1’. Though I do not want to pursue the syntax of
Japanese adverbial clauses here, the make-up of the two conjunctions is in line
with my proposal that peripheral adverbial clauses encode speaker-anchoring
in a way that central adverbial clauses do not. I return to topicalisation in
section 5.

3.4. If… then as MCP

It has been proposed in the literature (Iatridou and Kroch 1992) that the
if…then sequence is a MCP. As expected, the pattern is available in peripheral
adverbial clauses and it is not available in central adverbial clauses.

(30) a France will be expensive while if you go to England then you will
get value for money.
b I wouldn’t recommend Virgin, ‘cos if you travel with them then you
may have the problem of overbooking.
c *Mary does not enjoy a film on TV until if she has put her children
to bed then they sleep well.

75
3.5. A syntactic account? 12

The availability of root phenomena/MCP in a restricted set of embedded


clauses has been repeatedly signalled in the generative literature. However,
earlier discussions tended to imply that a fully syntactic account is unavailable:
As a positive environment we can say that [root] transformations operate
only on Ss that are asserted. …some transformations are sensitive to
more than just syntactic configurations. It does not seem possible to
define the domain of an RT in terms of syntactic structures in any
general way. However, …, even if it were possible to define in syntactic
terms the conditions under which RTs can apply, … the question of why
these transformations can apply in certain syntactic environments and
not others would still be unanswered (Hooper & Thompson 1973: 495,
italics mine)

In the next section I outline a syntactic account for the observation that
peripheral adverbial clauses do and central adverbial clause don't display
MCP.13

4. RT/MCP and the internal make-up of CP

4.1. 'Reduction'

My proposal is inspired by Hooper and Thompson's own observation,


reproduced in the citation below, that MCP are generally excluded from what
they call 'reduced' clauses:
Though RTs may apply in some complements that are full sentences
introduced by the complementiser that, they may never apply in any
complements that are reduced clauses. By reduced clauses we mean
infinitives, gerunds, and subjunctive clauses14, i.e. those complement
types which have uninflected verbs. (Hooper and Thompson 1973: 484-
5, italics mine)15

I will pursue this idea of structural truncation16. Following Hooper and


Thompson's suggestion, I elaborate the hypothesis that central adverbial
clauses are structurally deficient while peripheral adverbial clauses display the
full clausal structure. More specifically, I argue that the CP-domain of central
adverbial clauses lacks the functional projection encoding the speaker-role and
which is projected in root clauses (and in clauses embedded under speech act
verbs or under propositional attitude verbs). The speaker-related projection, I
contend, is available in peripheral adverbial clauses.17 Observe that the
syntactic distinction is semantically motivated. Peripheral adverbial clauses
express independent propositions that serve as the immediate discourse
background to the associated clause; central adverbial clauses are part of and
modify the proposition with which they are associated.

76
4.2. The periphery of the clause: the split CP (Rizzi 1997)

Various authors have proposed that the unitary CP-layer be replaced by a


hierarchy of functional projections. (31a) is the hierarchy proposed by Rizzi
(1997), on which I will base the discussion here.

(31) a Force > Topic > Focus > Fin18

Following Bhatt and Yoon (1992), Bennis (2000) and Rizzi (1997: note 6), I
distinguish the head Force, which encodes anchoring to speaker, from the
head Sub whose function is merely to hosts the subordinator and introduce the
subordinate clause.19 Subordinating conjunctions are inserted in ‘Sub’; Sub
serves to subordinate the clause, to 'make it available for (categorial) selection
independently of its force' (Rizzi 1997). The head Force encodes anchoring to
speaker, and I propose that this head is implicated in the licensing of, among
other things, illocutionary force and epistemic modality. Both central adverbial
clauses and peripheral adverbial clauses contain the position Sub, but only the
latter encode anchoring to the speaker, represented as Force. Root clauses
obviously also contain the functional head Force. We thus end up with the
following functional hierarchies in the left periphery of finite clauses20:

(31) b Central adverbial clause: Sub Fin


c Peripheral adverbial clause: Sub Force Top* Focus Fin
d Root clause: Force Top* Focus Fin

I propose that in addition to encoding illocutionary force, Force licenses speech


time. The independent encoding of temporal relations in a syntactic domain
depends on the availability of an independent Speech time in that domain,
which by hypothesis depends on the presence of Force.21 Epistemic modality,
which I take to be anchored to speaker and to speech time, requires the
presence of the head Force. Given (31b), epistemic modality will be
unavailable in central adverbial clauses and, given (31c), it will be available in
peripheral adverbial clauses. For an independent argument for relating
epistemic modality and illocutionary Force I also refer to recent work by Bayer
(2001: 14-15).

5. Argument fronting and adverbial clauses

5.1. Topicalisation in embedded domains in English

Let us now turn to the availability or absence of topicalisation in adverbial


clauses signalled in section 3.3. I exploit a correlation postulated by Bayer
(2001) for German embedded topicalisation. Bayer says:
… this form of topicalisation is the grammar’s reflex of the speech act to
be performed and is as such on a par with German constructions
involving modal particles like aber, denn, doch, ja etc. Modal particles
supply features which interact with other features such as [WH] yielding a
wide range of illocutionary forces. Bayer 2001: 14-15)

77
. …if emphatic topicalisation belongs to the class of grammatical means
of force projection in the sense of Rizzi (1997), its root clause property
and strict left peripherality [in Bavarian] are not surprising. ‘ (Bayer 2001:
14-15)

Let us assume that English topicalisation is like the German topicalisation


discussed by Bayer. Putting it very roughly, (32a) would have the reading
(32b): the speaker relates the topic to the clause that is predicated of it. In
other words, topicalisation is a 'speech act'.

(32) a This book, I don't like.


b Concerning this book, the speaker says that speaker does not
like it.

I assume that English does not have an alternative way of relating a fronted
topic to the associated clause, while, as we will see below, some Romance
languages have alternative mechanisms. Assuming that the head Force
encodes speech anchoring and that this head is available in peripheral
adverbial clauses will allow us to predict that such adverbial clauses allow for
argumental topicalisation. Conversely, the non-availability of the head Force in
central adverbial clauses will mean that fronted arguments in such clauses
cannot be related to the associated clausal domain. 22
Along minimalist lines, we might express the dependency between
argumental topicalisation and Force by associating a FORCE feature to the
head topic23. Topicalisation is then somehow represented as a type of speech
act. If there is no alternative way of licensing a topic in English, then, a Topic
can only be licit in a domain containing Force. An embedded clause containing
a topicalised argument but lacking Force will be ungrammatical (33d). In (33) I
assume without discussion that focalisation too is speaker-related and that
Focus can also be associated with a Force feature, though this remains a
speculation subject to future study. 24

(33) a Central adverbial clause:


Sub Fin IP
b Peripheral adverbial clause:
Sub Force TopForce Focus Force Fin IP
c Root clause:
Force Top ForceFocus Force Fin IP
d. :
*Sub TopForce Focus Force Fin IP

5.2. CLLD in Romance

Rizzi's (1997) account of topicalisation relies heavily on an assumed similarity


between argument fronting in English and clitic left dislocation (CLLD) in
Romance. Notably, Rizzi assumes that in both operations, the topicalised
argument is located in the specifier of TopP. The problem with this account is
that while argument topicalisation is excluded in central adverbial clauses in
English (and Japanese), as discussed above, this restriction seems to be

78
relaxed in Romance where clitic left dislocation (CLLD) is possible in at least
some types of central adverbial clauses. If central adverbial clauses lack the
projection Force, and if Force is crucial in licensing topicalisation, then we do
not expect this asymmetry. Exploring and adapting some accounts in the
literature, I will propose in section 6 that topicalisation in Romance need not
depend on Force but can be licensed by the head Finiteness.

5.3. Some data

The different distribution of English topicalised arguments and their Romance


counterparts has not passed unnoticed in the literature. For instance,
concerning Italian CLLD, Cinque (1990) points out that
The ‘left-dislocated’ phrase of CLLD [in Italian, lh] can occur at the front
of virtually any subordinate clause type:

(34) Da quando, al mercato, ci va lui, non mangiano piu bene.


since when to the market he goes there they don’t eat well any
more

CLLD contrasts with LD, which typically occurs in root contexts and (to different
degrees of marginality) in the complements of only a few classes of
propositional attitude verbs (Cinque 1990: 58, (41)= his (1b))

The data in (35) show that Italian central conditional clauses allow CLLD.
Italian subjunctive clauses also allow CLLD (35c). In this respect, Italian again
differs from English, where subjunctive clauses are incompatible with
topicalisation (Hooper and Thompson 1973: 484-5, see citation above).

(35) a Se gli esami finali non li superi, non otterrai il diploma. 25


if the final exams not them pass, you won’t obtain the degree
b Se questi tre esami riuscirai a superarli, avrai superato tutti i
requisiti per il diploma.
if those three exam you manage to pass, you will have satisfied all
the requirements for the diploma
c Non vorrei que lo stipendio lo prendesse anche una che non ne
ha bisogno26
I wouldn’t want the salary it-take-subjunctive also one that not of it
has need

For Spanish, it has also been noted that CLLD is not restricted to root
environments (Zubizaretta 1998: 187). Spanish subjunctive clauses also allow
CLLD:

(36) a Si este examen no lo apruebas con un cinco, perderás el curso


entero.27
If this exam you don't pass with a 5, you'll miss the whole year
b Prefiero que estos textos no te los prepares hasta la semana
próxima.
I prefer that these texts you do not prepare until next week

79
The Catalan analogues of these data, given in (37), are also grammatical28:

(37) a Si aquest examen no l'aproves amb un cinc, perdràs el curs


sencer.
if this exam you don't pass with a 5, you'll miss the whole year
b Si els exàmens de setembre no els corrigeix el professor, ho fa el
seu ajudant.
if the September exams the professor doesn't correct them, his
assistant will do it
c Quan aquesta cançó la vaig tornar a sentir al cap dels anys, em
vaig emocionar molt.
when this song I heard again after some years, I got really
emotional

The French data are slightly less clear cut, possibly due to the influence of
normative attitudes of speakers. Many of my informants did accept the
examples in (38), with CLLD in a central, event-related adverbial clause.29

(38) a Si ce livre-là tu le trouve à la Fnac, achète-le.


if this book there you it find in the Fnac, buy it
b Dès que ton texte, je l'aurai lu, je t'appellerai.
as soon as your text, I it will have read I will call you
c Quand ça, je l'ai appris, j'ai immédiatement téléphoné à la police.
when that I it heard I have immediately called the police

In French, CLLD is also allowed in subjunctive complements:

(38) e Je préfère que ce texte-là, tu ne le lises pas.


I prefer that this text there you don't read it
f J'aimerais que ce texte-là, tu le prépares pour la semaine
prochaine.
I would like that that text there you it prepare for next week30

Finally, to extend the comparison, in Modern Greek too, CLLD is not restricted
to root environments. Anagnostopoulou (1997: 163, her (26)) provides the
following:

(39) I Maria ine xaroumeni tora pu ton Janni ton paratise.


the Mary is happy now that the John him left-3sg
'Mary is happy now that she has left John.'

6. The landing site(s) of topicalised arguments

6.1. Topicalisation to Spec,IP

The Romance and the Greek data illustrated in section 5.3 are problematic if
we account for the absence of topicalisation in English and Japanese central
adverbial clauses by the non-availability of Force and if, as, for instance, in

80
Rizzi (1997) and Grewendorf (2002), CLLD also targets a Force-licensed
TopP. Since the syntactic distinction between the two types of adverbial
clauses is claimed to be motivated semantically, I do not want to simply
stipulate that while Force is missing in English central adverbial clauses, it is
available in the Romance ones.31 One solution is to propose that argument
topicalisation is not a syntactically uniform phenomenon and that in addition to
the argument fronting to TopP discussed above, fronted arguments may also
target a position that remains licensed in central adverbial clauses. 32
Such proposals have indeed been made in the literature. According to one
type of analysis, the landing site of topicalisation can also be IP-internal.
Iatridou and Kroch (1992) exploit the distinction between a CP topic and an IP
topic to account for the contrast between Yiddish and Icelandic on the one
hand and Danish and Frisian on the other. Their analysis is exploited with
respect to CLLD in Spanish by Zubizaretta (1998):
Spanish to some extent resembles some of the Germanic languages –
specifically, Yiddish and Icelandic (references omitted)…. Generalised
TP analysis. Languages with a generalised TP may be said to allow a
certain amount of feature syncretism. More precisely, in these languages
a discourse-based functional features, such as ‘topic’, ‘focus’, or
‘emphasis’, may combine with the feature T(ense), giving rise to the
syncretic categories T/’topic’, T/’focus’, T/’emphasis’. A topic, focused, or
emphatic phrase may therefore be moved to [Spec,T] for feature-
checking purposes … This of course is possible only to the extent that
the nominative subject can be licensed in these languages in some way
other than via specifier-head agreement with T. (Zubizaretta 1998:
100)33

If in addition to the movement to TopP deployed in English, Romance CLLD-ed


topics may also target an IP-internal position we predict that CLLD topics arise
in central adverbial clauses, regardless of the non-availability of TopP in the
CP-area.
The proposal that CLLD in Romance can target the IP-internal specifier of
TP raises two problems for the discussion at hand. First of all, assuming
uniqueness of specifiers34, the prediction would be that in Romance central
adverbial clauses in general either the subject is preverbal or a CLLD-topic is
but not both. This prediction is not fully borne out. Though judgements are not
homogeneous, in central adverbial clauses both subject and CLLD argument
may precede the verb, at least for some speakers (see also Zubizaretta 1999
though)

(40) a Cat Si els exàmens de setembre el professor no els corregeix a


temps, ho fa el seu ajudant.
if the September exams the professor doesn't correct them on
time, his assistant will do it
b Sp Si los exámenes de septiembre el profesor no los corrige a
tiempo, lo hace su ayudante.

A second problem concerns infinitival clauses. Following Hooper and


Thompson (1973), I assume that infinitival control complements are structurally

81
reduced and lack ForceP. This accounts for the fact, observed by Hooper and
Thompson, that English infinitival clauses resist argument fronting (40a) (see
also Emonds's (2000: 8) restriction of discourse projections to finite clauses).

(41) a * I have decided your book to read.

Again, the Romance data are different. Rizzi (1997) and Bianchi (2001) signal
that CLLD is possible in Italian control infinitives (41b-d). If the lower
topicalisation in Romance targets an IP-internal position such as, say, SpecTP,
we would indeed expect topicalisation to be possible in infinitival clauses.
However, we would also expect the fronted argument to follow the infinitival
conjunction di. This expectation is not borne out for Italian as shown by (41b-d):

(41) b Gianni pensa, il tuo libro, di conoscerlo bene. (Rizzi, 1997:


309)
Gianni thinks, the your book, di know-it well
c Mi sembra, il tuo libro, di conoscerlo bene. (Rizzi, 1997: 309)
Me seems the your book di know –it well
d Gianni sostiene, il tuo libro, di conoscerlo bene.
Gianni maintains, the your book, di know it well (Bianchi 2001:
29, her (69c)

As a way out, one might propose that the topicalised arguments in (41b,c,d) are
parenthetical constituents inserted at constituent boundaries but perhaps not
(fully) syntactically integrated in the structure. However, observe that such a
parenthetical analysis is not plausible. If Italian topics associated with infinitival
complements were simply parentheticals, we would probably expect them to
also be compatible with raising patterns35. Rizzi (1997: 309) signals that while
in Italian control complements allow for topicalisation, raising complements do
not:

(41) e *?Gianni sembra, il tuo libro, conoscerlo bene.


Gianni seems the your book know –it well

In French too, control infinitives marginally allow argument topicalisation and,


again, the fronted topic precedes the infinitival complementiser à:

(42) a Je cherche ton livre à l’acheter d’occasion. (Tellier 2001: 356-7)


I seek-for your book à it buy second hand
b *Je cherche à, ton livre, l’acheter d’occasion.
I seek-for à your book it buy second hand

There is also a contrast between control patterns and raising patterns, as


signalled by Rizzi (1997: 331, note 24), who says:
Speakers of French are reluctant to accept CLLD with infinitives.
Nevertheless, a detectable contrast exists between control and raising
(Ch. Laenzlinger p.c.):

Rizzi provides the following contrast:

82
(43) a ??Je pense, ton livre, pouvoir le comprendre.
I think, your book, to be able tito understand.'
b *Marie semble, ton livre, pouvoir le comprendre.
Marie seems, your book, to be able it understand
Let us assume with Rizzi (1997) that Italian infinitival di and French à occupy
the head Fin, the lowest functional head in the CP domain. Let us also continue
to assume that control complements lack Force. Given the grammaticality of
(41b-c-d, 42a, 43a) we conclude that topicalisation in Italian and French control
infinitives cannot be dependent on Force. Putting it differently, the fronted topic
is not related to the following clause by the speaker-anchoring. The assumption
would have to be that in Romance there is an alternative mechanism for
achieving the link between topic and the associated clause and that this
second device is not available in English.

6.2. CLLD as IP adjunction

To account for the different distributions of topicalisation in English and CLLD


in Greek, Anagnostopoulou (1997: 166) suggests that in Romance the fronted
topic is adjoined to IP.
[the difference] follows from general conditions under which Chains of
this type are licensed. Recall that CLLD is a predication construction.
The CLLD element is the subject of predication and the rest of the
clause is the predicate, the clitic chain being the predicate variable …i.e.
the open position which permits S (CP or IP) to qualify as a predicate.
…In CLLD, IP may qualify as the predicate because it properly contains
the predicate variable chain (i.e. the clitic chain). For this reason, the
subject of predication can be adjoined to IP. (1997: 166)

According to Anagnostopoulou, English topicalisation must be related to


Spec,CP, hence cannot adjoin to spec,IP (see her account for details).
Though this analysis would indeed achieve the desired result of making
topicalisation independent of Force in Romance, an IP-adjunction approach is
also problematic in the light of the fact that topicalisation arises in control
infinitives and not in raising infinitives. In particular, the fact that the CLLD-topic
precedes the infinitival complementiser, which we assume occupies Fin, is
problematic for IP-adjunction.

6.3. CLLD and FinP

In independent work (Haegeman 2002) I have shown that Rizzi's proposal that
fronted adjuncts target TopP must be modified to account for the difference in
distribution between long moved adjuncts and locally moved adjuncts (see
Haegeman 2001 for detailed discussion). Let us assume that locally fronted
adjuncts in English (and in Romance) target FinP, the lower projection in the
CP domain36. This allows us to predict correctly that they will be admitted in
central adverbial clauses.

83
(44) a If with all these precautions you don't succeed, you should try
again.
b If after lunch he's not there, you should call him up.

In order to account for the distribution of CLLD in Romance and in Modern


Greek I propose that in those CLLD patterns, the fronted argument can
licensed by Fin, and that this is not possible in English. The question obviously
arises why Fin is able to license argument topicalisation in Romance and why it
doesn't in English. The data in (40) suggest that it is not the mere fact that the
subject occupies a post-verbal position that is crucial to allowing topicalisation
in central adverbial clauses, since even with preverbal subjects adverbial
clauses allow for CLLD, at least for some speakers.
To account for the difference between English topicalisation and
Romance CLLD I adopt some version of Anagnostopoulou's account:
topicalisation is a predication relation established between the topic and the
predicate, the clause that expresses the comment. The licensing of the topic in
the Romance Spec,FinP is a function of the phi features of Fin. Observe that
we know independently that the phi features of Fin match those of I. In West
Flemish agreement between the inflectional features of Fin and those of I are
confirmed by the morphological spell-out:

(45) a WF dan-k (ik) goan


that -sg -I (I) go
b da-se (zie) goat
that-sg she (she) goes
c dan-ze (zunder) goan
that-pl-they (they) go

In finite clauses, Fin agrees with I in terms of the subject phi features. In
addition Fin also has a tense feature, possibly encoding a Reference time (see
Bianchi 2001 for some discussion of the encoding of temporal values in C, and
for the proposal that Fin licenses person features, see Carstens (20020 for
discussion of complementiser agreement in minimalist terms).
By virtue of the clitic on I, languages with CLLD spell out object phi
features on the head I. If Fin and I agree in phi features, then it is a natural step
to propose that in Romance CLLD structures Fin agrees with I not only for
subject phi features but also for object phi features. This accessibility of the
object phi features on Fin (via their presence on I by virtue of the clitic) can
license the 'low' CLLD- topic in Spec,Fin.
If raising infinitives involve bare IP, they lack Fin, hence here is
possibility for licensing a CLLD-topic even in Romance. If, on the other hand,
control infinitives, though reduced, comprise at least FinP they will be
compatible with CLLD. (cf. Rizzi 1997: 309). 37
In English or in Japanese topicalisation does not involve the spell out of
the fronted argument as a clitic on I; object phi features are not encoded on I
and hence they cannot be picked up by Fin under agreement. Such languages,
which lack object phi features on Fin, cannot link the fronted argument to the
clause via the agreement relation in Fin and they have to have recourse to a
separate strategy for licensing topics. In the spirit of the discussions in Bayer

84
(2001) and Whitman (1989) I suggested above that in these languages
argument fronting depends on speaker anchoring; it always targets the TopP
with the feature Force.
Note that though the presence of the object clitic on I in Romance
contributes to the availability of object fronting in central adverbial clauses the
presence of object clitics as such is probably not essential to the availability of
lower topicalisation. Alternative mechanisms might achieve the same effect as
shown by the fact that Hungarian allows fronting of arguments in central
adverbial clauses while lacking the clitic doubling process (Puskas pc).

(45) Ha az orvost ' nem talàlja, hivja fell az àpolonöt.


if the doctor neg find-you call up the nurse
"If the doctor you don't find, call up the nurse".

However, Hungarian displays object agreement, so arguably in spite of the


absence of an object clitic, the head I still encodes object phi features and Fin
can inherit these features and license a low topic.
The intuition that FinP is involved in the lower topicalisation is also echoed in
recent work by Grewendorf (2002), though he exploits it differently. Simplifying
his account here for expository reasons, he arrives at the partial structures in
(46) for (i) German topicalisation of an argument (2002:48, 68)38, (ii) German
left dislocation of an argument with a D-pronoun, and (iii) Romance CLLD. In
(46b) and (46c) the topicalised DP is the specifier of a 'Big DP' and the d-
pronoun/clitic is the head of the BigDP. I refer to his discussion for details
(2002: 68).

(46) a [ForceP [TopP [FocP [FinP den Studenteni kenne [IP ich ti]]]]]
the student know I
b [ForceP [TopP den Studenteni [FocP [FinP [BIGDP ti [D den] kenne [IP ich t
BIGDP]]]]]]
the student that know I
c [ForceP [TopP Maria i [FocP [FinP [IP Leo la incontra [BIGDP ti [D tla] t BIGDP
]]]]]]
Maria Leo her meets

Grewendorf assumes that both German left dislocation and Romance CLLD
target TopP, but it is not clear how he accounts for the fact that in certain
deficient structures CLLD is possible. Within his account, the Romance CLLD
element might arguably also move to FinP, which he claims for German
topicalisation, in which case the truncated structures could still locate topics.

6.4. A lower TopP in the periphery?

An alternative to the proposal that Fin licenses topics is to adopt a more


articulated structure for the left periphery in which there is a lower specialised
position for fronted adjuncts and for lower topics. Rizzi (1997, 2001), for
instance, proposes a lower Topic position, which dominates Fin. If we assume
that the lower topic position is available in Romance and not in English the
proposal will also capture the difference between Romance and English. This

85
is the account in Haegeman (2002). However, then the question remains why
the lower position is available in one language group and not in the other. Here
again one might appeal to the properties of Fin. There is also some
independent debate as to the plausibility of postulating a lower specialised
TopP (see Benincà 2000, Grewendorf 2002: 46).

7. Summary and further questions

This paper first examines the syntax of adverbial clauses and its relevance for
argument fronting. We distinguish between peripheral and central adverbial
clauses in terms of their external syntax. Based on a number of diagnostics, I
propose that the former are merged in the derivation later than the latter,
leading to different scope relations with operators in the associated clause.
Peripheral adverbial clauses show evidence for the head Force in their CP
domain while central adverbial clauses have a reduced CP structure, lacking
Force.
In section 5, I examine the distribution of topicalised arguments in central
adverbial clauses. While fronted topics are excluded in English central
adverbial clauses, Romance CLLD is allowed in central adverbial clauses. In
section 6 I propose that while fronted topics in English are anchored to the
clause via the head Force, CLLD arguments in Romance are related to the
clause via a specifier head relation with Fin. This relation is made available by
virtue of the object phi features in Fin, inherited from the feature composition of
I.
A question that is obviously of interest is whether, if there are indeed two
processes of topicalisation in the left periphery, these lead to interpretive
differences. One suggestion that one might explore is that the Force-licensed
topic is temporally related to Speech -time and that the Fin-related topic is
temporally related to the reference time. Klein (1991) introduces the concept
Topic time for what seems to be the Reference time. I hope to explore this
question in future work.
Assuming that indeed topics are not only licensed in the specifier of TopP,
then the question raised above extends to the more general issue of
topicalisation as a uniform process: it would seem desirable to bring out some
kind of commonality between the licensing conditions of these topics.

References

Anagnostopoulou, E. 1997. "Clitic left dislocation and contrastive left


dislocation." In Materials on Left Dislocation, E. Anagnostopoulou, H.
van Riemsdijk & F. Zwarts (eds), 151-192. Amsterdam: John Benjamins.
Ashby, W. 1988. "The syntax, pragmatics, and sociolinguistics of left- and right-
dislocations in French. Lingua, 75, 203-229.
Authier, J.-M. 1992. "Iterated CPs and embedded topicalisation." Linguistic
Inquiry 23: 329-336.
Barnes, B. 1985. Left Detachment in Spoken Standard French. Amsterdam:
John Benjamins.
Bayer, J. 2001. "Asymmetry in emphatic topicalisation." Ms. University of
Utrecht.

86
Benincà, P. 2001. "The position of topic and focus in the left periphery." In
Current Studies in Italian Linguistics Offered to Lorenzo Renzi, C.
Cinque & G. Salvi (eds.), 39-64. Dordrecht: Foris.
Benincà, P./Poletto C. 2001. "Topic, focus and V2: defining the CP sublayers."
Ms. University of Padova.
Bennis, H. 2000. "On the interpretation of functional categories." In Interface
strategies, H. Bennis, M. Everaert, & E. Reuland (eds). KNAW
publications
Bhatt R, and J. Yoon 1992. "On the composition of Comp and parameters of V-
2." In Proceedings of WCCFL, D. Bates (ed.), Stanford CSLI. 10: 41-53.
Bianchi, V. 2001. "On Person Agreement." University of Pisa, ms.
Butler, J 2002. "A minimalist treatment of modality." www-
users.york.ac.uk/~jrcb100/.
Carstens, V. 2002. "Agreement in Comp." Ms. University of Missouri-Columbia.
Cinque, G. 1990. Types of A’ Dependencies. Cambridge, Mass: MIT Press
Declerck, R. & S. Reed. 2001. Conditionals: a Comprehensive Empirical
Analysis. Berlin: Mouton de Gruyter.
Doherty, M. 1985. Epistemische Bedeuting. Berlin: Akademie-Verlag.
Enç, M. 1987. "Anchoring conditions for Tense." Linguistic Inquiry 18 (4): .633-
657.
Emonds, J. 1970. "Root and Structure-preserving Transformations." Ph.D.
diss., Massachusetts Institute of Technology.
Emonds, J. 2000. "Unspecified categories as the key to root
constructions." Paper presented at the Peripheries conference,
University of York (2000).
Escobar, L. 1997. "Clitic left dislocation and other relatives." In
Anagnostopolou, van Riemsdijk and Zwarts, 233-274.
Frascarelli, M. 2000. The Syntax Phonology Interface in Focus and Topic
Constructions in Italian. Dordrecht: Kluwer.
Grewendorf, G. 2002. "Left dislocation as movement." In Georgetown
University Working Papers in Theoretical Linguistics, Volume 2, Fall
2002, S. Mauck and J. Mittelstaedt (eds), 31-81.
Haegeman, L. 1984a. "Parasitic gaps and adverbial clauses." Journal of
Linguistics 20: 229-232.
Haegeman, L. 1984b. "Pragmatic conditionals in English." Folia Linguistica 18:
485-502.
Haegeman, L. 1984c. "Remarks on adverbial clauses and definite anaphora."
Linguistic Inquiry 15 (4): 712-715.
Haegeman, L. 1984d. "Pragmatic conditionals in English." Studia Anglica
Posnaniensia, xviii, 485-502.
Haegeman, L. 1987. "Complexity and literary prose: some suggestions for
formalisation." Language and Style 20: 214-222.
Haegeman, L. 1991."Parenthetical adverbials: the radical orphanage
approach." In Aspects of Modern English linguistics: Papers Presented
to Masatomo Ukaji on His 60th Birthday, S. Chiba, A Ogawa, Y. Fuiwara,
N. Yamada, O. Koma & T Yagi (eds.), 232-53. Tokyo: Kaitakushi.
Haegeman, L. 2001. "Speculations on adverbial fronting and the left periphery."
Paper presented at the Tournesol conference, Temps et Point de Vue/
Tense and Point of View, France/Flanders (GOA-UIA Anvers, Jeune

87
Equipe Syntaxe – U. Paris), December 2001. To appear in the
proceedings.
Haegeman, L. 2002. "Anchoring to speaker, adverbial clauses and the
structure of CP." In Georgetown University Working Papers in
Theoretical Linguistics, Volume 2, Fall 2002, S. Mauck and J.
Mittelstaedt (eds), 117-180.
Heycock, C. 2002. "Embedded root phenomena." Ms. University of Edinburgh
Hooper, J. and S. Thompson. 1973. “On the applicability of Root
Transformations.” Linguistic Inquiry 4: 465-97.
Hornstein, N. 1990. As Time Goes By. Cambridge, Mass: the MIT Press.
Iatridou, S. and T. Kroch 1992. 'The licensing of CP recursion and its relevance
to the Germanic Verb-Second phenomenon." Working papers in
Scandinavian Syntax, 50, 1-24.
Jespersen, O. 1922. Language, Its Nature, Development and Origin. London:
G. Allen & Unwin.
Klein, W. 1991. Time in Language. London: Routledge.
Kuno, S. 1973. The Structure of the Japanese Language. Cambridge, Mass:
the MIT Press.
Lambrecht, K. 1981. Topic, Antitopic and Verb Agreement in Non-Standard
French. Amsterdam: John Benjamins.
Lopez, L. 2002. 'Toward a grammar without Topp and FocP. In Georgetown
University Working Papers in Theoretical Linguistics, Volume 2, Fall
2002, S. Mauck and J. Mittelstaedt (eds), 181-209.
Maki, H., L. Kaiser, and M. Ochi 1999. "Embedded topicalisation in English and
Japanese." Lingua 109: 1-14.
Meinunger, A. (2000) Syntactic Aspects of Topic and Comment. Amsterdam:
John Benjamins.
Nissenbaum, J. 2000. "Investigations of covert phrase movement." Ph.D. diss.,
Massachusetts Institute of Technology.
Paul, I. 2002. "The syntactic encoding of topic and focus." In Georgetown
University Working Papers in Theoretical Linguistics, Volume 2, Fall
2002, S. Mauck and J. Mittelstaedt (eds), 247-261.
Postal, P. 2001. "Parasitic and pseudo parasitic gaps." In Parasitic Gaps, P.
Culicover & P. Postal (eds), 253-313. Cambridge, Mass: the MIT Press.
Poutsma, H. (1926) A Grammar of Late Modern English, Part I, The Sentence.
Second half, Groningen: Noordhoff.
Quirk et al (1985) A Comprehensive Grammar of the English Language.
London: Longman.
Reichenbach, H. 1947. Elements of Symbolic Logic. New York: Free Press.
Rizzi , L. 1997. "The fine structure of the left periphery." In Elements of
Grammar, L. Haegeman (ed), 281-337. Dordrecht: Kluwer.
Rizzi, L. 2001. 'Locality and left periphery." Ms. University of Siena.
Roussou, A. 2000. "On the left periphery. Modal particles and
complementisers." Journal of Greek Linguistics 1: 65-94.
Rutherford, W. 1970. "Some observations concerning subordinate clauses in
English." Language 46: 97-115.
Sabel, J. 2002. "A minimalist analysis of syntactic islands." The Linguistic
Review, 19 (3): 271-315.
Sperber, D. and D. Wilson 1986. Relevance. Oxford: Blackwell.

88
Tellier, C. 2001. "On some distinctive properties of parasitic gaps in French." In
Parasitic Gaps, P. Culicover & P. Postal (eds), 341-367. Cambridge,
Mass: the MIT Press.
Thompson, E. 1994. "The syntax and semantics of temporal adjunct clauses."
Studies in the Linguistic Sciences 24.
Verstraete, J.-C. 2002. "Interpersonal grammar and clause combining in
English." Ph.D. diss, University of Leuven.
Zubizaretta, M-L. 1998. Prosody, Focus and Word Order. Cambridge, Mass.:
the MIT Press.
Zubizaretta, M-L. 1999. "Word order in Spanish and the nature of nominative
case." In Beyond Principles and Parameters. Essays in Memory of
Osvaldo Jaeggli, K. Johnson & I. Roberts (eds), 223-250. Dordrecht:
Kluwer.

Notes
* This paper was presented at the conference Linguistique Comparée des Langues Romanes.
Hommage à Liliane Tasmowski, held in September 2002 at the University of Antwerp. I thank
the participants of the conference for their comments and suggestions. Special thanks go to
Liliane Tasmowski for her many inspiring suggestions, to Jacqueline Guéron for clarifying some
crucial areas and to Caroline Heycock for her very useful comments on several versions of this
work. Thanks to Caroline Heycock, Ruth Huart, Brenda Laca, Anne-Marie Le Boles, Hideki
Maki, Philip Miller, Josep Quer, Luigi Rizzi and Marleen van Peteghem for help with the data.
None of those mentioned can be held responsible for the way I have used their help.
1
For a recent illustration, see Sabel's (2002) reworking of subjacency in minimalist terms. In
his paper, all adverbial clauses are grouped as giving rise to CED effects; no mention is made
of
2
the distinction discussed here.
In this paper, I sometimes use attested examples because the desired reading is more
directly accessible than would be the case with constructed examples. This is particularly useful
to illustrate the discourse-related adverbial clauses. By using attested examples there is not
need to create a discourse as well as an example, the discourse being the one associated with
the
3
original example.
It is not clear what determines whether a conjunction has dual behaviour. I leave this for
future
4
study.
As I will be assuming that the relation between the peripheral adverbial clause and the
associated clause is not really one of genuine subordination (cf. section 2.6. (20)), I sometimes
use the term 'associated clause' rather than the term 'main/matrix clause'. The latter terms,
though appropriate for the clauses modified by central adverbial clauses are less felicitous for
the clause associated with peripheral ones.
5
The conjunction although does marginally seem to allow for subject ellipsis as shown by the
following attested example.
(i) If you prefer a warmer surface, wood is resilient and will mellow beautifully with
age, but can be pricey. Laminate flooring is less expensive, although ∅ isn’t
suitable for areas that might get wet. (Ideal Home, July 2001 , page 68)
Possibly, clauses introduced by although are indeed near-co-ordinate. An additional argument
for this is provided by Verstraete (2002), who points out that although clauses, unlike while
clauses, may contain imperatives:
(ii) a The students should have enough money, although remember we are
expecting a drop in the department funding.
b ??The students should have enough money while remember we are expecting
a drop in the department funding.
(iii) a. I wouldn't worry too much about it, although do prepare for words with this
enemy ( Verstraete 2002: 147)
b ??I wouldn't worry too much about the syntax exam, while do prepare carefully
for the semantics.
Liliane Tasmowski (pc) points out similar contrasts in French: parce que ('because') allows for
an imperative but puisque ('since') does not.

89
One might use such contrasts in support of a further subclassification of adverbial clauses.
Based on work by Verstraete (2002) and pursuing a suggestion by Jacqueline Guéron, it might
be proposed that some peripheral adverbial clauses have Mood though not Force, while others
have
6
both Mood and Force. See also Roussou (2000) for Mood and the left periphery.
This contrast between the two types of clauses is not very sharp, as pointed out by Caroline
Heycock (pc). Judgements are difficult because peripheral adverbial clauses, being near-
coordinate will allow (marginally) for ATB extraction. See also the discussion in Postal (2001).
7
8
Cf. I enjoyed the conference very much while I disliked his paper.
One might further explore this analysis using recent proposals by Butler (2002). He argues
that just like IP is associated with a periphery (CP) VP is associated with a peripheral domain.
The CP associated with VP would be the domain that hosts central adverbial clauses. For
reasons of space I cannot go into this here.
9
As discussed in section 2.2.1 (examples (5), (6)) I assume that the patter is not identical to
that of co-ordination.
10
But locally fronted adjuncts are possible
(i) If with all these precautions things still go wrong, you should call the police
This shows clearly that argument fronting and adjunct fronting are not identical. See section
6.3.
11
12
Thanks to Hideki Maki (pc) for the Japanese data.
See also discussion in Heycock (2002).
13
Emonds (2000) offers an account according to which MCP occur in what he calls 'Discourse
Shells'. Discourse Shells are categorically unspecified, and 'may immediately dominate (only)
IPs specified as 'Discourse projections'. Unembedded clauses are always potential discourse
projections; particular languages may also specify progressively larger classes of finite clauses
as discourse projections. (Emonds 2000:8).
In my account the head Force licenses what would probably correspond to Emonds's
Discourse Shells. I do not assume, though, that Force may be arbitrarily available in embedded
domains depending on language specific properties. Rather, I would assume there is
interpretive basis for the availability of Force. In section 6, I propose that languages vary as to
the functions Fin can assume, one of which may be topic licensing. This account should not be
taken
14
to imply that Fin acquires the feature Force in specific languages.
There is an important difference between English and Romance; this is discussed in
sections
15
5 and 6.
That non-finite clauses cannot contain MCP is also endorsed by Emonds (2000: 8).
16
For similar proposals to differentiate the complements of bridge verbs and non-bridge verbs
in terms of structural truncation see Benincà (2001), Benincà/Poletto (2001), Grewendorf
(2002).
17
For reasons of space I cannot show in detail how my account deals with temporal
subordination
18
in central adverbial clauses. I intend to elaborate this in future work.
Mainly on the basis of Romance data, Rizzi (1997) introduces a second topic position to the
right of Focus. It is not clear that this is generally available and may well be specific to
Romance
19
(cf. discussion in Grewendorf (2002) and also section 6.4).
Alternatively, FORCE could be seen as a feature on the subordinating head Sub, but this
raises question of root clauses which do have illocutionary force but presumably normally lack
the
20
subordinating head.
21
I discuss non-finite infinitival clauses in section 6.1..
Conceivably, the dependency should be stated the other way and Force depends on Speech
time. What is distinctive in peripheral adverbial clauses and what licences MCP would then be
the syntactic encoding of Speech Time (S). Such a position would entail the reworking of the
claims made below but is not incompatible with the main argumentation of this paper. For
instance, epistemic modality might be argued to depend on S. I hope to look into different
alternative
22
formulations in future work.
Whitman (1989) postulates a link between topicalisation and the availability of modal
markers. If epistemic modality depends on Force (as suggested above), and if topicalisation
also may depend on Force, this is expected.
The restrictions could also be restated in terms of the alternative proposal briefly introduced
in note 21: if topicalisation depends on epistemic modality and if the latter is anchored to
speech time, we predict topicalisation will be excluded from non-root environments.
23
In (33), ForceP dominates TopP, but the ordering may well be the other way round as
discussed in Haegeman (2002). This would mean that (33) is replaced by (i).
(i) a Central adverbial clause: Sub Fin IP
b Peripheral adverbial clause: Sub TopForce Focus ForceForce Fin IP
c Root clause: Top ForceFocus ForceForce Fin IP
d : *Sub TopForce Focus ForceForce Fin IP

90
Along Minimalist lines, Top and Focus would have an uninterpretable FORCE feature, which
is deleted by Force to Top movement or by agreement.
24
Crucially, I do not assume that fronted adjuncts are invariably topics. See Haegeman (2001)
and section 6.3. for discussion. I leave aside the question whether English topicalisation
involves
25
a null operator in the CP domain (cf. Rizzi 1997).
(a,b): judgements Rizzi (pc), (c) Frascarelli (2000: 151).
26
27
Frascarelli 2000: 151, her (183a)
Judgements Josep Quer (pc). The data are apparently not uniform across speakers as
shown by the following citation from Escobar (1997):
For Spanish, left-dislocated phrases with CLLD cannot appear embedded with
subjunctive mood which otherwise seem to facilitate the most clear cases of
embedding in Spanish:
[i] ??/*Ella prefiere que a Luis, el médico lo examine
She prefers the that a Luis, acc-cl the doctor examines
…we may conclude that CLLD is a root phenomenon. (Escobar 1997: 248, italics
mine)
28
29
Judgements Josep Quer.
For the use of dislocation in spoken French, see also Ashby (1988), Barnes (1985),
Lambrecht
30
(1981).
Thanks to A Le Boles and Philip Miller for help with the French examples. Again there is
variation in judgements, but this may well be due to prescriptive attitudes (Marleen van
Peteghem
31
p.c.).
32
Such an account would be along the lines of work by Paul (2002).
Pushing the account in the text, note that it may no longer be clear whether one should
actually postulate a designated position TopP (as assumed by Rizzi 1997). It could be that the
operation of topicalisation is parasitic on some other property of the left periphery, say, for
instance, the encoding of speech time or of reference time, or the availability of phi features.
Meinunger (2000) interprets IP-internal scrambling in German as topicalisation. This also
means that topicalisation is an operation not specifically associated with the left periphery, and
presumably it depends on the availability of other properties. Meinunger relates topicalisation
and agreement features.
33
For another alternative approach in terms of multiple specifiers see also López (2002).
34
See Zubizaretta (1999) for a slightly different implementation.
This is not generally accepted , though, cf. Sabel (2002), López (2002) and many others.
In English (i) a parenthetical element can be inserted to the immediate left of the raising
infinitive:
(i) a John seems, as far as I can tell, to be completely unaware of the problems.
b John seemed, however, to be completely unaware of the problems.
36
In fact, in Haegeman (2002) and following Rizzi (2001), I postulate a lower TopP and a
lower ModP for locally fronted adverbial adjuncts. See the discussion of this proposal in section
6.4.
37
Zubizaretta (1999: 241) gives the following Spanish example as ungrammatical. Observe
that there is no overt spell out of an infinitival complementiser:
(ii) Maria piensa la carta escribirla
Maria thinks the letter write+it
The absence of the complementiser might be taken to indicate either a further structural
reduction or a weakening of the featural composition of Fin. The same account could be used
to account for the degradation of French (36).
38
My summary is based on his discussion on page 48 but topicalisation in German is not
developed in detail.

91

You might also like