15.1 Introduction
Role and Reference Grammar (RRG) holds the view that a sentence is not just a concatenation of words into a sequence but the incarnation of a set of formal templates that are stored in syntax.Footnote * Syntactic templates have two effects that are crucial for the understanding of the nature of linguistic symbols. First, they are a key factor in determining the pattern followed by the influx of information contributed by each word into the new symbol (i.e. the phrase). Second, the phrase is an enclosure where certain relations occur among its constituents that are shielded from some external influences. Cleft sentences and relative clauses make this encapsulation of the phrase evident. Sentence (1) illustrates a typical instance of a relative clause structure (hereafter, RC).
The student who failed the test quit the program.
This sentence is an instance of an externally headed relative clause (EHRC), which will be analysed in Section 15.2. It contains a relative clause (‘who failed the test’) that modifies the head noun (i.e. ‘student’) of the referential phrase (RP) – in which the relative clause is embedded. This head noun is the antecedent that fixes the reference of the relative pronoun (‘who’) in the relative clause and, in doing so, it might also demand agreement on nominal features like person, gender and number depending on the morphology of the language. The head noun functions in (1) as the privileged syntactic argument (PSA) of the main clause; however, relative clauses can modify any non-PSA direct core argument and even adjuncts modulo Keenan and Comrie’s (1977) Accessibility Hierarchy.1 In the same sense, the co-indexed wh-form can have the function of any direct core argument or it can even be an adjunct in the relative clause (though see Section 15.2 for some relevant restrictions in Malagasy, a Western Austronesian language spoken in Madagascar). Therefore, the shared participant is not required to play any specific syntactic or semantic role in either the matrix clause or the relative clause. This has important consequences for the syntax of RCs. The relative clause is safeguarded from any requirement or influence from the matrix clause by the RP in which it is embedded. Operators like tense, grammatical aspect or negation affect each clause (matrix and relative) independently. All the formal properties imposed on the relative pronoun come from the head noun – and are of a referential nature (person, gender and number) – or from the matrix verb of the relative clause. These properties are captured by an analysis of relative clauses as RP nuclear modifiers (Pavey Reference Pavey2004; Van Valin Reference Van Valin2005: 220–223; 260–266; Van Valin and LaPolla Reference Van Valin and LaPolla1997: 497–499). This means that RCs, just like English attributive adjectives, are elements in the periphery that describe either a property of the referent or an event in which it participates. This semantic dependency is reflected in the syntax and further justifies the nuclear periphery analysis: for those languages with a rich nominal morphology, the noun imposes agreement relations on the attributive element.
The semantics of RCs displays properties that illustrate the insufficiency of truth conditional semantics to account for linguistic meaning. The meaning of the relative clause and the head noun in (1) can be expressed by the following simple sentence.
The student failed the test (and quit the program).
Thus, the information about the world is the same in both cases. However, the meanings are quite different because the composition of the respective semantic representations is different. This will be captured later in this chapter (see the representation in (10)) by embedding the logical structure (LS) of the RC as an argument of the attributive LS, which is consistent with the RC syntax, in that the relative clause is a modifier of the RP nucleus. In addition to EHRCs, Section 15.2 will also discuss Internally Headed Relative Clauses (IHRCs), free relatives and non-restrictive relatives. In IHRC, the head noun is inside the embedded clause, which we know is embedded due to the presence of a complementizer. The so-called ‘free relatives’ are not really RCs, since they lack a head noun as well as a relative pronoun in many languages. Finally, in non-restrictive RCs, the embedded clause is separated from the head noun by a pause, and it does not fulfil the same pragmatic function characteristic of relative clauses, namely, the identification of the referent of the head noun.
This chapter also discusses cleft sentences, specifically, it-clefts as exemplified by (3), which will be the focus of Section 15.3.
It was John who asked the question.
The components of it-clefts are a matrix clause with a ‘dummy’ ‘it’ pronoun, a copula clause and a clefted noun phrase (John in (3)), which is the antecedent that fixes the reference of either a relative pronoun or, if this is absent, an argument or adjunct slot in the embedded clause, as in (4).
It was in 1996 Mary had her baby!
In (3) the subordinated clause is introduced by a relative pronoun but it can also be subordinated by a complementizer, as in (5).
It was in 1996 that Mary had her baby!
The analysis offered in this chapter adopts a perspective that highlights the interfaces that constitute the internal structure of clefts. From a truth-conditional perspective, sentence (3) carries the same content as a simple sentence like John asked the question. In fact, (3) describes a single event whose description is contributed by the LS of the embedded verb (i.e. ask). In this sense, the embedded clause provides the bulk of the semantic content carried by the whole sentence, whereas the matrix clause contains almost no truth-conditional information. This is the reason why this chapter introduces the notion of clefts as a structure that materializes an ‘inverted’ linking. By ‘inverted linking’ I mean that the stronger semantic clause, or the clause that provides the meaning for the whole sentence, is syntactically dependent on a semantically weaker matrix clause (see Section 15.3). This linking is not accidentally associated with a particular information structure. In clefts, matrix clauses carry the pragmatic information weight, since they contain the narrow focus of the sentence, namely, the clefted noun (i.e. John in (3)). Further, this information is also the reference value of the relative pronoun in the dependent clause. Therefore, the matrix clause is semantically weaker, yet informatively stronger, than the dependent clause, so that the syntax of clefts is clearly driven by pragmatics. This conclusion is consistent with the view held in RRG on the depth of the insertion of communicative needs in grammatical structure. Information structure digs down into the core of the grammatical engine.
There are strong similarities between RCs and clefts. The most salient one is that both constructions involve an embedded clause that can be introduced by a relative pronoun, which might be absent in some contexts. In addition, both the embedded clause in clefts and those in RCs are presupposed. These resemblances motivate some proposals in which cleft clauses are analysed as relative clauses. In contrast, the view pursued in this chapter – in particular in Section 15.4 – focuses rather on the differences between the two types of embedded clause. One of the properties of relative clauses as defined here is that they are modifiers of the nominal nucleus of a noun phrase. There is clear evidence that this is not the case with clefts. For instance, as will be discussed in due course, in RCs, quantifiers in the matrix clause range over the reference of the noun as restricted by the relative clause, while in clefts they have scope over the meaning of the antecedent noun without any restriction imposed by the embedded clause. This evidence leads to a representation of clefts in terms of a clause-linkage relation in which the embedded clause modifies the core of the matrix clause. Section 15.4 will also explore fundamental pragmatic differences between the two constructions. The embedded clause in an RC is a presupposition that helps the listener to identify an otherwise unidentifiable participant. In contrast, the activation status of the referent of the clefted noun in clefts is irrelevant; what is crucial here instead is that this noun is the focal element.
The chapter is organized as follows. Section 15.2 presents an analysis of RCs: their syntax, semantics and syntax–semantics linking. Section 15.3 introduces cleft sentences and subsection 15.3.1 the relevant portion of the linking algorithm that accounts specifically for clefts. Section 15.4 offers a contrastive analysis of clefts and RCs. The chapter ends with Section 15.5, which presents a short conclusion.
15.2 Relative Clauses
The ingredients of a typical English RC like (1), repeated in (6a) for ease of exposition, are a clause introduced by a relative pronoun (i.e. who) co-indexed with a head noun outside the RC and preceding it (i.e. student) and, finally, a matrix clause, which the head noun belongs to. The RC is embedded within the RP headed by the antecedent noun – student – as can be seen, for example, by the fact that a proform can replace the entire RP (see 6b).
a. The student who failed the test quit the program.
b. She quit the program.
The events described by the RC – the failing event – and by the matrix clause – the quitting event – might not be connected beside the fact that they share the same participant. A central relation within the RP is that of obligatory co-indexation between the head noun and the relative pronoun. As long as it belongs to the RP, there may be intervening material – all within the RP that contains the relative clause – between the head noun and the relative pronoun, as in (7).
The student of the teacher who called you yesterday is sick.
However, the two nouns have the right features to function as antecedent of the pronoun – they are both [+human] – and the interpretation is ambiguous. Clearly, the preferred interpretation takes the closer antecedent (i.e. teacher). In addition, if there is more available intervening material between the head noun and the relative pronoun, the chances of acceptability of the sentence are proportionally lowered.
The fact that there is no restriction on the function that the head noun can have in the matrix clause is illustrated below. It can be a direct core argument like in (6a) or it can even be an adverbial as in (8).
I saw Mary at the bar where you met her.
In this case the relativizer is not a pronoun but a wh-form with an adverbial function in the RC; in other languages, it could be any pro-adverbial form.2 In English as well as many other languages there are no restrictions on the function of the co-indexed proform in the relative clause, as suggested by (8). However, some languages pose specific requirements. As noted by Keenan (1976), Malagasy allows only the co-indexed element to be the subject of the relative clause (9).
(9)
a. Na-hita ny vehivavy [(izay) nan-asa ny zaza] Rakoto. pfv.atv-see det woman comp pfv.atv-wash det child Rakoto ‘Rakoto saw the woman that washed the child.’3 *‘Rakoto saw the woman that the child washed.’ b. Na-hita ny zaza (izay) nan-asa ny vehivavy Rakoto. pfv.atv-see det child comp pfv.atv-wash det woman Rakoto ‘Rakoto saw the child that washed the woman.’ *‘Rakoto saw the child that the woman washed.’ c. Na-hita ny zaza (izay) sas-an’ ny vehivavy Rakoto. pfv.atv-see det child comp wash-pass det woman Rakoto ‘Rakoto saw the child that was washed by the woman.’
RCs require a matrix clause with an available antecedent noun that heads an RP that contains an embedded clause via a proform co-indexed with the antecedent noun. The lexical category of the antecedent is fixed: it can only be a noun. This is easily relatable to the communicative function generally attributed to (restrictive) RCs, namely that they help identify the referent of the antecedent noun by relating it to some already known event.
The syntactic representation of the EHRC in (1) is given in Figure 15.1. Just like any adjectival modifier, the relative clause is in the Periphery of the nominal nucleus of the Referential Phrase.
Figure 15.1 Externally headed relative clauses
The semantic representation expresses the contribution of the RC as an attributive modifier (i.e. be′ (…)) of the head noun. This predication is inserted into the matrix LS slot that corresponds to the first argument, that is, the first argument of quit′, which should be the position occupied by the head noun. The participant referred to by the head noun student is shared with the predicative LS (be′(student, …). In short, (10) represents that the semantics of the RC fills in the argument slot in the main LS that corresponds to the head noun. In particular, this is not a co-indexation relation, but a position that belongs to two LSs (the one that corresponds to quit′ and the one that corresponds to be′). The slot of this shared participant is shown in (10) by a thick broken underline (Van Valin 2012). In addition, this position is co-indexed with the argument position filled with the wh-word in the relative clause LS.4 The superscript ‘¬i’ refers to the activation status of the participant; in particular, the head noun of the RC is unidentifiable, which mean that it is neither active or accessible nor inactive, it is unidentifiable and either anchored or non-anchored (Chafe Reference Chafe and Tomlin1987; Pavey 2004: 5).
INGR quit′ ([be′ (student1¬i, [INGR fail′ (who1, test)])], program)
The regular syntax-to-semantics linking algorithm applies to relate Figure 15.1 to (10) but, in addition, we need the following further restrictions (Van Valin 2012: 57).
(11)
a. Retrieve from the lexicon an attributive LS (i.e. be′ (x, [pred′])) and substitute the LS of the verb in the relative clause (i.e. [INGR fail′ (x, y)]) for the second argument of the attributive LS (i.e. [pred′]). b. Co-index the first argument in the attributive LS with either the unlinked argument position in the relative clause LS, or, if there is a relative pronoun, to the argument position linked to the relative pronoun. c. Insert the attributive LS into the argument position in the matrix LS occupied by the head noun, replacing the variable in the first argument position in the attributive LS with the head noun.
There is still the question whether this structure complies with the Completeness Constraint in cases like (12), where there is no overt relative pronoun. The absence of an overt relative pronoun is possible in all cases except in those where the pronoun is linked to the argument in the LS that should be the PSA/subject.
The student I saw in the library was not screaming.
In this case, the undergoer of see is not realized within the embedded clause. This requires the reduction of a slot in the verb core, although the structural conditions for this reduction are entirely different from those that characterize control structures (Van Valin 2005: 255). Control involves a non-subordination nexus at a core juncture, in which the relevant core is a semantic argument of the matrix core. Quite differently, the relative clause in RC is a subordinated optional modifier of a noun nucleus.
The answer to this problem is to resort to the formulation of the Completeness Constraint in Van Valin (2005: 233). Under this principle, the argument of an LS that is not syntactically realized within the clause can be satisfied anywhere within the sentence that contains that clause. In this way Completeness is respected. Still, as is the case with control structures, we need to determine the specific structural conditions that license this slot reduction in RCs. The following is an attempt to capture the relevant restrictions.
The syntactic slots of the core of a finite clause may be reduced by 1 if the argument to which the missing slot should have been linked is fulfilled by a variable co-indexed with an antecedent noun that is the nucleus modified by the phrase containing the empty slot.
In this way we identify the specific condition that allows the optional reduction of an argument slot in some English EHRCs.
Internally headed relative clauses are characterized by the head noun being part of the relative clause. In particular, the head noun is expressed in a position within the relative clause, while it is omitted in the matrix clause. This type of RC is found in Bambara, an SOV language.
(14)
Bambara (Mande, Africa; Bird Reference Bird1968, cited in Van Valin 2012) [Ne ye so min ye] tye ye san. 1sg pst horse rel see man pst buy ‘The man bought the horse that I saw.’
The relevant question is how the hearer recovers the semantic function of the co-indexed participant in the matrix LS. Upon hearing (14), a transitive syntactic template is retrieved because of the verb buy. There is a relativizer which indicates the presence of an embedded clause headed by the verb see, which contains two arguments: ne (‘I’) and so (‘horse’). The word order of the language (SOV) indicates that the first one is the Actor (and, hence, the PSA) while the second is the Undergoer and, thus, the non-PSA macrorole argument. In turn, the matrix clause nucleus is buy, which requires two arguments, although there is only one: tye (‘man’). The fact that the verb buy selects human actors is the only cue that indicates, first, that man is the Actor and, second, that there is a missing Undergoer argument, which has to be found in the relative clause. There are two candidates, one is I and the other is horse. World knowledge indicates that people are less likely to be sold than horses and this might motivate the choice of horse in this case. In other cases, the structure may be disambiguated on the basis of contextual clues. Taking into account the preceding considerations, the rules governing linking from syntax to semantics in IHRCs are the following (Van Valin 2012: 60):
(15)
a. Retrieve from the lexicon an attributive LS and substitute the LS of the verb in the relative clause for the second argument. b. Co-index the first argument in the attributive LS with the argument in the relative clause LS identified as the head noun. c. Insert the attributive LS into the open argument position in the matrix LS.
The constituent structure of this construction, shown in Figure 15.2, is extremely interesting. One of the salient features is that it allows the antecedent of a co-indexation relation to be more embedded than the co-indexed variable. Typically, the antecedent precedes the co-indexed element in the semantic representation and the antecedent is either equally or less syntactically embedded than the co-indexed element. However, this is not the case with IHRCs.
Figure 15.2 Internally headed relative clauses
In IHRC, the Completeness Constraint is satisfied since the argument linked to a reduced slot is satisfied somewhere else within the sentence. However, it is worth pointing out the structural difference between reducing an argument slot in the matrix core and reducing it in an embedded clause. In particular, binding a variable with an argument nested more deeply in the LS is not at all frequent in languages like English, if possible at all, but this is exactly the case with [be′ (xi, [see′ (ne, soi)])], where the lexical element so is an argument of an LS contained within another LS with a variable bound to the embedded lexical argument. This relative clause is not a modifier of an RP, but rather the relative clause and the co-indexed element are arguments of the same clause.
Another issue raised by IHRCs concerns co-indexation. Typically, obligatory co-indexation like the one exhibited in RCs is governed by structural restrictions. In the case of obligatory reflexives, the antecedent must hold a semantic role that is higher on the macrorole hierarchy than the one held by the co-indexed element. Ultimately, this principle is motivated by the semantic role hierarchy, and therefore, it is independently motivated (Van Valin 2005: 61). Control structures are another example: the controller is an argument of the matrix verb and the controllee is an argument in the LS embedded in the matrix verb LS.
In EHRCs the antecedent is higher than the proform both syntactically and semantically. Syntactically, the antecedent is an argument of the matrix clause that contains the clause linked to the LS with the co-indexed proform. Semantically, the higher status of the antecedent does not need to match a higher ranking on the Actor-Undergoer Hierarchy, since they are mapped to macroroles within different LSs.
The dog bit the boy that killed the spider.
The antecedent boy is an Undergoer in the matrix clause. Yet, the co-indexed relative pronoun is an Actor in the embedded clause. Therefore, the higher status of the antecedent bears on the LS that contains it in relation to the LS that includes the proform.
The Bambara IHRC is quite different in this regard. The antecedent is within a clause that is at the same syntactic level as the reduced slot. In turn, the semantics contains the bounded variable in the LS of the dependent verb, which is nested within the attributive structure that contains the antecedent as an argument. This contrast – plus the fact that the relative clause precedes the matrix clause – strongly suggests that the co-indexing relation in Bambara IHRCs is driven by pragmatics and world knowledge, as we argued above.5
As for non-restrictive relative clauses, these are clearly different from restrictive ones, in that there is a pause separating the relative clause from its antecedent. In addition, the attribution does not serve the purpose of helping the hearer to identify the referent. Indeed, as can be seen in (17), the antecedent can be a proper noun. Consequently, the information it provides does not need to be presupposed. These properties are well taken care of by the analysis in Figure 15.3, in which the relative clause is a peripheral modifier of the RP (Van Valin 2005: 222).
Chris, who loves soccer, flew to Barcelona.
Figure 15.3 Non-restrictive relative clauses
Finally, we should briefly analyse the so-called ‘headless relative clauses’ or ‘free relative clauses’ as exemplified in (18) and (19). The embedded clauses in those sentences (what John said and los que rompieron el libro ‘those who tore the book’) are introduced by wh-words (what and los que) which satisfy argument positions in the matrix LS. The wh-words are in turn modified by an embedded clause.
What John said surprised Sandy.
(19)
Los que rompieron el libro corrieron a sus casas those.m.pl that tear.pst.3pl the-m.sg book run. pst.3pl to their house.pl ‘Those who tore the book apart ran to their houses.’
From a purely semantic standpoint, these examples are like relative clauses, since the argument position in the matrix clause is filled in by the entity referred by the wh-word. This entity is in turn modified by the LS of the embedded verb. In particular, it is not the case that the LS of the embedded clause is an argument of the matrix LS, but rather it is part of the information that fills in that argument position. This meaning is captured by (20), an LS that is structurally identical to the one assigned to the relative clauses previously analysed. There is a crucial difference, though. In relative clauses the underlined argument is filled in by a full noun that is the head of the relative clause and functions as the antecedent of the relative pronoun. In contrast, in (20) the position is occupied by a pronoun and the typical slot of the relative pronoun in the attributive LS is filled by a variable co-indexed with the pronoun. It is clear from (20) that the entity that caused the surprise is the referent of what, and the LS that corresponds to the embedded clause helps to restrict the possible referents of that pronoun, which is just what plain restrictive relative clauses do. In addition, the description in (20) derives the fundamental property of the pronoun in (18) and (19), namely that they do not have antecedents and are thus free to any (contextually appropriate) interpretation.
[do′(whati [be′[do′ (John, say′(John, xi)])]) CAUSE [BECOME surprised′(Sandy)]]
Is there any formal correlate of this semantics? The most prominent one is that the nominative inflexion on the main verb in (19) is plural and is being controlled by the pronoun los que ‘those that’. Crucially, the inflexion is not determined by the whole embedded clause los que rompieron el libro but only by the pronoun. In addition, the nominative case of the pronoun is also determined by the matrix clause. In a sentence where the pronoun plays a non-PSA role in the matrix clause, it has to be expressed with an accusative/dative marker (i.e. a ‘to’) like in Pedro reemplazó a los que llegaron tarde ‘Peter replaced those who arrived late’. Had the embedded clause determined the shape of the pronoun, it would have been just los que (nominative) not a los que (accusative). This is also true in cases where the pronoun has an Actor role in the matrix clause and an Undergoer role in the embedded clause (for example, El que invitaste llegó tarde ‘The one you invited arrived late’). In such a case the pronoun should be realized as a nominative form consistent with its meaning in the matrix clause.6 In contrast, the relative pronoun in plain relatives takes the shape that is consistent with the semantics it takes from the embedded clause. For example, in El niño al que le regalé el juguete sonrió ‘The boy whom I gave the toy smiled’: the pronoun expresses the Beneficiary of the verb regalar ‘to give’ and the Actor of the main verb sonreír ‘to smile’, but its formal realization is consistent only with its Beneficiary role in the relative clause.7
The generative tradition assumes either a ‘matching’ or a ‘raising’ analysis of free relatives which subsumes them within plain relative clauses (Cinque Reference Cinque2015). Both analyses assume the presence of an empty position outside the relative clause (the position of the head/antecedent), which is the landing site for the pronoun of free relatives. One of the problems of subsuming the free relatives to plain relatives is that they do not behave in the same way in the so-called ‘extraction’ phenomena. The antecedent of the plain relative El discurso que diste sorprendió a Sandy ‘The talk you gave surprised Sandy’ can be replaced by a wh-word as in ¿Qué que diste sorprendió a Sandy? ‘What that you gave surprised Sandy?’. However, we cannot ask for the identity of the agent of the matrix clause in (19) as in *¿Quiénes rompieron el libro corrieron a sus casas? ‘*Who tore the book apart ran to their houses?’. The sentences in (18) or (19) cannot be reduced to regular relative clauses by claiming an empty position for the antecedent that is filled in by the pronoun coming from the relative clause. They need a different and specific treatment.
The evidence presented so far suggests, albeit not conclusively, that the pronouns in (18) and (19) are not relative pronouns. They are more properly analysed as arguments of their matrix clauses, as in Figure 15.4. We can call the embedded clause ‘pronounless relative clause’ in order, first, to emphasize the opposition with a headless relative clause analysis and, second, to point to the exceptional nature of this ‘relative clause’. In sum, this is a construction of its own that cannot be understood by just applying an old label to it. The linking algorithm has to incorporate a statement that guarantees that the interpretation of the non-syntactically realized argument of the verb in the embedded clause is driven by the co-indexation with the pronoun in the matrix clause.
Figure 15.4 Pronounless relative clauses
15.3 Cleft Sentences
Cleft sentences are as complex as they are interesting and, similarly to RCs, they involve a matrix and a subordinate clause. There are different subtypes of cleft, but in this chapter the focus will be on ‘it-clefts’, as represented by sentence (21). This example includes a cleft clause (who blamed the bus driver) introduced by a wh-pronoun that obtains its reference from the antecedent noun (teacher) in the matrix clause, also called ‘clefted noun’. This matrix clause consists of a copula with dummy it.
It was the teacher who blamed the bus driver.
The wh-form that introduces the cleft clause can fulfil any direct core argument position. In fact, it can even express adjuncts, as can be seen in (22).
It was in Buffalo where/that Shelly met her husband.
Clefts have properties that illustrate the complexity of the grammatical interfaces in natural languages. The same semantic information carried by (21) and (22) can be expressed by the simple sentences in (23) and, respectively, (24) without any loss in the description of the relevant events.
The teacher blamed the bus driver.
Shelly met her husband in Buffalo.
The unmarked expression of a proposition is a simple sentence, but clefts involve a bi-clausal structure with an unequal distribution of the content. The semantic weight of (23) and (24) is conveyed by their respective verb LSs. Yet, in clefts like (21) or (22), these content verbs are expressed in embedded clauses. In some sense, this is an ‘inverted’ linking, in that the semantic asymmetry by which one clause is richer in content than the other is not reflected by the syntactic asymmetry. The dependent clause expresses the semantically richer content, while the matrix clause conveys a weaker meaning. This ‘inverted’ syntax–semantics interface is pragmatically driven. It will be shown later that, while being semantically weaker, the matrix clause expresses the element that bears the focus of the sentence. Furthermore, the syntactic realization of the clefted noun in the matrix clause is determined by the semantic role this participant plays in the embedded cleft clause, as clearly shown by (25). In particular, the PP to the clerk needed to introduce this participant into the matrix clause is motivated by the syntax–semantics interface of the embedded clause. This participant is the third argument –Beneficiary – of a three-place predicate (give), therefore, it is a non-macrorole argument that can thus be expressed by a to-PP. There is no motivation for the formal properties of this argument (i.e. it being a prepositional phrase) in the matrix clause.
It was to the clerk that I gave my application.
Clefts are a primary example of a syntactic structure that is not entirely semantically motivated. Indeed, we can speak metaphorically of a syntax–semantics interface that is twisted for the expression of a particular kind of information structure. This is a point that might contribute to our understanding of linguistic systems. At every moment speakers are faced with the daunting task of capturing new situations and this means that every context requires to some degree novel linguistic responses. The lexicon and the rules that regulate the syntax–semantics interface are highly structured and, yet, clefts show that they are flexible enough to accommodate to the demands of a particular speech situation.
We can think of the communicative situation as the encounter of two vectors coming from opposite directions in a sort of ‘pincer effect’: a structuring force that holds the pieces together and the need to adapt to the particular demands of a context. Clefts show that the supposedly more rigid part of the system, syntax, has built-in templates specifically designed to serve information structure, one of the subsystems that deals with adjustment to the dynamics of contexts. There is no part of grammar that is immune to the demands of communication.
Clefts allow speakers to highlight the particular insertion of the content contributed by an isolated element – the clefted constituent – into the incomplete content of the cleft clause. The single proposition has been divided into an open proposition – a representation that comes with an open slot or variable – expressed by the clefted clause and a matrix clause that provides the value for that variable. The cleft clause is part of the presupposition and, hence, outside the potential focus domain in a strong sense: neither its internal constituents taken individually nor the entire clause as a unit can be the focus. Example (21), repeated in (26a) for convenience, cannot be a felicitous answer to the question What happened?. The potential focus domain only has scope over the matrix clause as confirmed by focus sensitive-particles. The negative operator in (26b) ranges only over the matrix clause and, consequently, on the co-indexation of the teacher as the Actor of the embedded clause.
a. It was the teacher who blamed the bus driver
b. It wasn’t the teacher who blamed the bus driver.
Further, a yes/no question can only ask for the clefted RP. Was it the teacher who blamed the bus driver? can be answered by (26a) whereas Was it the bus driver who was blamed by the teacher cannot. The appropriate context for an it-cleft sentence is a contrastive one; it had been asserted that somebody else other than the teacher (for example, a student) had accused the bus driver. This contrastive focus typically conveys an exhaustive reading of the clefted RP; namely, it was the teacher and nobody else who accused the bus driver (Kiss Reference Kiss1998).
Languages like French have constraints preventing PSAs – subjects – to be part of the actual focus domain in a sentence focus structure (SF). Since French also has rigid SVO word order, preverbal elements cannot normally be in focus. In consequence, French might use clefts as in (27) or presentational clefts as in (28) as pragmatically felicitous in a context that demands SF (Lambrecht Reference Lambrecht1994: 226, Reference Lambrecht2000: 653; Van Valin Reference Van Valin, Raxilina and Testelec1999: 519).
(27)
J’ai ma voiture qui est en panne. I have.prs.1sg my car that be.prs.3sg in breakdown ‘My car broke down.’
(28)
Y a Jean qui a téléphoné. there have.pst.3sg Jean who have. pst.3sg call.ptcp ‘Jean called.’
These sentences can be a felicitous answer to a What happened? type of question. The rather fixed French word order and the pragmatic restriction of subjects as topics motivates the use of clefts.
In RRG the presence of a bi-clausal structure like the it-cleft raises the issue of nexus and juncture linkage. ‘Nexus’ refers to the nature of the dependency between the two clauses, specifically, whether one entirely depends on the other (subordination) or they are independent of each other (coordination) or, finally, they are co-dependent (cosubordination). In addition, the theory of ‘juncture’ individuates the layer in which the combination occurs: nuclear, core or clausal (see Chapter 13). Since clefts contain an obligatory co-indexation relation among members of different clauses, the corresponding juncture is at the core layer. Regarding nexus, we can rule out daughter subordination since the subordinate clause does not fill an argument slot in the matrix clause and the co-indexation relation might involve adjuncts in the embedded clause. Furthermore, the embedded verb inflexion agrees with the wh-form and the clefted element in the matrix clause does not control agreement. This can be seen in Spanish where the antecedent noun is a first-person pronoun while the embedded verb agrees with the third person that corresponds to the embedded pronoun (Pavey 2004: 29).
(29)
Fui yo el que chocó el auto. be.pst.1sg I the.m.sg that crash.pst.3sg the.m.sg car ‘It was me who crashed the car.’
This should be another piece of evidence that supports a core peripheral subordination in a core juncture analysis. In the RRG account, the nucleus of a clause with a copula is the attributive expression, in this case, the RP. The embedded clause is a modifier of a core with a nominal nucleus. In this way the similarity between it-clefts and RCs is also captured (see Figure 15.5).
Figure 15.5 Cleft sentence
The embedded clause in it-clefts is a peripheral modifier at the core layer and, hence, cannot be part of the focus. There are some exceptions to this, but they are other subtypes of clefts. For instance, there-clefts like the one in (30) might be appropriate answers to What happened?
There was one man that kept interrupting.
This is an existential cleft with a lightly informative antecedent noun; the cleft clause might be part of the focus (Bentley et al. Reference Bentley, Ciconte and Cruschina.2015). Another example is offered by it-clefts with ‘this’ or ‘that’ in place of the ‘it’ pronoun, like (31). The information in the matrix clause retrieves already known information so that the assertion falls directly on the identity relation between the copula and the cleft sentence, which is the focus of the assertion.8
(31)
a. That’s the reason I don’t want to go to Miami! b. Yeah. Wasn’t that somewhere in Southern Florida where they thought those people got AIDS from bug bites…
Spanish does not have It-clefts but it does have a construction that is structurally similar to the English one. It places the focus on the cleft constituent in a clause with a copula followed by an embedded clause introduced by a relative pronoun.
(32)
Fueron los niños los que pidieron leche. Be.pst.3pl the.pl child. pl the.pl that ask.pst.3pl milk ‘It was the children who asked for milk.’
A difference with English is that in the Spanish cleft, the clefted noun might be the subject of the matrix clause – as in (32) – since it controls the agreement with the verb inflexion (i.e. the copula). Hence, the absence of an ‘it’ pronoun cannot be attributed to the Spanish ‘pro-drop’ status. Despite the variety of relative clauses existing in Spanish, here the relative pronoun has to be of a certain kind – that is, lo que (‘the.m.sg that’), la que (‘the.f.sg that’), los que (‘the.pl that’) and so on. As in English, the morphosyntactic realization of the antecedent noun is determined by the meaning of the embedded verb. In (33) the subject is impersonal and the antecedent is the nominal element of a non-predicative PP that marks indirect objects.
(33)
Fue a los niños a los be.pst.3sg to the.pl child.m.pl to the.m.pl que les dio la leche. that them give.pst.3sg the.f.sg milk ‘It was to the children to whom he gave the milk.’
Furthermore, if the co-indexed argument plays the role of an adjunct in the cleft clause, it should be preceded by the relevant predicative preposition, as shown in (34).
(34)
Fue en el baño donde encontré la llave. Be.pst.3sg in the.m.sg bathroom where find.pst.1sg the.f.sg key ‘It was in the bathroom where I found the key.’
In spite of being syntactically embedded, the verb LS of the cleft clause determines the morphosyntactic realization of the matrix participant. This is a key property that justifies a different analysis of RCs and clefts. The syntax–semantics linking in clefts displays properties of an ‘inverted’ interface. Why is it, then, that the cleft clause is not represented as the matrix clause? The presence of a subordinator – either a relative pronoun or a complementizer – is the crucial piece of evidence for the embedding of the cleft clause. In addition, an antecedent that controls the reference of a pronominal expression typically occupies the syntactic dominant position, which in this case is the matrix clause.
15.3.1 Linking Cleft Sentences
I shall follow Pavey (2004: 215) in assuming that the LS that corresponds to it-clefts is based on the be′ predicate structure that expresses the meaning of clauses with a copula, as shown in (35). This predicate captures the specificational nature of the predication, one of the meanings that can be expressed by the copula in English and Spanish, the one that corresponds to clefts. In particular, the specificational meaning highlights the ‘exhaustive reading’ in the sense that it was the teacher and nobody else who blamed the bus driver.9
be′([INGR blame′(who1, bus driver)], teacher1)
There is an important assumption underlying (35) which is that LSs are not merely registering propositional contents (interpreted either as a cognitive state or as truth-conditional content). In fact, if this were the case, this LS should also correspond to The teacher blamed the bus driver since, as I pointed out earlier, this sentence and the one in (21), (26a) have the same propositional value (see also Lambrecht 1994: 22). On the contrary, it is assumed that LSs should also be sensitive to the way symbols are combined which, in addition, highlights the systematicity of the syntax–semantics–pragmatics interface. The salient property of clefts rests on the satisfaction of communicative needs. The LSs of clefts should be consistent with the fact that a speaker is interchanging information with an addressee who is in a different cognitive state. It-clefts are nested in a communicative context where one interlocutor had asserted that an individual ‘x’ was the Actor of a blaming event. The speaker utters the it-cleft to contradict such a statement and offer a different identity for that participant.10
The syntax-to-semantics linking proceeds as follows. The clause with the copula demands that we retrieve a specificational structure and the LS of the verb in the embedded clause. Since it has no recoverable referential value, the argument slots of the specificational LS are filled in by the LS of the embedded clause and the full RP. This RP is the specificational element; hence, it has to fill in the second argument position. Then, we proceed to co-index the wh-pronoun in the embedded verb LS with the specificational RP. If there is a complementizer, then, the co-indexation involves the unspecified argument slot of the embedded verb LS. Finally, the bus driver is an RP in postverbal position. The Undergoer is, in turn, linked to the lowest-ranked argument in the verb LS. The highest-ranked argument is linked to the remaining wh-word or, if there is no available wh-word, it remains syntactically unexpressed.
15.4 Contrasting Clefts and Relative Clauses
There is a sense in which the central meanings of it-clefts and RCs are equivalent in terms of truth conditions. Specifically, in the sentence The student that failed the test quit the program, the truth-conditional content of the RC can also be expressed by a simple sentence (The student failed the test) and this exact content can be expressed by an it-cleft (It was the student who failed the test). The semantic difference in truth-conditional terms is reduced to the minimum; it-clefts add to the interpretation of the relevant noun the further assumption that the student was the only member who performed the action (exhaustive reading). The bulk of the semantic contrast lies in the way the semantic representation is assembled and these different assemblages are the encodings of different communicative functions. Speakers use RCs to provide further information that will help hearers to identify the referent of a head noun. It-clefts, instead, highlight the identity of one participant with a particular role in one event.
RRG captures this divergence through an independently motivated distinction. It is assumed that the be′ predicate includes different kinds of predications (Van Valin 2005: 48), even though they might all be expressed via a copula verb in a particular language. Two of them are relevant here; one is attributive predication and the other is specification (Declerck Reference Declerck1988: 47; Pavey 2004: 29), represented in (36) and (37), respectively. The former states a predication relation between a predicate and a referential expression. It subsumes attributive adjectives and, thus, the relation between relative clauses and their head nouns. In contrast, the specification relation co-indexes two referential expressions; one might be a constant (i.e. Chris) and the other is a definite description (the winner), a representation that allows us to identify a unique individual in the world.
(36)
a. Pat is tall a′. be′(Pat, [tall′])
(37)
a. Chris is the winner a′. be′(Chris, [the winner])
For the purpose of representing it-clefts, the definite description of structures like (37a′) is replaced by the clefted noun. The LS that captures the meaning of the sentence is thus shown in (38).
be′([INGR blame′ (who1, bus driver)], teacher1)
It is the head noun that provides the specifying information and, hence, it plays the role of the predicative element. Indeed, it is predicated of the embedded logical structure (i.e. ‘… blame′ …’) which constitutes the meaning of the embedded clause. The element that receives this predication is the co-indexed argument in the embedded verb LS (i.e. the wh-form).
In contrast, the matrix verb of the relative clause in RC provides the LS for the whole sentence, as can be seen in (10), which is repeated below in (39). The predicative structure (i.e. be′) that corresponds to the semantic relation between the head noun and the relative clause is inserted into one of the matrix LS argument slots (i.e. the first argument of quit′).The participant that fills in a semantic slot is literally shared with the predicative LS. This is not a typical co-indexation relation but a unique slot that belongs to two LSs, which is shown with a thick broken underline, as was pointed out above. In addition, this position is co-indexed with the argument slot filled by the wh-word in the relative clause LS.
INGR quit′([be′ (student1, [INGR pass′(who1, test)])], program)
The site of the syntactic contrast resides in the linkage of the embedded clause with the matrix clause. It-clefts involve a direct relation between two clauses, while the relative clause in RC acts as a modifier of the head noun and it is embedded into an RP in RC. This difference is attested, for example, by the different scope of the universal quantifier in (40) and (41) (Davidse Reference Davidse2000: 1114, cited in Pavey 2004: 193).
It was all the passengers who had committed the murder.
All the students who attended will receive a bonus point.
In (40) the quantifier has scope over only the clefted noun, and the co-indexed pronoun in the cleft clause inherits this interpretation. The full RP denotes the set of ‘all passengers’ and all of them committed the murder. In other words, the cleft clause does not affect the interpretation of the noun phrase in the matrix clause. By contrast, in (41), the quantifier has scope over the RP the students, which has already been restricted by the relative clause. Therefore, the relevant set does not include all the (contextually possible) students but only the subset of them that attended. Indeed, it is implicated that some students did not attend the class. The relative clause restricts the set of the noun to a subset and it is only to this subset that the universal quantifier applies. These properties can be derived naturally by representing relative clauses within the RP whereas cleft sentences are outside the RP, as shown in Figure 15.6.
Figure 15.6 The contrast between cleft sentences and relative clauses
With respect to information structure, neither the relative clause in RC nor the cleft clause in it-clefts are part of the assertion, but are instead presupposed. They both carry a proposition with a variable, namely, an open proposition. However, these embedded clauses differ in subtle ways. The relative clause helps the interlocutor to identify the reference of the head RP, which is necessarily a non-identifiable participant. In contrast, the reference of the clefted noun could be identifiable or not, but it receives an exhaustive reading and, crucially, its pragmatic function as focus is to saturate the open proposition introduced by the embedded sentence. The cleft clause provides the event in which this entity was involved. Understanding focus in Lambrecht’s sense, the embedded clause in clefts carries information that is necessarily part of the focus. In the example we have been analysing, this is the semantic role ‘Actor’ in the blaming event of the teacher (clefted noun). The teacher is the focal element but the focus is not only the identity of the teacher but also its role (Actor) in the event described by the embedded clause. Let’s call the kind of open propositions carried by embedded clauses in clefts ‘operative presuppositions’ since they carry information that is necessarily in focus. The embedded clause in RCs also carries an open proposition, but it does not function in the same way in information structure. Its information helps in the identification of a participant that might not be in focus at all. In sum, RCs are related to the issue of identifying a referent whereas clefts are centred on focus structure and, consequently, their respective embedded open propositions have different functions: one is an ‘operative presupposition’, whereas the other is just a plain presupposition. Their syntactic templates are consistent with their different informative roles since the relative clause is embedded into the RP, whereas the cleft clause is a modifier of the matrix core and it does not play any role within the RP.
In some sense, any sentence with a pronominal form is associated with an open proposition. The pragmatic notion of open proposition that is relevant here is that of an open proposition that is specified by the focus. The embedded clauses of both constructions are open propositions, but only the cleft clause is an open proposition in a pragmatically relevant way, namely, it is the presupposition specified by the element in focus.
One condition that makes it possible for the relative clause in RC to be presupposed, even though it carries specifying information, is to be found in its syntactic layout. While the cleft clause is outside the clefted RP, the relative clause is a modifier within the RP in RC. The embedding of the relative clause in the RP leaves it out of the potential focus domain in a weak sense, that is, it can be the focus as a whole, but its internal constituents cannot be in focus on their own. This is shown in (42a), where the wh-form targets the whole RP in (42b).
a. Which student quit the program?
b. The one that failed the test.
Another contrasting behaviour related to the inherent pragmatic functions of the constructions is that the head noun in it-clefts can be a proper noun whereas this is not the case with RCs, as shown in (43) and (44), respectively.
It was Peter who sold the house.
*Peter who sold the house came to see you.
In sum, the relative clauses in RC can be the focus as a whole because they are nuclear modifiers of the RP, which might be within the potential focus domain of the sentence. If the relative clause modifies a noun head that is outside the potential focus domain – for example, an RP within an adverbial clause in a clause juncture – the relative clause loses the possibility of being part of the focus.
15.5 Conclusion
The study of RCs and clefts is enlightening in different ways and at different levels. From a very abstract perspective, RC is a primary example of the need for phrases in any grammatical theory. It is a structure – not just a mere concatenation– that results from a meaningful combination of symbols. The relative clause in RC is semantically and syntactically independent from the matrix clause because it is embedded within a noun phrase. The RP is an enclosure that fosters relations among the symbols inside, preventing them from having certain relations with symbols outside. In addition, RC reveals the structural richness of human languages. As defined here, relative clauses are embedded into noun phrases as peripheral modifiers of the nucleus. It turns out that different subordinations are possible even in a single language, through a relative pronoun or no subordinator at all. Furthermore, a typological perspective instructs us about different structural conditions for the antecedent–pronoun co-indexation relation. In the typologically more frequent externally headed relative clause (Van Valin 2005; Comrie Reference Comrie1998), the antecedent noun is in the matrix clause whereas the referentially dependent form or argument slot belongs to the dependent clause. In internally headed relative clauses the situation is the opposite, in that the full noun is part of the dependent clause whereas the argument slot that takes its value from that noun is in the matrix clause.
Clefts too illustrate very general properties of linguistic systems. The perspective I have taken here to describe clefts bears on the question: how deep does information structure carve into the grammatical system? In other words, the issue is whether a system that is designed to satisfy communicative needs operates at the grammatical core. A functionalist linguist would predict that information structure goes all the way down to the centre of the grammatical system, and this is precisely what it-clefts show. The matrix clause does not carry the semantically rich content of the sentence, which is in fact contributed by the embedded clause. This embedded clause determines even the morphosyntactic realization of the attributive expression (typically a noun phrase but also a prepositional phrase) in the matrix clause. This is what, in this chapter, has been called ‘inverted’ linking. However, the matrix clause conveys the focal element of the assertion, that is, it provides the salient information structure content. In short, information structure reverses the dependency relation that would be predicted from mere semantic considerations.
The contrast between clefts and RCs highlights how two structures that share striking similarities can be at the same time extremely different. Both clefts and RCs contain an embedded clause that can be introduced by a relative pronoun. Both of them hold a co-indexation relation between an element in the matrix clause and another in the embedded clause. However, the relative clause in RC is a modifier of the antecedent noun, whereas clefts involve an interclausal relation such that the cleft clause is dependent on the matrix clause; it is in the periphery of the matrix core. This analysis explains among other facts the different scope of quantifiers in the matrix clause. Their different structural shape is systematically tied to their different communicative function. The function of RCs revolves around the identity of the participant expressed by the head noun. The listener has to identify it with the help of the information carried by the embedded clause, information that qualifies as a presupposition. The function of clefts is centred on information structure. The matrix clause conveys a participant whose identity, together with the role it plays in the event introduced by the embedded clause, constitutes the focus of the sentence. In this sense, the embedded clauses in RCs and clefts both carry open propositions but the one in RC counts as a plain presupposition while the one in clefts is an ‘operative presupposition’, which means that part of it is in focus.
