Colloquial English: Structure and Variation
Metadaten – PDF
Radford, Andrew. 2018. Colloquial English: Structure and Variation. 1st edn. Cambridge University Press. https://doi.org/10.1017/9781108552202.
- Zotero öffnen
Imported: 2023-05-15 18:36
Such structures are interesting for (at least) five different reasons. One is that they raise the question of whether (as prescriptivists would have us believe), non-canonical structures like those above are simply instances of ‘sloppy grammar’ produced by people who have an inadequate mastery of the syntax of ‘proper English’, so that sentences like those in (1) have no real structure (or have a ‘wild’ structure not conforming to principles of Universal Grammar (UG)). In this monograph, I shall argue strongly against this view, and instead maintain that they have a UG-compliant structure of their own, and that studying this structure closely tells us a great deal about parametric variation in syntax.
two principles widely considered to be universal. One is the Phase Impenetrability Condition/PIC of Chomsky (1998) under which a constituent c-commanded by a phase head P is impenetrable to any constituent c-commanding the maximal projection of P. If all finite clauses are CPs and all CPs are phases (as Chomsky claims), PIC would bar T-has from attracting what to move from being the specifier of T-is to becoming the specifier of T-has, because there is a CP phase boundary
Highlight ( p. 15)
Inactivity Condition of Chomsky (2008: 150) which makes an A-chain inactive for further syntactic operations once its uninterpretable features have been valued
the clause periphery – i.e. that part of the structure on the left edge of the clause, preceding the subject.
In early work in generative grammar in the 1950s and 1960s (e.g. Chomsky 1955, 1957, 1964, 1965, 1966; Lees 1960; Rosenbaum 1965, 1967; Ross 1967) clauses were taken to be S-constituents generated by a phrase structure rule such as S → NP AUX VP (Chomsky 1965: 68) which can be seen in contemporary terms as specifying that S is formed by merging (i.e. combining) NP, AUX and VP.
Highlight ( p. 20)
complementisers (in the analysis of Rosenbaum 1967) were absent from the underlying structure, and inserted on the left edge of S via a Complementiser Placement transformation.
Highlight ( p. 21)
peripheral constituents in work of this era were generally treated as positioned on the edge of S, serving as daughters of S and sisters of the subject. 50s/60s S-analysis
A step change in research on the periphery came with work on the syntax of questions by Katz & Postal (1964), refined in later work by Chomsky (1968), Jacobs & Rosenbaum (1968) and Baker (1970). They posited that wh-questions contain an S-initial question morpheme (designated as Q), generated by a phrase structure rule of the form S → (Q) NP AUX VP (Chomsky 1972: 144).
Baker (1970: 209) argues that the Q-replacement analysis provides ‘an explanation for the fact that only one questioned constituent may be moved’. He further posits (1970: 207) that the complementisers if/whether are ‘lexical realizations of the Q morpheme’
Highlight ( p. 23)
A key characteristic of the S-analysis of clause structure was the assumption that all peripheral constituents are positioned on the edge of S and thus are immediately contained in the same S-projection as (and hence are sisters of) the subject and auxiliary. However, some work towards the end of the S-era treated phrasal peripheral constituents as adjoined to S…It also marks a move away from the earlier assumption that clauses have a flat structure, towards a more hierarchical structure.
The assumption that the clause periphery has a hierarchical structure was taken a step further in seminal work by Bresnan (1970, 1972, 1976, 1979). She argued that COMP is contained within a higher clausal projection termed S’/ S-bar, introduced into the derivation by the phrase structure rule S’ → COMP S (Bresnan 1972: 29)
Evidence in support of postulating two different types of clausal projection (S and S-bar) comes from data such as the following:
Bresnan takes Wh-Movement to be a ‘COMPattraction transformation’ (1972: 42) which serves the function of ‘moving a constituent … into COMP position’
Chomsky & Lasnik (1977: 434) hypothesised that ‘Wh Movement places a wh-phrase to the left of the complementiser in COMP’ (1977: 446).
work of this era generally assumed that all peripheral constituents could be accommodated within an analysis which posited two different clausal projections (an S-nucleus and an S-bar periphery)
Chomsky (1977b), who argued for the postulation of a third clausal projection: more specifically, he argued that fronted and dislocated topics are contained within a superordinate S”/S-double-bar constituent immediately containing S’.
Theoretical developments in the early 1980s sounded the death knell of the S-, S’- and S”-analyses of clause structure outlined in the two previous sections, because they fell foul of a range of different constraints.
One such was an Endocentricity Constraint to the effect that all phrases and clauses are projections of a zero-level head (i.e. a lexical item/word or inflectional affix), and conversely all zero-level heads (including functors) project into phrases.
A second constraint which had important repercussions for the analysis of the clause periphery was the Binarity Constraint of Kayne (1984), requiring all syntactic structure to be binary-branching.
A third constraint which had radical consequences for the structure of the periphery was the Like-Attracts-Like Constraint/LALC of Baltin (1982: 2), requiring that ‘When they move, phrasal categories adjoin to phrasal categories, and non-phrasal categories adjoin to non-phrasal categories.’ Since only heads and maximal projections can undergo movement (not intermediate projections), LALC allows only adjunction of one head to another, or of one maximal projection to another – and indeed this assumption is adopted in Chomsky (1986: 71, 73).
A fourth constraint which had important implications for the analysis of the periphery was the No-Adjunction-to-Arguments Constraint posited by Chomsky (1986: 6), barring adjunction to arguments (and reformulated by McCloskey 1992: 9 as barring adjunction to a phrase s-selected by a lexical head).
A fifth constraint which undermined earlier analyses of clause structure was the requirement imposed in Trace Theory (Fiengo 1977) that a moved constituent should c-command its trace (i.e. the gap that it leaves behind when it moves).
A sixth constraint with radical implications for the analysis of clause structure was the Linear Correspondence Axiom/LCA of Kayne (1994), specifying that linear ordering at PF is determined by asymmetric c-command relations between constituents.
A seventh constraint (or more properly, PF filter) which impacted on the structure of the periphery was the Doubly Filled COMP Filter of Chomsky & Lasnik (1977), which Chomsky (1981: 243) formulated as in (21) below (where α and β are overt constituents):
Constraints such as those mentioned above undermined earlier analyses of clause structure and led to the development of a new binary-branching CP/IP analysis of clause structure in which the clause periphery is treated as a CP headed by a complementiser constituent which came to be abbreviated as C rather than COMP, and the clause nucleus is treated as
Highlight ( p. 29)
an IP constituent headed by an INFL(ection) constituent which came to be abbreviated as I. Under this analysis, peripheral constituents were generally taken to be positioned on the edge of CP
Highlight ( p. 30)
CP recursion…in which each of the complementisers heads a separate CP projection and the underlined subordinate clause
Highlight ( p. 34)
From the end of the 1980s on, awareness developed of potential problems with the CP recursion analysis of the periphery.
a peripheral adverb like tomorrow can intrude between the declarative complementiser that and the subject John, but not between the infinitival complementiser for and John. This is difficult to account for in a principled fashion under a CP recursion analysis in which each of the highlighted constituents is contained in a separate CP. By contrast, the data can be accommodated in a relatively straightforward fashion if the two types of complementiser represent two different types of peripheral head, with that marking declarative force (or clause type10), and for marking (non-)finiteness.
We can then suppose that a force-marking head like that occupies the highest position in the periphery (as the head of FORCEP)
Rizzi treats force and type as equivalent terms,
The key insight underlying the idea of splitting up CP into a number of distinct types of projection is that each different type of peripheral constituent occupies a specific position on the edge of a dedicated functional projection (i.e. a functional projection which is dedicated to housing a constituent of the relevant type on the edge of the projection).
this has been dubbed the ‘split CP’ analysis, because it involves splitting what had previously been treated as a single type of peripheral constituent (= CP) into a number of different types of projection (a topic phrase, focus phrase etc.)
It has been termed cartographic since its aim is to devise a ‘map of the left periphery’ (Rizzi 1997: 282).
Highlight ( p. 39)
Rizzi argues that each dedicated peripheral head typically12 carries an edge feature which requires it to project a particular kind of specifier on the edge of the relevant projection. If we use the traditional label C to denote a peripheral head, we can say that a C with a topic feature (TOP) houses a topic (i.e. requires a topic to be projected on the edge of the projection as its specifier),13 a C with a focus feature (FOC) houses a focused constituent, a C with a question feature (Q) houses an interrogative constituent,14 a C with a relative feature (REL) houses a relativised constituent, a C with an exclamative feature (EXCL) houses an exclamative constituent, a C with a modifier feature (MOD) houses an adverbial modifying the clause, and so on
Highlight ( p. 41)
Once a constituent is in its criterial position, it is thereby frozen in place in consequence of a constraint termed the Criterial Freezing Condition in Rizzi (2005, 2006a, 2010, 2014a, 2014b) and Rizzi & Shlonsky (2006, 2007).
Rizzi has used a template as an informal descriptive device designed to represent the relative positions occupied by different functional heads in the clause periphery.
where TOP = Topic, FOC = Focus, INT = Interrogative, MOD = Modifier, FIN = Finiteness, and a star is used to mark the possibility of having one or more heads of the relevant type
no clause can contain more than one FORCEP projection
no more than one FOC projection
FORCEP will be the highest projection in embedded clauses
Highlight ( p. 43)
FINP must be the lowest projection
Highlight ( p. 44)
Rizzi & Shlonsky (2007) and Rizzi (2014a) posit that bare declarative complement clauses are truncated at the FINP level (in the sense that they project only as far as FINP)
A bridge verb is a verb (like say) which allows (italicised) material to be extracted out of its (bracketed) complement (thereby bridging the boundary between the two clauses), whereas a non-bridge verb (like quip) does not – as illustrated below (i) What did Mike say/*quip [that she had done —]?
2 Topics .h1
This chapter looks at the syntax of peripheral topics in colloquial English. I shall argue that colloquial English differs from more formal styles of English in making less use of fronted topics, and more use of dislocated and orphaned topics.
2.2 Three Types of Topic in Colloquial English .h2
In colloquial English, we find three superficially distinct types of structure containing a peripheral topic followed by a comment clause. One type involves an italicised topic associated with a gap (—) in the comment clause
The topic is interpreted as originating in the gap position
The gap is traditionally taken to arise through movement of the topic phrase from the gap position into the italicised position in the clause periphery via a displacement operation called Topicalisation. this is also called topic fronting
A second type of peripheral topic structure involves an (italicised) topic associated with an (underlined) resumptive constituent:
the topic is said to be dislocated in the sense that it is in a separate intonation group from the comment clause
such structures are traditionally said to be instances of (Hanging Topic) Left Dislocation.
A third type of topic structure is that illustrated below:
The topic is once again dislocated,
However, the dislocated topic in structures like (3) is orphaned, in the sense that there is no apparent syntactic or lexical link between the topic and the comment clause
Lambrecht (1994: 193) refers to this type of constituent as an ‘unlinked topic’. the link is not overtly expressed, it only exists by pragmatic means
Structures like (1) contain a gap-linked (or G-linked) topic; structures like (2) contain a resumptive-linked (or R-linked) topic; and structures like (3) contain a pragmatically linked (or P-linked) topic.
English also has clauses containing a peripheral focused constituent which serves to introduce new information
topics represent old/familiar information and tell us what a sentence is about, and are often paraphraseable by ‘as for … ’; by contrast, focused constituents represent new/unfamiliar information, and are often paraphraseable by a cleft or pseudo-cleft sentence.
there is typically a slight pause immediately after a topic (marked by the comma in 3), but not after a focused constituent
a topic can be positioned on the edge of a clause either by Move or by Merge, whereas a focused constituent can only get into the clause periphery by Move
topics (by virtue of referring to a specific entity assumed to be familiar to the hearer) are referential expressions, and consequently non-referential/nonspecific pronouns like something/someone/somebody, anything/anyone/anybody, or everything/everyone/everybody can only be focused, not topicalised
2.3 The Derivation of Topics .h2
In this section, I explore the question of whether topics are directly merged in situ, or whether they are moved into their superficial position in the clause periphery from some lower position.
Ross (1967: 209) claims that gap-linked topics undergo a movement operation called Topicalisation by which a constituent like that bold-printed in (8a) below is moved into the italicised position in the clause periphery in (8b) in order to mark it as a topic; likewise Ross (1967: 422) claims that resumptivelinked topics are also moved to the clause periphery via a copy-movement operation termed Left Dislocation, which yields the outcome that the fronted topic ‘is not deleted but remains behind in a pronominal form, as a kind of place-marker’ (Ross 1967: 421) – the relevant resumptive pronoun being underlined in (8c):
Highlight ( p. 59)
The fact that Topicalisation gives rise to intervention effects but Left Dislocation does not suggests that Topicalisation involves the operation Move whereas Dislocation involves Merge.
Highlight ( p. 60)
If imperatives are introduced…a gap-linked topic which moves across the operator (or vice versa) will induce an intervention effect, but a resumptivelinked topic which is generated in situ will not.
Highlight ( p. 61)
This follows if gap-linked topics involve movement, but resumptive-linked topics do not.
an operator can bind a pronoun inside a fronted topic (with the pronoun being interpreted as a variable), but not inside a dislocated topic;
As noted by Ott (2012, 2014, 2015), a gap-linked topic licenses a parasitic gap (= pg) in a sentence like (22a) below, but a resumptive-linked topic in a sentence like (22b) does not:
Highlight ( p. 63)
The more general point to note is that dislocated topic structures in which there are case differences between topic and resumptive provide evidence against a movement analysis, and in favour of treating dislocated topics as directly merged in situ.
Note ( p. 65)
if the resumptive link is not a copy of its topic, they have to be merged in separate locations
Highlight ( p. 69)
The overall conclusion we reach from our discussion in this section and the previous one is the following. English has three types of peripheral topic constituent: gap-linked topics, resumptive-linked dislocated topics, and orphaned (= unlinked/pragmatically linked) dislocated topics. While gaplinked topics can be argued to involve movement of a topic to the clause periphery, dislocated topics (whether resumptive-linked topics, or orphaned topics) are directly merged in situ in the appropriate position in the periphery.
2.4 Multiple Topic Structures .h2
A number of linguists (including Rodman 1974; Cinque 1983; Shaer & Frey 2004: 490) have claimed that topic recursion is barred, in the sense the periphery of a given clause cannot contain more than one topic of a given kind.
Emonds (2004: 107) claims (on the basis of contrasts like that below) that the clause periphery can contain two different types of topic (e.g. a fronted and a dislocated topic), but that when the two co-occur an (italicised) dislocated topic must always precede a (bold-printed) fronted topic:
Highlight ( p. 71)
Bianchi & Frascarelli (2010: 15) claim that dislocated and fronted topics occupy different (but unique) positions in the clause periphery, and that this reflects their different semantic and pragmatic properties. They argue that a dislocated topic in English serves to mark ‘a shift with respect to the aboutness topic of the previous sentence’ – i.e. a change in what is being talked about: they refer to this type of topic as an A-topic. By contrast, fronted topics in English generally have an implicit or explicit contrastive function; hence they refer to this type of topic as a C-topic.
Highlight ( p. 72)
Thus, the key claims made in the research reported above are that English has two types of peripheral topic (dislocated aboutness-shift topics, and fronted contrastive topics), that each is non-recursive, and that the former always precedes the latter.
2.5 Topics and Complementisers .h2
In this section, I explore the position of topics relative to complementisers in colloquial English. We shall see that topics generally follow the declarative complementiser that, but can precede or follow the interrogative complementisers if/whether.
Highlight ( p. 114)
In §2.4, I examined traditional claims that topic recursion is not permitted in English, and the weaker claim that no clause can contain more than one topic of a given type. However, I argued against these claims, adducing data which show that both fronted topics and in situ topics can be recursively stacked in appropriate contexts (as long as no strong constraint violation is thereby incurred).
I noted (and illustrated) the traditional claim that declarative clauses only allow a topic if introduced by that, and that the topic must always follow that. I went on to show that in embedded finite interrogative clauses, the complementisers if/whether occupy an INT position between FORCE and FIN, and allow a topic to be positioned before or after them. I further noted that nonfinite clauses generally do not license topics in English (though I observed that there are some potential exceptions). 2.5
In §2.6, I looked at the position of topics with respect to focused constituents and clause-modifying adjuncts, and noted that topics can either precede or follow FOCP and MODP: more generally, I suggested that MODP can in principle occupy any medial position in the periphery (i.e. any position between the highest FORCE head and the lowest FIN head).
3 Complementisers .h1
This chapter looks at aspects of the syntax of complementisers, and more specifically at how a range of different peripheral heads come to be spelled out as the complementiser that in indicative clauses in colloquial English.1
(1a) and (1b) suggests that indicative that can only be used in a subordinate clause like that bracketed in (1b), not in a main clause
that is obligatory in complement clauses when followed by a peripheral constituent.
(1c) suggests that has to be the first overt constituent in its clause:
the overall picture painted by sentences like (1) is that (in canonical uses in standard varieties of English), indicative that is a subordinate, declarative,2 clause-initial complementiser.
3.2 Primary Spellout
In this section, I shall examine what I shall take to be the primary use of the complementiser that (because it is varietyand register-neutral), namely to…introduce an embedded finite clause (whether indicative or subjunctive
we shall see that the complementiser that can be used to spell out a range of different types of peripheral head in finite embedded clauses when it is the first overt constituent in its clause: more specifically, it can spell out a FORCE head, a REP(ort) head, a REL(ative) head, a SUB(ordinator) head, and even a FIN head in a truncated clause.
the most familiar use of the complementiser that is to serve to introduce an embedded declarative clause
that can be followed by one or more peripheral constituents,
including fronted topics, fronted VPs, fronted negative phrases, dislocated topics, and adjuncts which comprise PPs, adverbs or adverbial clauses
This is consistent with the claim made by (Rizzi 2014a) that clauses which can contain a full range of peripheral constituents have the status of FORCEP constituents, and that the head FORCE constituent of declarative complement clauses is spelled out as that.
Since FORCEP is the highest projection in the clause periphery, the complementiser that will therefore precede all peripheral constituents in the clause.
there are a class of defective clauses whose periphery is truncated in the sense that it projects only as far as FINP. These are declarative complement clauses which are ‘bare’ in the sense that they are not introduced by the complementiser that. Such clauses are extremely frequent in colloquial English:
Rizzi (2014a: 30) posits that ‘finite clauses with null complementisers in English involve truncation of the higher structure of the C system, of Force (hence the absence of the declarative force marker that) and of the Topic-Focus system’
Bare complement clauses differ from complete complement clauses not only in respect of having a more limited structure (and not containing the complementiser that), but also in respect of having a more limited distribution. Thus, bare complement clauses are traditionally said to be restricted to occurring in a position immediately adjacent to and immediately ccommanded by an appropriate selector.
The selector has to be a so-called
It is widely claimed that (unlike their indicative counterparts) subjunctive clauses cannot be bare, but rather require the use of the complementiser that: if so, one might conclude that they are unable to have a truncated FINP structure.
‘bridge’ predicate like say, and cannot be (e.g.) a non-bridge verb like qui ![[email@example.com]]
Highlight ( p. 122)
Furthermore, according to Bošković & Lasnik (2003), bare clauses cannot be used in structures in which an embedded clause is separated from its selecting predicate (or has none), as in topic structures like (8a), pseudo-clefts like (8b), extraposition structures like (8c), right node raising structures like (8d), and gapped clauses like (8e):5
Highlight ( p. 133)
One property which all five uses share in common is that each of the peripheral head positions occupied by that is non-verbal, in the sense that it is not a position which can be occupied by a verbal constituent like an inverted auxilary. A second property is that in each of the five cases, that is the highest head/first overt word in the periphery of the clause containing it.
3.3 Secondary Spellout .h2
in this section I turn to look at the secondary use of that to spell out lower peripheral heads.
complementiser recursion (sometimes termed recomplementation) has been widely reported to occur in a number of Romance languages
and indeed my broadcast English data contain 78 examples of embedded clauses containing two different occurrences of that with one or more other peripheral constituents positioned between them.
intervening constituent is a local prepositional or adverbial adjunct (i.e. one modifying the clause it is contained in)
My hope is [that by the time we meet that we’ll have made some progress] (President Obama, press conference, BBC Radio 5)
a local subordinate adverbial clause,
dislocated constituent reprised by an (underlined) resumptive pronoun
both an intervening (italicised) dislocated topic and an intervening (underlined) phrasal adjunct between the two (boldprinted) occurrences of that:
In analyses within the CP framework, recomplementation was treated as a case of CP recursion (Escribano 1991; Iatridou & Kroch 1992; McCloskey 1992, 2006; Fontana 1993).
CP recursion would lead to the undesirable outcome that the declarative force of the clause ends up being multiply marked
This problem can be circumvented if we suppose that the two different occurrences of that correspond to two different types of head with two different sets of properties:
that corresponds to że in Polish, and the second occurrence corresponds to the different item to
an important difference between the two types of complementiser (noted by Villa-García 2010, 2012c, 2015 for Spanish and by Haegeman 2012: 26, 84 for English) is that a non-initial complementiser is a barrier to extraction so if there’s a that in first position, you can’t pull out the intervening bit to the front of the clause… right? that’s it?
adopt a cartographic approach to the syntax of the periphery and suppose that the initial occurrence of that marks declarative force/clause type, the second marks finiteness, and that the latter (but not the former) is a barrier to movement.
some speakers also have a secondary that-spellout condition allowing that to spell out a FIN head in an embedded finite clause in which FORCE and FIN are separated by one or more intervening projections.
This would mean that the higher (bold-printed) occurrence of that is in FORCE and the lower (underlined) occurrence is in FIN.
Note ( p. 138)
can’t account for examples where secondary that should be FIN but is actually also followed by adjunct
However, the bold-printed occurrences of that in all three examples in (40) are unlikely to be exponents of FIN, since FIN is the lowest head in the periphery and yet the bold-printed occurrences of that are followed by an italicised local adverbial or prepositional adjunct which can be taken to be positioned in MODP.
would be consistent with taking lower occurrences of that to lexicalise FIN if we were to follow Rizzi (2014a) in allowing FINP recursion structures of the form FINP+MODP+FINP (see §1.5)
Note ( p. 139)
so what stops us from going “all of these are FINs! FINs all the way down!”
economy considerations (in particular, the ‘Avoid structure’ principle posited by Rizzi 1997: 314) would argue against the postulation of three separate FINP constituents in (42)
a fatal blow is struck to any lingering hopes of maintaining that the complementiser that can spell out only FORCE or FIN by (constructed) sentences such as the following:
a key assumption in Rizzi’s FINP recursion analysis is that only a MODP projection can intervene between two FIN heads – not a FOCP or TOPP projection
A plausible answer is that in such cases, that is the head of the projection containing the constituent immediately preceding it.
complementiser is preceded by an italicised dislocated topic, raising the possibility that the bold-printed complementiser spells out the TOP head.
a topic analysis of adverbial adjuncts is incompatible with the claim made by Cinque (1983) that hanging topics are canonically nominal in nature – and indeed with the similar claim made by Villa-García himself (2015: 164) that a ‘hanging topic can only be a DP or NP’. In work since Rizzi (2004b), local peripheral adverbials have instead been treated as clausal modifiers housed in a MODP projection.
a complementiser following a circumstantial adverbial clause spells out the head of the projection containing the adverbial clause.
a secondary complementiser ‘can only occur where its specifier is filled’.
The assumption that only a constituent in a criterial specifier position can license secondary spellout of that embodied in (48ii) will preclude the possibility of (e.g.) FIN being licensed to be spelled out as that
In each case, the bold-printed FORCE heads will be spelled out as that via the primary that-spellout condition (48i). In addition, the underlined MOD heads in (49a, c, d) and the underlined TOP head in (49b) will be lexicalised as that via the secondary that-spellout condition (48ii), because each has an italicised specifier in a criterial position.
the analysis outlined here can be extended to deal with cases of recomplementation in subjunctive clauses
Note ( p. 142)
we could use this to let someone annotate it or do it step by step
Thus far, I have examined use of secondary that in clauses which also involve primary use of that (i.e. in clauses introduced by that).
bare declarative complement clauses (i.e. declarative clauses not introduced by an initial complementiser) like those below which show secondary use of that after an (italicised) local phrasal or clausal circumstantial adjunct:
assumptions made by Rizzi (2014a) that bare clauses are FINP constituents, and that FINP can be recursively projected into a FINP+MODP+FINP structure
there are other bare clauses containing secondary that for which a FINP recursion analysis seems implausible. These include relative clauses like those in (54a–c) below, and predicative clauses like those in (54d, e):
relative clauses are RELP constituents which contain a separate FORCE constituent which is generally declarative
If only complement clauses immediately adjacent to their selector can have a truncated FINP structure (as suggested in §3.2), it follows that predicative clauses like those in (54c–d) must be complete clauses which project as far as FORCEP, with the relevant FORCE head having a null spellout.
Note ( p. 144)
i don’t get this
A further class of bare clauses which show secondary use of that are clauses which involve a dislocated constituent immediately followed by that (and reprised by an underlined resumptive pronoun).
Given Rizzi’s (1997) claim that only complete (FORCEP) clauses can contain a topic in their periphery, it follows
The TOP head can be spelled out as that via the secondary that-spellout rule (48ii) by virtue of having an (italicised) topic specifier; the FORCE and FIN heads will receive a null spellout by default.
My data also contain the following examples of (bracketed) bare complement clauses containing both an (italicised) dislocated topic and and one or more (underlined) adverbial adjuncts:
On the assumption that only complete (FORCEP) clauses can contain a topic,
The lower MOD head can then be spelled out as that in accordance with the secondary that-spellout rule
To summarise: in this section, we have seen that English has a secondary use of the complementiser that to lexicalise a peripheral head with a criterial specifier. As we have seen, this secondary spellout is found not only in clauses beginning with that, but also in bare complement clauses.21 And in the next section, we will see that some speakers also extend the use of secondary that to wh-clauses.