Clausal Subordination and the Structure of the Verbal Phrase

Abstract

In his first approach to recursion in clausal embedding, Chomsky (1957) postulates a proform in the matrix clause linked to an independently constructed clause that, via an application of the generalised transformation, eventually becomes the matrix verb’s complement. Chomsky (1965) replaces this with a direct clausal embedding analysis, with clausal recursion in the base component of the grammar. I argue here that, while direct clausal recursion is certainly needed, an update to the Chomsky’s (1957) approach (minus the application of the generalised transformation) deserves a prominent place in syntactic theory as well. The discussion is based on data from Dutch, German, and Hungarian. This paper addresses the role of presuppositionality in the context of clausal coordination, the analysis of the so-called wh-scope marking construction, and the importance of Agree in connection with a subordinate clause’s transparency or opacity to extraction. Central in the analysis is a perspective on the structure of the verbal phrase which accommodates two discrete structural positions for the object.

1. Introduction

In the first approach to recursion in clausal embedding in the transformational-generative framework (see [1]), the matrix clause contained a proform linked to an independently constructed clause that, after association with the proform and subsequent elimination of this proform via the generalised transformation, ended up serving as the subordinate clause. (1) sums this up.

1.	a.	John believes that Mary is pregnant
		John believes that Mary was kissed by Bill
	b.	[S John believes it]
		[S’ that [S Mary is pregnant/was kissed by Bill]]
	c.	[S John believes [S’ that [S Mary is pregnant/was kissed by Bill]]]

In [2], Chomsky abandons this proform-based approach to clausal recursion, and substitutes it with a direct clausal embedding analysis: the matrix verb selects the subordinate clause directly as its object, in the base component, which includes a base rule rewriting VP as V + S’. Appropriately updated, the Aspects approach is a staple of mainstream generative theory today—and there can be no doubt that we need it. But there is reason to believe that an update of (1b), the hallmark of the Syntactic Structures approach [1], deserves a prominent place in current syntactic theorising as well.

In this paper,1 I will present an analysis of clausal subordination that mobilises both direct recursion and a proform-based strategy, each addressing different subspecies of embedding of a clause within a larger clause. The empirical discussion will be based on data from Dutch, German, and Hungarian. This last language is particularly informative thanks to the fact that (to recycle the old adage used frequently with reference to Hungarian in other contexts) it wears (1b) on its sleeve.

This paper is structured as follows. In Section 2, I begin by looking at the role of presuppositionality in the context of clausal coordination, showing that in Hungarian the distribution of a proleptic pronoun associated to the clause (very much as in (1b)) plays a key part in determining whether the subordinate clause gets a presuppositional reading or not, whereas in Dutch it is the placement of the subordinate clause vis-à-vis the verbal cluster that determines this. Section 3 provides a structural perspective on object positions and presuppositionality which allows us to understand the empirical picture emerging from Section 2. In Section 4, I subsequently apply the insights gained from Section 2 and Section 3 to an analysis of the so-called wh-scope marking construction, with particular emphasis on German and Hungarian. Section 5 adds a note on the role of Agree in determining whether a subordinate clause is transparent or opaque to extraction. Section 6 concludes.

2. Clausal Subordination and Presuppositionality

Hungarian clausal subordination constructions participate in three different syntactic patterns. In one, which looks very much like what we find in English, the matrix verb combines directly with the subordinate clause. We see this in the a-examples in (2) (for hisz ‘believe’, a bridge verb in the sense of [4]) and (3) (for beismer ‘confess’, a factive verb). In the second pattern, the verb combines with an accusative pronoun (azt ‘it. acc’) in addition to the subordinate clause. This is illustrated in the b-examples in (2) and (3). The third pattern differs from the second in featuring the proform úgy ‘so’ in lieu of azt, and in being available only for bridge verbs: (3c) is ungrammatical.

2.	a.	János hiszi,	hogy	Mari terhes	(Hungarian)
		János believes	that	Mari pregnant
	b.	János azt hiszi,	hogy	Mari terhes
		János it-acc believes	that	Mari pregnant
	c.	János úgy hiszi,	hogy	Mari terhes
		János so/thus believes	that	Mari pregnant
		‘János believes that Mari is pregnant.’
3.	a.	János beismeri,	hogy	Mari terhes	(Hungarian)
		János admits/confesses	that	Mari pregnant
	b.	János beismeri azt,	hogy	Mari terhes
		János admits/confesses it-acc	that	Mari pregnant
	c.	* János beismeri úgy,	hogy	Mari terhes
		János admits/confesses so/thus	that	Mari pregnant
		‘János admits/confesses that Mari is pregnant.’

The interpretation of (2a) is not necessarily the same as that of (2b) or (2c): when azt or úgy is absent, the hogy ‘that’ clause in bridge-verb constructions has a strong tendency to be interpreted as denoting “a proposition (without a necessary commitment to its truth) about which the complex sentence makes an assertion”, in the words of de Cuba and Ürögdi [5] (p. 37). De Cuba and Ürögdi show convincingly that the proposition denoted by the embedded clause in sentences of the type in (2a) is not presupposed to be true (it can in fact be known to be a lie), so the effect of omitting azt or úgy in (2) is not equivalent to “factivity” (cf. (3)). But the content of this proposition does have to be taken to be part of the common ground—and it is in this sense that I will use the term “presuppositional” in this work (close to de Cuba and Ürögdi’s “referential”, though by their own admission that their evidence that the CP in sentences such as (2a) is a referring expression is “impressionistic at best” [5] (p. 45)). Thus, the hogy ‘that’ clause in bridge-verb constructions lacking proleptic azt/úgy shows a strong tendency to be interpreted presuppositionally, similar to but not quite in the same way as in factive-verb constructions. In factive-verb constructions, azt can be absent or present (cf. (3a,b)); but the other proform, úgy, cannot be used.2

So in Hungarian, though both bridge verbs and factive verbs allow their subordinate clause to co-occur with the proform azt, the absence of this pronoun usually has a presuppositionalising effect on the embedded clause in the case of bridge verbs. In Dutch, we see a similar presuppositionalising effect with bridge-verb complements, this time manifesting itself in terms of linear order. When a subordinate clause is placed in the Mittelfeld, to the left of the verbal cluster, it is obligatorily interpreted presuppositionally: thus, the interpretation of (4a) is on a par with that which reference [5] describes for Hungarian (2a); to get the “ordinary” non-presuppositional interpretation, the subordinate clause has to be placed in the Nachfeld, following the verbal cluster. For factive verbs (5), placement of the subordinate clause to the left or to the right of the cluster has no interpretive effect. Although placing a finite clause in the Mittelfeld is always a marked option in Dutch, the result is typically less marked with factive verbs than with bridge verbs, precisely because a presuppositional interpretation for the subordinate clause is guaranteed with factives.

4.	a.	dat	Jan	[dat Marie zwanger is]	gelooft	(Dutch)
		that	Jan	that Marie pregnant is	believes
	b.	dat	Jan	gelooft	[dat Marie zwanger is]
		that	Jan	believes	that Marie pregnant is
5.	a.	dat	Jan	[dat Marie zwanger is]	betreurt	(Dutch)
		that	Jan	that Marie pregnant is	regrets
	b.	dat	Jan	betreurt	[dat Marie zwanger is]
		that	Jan	regrets	that Marie pregnant is

These contrasts between bridge-verb and factive constructions point to an important structural difference between factive and non-factive constructions with regard to the way the syntax integrates the subordinate clause into the structure of the complex sentence. A proper syntactic understanding of how this comes about will tell us a lot about the way clausal subordination works in natural language, and how the grammar should operate in order to deal adequately with the syntax of clausal hypotaxis.

3. Object Positions, Proforms and Presuppositionality

In this section, I argue that the observations about Hungarian and Dutch made in Section 2 translate structurally into a syntactic analysis in which the verb phrase accommodates two object positions—much like the way the clause accommodates (at least) two subject positions. One object position is the familiar complement-of-V position; the other is a specifier position—either the specifier of the projection of the verbal root (VP) or that of a functional category (e.g., AspP) projecting between v and VP. For simplicity (and since nothing hinges on the choice between these options here), I will work in this paper with the structure in (6), where the higher of the two object positions is SpecVP (see [6,7,8,9,10,11,12], i.a., for relevant discussion).

6.

a.

[vP subject [v’ v [VP <object> [V’ V <object>]]]]

Whenever the occupant of SpecVP is an argument (i.e., receives a θ-role in this position), it is interpreted presuppositionally—i.e., outside the domain of existential closure (see [13]), which, for the object in (6), is the V’ (see [3] (Section 2.3.1.2) and [12] for discussion).3

This link between the thematic object’s occupancy of SpecVP and its presuppositional interpretation immediately addresses the contrast between (4a) and (4b) in Dutch. When the subordinate clause appears in SpecVP, it surfaces to the left of the verb (which, in non-root contexts, is spelled out at V in Dutch) and occurs in the Mittelfeld; it obligatorily receives a presuppositional interpretation, thanks to being outside V’. On a non-presuppositional interpretation, the subordinate clause is in the complement-of-V position, which, for clauses, is linearised to the right of the verb in Dutch.4

The structure in (6), with its two positions for objects, also helps us solve the puzzle posed by the Hungarian examples in (2) and (3). The proposal runs as follows.5

With factive verbs such as beismer ‘admit/confess’, SpecVP is always projected, because the complement-of-V position is occupied by a secondary predicate headed by “fact”—a development of the classic approach to factives in [16], but with fact now serving as a secondary predicate at the level of the VP (rather than as the head of a complex noun phrase, as in Kiparsky and Kiparsky’s original proposal).6 The occupant of SpecVP is the subject of V + fact. Just as in copular sentences, this subject can be either the subordinate CP itself (cf. [that Mary is pregnant] is a fact) or a proform linked to an extraposed CP outside VP (cf. it is a fact [that Mary is pregnant]).7 Whenever it occurs in factive-verb constructions, the pronominal proform is a thematic argument of the complex predicate formed by the verb and the secondary predicate fact.

Since with factive verbs the complement-of-V position is always occupied by the abstract secondary predicate fact, this position is never available for the subordinate CP. In (7a), the CP occupies the SpecVP position. The secondary predicate fact, occupying the complement-of-V position, forms a complex predicate with V and introduces its argument in SpecVP.8 In (7b), the argument of the complex predicate V + fact is the proform occupying SpecVP, coindexed with the CP in extraposed position. So in factive-verb constructions, the CP is always associated with the θ-role assigned to SpecVP, which ensures that it is always given a presuppositional interpretation, regardless of which of the two structures in (7) is built.9

7.	a.	[vP subject [v’ v [VP <object> [V’ V <object>]]]]
	b.	[vP subject [v’ v [VP proformi = arg [V’ V [PRED fact]]]]] [CP ... ]i

With verbs such as hisz ‘believe’, when azt is not present, the subordinate clause has a choice of positions: it can either take the complement-of-V position, as in (8a.i), or be mapped into the SpecVP position, so that it comes to behave exactly like the object clause of a factive verb (cf. (7a) and (8a.ii)). (Of course a blend of (8a.i) and (8a.ii), with a CP in each of the two positions in the VP, is uninterpretable: V’s argument structure accommodates no more than one propositional argument).

8.	a.i	[vP subject [v’ v [VP V [CP ... ]]]]
	a.ii	[vP subject [v’ v [VP [CP ... ] [V’ V]]]

Hungarian speakers, for whom the presuppositional interpretation of the hogy-clause in (2a) is categorical, select (8a.ii) over (8a.i) whenever no proform is present.10 But categorical presuppositionality for (2a) is not the norm: it is, for most speakers, merely a strong tendency (as I noted in the paragraph below (3)). With verbs such as hisz ‘believe’, it will usually be possible to map the subordinate clause into the complement-of-V position.

Whenever azt is present with verbs such as hisz ‘believe’, it once again occupies SpecVP, just as in constructions with factive verbs. But whereas in factive-verb constructions azt is an argument of the complex predicate V + fact, in bridge-verb constructions this azt is not an argument but a secondary predicate for the complement clause occupying the complement-of-V position, as in (8b), which recalls Moro’s analysis of sentences such as it’s that she’s pregnant, for which he treats it as the predicate of the that-clause [21].11

8.

b.

[vP subject [v’ v [VP proform=pred [V’ V [CP ... ]]]]]

In (8b), even though the SpecVP position is occupied, we get no presuppositional reading for the subordinate clause. That is because the subordinate clause itself occupies the complement-of-V position, which is within the nuclear scope (V’), and SpecVP is occupied by a predicate, not by the CP or a placeholder for it. Even with the predicate being assigned a presuppositional interpretation in the semantics, this does not accrue to its subject: in a copular inversion construction such as the winner must be someone from New York, the fact that the predicate nominal occupies SpecTP, a position outside the nuclear scope, does not prevent the notional subject someone from New York (which is inside the nuclear scope) from being interpreted non-specifically.

As far as the distribution of presuppositional readings in bridge and non-bridge constructions is concerned, the analyses in (7) and (8) make exactly the right predictions. This analysis of the bridge/factive dichotomy also directly explains the fact that azt alternates with úgy ‘so’ in bridge-verb constructions but not with factives (recall the c–examples in (2) and (3)). This falls out once we realise that úgy ‘so’ is always a predicate: it has no argumental functions. Placing úgy in SpecVP in (7b) would cause a clash between the fact that a θ-role is assigned to this position by the complex predicate V + fact and the fact that úgy tolerates no θ-role. In (8b), by contrast, azt plays a predicational role; replacing it with úgy should be perfectly fine, and indeed it is, as we saw in (2c). We see the same alternation between az and úgy with semi-copulas such as látszik ‘seem/appear’, as in (9). Here, since there is, in fact, no other predicate around (látszik is merely a copula), az and úgy play the role of primary predicate for the hogy-clause.

9.	a.	az	látszik,	hogy	Mari terhes	(Hungarian)
		it	seems	that	Mari pregnant
	b.	úgy	látszik,	hogy	Mari terhes
		so	seems	that	Mari pregnant
		both: ‘it seems/appears that Mari is pregnant.’

A reviewer points to two respects in which the distributions of the proleptic proforms azt and úgy are different in the realm of bridge-verb constructions. One is that insertion of sentential negation in (2c) is ungrammatical while it is perfectly fine in (2a). The examples in (10) (provided by the reviewer and verified by me with another native speaker) illustrate this:

10.	a.	azt	nem	gondolta,	hogy	Mari terhes	(Hungarian)
		it.acc	not	thought	that	Mari pregnant
	b.	* úgy	nem	gondolt,	hogy	Mari terhes
		so/thus	not	thought	that	Mari pregnant
		‘he doesn’t think that Mari is pregnant.’

The other difference between azt and úgy with bridge verbs is that the string azt+hogy-clause can be fronted as a unit (either via ordinary topicalisation or via left dislocation; in the latter case, János is immediately preceded by a resumptive proform—the second azt in (11a)) while fronting of the string úgy+hogy-clause is impossible:

11.	a.	azt,	hogy	Mari terhes,	(azt)	János	is	mondta	(Hungarian)
		it. acc	that	Mari pregnant	it. acc	János	also	said
	b.	* úgy,	hogy	Mari terhes,	(azt/úgy)	János	is	mondta
		so/thus	that	Mari pregnant	it. acc/so	János	also	said

These two observations are probably relatable to a single factor: movement of úgy. The grammaticality of (11a) suggests that azt can stay within the node in which it forms a constituent with the hogy-clause;12 the ill-formedness of (11b) suggests that úgy cannot. If, indeed, úgy must leave its base position, the ungrammaticality of (10b) falls out as a case of intervention (an “inner island” effect in the terminology of [22]): úgy, a non-referential expression (a predicate), must move out of the verb phrase but in so doing crosses the negation operator, which blocks the relation between the moved úgy and its trace. The grammaticality of (10a) (which has the same linear order as (10b): azt, too, is to the left of nem “not”) is then left to be dealt with. One possibility would be to assign this azt referential (hence argumental) status (so that it is immune to the inner island)—which would lead to a presuppositional interpretation of CP. Alternatively, azt is, like úgy, a predicate, but unlike úgy in (10b), the azt in (10a) has not crossed over the sentential negation operator: though it occurs to the left of nem, the morphological marker of sentential negation, it remains below the abstract operator (outside TP) supplying the semantics of negation (see [3] (chapter 3) for some relevant discussion). These remarks are tentative. There clearly is much more to be said about these examples. They should be revisited in future research.

The analyses of clausal subordination in (7) and (8) and the brief remarks about (10) provide a natural launching pad for an approach to so-called wh-scope marking constructions that finds a natural home for the “wh-expletive” that occurs in them.13 I turn to this next.

4. On the Syntax of wh-Scope Marking

In the wh-scope marking constructions (also known as partial wh-movement constructions) in (12a) (from German) and (13a) (from Hungarian), we are dealing with root wh-questions in which the wh-constituents wer and ki have matrix scope, just as in the long-distance wh-fronting constructions in (12b) and (13b) (which are dispreferred, to a greater or lesser degree, whence the “^%”). But in (12a) and (13a), the wh-operators wer and ki are not at the left edge of the matrix clause: their scope is marked by an “expletive” wh-element corresponding to English what.

12.	a.	was	glaubt	Hans,	wer	schwanger	ist?	(German)
		what	believe	Hans	who	pregnant	is
	b.	% wer	glaubt	Hans,	dass	___	schwanger	ist
		who	believe	Hans	that		pregnant	is
		both: ‘who does Hans believe is pregnant?’
13.	a.	mit	hisz	János,	hogy	ki	terhes?	(Hungarian)
		what	believe.3sg.indef	János	that	who	pregnant
	b.	% ki	hiszi	János,	hogy	___	terhes?
		who	believe.3sg.def	János	that		pregnant
		both: ‘who does János believe is pregnant?’

In Hungarian, it is clear that the verb agrees with this “wh-expletive” and assigns case to it: mit in (13a) bears the accusative case particle -t, and the inflectional form of hisz ‘believe’ is from the indefinite/subjective agreement paradigm, unlike what we see in (13b), where hiszi agrees in definiteness with the finite subordinate clause (which the grammar of Hungarian treats as definite). This property of (13a) makes it plausible to assume that upstairs, (13a) is the wh-counterpart to (2b), with mit replacing azt (see [23] for the original insight): while accusative azt is definite and triggers a form of the matrix verb from the definite/objective conjugation, accusative mit is indefinite and cooccurs with indefinite/subjective inflection. Thinking of (13a) along these lines, and bearing in mind the treatment of azt in (2b) presented in (7b), we immediately procure an analysis of the wh-scope marking construction that finds a home for the “wh-expletive”: it originates in the SpecVP position, with the subordinate clause occupying the complement-of-V position.

Rizzi [24] notes that German wh-scope marking constructions with bridge verbs resist the presence of a sentential negation in the upstairs clause (see also [25], [26] (p. 378)), in contradistinction to their long wh-fronting counterparts:

14.	a.	* was glaubst du nicht, mit wemHans sich dort treffen wird?				(German)
		what believe you not with whom Hans refl there meet will
	b.	mit wem glaubst du nicht, dass Hans sich dort treffen wird?
		with whom believe you not that Hans refl there meet will
		‘who don’t you think that Hans will meet there?’

In Hungarian, this effect of negation also manifests itself clearly in the wh-scope marking construction: (15) is systematically rejected when nem is included in it (see [23]). What is particularly interesting in the context of the discussion earlier in this paper, however, is that Horvath points out that the wh-scope marking construction in (16a) is immune to the presence of matrix negation while long A‘–fronting of the meaningful wh-constituent across the negation leads to an ill-formed result, as shown in (16b) [23] (p. 536).

15.	mit (* nem) gondolsz, hogy ki fog elmenni?	(Hungarian)
	what-acc not think-2sg.indef that who(nom) will pv-go

16.	a.	mit nem ismert be János, hogy hányszor hamisította az aláírásodat?	(Hungarian)
		what not admitted János that how.many.times forged the signature-2sg-acc
	b.	* hányszor nem ismerte be János, hogy hamisította az aláírásodat?
		‘how many times didn’t János admit that he had forged your signature?’

The ungrammaticality of (16b) is an inner island effect induced by overt-syntactic fronting of a non-argumental wh-expression.14 The intervention effect seen in (15) and also in German (14a) can be assimilated to (16b) if the wh-scope marker (German was, Hungarian mit) is a non-argumental wh-operator in these cases. By this logic, mit in (16a) should be an argumental wh-expression: otherwise, it would be difficult to account for its immunity to nem-intervention.

The difference between the intervention-sensitive examples of the wh-scope marking construction in (14a) and (15), on the one hand, and the sentence in (16a), on the other, lies in the nature of the matrix verb (non-factive glauben, gondol ‘think’ versus factive beismer ‘admit, confess’)—and, concomitantly, in the (non-)presuppositional nature of the complement clause. This is systematic: whenever a matrix verb is used whose CP complement is presuppositional (or D-linked, in Horvath’s terms [23]), no intervention effect manifests itself in the wh-scope marking construction.

We can make immediate sense of this in light of the discussion in Section 3. There, I argued that the complement position of a factive verb is always taken by an abstract secondary predicate fact, and that the SpecVP position is occupied by the argument of that secondary predicate—the subordinate clause itself, or a proleptic object (in which case the clause is merged as a satellite; see fn. 7). An argumental expression in the SpecVP position receives a presuppositional interpretation. Thus (7a), above, directly accounts for the presuppositional status of the factive object clause. And (7b) does so indirectly, by interpreting the proform as presuppositional/D-linked, and having the clause associated to it via a relationship of apposition. It is (7b) that, by realising the proform as the wh-element mit, gives rise to the wh-scope marking construction in (16a). The thing to note is that mit here is an argumental wh-expression, immune to the inner island set up by the negation.

In the syntax of bridge-verb constructions with a place-holder for the complement clause, the proform is once again in SpecVP, as in (8b). But here, the proform in SpecVP does not play the role of an argument: it is the CP in the complement-of-V position that serves as the argument; the proform is a secondary predicate of this CP, in a reverse predication structure. The hypothesis that the proform in the SpecVP position of bridge-verb constructions is not an argumental expression had already accounted for the non-presuppositional interpretation of the complement clause in bridge-verb constructions—and it now also derives the intervention effect seen in German (14a) and Hungarian (15): building a non-argumental bare wh-dependency across a scope-taking element is impossible (see also (16b)).

5. A Note on wh-Dependencies across a Subordinate Clause

In wh-scope marking constructions such as Hungarian (13a), movement of the “real” wh-constituent into the matrix clause, across the CP “associate” of the wh-scope marker, is never allowed, regardless of how the “real” wh and the wh-scope marker are ordered vis-à-vis one another:

17.	a.	* ki	mit	hisz	hogy	___	terhes?	(Hungarian)
		who	what	believe.3sg.indef	that		pregnant
	b.	* mit	ki	hisz	hogy	___	terhes?
		what	who	believe.3sg.indef	that		pregnant

The ungrammaticality of the examples in (17) is arguably a consequence of the fact that the hogy-clause here is not an Agree-goal for the verb: the verb hisz in these sentences, as in (11a), bears indefinite/subjective inflection, agreeing with the scope marker mit rather than the finite hogy-clause (which is definite). As shown forcefully by Rackowski and Richards [27] and Van Urk and Richards [28], subordinate domains are transparent for extraction only if they are Agree-goals.

For speakers who are tolerant of long-distance-fronting in Hungarian, (13b) is grammatical. This falls out from the fact that the matrix verb in (13b) bears definite/objective inflection (hiszi), agreeing in definiteness with the hogy-clause from which extraction is taking place. In [3], I show in depth that the hypothesis that filler–gap dependencies can be established only across domains that serve as Agree-goals for higher probes gives us a purchase on the entire spectrum of “strong island” effects. I refer the reader to this work for details and discussion.

6. Conclusions and Consequences

A non-presuppositional subordinate clause can be generated as the verb’s complement. But the grammar also countenances the possibility of base-generating a proform in the higher of the two object positions (SpecVP) and associating the subordinate clause to this proform, in the spirit of (1b): this is what happens in (7b), with factive matrix verbs. Bridge-verb constructions can mimic the structure in (7b), but the proform in (8b) has properties that are very different from those of the occupant of SpecVP in (7b): instead of being an argument, it serves as a secondary predicate for the CP, which in (8b) (unlike in (7b)) occupies the complement-of-V position.

Chomsky’s original proposal for the syntax of clausal subordination (see [1]) has now morphed into an analysis that encompasses both bridge-verb and factive-verb constructions, makes sense of the distribution and form of the proforms, accounts for extraction, and takes care of the case and agreement facts. In Chomsky’s original proposal, the clause associated to the proform is not merged into the structure of the complex sentence as an independent constituent, alongside the proform: rather, the subordinate clause is merged in via an application of the generalised transformation, which effaces the proform and turns the embedded clause into the verb’s object. The proposal advanced in this paper has no business with the generalised transformation: the proform, whenever present, never gets replaced; the proform and the CP each occupy their own positions in the tree, with the CP to which the proform is associated sitting in the complement-of-V position in bridge-verb constructions, and in a clause-peripheral position in factive-verb constructions with an overt object pronoun.

In the structure in (8b), the proform must be merged into the structure before the subordinate CP is merged in the complement-of-V position. This is so because the predicative proform and the verbal root must form a complex predicate that takes the CP in the complement-of-V position as its subject. Such a complex predicate can only be formed, in the structure in (8b), if this structure is built from the top down. On a bottom-up derivation, CP is first merged directly with the verb, at the V’ juncture. At this point, CP is interpreted as an argument of the verb. Upon the subsequent arrival of the predicative proform in the SpecVP position, we could countenance a predication relation between the complex predicate “proform + V” only by revising the conclusion, drawn at the V’ juncture, that CP is an argument of V alone. Such a revision would amount to a derivation that is not strictly cyclic.

The top-down approach, by contrast, delivers the complex predicate “proform + V” before CP is merged into the structure. Upon merger of CP in the complement-of-V position (the last position reached in the course of the top-down structure-building process), CP is interpreted right away as the argument of the complex predicate formed by the proform in SpecVP and the V-head. No revision of a conclusion drawn earlier is necessary—the derivation proceeds strictly cyclically.

The outcome of this discussion of clausal subordination thus bears, in an important way, on the directionality of structure building. In [3], the pros and cons of top-down and bottom-up structure building and syntactic derivation are discussed at much greater length. The interested reader is referred to this work for further details.