Cº realizations along the left edge across English and Spanish

Abstract

This paper investigates the lexicalization of the complementizer that/que in English and Spanish varieties in different contexts along the left edge of the clause. This is performed through discussion of a range of constructions traditionally attributed to the CP domain/left periphery, primarily (but not only) in certain embedded clauses. The ubiquity of that/que, that is, the lexical realization of that/que in subordinating environments, exclamative clauses, interrogative contexts, and subjunctive clauses, amongst others, sheds light not only on the characterization of the relevant constructions but also on the make-up of the left edge of the clause. The fact that such realizations can be obligatory, optional, or, on occasion, impossible, sometimes depending on the variety in question, furthers our understanding of head lexicalizations while contributing to macro and microvariation studies in syntactic theory. In so doing, this paper paves the way for holistic investigations devoted to complementizer realization in the head position of different left-edge-related constructions and in different linguistic varieties.

1. Introduction

Complementizers offer a valuable window into the architecture of the leftmost part of clauses, the demarcation of the limits between the left edge (Complementizer Phrase, CP) and the inflectional/tense domain (IP/TP), and the analysis of a range of constructions traditionally attributed to the leftmost portion of the clause (i.e., the left periphery).

Beyond merely heralding an upcoming subordinate clause, as in I think [that complementizers are mysteriously fascinating], recent research has unearthed naturalistic data pointing to the conclusion that complementizers in languages like English and Spanish can occupy a host of positions along the clausal left-peripheral spine, contingent on the specific constructions at issue, as well as on the root vs. embedded dichotomy, in different varieties of said languages. Radford (2018), for instance, shows through numerous examples from spontaneous speech that spoken English is replete with what seems like different instances of that, boldfaced examples of which appear in (1):

(1)	a.	I think that you are nice.
	b.	They told me that given the current crisis, that the building will remain closed.
	c.	Obviously that the Achilles was giving him a bit of a problem (Ian Chappell, BBC Radio 5, cited in Radford (2018, p. 162)).
	d.	I am not sure what kind of ban that FIFA has in mind (Bert Millichip, BBC Radio 4, cited in Radford (2018, p. 159)).
	e.	What a mine of useless information that I am! (Sir Terry Wogan, BBC Radio 2, cited in Radford (2018, p. 159)).
	f	Please, ensure that if your faculty commit to permitting candidates to attend their classes, that there be sufficient diversity of courses and that syllabi permit visitors to attend (Official university communication, Pennsylvania, 20 November 2013, cited in Villa-García (2015, p. 96)).

In (1)a, we observe a classical example of subordinating that, which can typically remain silent in non-formal contexts. (1)b illustrates the double-that (cf. recomplementation) configuration involving a sandwiched element flanked by overt instances of the complementizer (Villa-García 2015, 2019; Radford 2018; Villa-García and Ott 2022, among others). (1)c displays an initial adverb followed by that. (1)d and (1)e, respectively, feature interrogative and exclamative phrases followed by an instance of that. Lastly, (1)f exemplifies an instance of a secondary that termed ‘jussive/optative’ that by Villa-García (2015). This complementizer, which shares much in common with the recomplementation that (cf. (1)b), as we shall see, is associated with the subjunctive mood. Note that in all the cases in (1), the different occurrences of that can in principle be silent for most speakers without obvious semantic consequences, an issue to which we return in due course.

Sentences featuring putatively different ques are indeed also attested in (certain varieties of) Spanish (see, for a subset of cases, Villa-García (2015)), as indicated by the data in (2). Note that not all sentences are attested in all varieties of the language.

(2)	a.	Cree			que		llueve
		believes			that		rains.
		‘S/He believes (that) it’s raining.’
	b.	Me		contaron	que	a	María,	que	no	la	llaman.
		cl.		told	that	acc.	Mary	that	not	cl.	call
		‘They told me that Mary, that they are not calling her.’
	c.	Ahí		sí	que		no	voy.
		there		yes	that		not	go
		‘I’m certainly not going there.’
	d.	¿	Por	qué	que	viniste?	(Diego Gibanal Faro, pers. comm. 2023)
			for	what	that	came
		‘Why did you come (here)?’
	e.	¡	Qué	alto	que		eres!
			what	tall	that		are
		‘How tall you are!’
	f.	¡	Vaya		que	si	voy!
			vaya		that	if	go
		‘Of course I am going!/How can you even wonder if I’m going?’
	g.	¡	A	Madrid	que	me	piro!
			to	Madrid	that	cl.	piro!
		‘I am off to Madrid!’
	h.	Por		supuesto	que	no	me	quedo.
		of		course	that	not	cl.	stay
		‘Of course I am not staying!’
	i.	¡	Ojalá		que	ganemos		Eurovisión!
			God-willing		that	winSubjunctive		Eurovision
		‘I hope we win the Eurovision contest.’
	j.	Gritó			que	a	la	fiesta,	que	vaya	Marta.
		shouted			that	to	the	party	that	goSubjuntive	Marta
		’S/He demanded by shouting that Martha go to the party.’

The Spanish paradigm displays cases akin to the English ones, including the subordinating complementizer, (1)a and (2)a, recomplementation cases, (1)b and (2)b, interrogative and exclamative complementizers, (1)d,e and (2)d,e, adverbial cases, (1)c and (2)h, and ‘jussive/optative’ cases, (1)f and (2)j. Additionally, the Spanish paradigm also includes sí (que) ‘yes that’ cases, (2)c, investigated from the standpoint of microvariation by Villa-García and González Rodríguez (2020); (2)f, which shows the exclamative particle vaya plus que plus si ‘if/whether’; (2)g, the emphatic construction involving a fronted constituent plus que plus sentence; and (2)i, which instantiates the desiderative/optative construction with ojalá (from the Hispanic Arabic expression law šá lláh ‘if God wants’) plus que plus a verb in the subjunctive.

Multiple-complementizer sentences like the ones in (1) and (2) raise several intriguing questions, such as:

(i)

Is that/que a mere overt manifestation of an otherwise null/silent head (i.e., [_XP YP [_X’ that/que vs. Ø …]])? If so, is that/que always present but deleted in the phonology (PF), that/que, or else inserted when phonologically realized?

(ii)

Is the presence of that/que indicative of a more complex underlying structure instead?

(iii)

Whatever the case may be, what determines the (non-)realization of the complementizer in different positions? Is it dialect-based? If so, are some dialects more prone to lexicalizing complementizers in different positions than others? Is complementizer lexicalization processing-based? Are there any other factors that play a role in determining the presence or absence of the complementizer, such as formal vs. informal contexts?

(iv)

Does the presence vs. absence of that/que have a bearing on the semantics (LF)? Put another way, is complementizer realization just a PF matter, or are there LF reflexes as well?

Set against this background, the present paper aims to constitute a first step towards partially answering the questions in (i)–(iv), in the hope that future studies will further advance our understanding of the puzzle presented herein. In the course of the ensuing discussion, additional questions will be posed which are relevant to various ongoing debates in syntax. Similarly, old and new data will be provided throughout, and previously unnoticed dialectal contrasts will be brought to light. In this sense, therefore, this paper aims to contribute to the diatopic mapping of the relevant constructions. This paper is thus meant to make both a descriptive and a theoretical contribution.

This paper is organized as follows: Section 2 presents existing accounts of the clausal left edge within the transformational generative paradigm; Section 3, which constitutes the bulk of the paper, returns to the data above and discusses a subset of the constructions in turn, in both English and Spanish, though mention of other linguistic varieties will be made when appropriate. In this connection, I concentrate on four major types of complementizers shared by English and Spanish, namely high complementizers, topic-related complementizers (cf. recomplementation), focal complementizers (exclamatives and questions), and low complementizers (‘jussive/optative’ complementizers). A critical review of existing accounts of the particular constructions is provided for each case; Section 4 turns to general extant accounts of inter- and intra-linguistic variation in relation to the presence vs. absence of complementizers and makes new analytical suggestions; and Section 5 is the conclusion.

2. The Analysis of the Leftmost Part of Sentences in the Generative Tradition

A complementizer such as that/que is assumed in the Chomskyan tradition to be a (functional) head, as shown abstractly in (3). I will take this conception of complementizers as heads as my point of departure (though see Vincent (2019) and references therein for a skeptical view).

(3)

[XP

[X’ that/que]]

Since the seminal work of Barriers (Chomsky 1986), complementizers have by and large been taken to occupy the head of CP, as follows:

(4)

[CP

[C’ that/que]]

Data along the lines of some of the sentences in (1) and (2) clearly indicate that a single CP projection may not suffice, as multiple left-edge-related constructions may occur concurrently in a given sentence/clause simultaneously. This in fact led to the postulation of CP recursion, namely the ability of multiple CPs to occur in a given clause (cf. (5)).

(5)

[CP

[C’

[CP

[C’

[CP

[C’

]]]]]]

Sentences displaying multiple instances of that/que, as in (1)b,f and (2)b,j appear to be ideal candidates for an analysis like that in (5), which is, in fact, the account pursued in Iatridou and Kroch (1992) for the English recomplementation case:

(6)

[CP

[C’ that/que

[CP XP

[C’ that/que

]]]]

The late 1990s witnessed the emergence of cartographic approaches to the leftmost portion of clauses: Rizzi (1997) made the highly influential claim that there is a fine structure of the left-periphery (referred to as a templatic structure by those arguing against it ever since), which in effect splits the old CP domain into dedicated projections, each of which is devoted to hosting different elements, which Rizzi argues display the relative order in (7) (note that in Rizzi’s system, TopicP is recursive in that multiple TopicP projections are allowed in the left-peripheral spine, if required by the presence of multiple topical phrases):1

This type of account is rather appealing at first sight, and in fact has been assumed in much work into a broad array of languages since then. For instance, on this view, high complementizers spell out Forceº; dislocated phrases (such as Clitic Left Dislocations, ClLDs) occupy the specifier of TopicP; wh-phrases and exclamative phrases (which are mutually exclusive) are the occupants of Spec, FocusP; low complementizers and elements related to mood and finiteness are associated with FinitenessP. For Rizzi (1997), TopicP and FocusP are only projected on an as-needed basis (that is, when topical or focal elements occur), and only the former is recursive, as noted. As Rizzi (2013, p. 200) observes, “the left periphery is populated by a system of functional heads dedicated to the expression of scope-discourse properties”, thus:

(8)	a.	Which car	Q	should I purchase?
	b.	This promise of lifelong service	TOP	I renew to all today (King Charles III, cited in Villa-García (2023, p. 2)).
	c.	THIS BOOK	FOC	you should buy	(, not that one).
	d.	What an incredible week	EXCL	we have had!

Accordingly, an exclamatory sentence featuring an exclamative phrase such as (8)d would receive the following simplified analysis under Rizzi’s approach (note that it is immaterial to the present discussion whether exclamatives target FocusP or a more specific ExclamativeP in the focus field):

(9)

[ExclamativeP

what an incredible week

[Exclamative’

EXCL

[IP/TP

we have had ]]]

A logical extension of the account in (9) for cases of exclamatives with a low that (cf. (1)d) would be to hypothesize that the complementizer that is adjacent to the wh-phrase is the spellout of the Exclamative head since it co-occurs with the focal phrase:

(10)

[ExclamativeP

what an incredible week

[Exclamative’

that

[IP/TP

we have had ]]]

A similar account can be adopted for the other types of heads (topic, Q, etc.), as we will see below. This will indeed be the null hypothesis adopted in what follows. In the remainder of this paper, I will try to determine whether a Rizzian analysis is tenable for (all) the cases under discussion (cf. (1)–(2)) and whether a unified account can be proposed for all the cases at hand, a non-trivial issue given the complex dataset this paper concerns itself with. We will then consider this type of account more generally in terms of inter- and intra-linguistic variation in Section 4.

3. Different Left-Edge-Related Constructions: To Spell That/Que or Not to Spell That/Que

The paradigm in (1) and (2) attests to the complexity of the left periphery of languages like English and Spanish, where different constructions can occur—and even co-occur—in the left portion of the clause, and on occasion these constituents may be accompanied by an overt instance of that/que. Below, I concentrate on a subset of those constructions, for the reasons discussed in the previous section. I follow the relative order claimed by a left-peripheral analysis like Rizzi’s (cf. (7)). I first look at the English case and then at the Spanish one. Different existing accounts are considered, and mention is made of micro- (i.e., dialectal) variation when appropriate. In Section 4, I return to overarching existing accounts of the type of variation observed. I begin by discussing high complementizers.

3.1. High That/Que

High complementizers (by assumption, the head of Cº or Forceº, as in (4)) constitute an area of the grammar where languages like English and Spanish stand in glaring contrast to one another.

English, for its part, typically allows the complementizer to remain silent, as shown in (11), a feature that spreads across dialects:

(11)	I think Ø complementizers should not be taken for granted.

The presence or absence of the overt complementizer in sentences like (11) does not have a semantic reflex, i.e., the truth-conditional meaning of the sentence remains intact. However, the presence of the complementizer is typically associated with academic or written registers. Certainly, using a complementizer facilitates reading and processing, and on occasion, it resolves potential ambiguities. This is shown by (12):

(12)	Johan said yesterday he accepted ten papers.

In (12), yesterday can modify either said or accepted. Realizing the complementizer does away with this potential confusion:

(13)	a.	Johan said that yesterday he accepted ten papers.
	b.	Johan said yesterday that he accepted ten papers.

The optionality of that is not always such. For instance, topicalized clauses require that (Bošković and Lasnik 2003, among others):

(14)	a.	*Ø he accepted ten papers John believes.
	b.	That he accepted ten papers John believes.

As for the subjunctive, conservative English speakers tend to disfavor the omission of that, contrary to what we observe in Spanish (cf. see the discussion surrounding (23) below) (Radford 2016), as in (15):

(15)	I demand that the children be here at ten.

However, that-drop in subjunctives has also been documented in present-day English, even in writing, as shown by the below examples:

(16)	a.	Brexit Secretary David Davis was also able to meet business leaders demanding Ø the UK stay in the single market immediately after leaving the EU (The Independent, 10 July 2017, cited in Villa-García (2019, p. 26)).
	b.	Following further evaluation this morning, The Queen’s doctors are concerned for Her Majesty’s health and have recommended Ø she remain under medical supervision (statement from Buckingham Palace, 8 September 2022).
	c.	Díaz evoked Dorado again on Monday, demanding Ø Feijóo explain how he knew the man (politico.eu, 18 July 2023).

Needless to say, other well-known cases, such as comp-t(race) effects, force the complementizer to be absent in most dialects of English:

(17)	a.	Who do you believe rocks?
	b.	*Who do you believe that rocks?

Furthermore, certain predicates have been claimed to require an obligatory that (Hegarty 1992; Adger 2003; Franks 2005; Radford 2016; Llinàs-Grau and Bel 2019, among others). The list includes verbs like whisper, quip, judge, and conjecture.2 Nonetheless, a simple Google search yields examples featuring such verbs and no overt complementizer, as illustrated for quip in (18):

(18)	He correctly replied 14, then quipped Ø he wished it was 15 (The Sun, 2011, cited in www.collinsdictionary.com).

Early theoretical accounts of the overt/null contrast in complementizer lexicalization in languages like English in the Chomskyan tradition deemed the presence vs. absence of that a phonological phenomenon. More specifically, Chomsky and Lasnik (1977) advanced an analysis whereby that is deleted when absent (i.e., our null complementizer above would be the result of phonological deletion: that). By contrast, Radford (2016) argues, on the grounds that the overt complementizer that and its null homolog Ø are not interchangeable in all contexts (e.g., (17)), that that and Ø are actually distinct lexical items exhibiting distinct properties (rather than a single item with two different spellouts, in the spirit of Chomsky and Lasnik (1977)).

According to Bošković and Lasnik (2003), null C (i.e., our Ø above) is a PF-affix that requires PF-adjacency with the verbal (or nominal) host. This requirement is fulfilled in (11) above, since the complement clause headed by C is PF adjacent to the V head, but not in (14)a, where the clausal complement has been fronted, which disrupts the necessary adjacency between the verbal head and C, rendering null C impossible and thus enforcing that, as in (14)b.

It is worth mentioning that a different line of analysis, adopted in the work of Pesetsky and Torrego (2001), takes that-less clauses to be the result of the subject moving to Spec, CP. In their view, that is the outcome of Tº moving to Cº for purposes of feature checking; this operation is satisfied if the subject moves to Spec, CP instead, resulting in that-less clauses. Put another way, for these authors, in I think syntax is fascinating, the subject syntax ends up in Spec, CP.

In stark contrast to what we observe for English, the high complementizer is generally mandatory in present-day Spanish, as shown by (19), a well-known cross-linguistic difference (Torrego 1983; Etxepare 1996; Brovetto 2002; Antonelli 2013; Cerrudo-Aguilar 2014; and Rodríguez-Riccelli 2018).

(19)	Creo	que/*Ø	los	complementantes	no	se	pueden	dar	por sentados.
	believe	that	the	complementizers	not	cl.	can	give	for seated
	‘I believe that complementizers cannot be taken for granted.’

On the other hand, diachronic and synchronic dialectal evidence suggests that the high complementizer is not obligatory at all times. For one thing, classical Spanish from the 17th century displayed null complementizers with certain predicates:

(20)	Les	dijo	Ø	tenía nuevas	de	que	en	el	cielo
	cl.	said		had news	of	that	in	the	heaven
	se	había		muerto	el	arcángel		san	Gabriel
	cl.	had		died	the	archangel		St.	Gabriel
	‘S/He said to them that s/he had news that the archangel St. Gabriel had died in
	heaven’ (Abarca de Bolea, Vigilia, 17th century, cited in RAE-ASALE (2009, p. 3232)).

As far as diatopic variation at present is concerned, varieties that include Mexican Spanish have been reported to omit the complementizer with thinking and judgment predicates (Silva-Corvalán 1994 et seq.; Rodríguez-Riccelli 2018), as in (21).

(21)	Creo	Ø	llamará.
	believe		will-call
	‘I believe s/he will call’ (RAE-ASALE 2009, p. 3232).

Using Tweeter data from Mexican and Los Angeles Spanish, Rodríguez-Riccelli (2018) concludes that verb modality and embedded subject position are the strongest predictors of que-drop in these varieties. Moreover, Rodríguez-Riccelli (2018) reports a higher percentage of que-drop in Mexico City (12.5%) than in Los Angeles (10%), suggesting that a contact-induced language change (from English) is not apparent. English contact, however, may allow the broadening of the contexts in which the omission of the complementizer occurs (e.g., stative verbs).

However, deletion can occur more generally across varieties under certain conditions. For instance, deletion of subordinating que occurs across Spanish varieties in formal-register contexts in cases where a mark of subordination (e.g., another que, an expression which includes que, such as porque ‘because,’ etc.) appears (Subirats-Rüggeberg 1987, pp. 170–71; Etxepare 1996; RAE-ASALE 2009):3

(22)	Recurrirán	la	sentencia,	porque	dicen Ø	no
	contest	the	sentence	because	say	not
	es	ajustada.
	is	fit
	‘They will contest the sentence, as they claim it is not appropriate.’ (Telediario 2, RTVE, Spain, 16 April 2018)

In fact, across Spanish we find that subjunctive cases involving verbs such as rogar ‘beg’ tend to appear in formal and written contexts without que (the que counterpart of (23) being grammatical but less formal):

(23)	Rogamos	Ø	nos	envíen	el	certificado	a	la	mayor	brevedad.
	beg		cl.	sendSubjunctive	the	certificate	at	the	bigger	brevity
	‘We would like to ask you to send the certificate to us at your earliest convenience.’

At this juncture, two questions arise in light of data like (23):

(i)

Are such cases instances of Forceº (high complementizers) or Finitenessº (low complementizers), since they are related to mood (associated under Rizzi’s proposal with FinitenessP)?

(ii)

Is complementizer deletion the result of the verb moving all the way to Cº (Forceº or Finitenessº, under Rizzi’s assumptions)? (See, in this connection, the related claim noted above by Pesetsky and Torrego that that-less clauses in English arise from subject movement to CP).

A relevant question is also whether the non-appearance of que points to the absence of a left periphery altogether (so that such clauses are analyzed as bare IPs/TPs), a claim that would also extend in principle to English that-less clauses (Bošković 1997; Brovetto 2002; Antonelli 2013).4 Antonelli (2013) argues that the left periphery of Spanish is present even in the absence of que, but that in cases like (23), the verb moves to a syncretic ForceP/FinitenessP projection, thus rendering the complementizer impossible. One problem with this type of account is that clitics show up preverbally (cl.+V—les rogamos, as in (23)), while in imperatives, which are standardly assumed to involve Tº-to-Cº movement, clitics show up postverbally (V+cl.—ruégales ‘beg_imperative them’). The answers to (i) and (ii) await further research.

In sum, we observe that English and Spanish high complementizers behave quite differently, with the Spanish high-complementizer drop being much more restricted than its English counterpart.5 More specifically, a (diachronic and synchronic) dialectal split emerges from our discussion surrounding Spanish: in Old Spanish, que could be dropped with verbs like decir ‘say’; in present-day Spanish, que can only be absent in a very limited set of contexts (e.g., subjunctives with verbs like rogar ‘beg’); in varieties such as Mexican or Los Angeles Spanish, que-less examples are confined to certain thinking and judgment predicates. Moreover, from an analytical perspective, the debate as to the analysis of complementizer-less sentences remains alive at present. We now turn our attention to what has widely been regarded so far as instantiations of non-high complementizers.

3.2. Recomplementation That/Que

The phenomenon of double-complementizer sentences, also known as recomplementation, illustrated again for English and Spanish in (24), has been subject to much debate in the recent literature (Escribano 1991; Iatridou and Kroch 1992; Demonte and Fernández-Soriano 2009; Villa-García 2012, 2015, 2019; Radford 2018; Villa-García and Ott 2022, amongst many others).

(24)	a.	Note that if you have already taught in Semester 1, that you are not required to resubmit paperwork to HR Services (official university communication, UK, January 2019, cited in Villa-García (2019, p. 2)).
	b.	Dice	que	si	llueve,	que	se	quedan	encamados
		says	that	if	rains	that	cl.	stay	bedded
		‘S/He says that if it rains, that they will stay in bed.’

As regards syntactic microcomparison in the realm of recomplementation, no noteworthy differences are reported in the existing works in relation to English; the data provided in the literature come from both American and British English (Radford 2018; Villa-García 2019), but little or no attention has been paid to whether there exists dialectal variation in English recomplementation, a gap in the literature at present.

As for Spanish, whereas Demonte and Fernández-Soriano (2009) note that recomplementation is found across Spanish varieties, Martínez-Vera (2019) claims that recomplementation is absent in American Spanish (its counterpart being a prolonged intonational break); however, Frank (2020) provides experimental evidence from Colombian and bilingual heritage US Spanish indicating that recomplementation is not impossible in such varieties. Similarly, linguists such as Andrés Saab and Carlos Echeverría (pers. comm. 2022), who are speakers of Argentinean and Chilean Spanish, respectively, use and accept recomplementized structures. Fontana (1993) and Echeverría and López Seoane (2019), for their part, observe, based on written evidence, that Old Spanish frequently featured recomplementation, the sandwiched elements being typically long if-clauses. Needless to say, the foregoing discussion strongly suggests that the dialectal map of recomplementation in Spanish is likewise in dire need of further research.

Be that as it may, early proposals (e.g., Iatridou and Kroch 1992) assumed that the different that complementizers featured in what looks like a single sentence whose embedded clause displays a complex left periphery are instances of Cº in a recursive CP, as in (25) (see also (6) above):

(25)

[CP

[C’ that/que

[CP XP

[C’

that/que

]]]]

In the wake of the Rizzian approach, the question soon arose as to which head is spelled out by doubled, secondary complementizers. A myriad of proposals arose (on which see Villa-García (2015)); I will concentrate on two here for the sake of illustration. Authors like Demonte and Fernández-Soriano (2009), López (2009), and Antonelli (2013) have advanced the hypothesis that the high and the low complementizers delimit the beginning and the end of the left-peripheral space, hence populating Forceº and Finitenessº, respectively:

(26)

[ForceP

[Force’

that/que

[TopicP

XP

[Finiteness’

that/que

]]]]

However, such an analysis runs into a number of empirical problems, as argued by Villa-García (2012, 2015, 2019). Instead, this author proposes to treat the second instance of that/que as a topic marker, hence the head of TopicP (see also Rodríguez-Ramalle (2003); see Ledgeway (2005) for an analysis of this type which assumes that the different complementizers are separate realizations of a complementizer that starts in Finitenessº and moves up to Forceº in a head-to-head fashion):

(27)

[ForceP

[Force’

that/que

[TopicP

XP

[Topic’ that/que

[FinitenessP

[Finiteness’

]]]]]]

This analysis is prima facie appealing both on empirical and theoretical grounds (it accounts for why it is typically topical phrases that appear in a sandwiched position, and it assumes that the different left-peripheral heads proposed by Rizzi (1997) can be spelled out).

Nevertheless, in marked contrast to monoclausal proposals like the ones just reviewed, recent research has convincingly argued that recomplementation is not bound to occur only with topical phrases and that in fact what doubled complementizers mask is two separate sentences that superficially look like one, rather than an elaborate clausal left edge (see, especially, Villa-García and Ott (2022) on this biclausal line of analysis), thus arguing against monoclausal accounts like those in (26) and (27). For these authors, the clausal portion headed by the second instance of that/que represents a restart in discourse (i.e., a new sentence, CP2, reprising the first one, CP1, begins), as shown abstractly in (28):

(28)

[CP1

subject V

[that/que … XP]]

[CP2

subject V

[that/que … ]]

This move accounts for issues including why the second sentence must be a syntactically complete sentence (i.e., [that he/*Ø hates seafood] in (29)a), why a non-topical phrase (e.g., a discourse marker, as in (29)b), can occur in between overt complementizers, and why even a focal phrase can appear in between thats/ques (cf. (29)c):

(29)	a.	He	told	me	that	Peter,	that	he/*Ø	hates	seafood.
	b.	Dice	que		bueno,		que	no	vienen.
		says	that		well		that	not	come
		‘S/He says that well, that they are not coming.’
	c.	Me	dijo	que	jamás,	que	jamás	se	casará	con	nadie.
		cl.	says	that	never	that	never	cl.	will-marry	with	nobody
		‘S/He says that never ever, that never will s/he marry anyone.’

For Villa-García and Ott (2022), therefore, the second instance of the complementizer is a restart that mirrors what we see in the first clause (i.e., dice que/says that… — dice que/says that…), hence a repeated high complementizer, but not the realization of a left-peripheral head such as Topicº (or Finitenessº, for that matter), as assumed in (26) and (27). This actually goes a long way to explaining why we also find reduplicative complementizers other than declarative that/que (i.e., he asked me whether…— he asked me whether…):

(30)	He asked me whether, given the current assessment boycott, whether we are getting a salary raise.

In light of data like (29) and (30), the prospect that recomplementation que is the overt or null spellout of Topicº under monoclausality (that/que vs. that/que) loses plausibility, which casts doubt on the claim that Topicº can be occupied by an overt realization of the complementizer that/que. On the bisentential account, therefore, recomplementation that/que would be an instance of high that/que in disguise. Technically, then, that/que is the same high element in the two occurrences (in CP1 and CP2), in spite of outward appearances. In the next two subsections, I turn to the exclamative and interrogative that/que.

3.3. Exclamative That/Que

Exclamative wh-phrases that come in the company of that have customarily been attributed to Irish English (Zwicky 2002). However, Radford (2018) provides a large set of data suggesting that other varieties permit the co-occurrence of wh-phrases with an instance of that below them as well, as shown again in (31):

(31)	a.	How gorgeous that you look! (www.abbieeandeveline.com, cited in Radford (2018, p. 160)).
	b.	How quickly that people forget! (web, cited in Radford (2018, p. 159)).
	c.	What a job that he’s done so far! (Sam Matterface, Talksport Radio, cited in Radford (2018, p. 159)).

Of course, the that-less counterparts of the examples in (31) would be the canonical versions of the relevant sentences, which shows that that is once again not mandatory. Radford (2018) pursues an account in the spirit of Rizzi (1997) and Rizzi and Bocci (2017) wherein the wh-phrase in the specifier of ExclamativeP (or FocusP) licenses the head that:

(32)

[ForceP

[Force’

[ExclamativeP

XP

[Exclamative’

that

]]]]

An issue that any analysis needs to tackle concerns the rare occurrence of wh-items (or wh-words) in this context. For the most part, it is almost always a full wh-phrase that appears immediately above that in exclamatives in English. Authors like Bayer (2014) and Radford (2018) have argued that unlike wh-phrases, which move to the specifier, wh-words move to the head position of the wh-operator projection (e.g., FocusP), thus preventing a complementizer from occurring in such a position (i.e., wh-phrase + that vs. *wh-word + that).

Structures like those in (31) call into question longstanding claims in the generative literature, including the Doubly-Filled Comp Filter, which bans the simultaneous occurrence of a wh-element and an overt complementizer in CP, a prohibition observed in standard English (i.e., *[_CP wh-element [_C’ that … ]]). In his discussion of similar examples of exclamative + che ‘that’ examples in Italian, Rizzi (2013, p. 2009) actually refers to this configuration as “the only case of legitimate ‘[D]oubly[-]filled Comp’ in Standard Italian” (see also Bayer and Dasgupta (2016)). Rizzi goes on to note that “[c]learly, (the equivalent of) that is an unmarked, versatile complementizer form, capable of occurring in the highest C position, and also, in cross-linguistically variable manners, in lower positions”, as in (32).

Moving from root to embedded contexts, Radford (2018) provides data indicating that it is not impossible to have a wh-exclamative below an instance of what appears to be a high (Forceº) complementizer (a well-known property of Spanish, as we shall see):

(33)	He realized, I think, that how big this thing was (Film critic, BBC Radio 5, cited in Radford (2018, p. 114)).

Radford (2018) does not give any examples of embedded wh-exclamatives plus that, but such examples sound fine to his native ear:

(34)	He realized that how big that this thing was (Andrew Radford, pers. comm. 2023).

Spanish exclamatives behave similarly to their English counterparts reviewed above in most respects, with interesting dialectal differences. In many varieties of Spanish, a pleonastic complementizer immediately adjacent to the wh-phrase is not unusual (speakers notice that the que version is more emphatic):

(35)	a.	¡	Qué	guapa	que	está		tu	niña!
			what	beautiful	that	is		your	daughter
		‘Your daughter looks so good!’
	b.	¡	Qué	rápido	que	conducen		aquí!
			what	fast	that	drive		here
		‘They drive so fast here!’

A notable difference with present-day English is that exclamatives in (non-Caribbean) Spanish trigger obligatory S-V inversion regardless of the presence vs. absence of que (see Villa-García (2018, in preparation) and Villalba (2019) for recent discussion; as noted by an anonymous reviewer, Italo-Romance varieties follow a similar pattern):6

(36)	*¡	Qué	guapa	que	tu	niña	está!
		what	beautiful	that	your	daughter	is
	Intended: ‘Your daughter looks so good!’

Certain dialects (e.g., Asturian Spanish) disallow the presence of que with qué phrases (Villa-García 2018, in preparation). By contrast, the popular Spanish of Asturias allows que with the exclamative determiner vaya (on vaya more generally, see, e.g., Espinal et al. (2022)). According to Bosque (2017, pp. 18–19), vaya in Asturian Spanish (and in areas of León) can combine not just with nouns (which is what happens in other parts of the Spanish-speaking world), but also with adjectives and adverbs:

(37)	a.	¡	Vaya	casa	que	tiene!
			what	house	that	have
		‘What a house s/he has!’
	b.	¡	Vaya	sano	que	está!
			what	healthy	that	is
		‘How healthy he is!’
	c.	¡	Vaya	mal	que	lo	hiciste!
			what	bad	that	cl.	did
		‘You did it so badly!’

In Spanish, cases like (37)a involving nouns can optionally have que (RAE-ASALE 2009). Asturian Spanish, though, exhibits a more nuanced contrast with respect to vaya exclamatives, irrespective of the category of the word following vaya. In areas such as Avilés and Gijón, the que version is the norm, with the non-que version being deemed incomplete or even unacceptable. However, in Oviedo and the surrounding hamlets, the que-less version prevails (perhaps due to influence from the Asturian language). To summarize, in Asturias, qué + phrase + que is not an option; only the que-less version is used. When it comes to vaya + N/A/Adv, however, we find subdialects that require que and others that dispense with it. In other words, the form of the exclamative particle determines the possibility of having an accompanying que below the exclamative phrase in these varieties.

The evidence furthermore shows that there is a dependence between the presence vs. absence of que and the type of phrase in the specifier (Villa-García, in preparation): que is sensitive to the sort of exclamative element to its left, which confirms that the two stand in a spec-head configuration, exactly as predicted by accounts like (32) above, which invoke FocusP/ExclamativeP. This analysis is substantiated by the obligatory inversion displayed by exclamatives, which is characteristic of focal phrases in (non-Caribbean) Spanish (on this issue, however, see Villa-García, in preparation). Thus, I will adopt the account in (32) for Spanish as well, as in (38):7

(38)

[ForceP

[Force’

[ExclamativeP

XP

[Exclamative’

que

]]]]

Beyond Peninsular Spanish, according to RAE-ASALE (2009, p. 3206), in Latin American Spanish, the presence of que with qué exclamatives is more restricted than in Spain, although examples occur in the River Plate area and less frequently in the Caribbean (see also Casas (2004, p. 268) for examples from Mexican Spanish).

As far as embedded clauses are concerned, pleonastic que is also acceptable in non-matrix contexts:8

(39)	Mira	qué	guapo		que	es	ese	podcáster.
	look	what	good-looking		that	is	that	podcaster
	‘Look at how good-looking that podcaster is’ (Antonio Cañas García, Raquel González Rodríguez, and Isabel Pérez-Jiménez, pers. comm. 2023).

Still, exclamative clauses selected by predicates other than pseudo-interjections like mira ‘look!’ cannot be construed with que, a poorly understood phenomenon to date (Ignacio Bosque, pers. comm. 2023):

(40)	Es	increíble	qué	cosas	(*que)	dice.
	is	incredible	what	things	that	says
	Intended: ‘The things s/he says are incredible’ (Bosque 1984, p. 287).

Villa-García (2015) furnishes embedded data with vaya under verbs of saying in Asturian Spanish, along the lines of (41):

(41)	Dice	mi	prima	que	vaya	rápido	que	conduce	tu	padre.
	says	my	cousin	that	what	fast	that	drives	your	father
	‘My cousin says that your father drives so fast.’

Qué-exclamatives plus que are also licit under decir ‘say’-like predicates. Note that English exhibits this pattern as well (cf. (33) and (34)):

(42)	Dice	que	qué	guapa	que	es	esa	niña.
	says	that	what	beautiful	that	is	that	girl
	‘S/He exclaimed that that girl is so beautiful’ (Raquel González Rodríguez and Isabel Pérez-Jiménez, pers. comm. 2023).

All in all, the English and Spanish evidence adduced here points to the conclusion that the exclamative phrase and que co-exist in the same projection. Nonetheless, dialect data from Spanish point out that the exclamative que is not really optional at all times. While in many areas of Spain exclamatives with qué and vaya seem to optionally co-occur with the pleonastic que, qué-exclamatives in Asturian Spanish occur without que; their vaya homologs require que in some parts of Asturias, but not in others. The data crucially corroborate that the licensing of que is sensitive to the nature of the wh-phrase in its specifier, which supports a spec-head analysis along the lines of (38). This account indeed gains cross-linguistic plausibility from the English data discussed above, which are not confined to Irish varieties of English, as used to be widely thought. Lastly, the evidence also reveals a high degree of variation across Spanish varieties, which further studies should certainly investigate in more depth. The facts are also relevant to non-trivial theoretical questions, including whether the cases at hand are compatible with a relative-clause analysis (on which see fn. 7). I now turn to wh-interrogatives with that/que.

3.4. Interrogative That/Que

Wh-interrogatives followed by that used to be believed to be a feature of regional varieties like Irish English, as in the Belfast English example in (43) (with some Belfast speakers accepting only wh-phrases –not wh-items– above that, much like in the wh-exclamative cases discussed above):

(43)	I don’t know when that he’s going (Henry 1995, p. 88).

Radford (2018) shows, by contrast, that wh-interrogative + that configurations transcend Irish varieties, as the data in (44) demonstrate:

(44)	a.	Definitions vary as to which of these types of criteria that are used (Member of the English Department, University of Göteborg, cited in Radford (2018, p. 137)).
	b.	I hadn’t realized just how many people that were there
		(Maxx Faulkner on WCBE, cited in Radford (2018, p. 138)).
	c.	This heat map shows just how active that Trippier was
		(Jermaine Jenas, BBC1 TV, cited in Radford (2018, p. 139)).

Radford (2018, p. 142) submits that the above data are amenable to a Rizzian account according to which the wh-interrogative is housed in the specifier of a WHP (or FocusP) below ForceP, with that in the head position of WHP/FocusP:

(45)

[ForceP

[Force’

[WHP

wh-interrogative

[WH’ that

…]]]]

One advantage of this account is that it can easily accommodate cases of embedded wh-interrogatives below a quotative element, reported in Radford (2018, p. 116), and which are used “to embed quoted speech into a matrix clause, with the quoted speech essentially being unmodified”, as in (46).

(46)	a.	He protested that how could he have known that his office was bugged? (Radford 2018, p. 113).
	b.	[ForceP [Force’ that [WHP wh-interrogative [WH’ … ]]]]

With verbs that intrinsically select a question as their complement, the secondary that is legitimate, but not the high that, as the below example, kindly provided by Andrew Radford (pers. comm. 2023), illustrates. This is a non-trivial dissimilarity between English and Spanish, where embedded interrogatives can be heralded by an instance of reportative que preceding either interrogative phrases or the interrogative complementizer (on which see Plann (1982) and much subsequent work; cf. (48)b).

(47)	I wonder (*that) what kind of party (that) he has in mind.

As far as Spanish is concerned, the literature notes an important paradigm gap owing to the non-existence of the low que with wh-interrogatives either in root or in embedded contexts, in sharp contrast to English and to what we observe in the case of Spanish exclamatives in the preceding subsection:

(48)

a.

¿

Cuántas

casas

(*

que)

se

ha

comprado?

how-many

houses

that

cl.

has

purchased

‘How many houses has s/he bought?’

b.

Preguntaron

(que)

cuántos

kilómetros

(*

que)

había

recorrido.

asked

that

how-many

kilometers

that

had

travelled

‘They asked how many kilometers I/s/he had travelled.’

The lack of que with wh-interrogatives in Spanish is widely considered to be the only formal mark distinguishing interrogatives from exclamatives as far as their structure is concerned (RAE-ASALE 2009, among others).9 As we shall see momentarily, however, there are dialectal data indicating that, on occasion, que surfaces in certain interrogative contexts, which refutes the widely held generalization that que never co-occurs with wh-interrogatives in Spanish.

Before we delve into the dialectal data in question, it is important to note there is a well-documented interrogative phrase that occurs with que in all dialects of Spanish, however. This is the cómo que ‘how come’ (lit. ‘how that’) construction. But unlike regular interrogatives, this one is a formulaic construction that triggers no inversion:

(49)	¿	Cómo	que	al	final	no	vienes	al	bodorrio?
		how	that	at+the	end	not	come	to+the	wedding
	‘How come you are finally not coming to the wedding party?’

Importantly, some varieties of Latin American Spanish permit certain interrogatives to occur with a low instance of que, contra standard contentions regarding the impossibility of having que immediately after a wh-phrase across Spanish, as in (48). The following data from CORPES XXI, generously furnished by an anonymous reviewer, show that this is in fact the case (see also Villa-García, in preparation):

(50)

a.

¿

Desde

cuándo

que

no

lo

ven?

since

when

that

not

cl.

see

‘When did you last see it/him?’ (Chile).

b.

¿

Cuándo

fue

y

dónde

que

ocurrió

ese

descubrimiento?

when

was

and

where

that

occurred

that

discovery

‘When was it and where was that discovery made?’ (Uruguay).

c.

¿

De dónde

que

alguna vez

en otra

vida

lejana,

había

of where

that

sometime

in other

life

far

had

pretendido

y

creído

ser

escritor?

intended

and

believed

be

writer

‘Where did you learn that, some other time, in a different, distant life, he had intended to be and believed himself to be a writer?’ (Cuba).

d.

¿

Por

qué

que

no

fuiste

a

rescatarnos?

x

for

what

that

not

went

to

rescue-cl.

‘Why did you not go to rescue us?’ (Colombia).

e.

¿

Y

por

qué

que

no

me

arriesgaría

a

algo

así?

and

for

what

that

not

cl.

risk

to

something

thus

‘And why wouldn’t I risk doing something like that?’ (Chile).

The wh-phrase-plus-que data just reviewed raise various questions, such as whether this pattern can be found with other wh-items or is confined to adjuncts, whether the interpretation of the sentences displaying que is different from that of their que-less homologs, and whether they can occur in subordinate environments. This pattern is actually well documented in other Romance languages like Brazilian Portuguese, Canadian French, and the Northern Italian dialect of Lamonat (Simone De Cia, pers. comm. 2023), as shown by the following example from Brazilian Portuguese:

(51)	Onde	que	você	mora?
	where	that	you	live
	‘Where do you live?’ (Oushiro 2011, p. 145).

Overall, we do find a subset of varieties of Spanish where the wh-interrogative + que configuration is legitimate. An analysis along the lines of that for English can therefore be advanced for these cases, as in (52). The question of course still remains as to why the presence of que below interrogatives in Spanish is so highly restricted.

(52)

[ForceP

[Force’

[WHP

wh-interrogative

[WH’

que

]]]]

Having discussed exclamatives and interrogatives, we now turn our attention to what has been assumed to be the lowest complementizer along the leftmost part of clauses: ‘jussive/optative’ that/que.

3.5. Low Complementizers: ‘Jussive/Optative’ That/Que

A final phenomenon that I will consider here is the so-called ‘jussive/optative’ complementizer (Villa-García 2012, 2015; Radford 2018), which I touched upon in passing in Section 3.1. This phenomenon is illustrated again for English below:

(53)	I am writing to ask that if you have not yet completed this training in this academic year that you do so as soon as possible and by the end of 14 July 2023 at the latest (Official university communication, United Kingdom, 30 June 2023).

That in these cases is deemed to be the lexical realization of the subjunctive. As noted, conservative speakers do not drop that in this context, as illustrated once more in (54), but present-day English allows that-omission (cf. (16)).

(54)	The University has ordered that a town be built in the premises.

The question which arises is whether ‘jussive/optative’ cases lexicalize Forceº or rather Finitenessº, which is connected to mood under Rizzi’s (1997, et seq.) system. One possibility, entertained by authors like Rizzi (1997), Villa-García (2012, 2015), Antonelli (2013), and Radford (2018), is that in the absence of sandwiched material, a conflated ForceFinitenessP is projected. In this context, we are no longer dealing with high or low that, since a unique realization would do the job under ForceFinitenessº (which is, in effect, equivalent to a Cº analysis like that in (4)):

(55)

[ForceFinitenessP

[ForceFiniteness’

that

]]

By contrast, when left-peripheral constituents occur, which is when multiple instances of que surface (i.e., that … XP … that), a split of the CP field is triggered (Rizzi 1997; Villa-García 2015; Radford 2018, among many others). Analytically, it is conceivable that this instance of low that may be a lexicalization of Finitenessº, as follows:

(56)

[ForceP

[Force’

that

[TopicP

XP

[Topic’

[FinitenessP

[Finiteness’

that

]]]]]]

This proposal receives empirical support from other linguistic varieties. For instance, Ledgeway (2005, p. 365) capitalizes on languages like Romanian, which exhibits a distinct (low) complementizer (să) in subjunctive clauses that appears to be a very low element in the left-peripheral spine, as witnessed by its mandatory proximity to the verb and any clitics that may come with the verb (that is, any left-peripheral phrase must precede să). Romanian (57) illustrates (as observed by an anonymous reviewer, other Balkan languages make the same point):

(57)	Vreau	(ca)	MÂINE	să	meargă.
	want	that	tomorrow	that	goSubjunctive
	‘I want him to go TOMORROW’ (Watanabe 1996, p. 44).

Such examples are ideal candidates for an analysis in the spirit of (56) above:

(58)

[ForceP

[Force’

ca

[FocusP

MÂINE

[Focus’

[FinitenessP

[Finiteness’

să

]]]]]]

Villa-García (2012, 2015, 2019) has made the contention that Spanish also possesses a Finitenessº complementizer lexicalized as que, which he dubs ‘jussive/optative’ que, exemplified by the following data:

(59)

a.

Juan

Carlos

también

le

dijo

a

su

hijo

que

si

tuvo

el

coraje

de

casarse

con

ella

Juan

Carlos

also

cl.

told

dat

his

son

that

if

had

the

courage

of

marry-cl

with

her

desobedeciendo

el

deseo

de

sus

padres

que

tuviera

el

carácter

disobeying

the

wish

of

his

parents

that

hadSubjunctive

the

temper

para

ponerla

en

su

sitio.

for

put-cl.

in

her

place

‘King Juan Carlos I also demanded that if his son had the courage to marry her, despite his parents’ wishes, that he muster the strength to put her in her place.’

(paraphrase of Jaime Peñafiel’s words, Lecturas, www.lecturas.com, 8 April 2018)

b.

Que

si

vas

a

salir

con

ella,

que

vayas

en

serio.

that

if

go

to

go-out

with

her

that

goSubjunctive

in

serious

‘I’m saying that if you are going out with her, you should get serious’ (RTVE, Servir y proteger, TV series, 4 April 2018).

Villa-García (2018) reports emphatic examples from naturalistic speech where indeed the two ques actually co-occur (which can arguably be analyzed as simultaneous realizations of Forceº and Finitenessº under Rizzi’s system):

(60)	a.	Que	te	ha	dicho	que	que	te	pires.
		that	cl.	has	said	that	that	cl.	goSubjunctive
		‘I’m telling you s/he told you to go away.’
	b.	Así	que	que	te	den.
		so	that	that	cl.	giveSubjunctive
		‘So go fuck yourself.’

An immediate question begged by the subjunctive data reviewed so far is whether this instance of low que is compulsorily overt or not, vis-à-vis recomplementation que in Section 3.2, which is optional (see Villa-García (2015) and Echeverría (2020) for much relevant discussion). A preliminary survey suggests that speakers prefer the realization of this low que, but not all speakers fully reject the que-less counterparts (a claim that can be extended to apply to the English cases above). Thus, when it comes to ‘jussive/optative’ cases in subordinate contexts, we are dealing with a matter of preference, rather than obligatoriness. Echeverría (2020, p. 48, fn. 24) furnishes the following example, which he attributes to the Royal Academy of the Spanish Language, indicating that ‘jussive/optative’ que can be silent:10

(61)	Dile	que	cuando	termine	venga	a	rendir	cuentas.
	say	that	when	finishes	comeSubjunctive	to	accounts	accounts
	‘Tell him/her to come (here) to give us an explanation when s/he’s done.’

There may be factors such as tense (present vs. past), the presence of negation, intrinsically subjunctive selecting predicates (pedir ‘request’) vs. communication verbs (decir ‘say’), or even diatopic variation at play here. In fact, speakers of Spanish in contact with Catalan seem more permissive in terms of low que-omission (Villa-García 2018). Echeverría (2020, p. 48, fn. 24) arrives at the conclusion that “[i]f Spanish optative and jussive sentences are overall more likely to include an extra complementizer, this might well be explained by the more general, historically increasing tendency to use que before verbs in the subjunctive”. Actually, this type of que also occurs in root clauses (Villa-García 2015, amongst others), in which case que is unquestionably compulsory:

(62)	¡Que	venga	a	verme	tu	hija!
	that	comeSubjunctive	to	see-cl.	your	daughter
	‘I demand that your daughter come to see me.’

Returning to the embedded cases in (59) displaying a doubled que, whether obligatory or not, the evidence is symptomatic that there exists a low subjunctive complementizer in languages like Spanish.

Nevertheless, an open question not addressed by Villa-García and Ott (2022) is whether ‘jussive/optative’ sentences in English and Spanish can also be reanalyzed as restarts, much like their recomplementation homologs. An analysis of this guise for the cases at stake would assume the following preliminary structure for subordinate ‘jussive/optative’ sentences:

(63)

[CP1

subject V

[that/que … XP

]]

[CP2

subject V

[that/que

Vsubjunctive

]]

This move would easily account for why the subjunctive que is more likely to be realized in this environment: the second occurrence would be a repeat of the same element in a restart configuration, with complementizers heralding subjunctive clauses being less omittable, as has been noted (see Section 3.2 for more details of this analysis when applied to recomplementation que). If this analysis ends up being the right account of reduplicative cases featuring “jussive/optative” that/que as well, then what I have referred to as medial (recomplementation) and low complementizers (‘jussive/optative’ que) would be underlying instantiations of a high que (see above on whether we are dealing with Forceº, Finitenessº or even a conflated ForceFinitenessº projection when no overt left-peripheral material occurs).

Having discussed a major subset of the putatively different positions in which complementizers can be realized in English and in Spanish, the following section explores the more general and crucial issue of parameterizing the various lexicalization possibilities observed hitherto.

4. Inter- and Intra-Linguistic Variation: (Micro-)Parameterizing the Presence vs. Absence of Complementizers

The preceding discussion has made it clear that the seemingly ubiquitous that and que complementizers in English and Spanish constitute a gold mine for the study of the geometry of the clausal left edge as well as for variation, including macro- and micro-variation. In the following subsections, I turn to potential accounts of the general variation observed, based on recent proposals in the generative tradition. I will divide the discussion into spell-out accounts and feature-driven accounts, with a final note on a potential consideration regarding the detectability of projections that may help explain the different realizations found. A more general question raised by the discussion so far to which we will return towards the end of the paper is in fact whether a unitary account is attainable.

4.1. PF Accounts

The work of Rizzi and his collaborators has advocated a Spell-Out (i.e., pronunciation) Parameter analysis to account for the (non)-overtness of left-peripheral heads. In the words of Rizzi (2013, pp. 201–2), we are dealing with “a spell-out parameter, a familiar and widely attested kind of low[-]level parameterization”. According to Rizzi and Bocci (2017, p. 13), the different left-peripheral criterial heads (Forceº, Topicº, etc.) may be “null…, but their presence may be detected indirectly”. The authors go on to say that “…the same relevant criterial [CP] heads are phonetically null, a familiar (and trivial) parametric difference”. I propose that this type of parameter could be extended to apply across varieties of the same language, effectively making it a micro-parameter (see also Villa-García and González Rodríguez (2020) on sí (que) ‘yes that’, exemplified in (2)c, across Spanish). An abstract visual representation of this binary parameter would be as follows, where strikethrough indicates PF deletion (note that the bracketings in (64) are highly simplified by only focusing on those projections that this work has specifically concentrated upon):

(64)	a.	[ForceP	[Force’	that/que	[TopicP	[Topic’	that/que	[FocusP	[Focus’	that/que	[FinitenessP
		[Finiteness’		that/que	IP/TP	]]]]]]]]
	b.	[ForceP	[Force’	that/que	[TopicP	[Topic’	that/que	[FocusP	[Focus’	that/que	[FinitenessP
		[Finiteness’		that/que	IP/TP	]]]]]]]]

Analogously, on the basis of English data akin to those in (1) in the introduction, Radford (2018, p. 170) puts forth the following conditions for complementizer realization in English (note that % signposts inter-speaker variation):

(65)	Complementizer spellout conditions (English)
	In a finite clause, a non-verbal peripheral head can be spelled out as that
	(i) if it is the first word in an embedded clause
	or (ii) if it (% is in an embedded clause and) has an adjacent superordinate (% edgemate) (% non-wh) licenser

It is of note that this author uses the modal can, which reflects the optionality of that in English across the board, although it is important to recall that in some contexts, high that must be pronounced (Bošković and Lasnik 2003, inter alia; see Section 3.1).

What the pronunciation accounts outlined above share is the assumption that left-peripheral heads (for us, the complementizers that/que) are always optional, their realization boiling down to a superficial parameter that decides whether the relevant head positions are PF realized (i.e., lexicalized) or not. In spite of the fact that such accounts are a priori theoretically appealing owing to their elegance and simplicity, they fall short of accounting for the facts in a satisfactory manner. For one thing, certain that realizations are more likely than others are (e.g., subjunctive that for conservative speakers) or even compulsory in some contexts (initial that in clause-fronting cases like (14)b). Similarly, once Spanish is brought into the picture, the wide range of variation brought to light herein demonstrates that limiting all variation to a mere (discretionary) pronunciation decision appears to be, at best, oversimplistic.

A broader issue concerns the conception of parameters as binary choices (Chomsky 1981). If truly a parameter, then [±spellout] should be set to one value (e.g., [+spellout] or [-spellout]) for a specific construction in one particular language/dialect, not to either option (i.e., a particular CP-related head would bear a specific value for the relevant feature in charge of parametric variation in this regard, but not varying values/settings; see below on the prospect of lexical parameters). Put another way, optionality should in principle be excluded, contrary to fact. Assuming that that/que may be null/overt when optional based on a parametric difference would be akin to saying that a given language can have the positive and the negative setting of the Null-Subject Parameter at the same time, on an optional basis. Unless ancillary stipulations are made, it is not at all clear how the Spell-Out Parameter option would work in practice for the cases at issue.

4.2. Feature-Based Accounts

In the wake of the Minimalist Program (Chomsky 1995), features acquired a prominent role in syntactic theory. Early works already hinted at the possibility of explaining differences in terms of complementizer lexicalization by means of features. For instance, in an attempt to rule out wh-interrogatives plus que in languages like Spanish (see Section 3.4), Uriagereka (1995, p. 160) observed that “although que can occupy the head…, in these languages [i.e., in Romance varieties such as Galician and Spanish, JVG] it does not have the appropriate features to agree with a focused phrase in its spec—much like that is not compatible with Wh-phrases [in standard English, JVG]” (see also Brucart (1993, p. 76) on a similar claim regarding exclamatives).

This is consonant with the prevalent idea that parametric variation is connected to features of functional heads (cf. the Borer-Chomsky conjecture; “lexical parameters”). In other words, this type of approach locates the relevant dimensions of (micro)variation in the properties of individual functional heads. As Ledgeway (2020, p. 31) puts it, “the locus of parametric variation lies in the lexicon, and in particular, in the (PF-)lexicalization of specific formal feature values of individual functional heads”. The Minimalist Program, therefore, paved the way to recast parametric variation as different featural configurations encoded in lexical heads, in such a way that distinct featural make-ups yield the differences noted.

With particular reference to microvariation, a number of studies (e.g., Smith and Adger 2005 et seq.; Thoms et al. 2019) have shown that the Minimalist Program is particularly well suited to deal with such micro-variation (see Green (2007) for an overview of approaches to syntactic (micro-)variation). For example, Smith and Adger (2005) note that the feature system allows for variable phonetic (i.e., audible) outputs with the same interpretation (by virtue of a particular head containing one additional formal feature, with visible phonetic/PF consequences but not semantic/LF ones). This move clearly accounts for those cases where an element can be present or absent without meaning consequences. Critically, the feature-driven approach can also capture those cases where the meaning may change, such as exclamatives with que in Spanish, where most speakers report added emphasis in the presence of the overt que. Under this feature-based-type of analysis, such arguably semantic differences would be owing to different semantic features in the relevant head. Thus, my proposal would be that all of the C-related heads share some feature (possibly C), which is responsible for the phonological uniformity of the realization (that/que), but that various dialects allow deletion/non-realization depending on the other features of the head, in the spirit of Smith and Adger (2005).

4.3. Projection Detectability

Beyond spellout conditions and features, Spanish complementizer realization may show sensitivity to whether a projection has or lacks an overt specifier (with ForceP and FinitenessP being excellent candidates for the latter type). It seems that in those projections lacking a specifier, the head is generally less omittable across dialects (i.e., high and low que), in keeping with the generalization that languages disallow XPs to be headed by silent heads and specs (Koopman 1997) or, more generally, that a phrase is only projected when overtly manifested (Roberts and Roussou 2003; An 2007; Bošković 2016, inter alia). By contrast, when a specifier (e.g., Topic, Exclamative) is present (in apparent violation of the Doubly-Comp Filled Filter, as noted above in passing), the head can more easily remain null across dialects. The co-occurrence of an overt specifier with an overt head then leads to a reinforced/even more emphatic construction (cf. the emphasis contrast in qué-exclamatives with and without que, with que adding emphasis according to my native-speaker consultants).

This move goes against a superficial pronunciation or spell-out parameter (see Section 4.1); nevertheless, it is also challenged by a number of non-trivial questions: why is que obligatory with vaya for a subset of Asturian Spanish speakers (see Section 3.3)? What about the English case, where spec-less Forceº can remain silent (see Section 3.1)?11 These and other questions strongly indicate that the ultimate account of the (non-)realization of left-peripheral heads in the form of that/que awaits and merits further investigation.

5. Conclusions

Since Rizzi’s (1997) split-CP hypothesis, the investigation of the mapping of the left periphery has commanded the attention of numerous researchers working in a vast number of languages. Much care has been taken to study the different constructions (topics, exclamatives, interrogatives, etc.) that are housed in the CP domain of the clause. Complementizers have generally not constituted the focus of attention, but their existence and importance in establishing the limits of the left-peripheral layer cannot be denied. Indeed, the data from English and Spanish investigated in this paper show that complementizers like that and que are ubiquitous, and they can in fact co-occur with virtually all the left-peripheral constructions investigated in the literature so far. Of course, not all speakers from all dialects (of English or Spanish) accept the presence/absence of that/que with every CP-related phenomenon, and it is precisely that high degree of variation that the present paper has tried to draw attention to. Importantly, the foregoing discussion has revealed that, on occasion, it is not even clear that a non-high complementizer necessarily evinces the presence of a medial left-peripheral head (e.g., recomplementation as a TopicP phenomenon or as a biclausal phenomenon; see Section 3.2). If the seemingly medial (recomplementation) that/que and low (‘jussive/optative’) that/que turn out to be repeats of the high that/que in a bisentential/restart configuration (à la Villa-García and Ott 2022), then the total number of distinct complementizer realizations would actually be significantly reduced in both languages, despite superficial appearances.

Throughout our discussion, we have seen cases where the complementizer can by and large remain silent (e.g., high that in English, as discussed in Section 3.1), configurations in which both the complementizer and the left-peripheral phrase are realized concurrently (e.g., exclamatives with vaya plus que in certain varieties of Asturian Spanish, on which see Section 3.3), and patterns where only the left-peripheral constituent can be realized (namely wh-interrogatives in Peninsular Spanish; Section 3.4). In short, there are cases in which the presence of the complementizer is (i) optional, (ii) obligatory, or (iii) impossible. This state of affairs raises four major questions, presented in the introduction and repeated here. These questions now receive partial answers, pending further investigation of the constructions studied herein (and others):

(i)

Is that/que a mere overt manifestation of an otherwise null/silent head (i.e., [_XP YP [_X’ that/que vs. Ø …]])? If so, is that/que always present but deleted in the phonology (PF), that/que, or else inserted when phonologically realized?

• The evidence adduced throughout suggests that merely reducing the presence/absence of that/que to a pronunciation parameter is rather simplistic. This would actually imply complete absence or obligatoriness in all contexts (assuming that the parameter is set to a particular value, not to either), contrary to fact, as amply demonstrated throughout. Put another way, the prospect of a unitary analysis that relies on PF realization seems untenable.

(ii)

Is the presence of that/que indicative of more complex underlying structure instead?

• At least for cases including recomplementation, which have been convincingly analyzed recently as restarts in discourse camouflaging two underlying sentences, the answer to this question appears to be positive: multiple thats/ques are symptomatic of a more intricate syntactic configuration behind the scenes (see also Villa-García, in preparation, for the claim that exclamatives with que likewise involve a more elaborated structure behind the scenes than their que-less homologs).

(iii)

Whatever the case may be, what determines the (non-)realization of the complementizer in different positions? Is it dialect-based? If so, are some dialects more prone to lexicalizing complementizers in different positions than others? Is complementizer lexicalization processing-based? Are there any other factors that play a role in determining the presence or absence of the complementizer, such as formal vs. informal contexts?

• In some cases, as we have seen, the complementizer is optional in some contexts, obligatory in certain environments, and mandatorily absent in others. This sometimes depends on the actual dialect in question. At this stage, it cannot be ascertained that a dialect omits complementizers more often than other dialects across left-peripheral constructions in either language. Regarding processing, the restart analysis of recomplementation (which, I argue, can be extended to subordinate ‘jussive/optative’ contexts) is compatible with this view, inasmuch as the restart contributes to facilitating the processing of the sentence. And lastly, factors such as formal vs. informal contexts do play a role in complementizer manifestation: for instance, omission of that in embedded declaratives in English has traditionally been attributed to colloquial registers; conversely, omitting que in requests is a feature characteristic of formal, written discourse in Spanish.

(iv)

Does the presence vs. absence of that/que have a bearing on the semantics (LF)? Put another way, is complementizer realization just a PF matter, or are there LF reflexes as well?

• If we put aside processing, emphasis, and the formal vs. informal distinction, the cases explored herein do not seem to manifest meaning differences (regarding, e.g., the truth conditions of the sentences in question) depending on the presence/absence of that/que.

Further investigations into the constructions at issue will hopefully provide fuller answers to the major questions posed in this paper.

Be that as it may, the paper has tried to parameterize (and micro-parameterize) the different left-peripheral head realizations observed across English and Spanish. The major proposals in the literature to account for microvariation rely either on superficial PF realization or on features, and in addition to extant accounts, I have adumbrated an account that capitalizes on whether the specifier positions of the relevant maximal projections are filled overtly.

At present, however, a unified account of the facts across the board appears to be far from reachable. Many questions remain at this point regarding both the diatopic extent of the variation observed in relation to head realizations (especially in English, but also in Spanish) alongside the analysis of such variation in English and Spanish. It is my sincere hope that the data, dialectal contrasts, and lines of analysis gathered in this paper will contribute to this important enterprise.