CLIPP Christiani Lehmanni inedita, publicanda, publicata€¦ · 1 We thank the Cabecar speakers named in §2.1 for their cooperation, the members of Martin Haspel-math’s PhD student

CLIPP

Christiani Lehmanni inedita, publicanda, publicata

titulus

The Cabecar relative clause(Guillermo González Campos coauctore)

huius textus situs retis mundialis

christianlehmann.eu/publ/gonzalez_lehmann_ cabecar_relative_clause .pdf

dies manuscripti postremum modificati

16.05.2020

occasio orationis habitae

Workshop der FG 1783 “Relativsätze”, 4.-6.5.2017, Universität Frankfurt

volumen publicationem continens

International Journal of American Linguistics 87

annus publicationis

2021

paginae

ignotae

http://www.uni-erfurt.de/%0Bsprachwissenschaft/personal/lehmann/CL_Publ/%0B*.pdf%20



The Cabecar relative clause

Guillermo González Campos & Christian Lehmann

Abstract

The Cabecar relative clause is a circumnominal relative clause without any markingof subordination, attribution or head selection. It is recognizable as such by struc-tural criteria only if its syntactic function in the matrix is signaled grammatically, andotherwise only by semantic and prosodic criteria. Eligibility of an internal nominalexpression as semantic head of the relative construction follows a complex decisionhierarchy of semantic and grammatical conditions. Diachronically, the relative con-struction is derivable from an asyndetic combination of two independent clauses.

Keywords: Cabecar, circumnominal relative clause, internal head, syntactic func-tions hierarchy, restrictiveness

1 IntroductionThis article presents the basic grammatical properties of the relative construction of theCabecar language [ISO 639-3: cjp] of Costa Rica,1 with an eye towards providing an ele-mental structural description and highlighting some of its important typologicalfeatures. The Cabecar relative clause has not been described previously; but relativeclauses of the Cabecar type are known to the typological literature.2 It consists of aninternally headed structure with no overt marker of subordination that has the syntacticfunction of a noun phrase, as illustrated in E1.3

1 We thank the Cabecar speakers named in §2.1 for their cooperation, the members of Martin Haspel-math’s PhD student seminar for fruitful discussion, as well as the journal editor and twoanonymous reviewers for substantial help with improving our text.2 The relative construction of the closely related Bribri language is similar in some respects andis the object of Wilson 1984, Villalobos 1994 and Coto-Solano et al. 2016.3 Letters have the same phonemic values as in Spanish orthography, with the following excep-tions: <l> is /ɺ/, <y> is /ʤ/, the digraphs <pj tj kj> represent aspirated stops. Vowels may bemarked by three diacritics: the underscore <_> marks nasality; the acute <´> marks high tone;the dieresis distinguishes particular vowels as follows: <aA> is /ɤ/, <eA> is /ɪ/, <oA > is /ʊ/.– The fol-lowing abbreviations are used in glosses: 1, 2, 3 first, second, third person, AD adessive, AM

autonomous motion, AND andative, APPP appropriate, APUD apudessive, ASC ascensive, AVERS aversive,CAUS causative, CL.ARBUST shrub class, CL.ELONG elongated class, CLM culminative, COP copula, D.AUD audi-ble demonstrative, D.MED medial demonstrative, D.MID dynamic middle voice, D.PROX proximal

González & Lehmann, Cabecar relative clause 2

E1 BaL amíL te chíLchi m-aL =sa Fernando ia su-aL yíLs te.[2SG mother ERG dog put/give-PFV=EGR Fernando DAT] see-PFV 1SG ERG

‘I saw the dog that your mother gave to Fernando.’ (FOM)

The relative clause contains the nominal expression – here chíchi ‘dog’ – which is itssemantic head; there is no external head. There are no morphological or syntactic prop-erties that would identify the (semantic) head. The relative clause has the samestructure as an independent declarative clause. It lacks any morphological or syntacticfeatures to mark the subordination. In the matrix clause, the relative clause occupies thesyntactic function of a noun phrase – in E1, it is the absolutive actant of the main verb.4

What is particularly interesting from a typological point of view is that such a relativeclause, although circumnominal, may easily be non-restrictive and that a non-restrictiverelative construction is in no way marked as differing from a restrictive one. This isrelated to a fact – likewise not in harmony with common expectations – concerning thedistribution of determiners: while the noun phrase formed by the relative clause – forexample, the absolutive actant in E1 – is not accompanied by a determiner, the internaldomain nominal (§3.1) may bear any determiner.

The core section of this article, §3, will analyze each of these features in some depth.In order to contextualize the properties of the relative construction in the rest of the lan-guage system, §2 will first present the basic grammatical features of Cabecar and someconstructions which are functionally or formally similar to relative constructions. §4complements the synchronic description by suggesting a diachronic path that leads tothe observed state of the modern language.

2 Background on CabecarIn this section, we describe briefly those properties of Cabecar grammar which areneeded to understand the formation of relative constructions. Additionally, we mentiona few constructions which are related to relative clause formation in many languages,but which are not, or only marginally so, in Cabecar. This is necessary in order to distin-guish relative constructions from these other constructions and to make it clear thatcertain expectations the reader may entertain are not met and certain generalizationsthat might suggest themselves cannot be drawn.

2.1 Language and data

Cabecar is a Chibchan language of the Isthmian branch. Its closest affiliate and geograph-ical neighbor is Bribri, its only sister in the Viceitic subbranch of the Isthmian branch. It

demonstrative, DAT dative, DSP dispositive, EMPH emphatic, EGR egressive, ERG ergative, EXCL exclusive,EXIST existence verboid, FIN final (= purposive), IN inessive, INDF indefinite, INSTR instrumental, INT

interrogative, IPFV imperfective, LAT lative, LOC locative, MNR manner, NEG negative, NNR non-orientednominalizer, NTR neutral, PER perlative, PFV perfective, PL plural, POS positional, PPV proprietive, PROG

progressive, PS person(al pronoun), RFL reflexive, S.PRF stative perfect, SBJ subjunctive, SG singular,SNR subject-oriented nominalizer, SPEC specifier, SPRR superior, SUPER superessive, TEL telic, TOT totalaffectedness, TRL translative, VEN venitive, VIS visible. In glosses, ‘:’ marks any morphologicalboundary, ‘+’ marks a compound boundary. In all of the examples, brackets enclose the relativeclause.4 A clause component provided for in the (structural) valency of its verb is the latter’s actant.


is the largest indigenous language of Costa Rica, spoken by approximately 14,000 peoplein several regions of the eastern interior of the country. The dialects are commonlydivided into northern and southern Cabecar. There are sizable numbers both of monolin-gual speakers and of people who are bilingual in Spanish. The language has been writtensince 1947. However, most speakers never use the language in writing, and only a lim-ited number of texts, including a bible translation, have been published. The mostvoluminous descriptive work published to date is Margery 1989. There is no grammar ofthe Cabecar language available yet. A dictionary (GonzaL lez & Obando 2020) and a gram-mar (GonzaL lez & Lehmann 2020) are underway.

Examples have been drawn from a corpus of Cabecar texts and sentences that com-prises a total of 45.955 running words representing both principal dialects.5 The sourcesof the examples reproduced here may be described as follows:

a) A set of Cabecar texts is listed in the references section. Of these, “¿ManeLpíL saL te díLyueAL roA ?” and “La historia de YeALbuleAL ” are unedited oral productions, while theother three were composed. “JíLshtaAL seL jeL duchíLwaLk kaL seLneLwaL ?”, “SeAL rikeAL paLkeAL ” and“La historia de YeALbuleAL ” represent the northern dialect, while the other two textsrepresent the southern dialect.

b) Several examples were contributed by Freddy Obando MartíLnez (FOM), a speakerof the northern dialect from ChirripoL , Cartago. A few of these were elicited as evi-dence for relative constructions, but most originated in different contexts. Somesimilar examples are drawn from GonzaL lez & Obando 2020, devoted to the samedialect.

c) Some examples were elicited from Fidelia ZuL niga HernaLndez (FZH) and RonaldoMayorga FernaLndez (RMF), both speakers of the southern dialect from UjarraL s,Puntarenas.

Thus, data from source (a) represent natural oral and written discourse. Data fromsource (c) were elicited by translation from Spanish. Data from source (b) comprise afew examples of the same kind. These will be easily recognized as “linguist’s sentences”.However, most sentences from source (b) were produced freely in order to illustratesome lexical item and are used here because of the syntactic structure that they happento display. To the extent possible, this article does not rely on grammaticality judge-ments. There are three passages in the text saying that a deviation from a certain rulewould render a sentence ungrammatical; and in all of these cases, this was the unani-mous judgement of the three consultants mentioned.

2.2 Clause structure

Cabecar has four major clause types in terms of the type of predicate, represented by E2– E5. The predicate of the verbal clause is headed by a verb, tuwa� in E2.

E2 Pedro te kalwaL tuw-aL .Peter ERG bench deal-PFV

‘Peter bought a bench.’ (RMF)

While the preverbal position of the absolutive actant is obligatory, order of other verbaldependents is free.

The predicate of an existential clause is headed by a form of the verboid tso� , as in E3.

5 The corpus is described in detail in GonzaL lez Campos 2016, §5.2.1.


E3 Chimo tsoL nala kjaAL jami skíL-blaAL .banana EXIST path edge AD five-CL.ARBUST

‘There are five banana plants at the edge of the road.’ (GonzaL lez & Obando 2020s.v. ski�bla� �)

Again, the locative adverbial here might alternatively introduce the sentence.The predicate of a positional clause is headed by a positional, kaldu in E4.

E4 YíLbaAL kaldu baL shabeAL kjane?who standing 2SG back APUD

‘Who is standing behind you?’

The predicate of an ascriptive clause is headed by a form of the copula dä, followed bythe predicate complement, as in E5.6

E5 jeL roA keAL geAL keAL goA loAD.MED COP major guardian.spirit‘he was a major guardian spirit’ (yer_15)

The verb does not inflect for person. Number conjugation is rudimentary and affordedby the postverbal clitic julu, which is limited to verbs of motion and indicates autono-mous motion and plural for their absolutive actant, as in E6.

E6 koAL chi-waL mineL=julupig-PL go:PFV=AM:PL

‘the pigs went away’ (⊂ E36)

All syntactic and semantic relations of noun phrases, with the exception of the absolu-tive, are marked by postpositions. Examples include the ergative in E2, the adessive inE3, the apudessive in E4 and the absolutive in all of these sentences. In constructionsinvolving an unmarked complement (absolutive of a verb, possessive dependent of anoun, complement of a postposition), constituent order is left-branching. The alignmentof the actants of the transitive and intransitive verb in terms of case marking (i.e., bypostpositions) follows the ergative system, while most “behavioral” aspects of grammati-cal relations follow the accusative system.

The system of verb-governed postpositions is complicated by wa DISPOSITIVE.7 This is apostposition (and a function marked by it) which combines the functions of the ergativeand the dative under complex conditions which are immaterial to relative clause gram-mar. It figures in several of the examples; E7 is offered for simple illustration.

E7 jeL kaL juL neL -r s’ waD.MED NEG know-D.MID(IPFV) 1SG DSP

‘that I don’t know ’ (⊂ E43)

In the simple verbal clause, the order of major constituents is rather free, except that theabsolutive actant of the verb appears immediately to its left, as already illustrated by E2and E4; this constraint is included in the left-branching syntax of unmarked comple-ments. E8 illustrates the typical disposition of actants in a transitive clause.

6 The copula is a verboid which inflects for a subset of verbal categories. The variants dä and räare conditioned phonologically, the variant rö (like other forms displaying <oA > instead of <aA>) isof the southern dialect.7 The term designates “the one who disposes”. Its Bribri equivalent is called agentive in JaraMurillo 2018, which term, however, does not fit the possessive function, as in E48 below.


E8 YíLs shtríL-waL taíL jíL taL chíL te.1SG tire-CAUS:PFV much D.PROX grandfather ERG

‘This grandfather has taxed my patience.’ (yer_18)

Verbal clause structure is schematized by Figure 1. W and Z may be anything, includingother actants of Y. X is obligatory except with a few verbal categories.

Figure 1 Verbal clause

[ (W) [ XNP YV ]VP (Z) ]S

absolutive predicate

Personal pronouns come in full and clitic variants. These also occupy actant positions,especially under anaphora.

2.3 Anaphora

By far the most common way of taking up an antecedent anaphorically is with thedemonstrative jé. This affords medial deixis and is the unmarked member of its para-digm. In textual anaphora across a sentence boundary, this pronoun may be deployed tofill an obligatory syntactic position. In the second sentence of E9, this is the position ofthe absolutive actant.

E9 JíL kaL +paLk-eAL raA kaL jíLraA kuL na ...D.PROX story+tell-NNR COP NEG now/today be.SBJ

saL te jeL paLk-eAL yaba-laL ia1.PL ERG D.MED tell-IPFV child-PL DAT

‘This story is not of our times … we tell it our children’ (ser_21)

In many cases, however, jé used as a resumptive pronoun follows its antecedent immedi-ately. In E10, it stands in for the preceding clause.

E10 SaL m-aL kuoAL tk-oA jeL roA ks-eAL ta.[1.PL go-PROG corn set-SNR] D.MED COP sing-NNR PPV

‘When we go to sow corn, we have a song.’ (lit.: ‘our going to sow corn, that is pro-vided with a song’; chicha_7.5’)

Even inside one clause, jé works as a resumptive following a semantically definite orgeneric NP. This use is optional, but frequent in oral discourse. E11 is an example of this;others include E23 and E63 below.

E11 BusuL bulu jeL roA taL chíL yuleAL+skilaBusubulu D.MED COP grandfather small+disfigured‘Busubulu(, he) was a gnomish grandfather’ (yer_14.1)

This construction originates in the left-dislocation of the antecedent as suggested by thetranslation. Its grammaticalization leads to the insertion of jé to mark the role of the pre-ceding constituent as a nominal expression in the ensuing context (GivoL n 1976). We willreturn in §4.2 to this resumptive in other syntactic constructions.


2.4 Attribution

There are two kinds of attributes, nominal (or possessive) attributes (E12f) and modi-fiers formed by adjectives (E14) and some other categories that behave syntactically inways similar to adjectives. The latter will simply be called “adjectival attributes”. Thenominal attribute precedes its head, as in E12.

E12 konoL yeAL riapaca hunter‘hunter of pacas’ (yer_00)

The anaphor jé shown in §2.3 above occurs optionally in the nominal attribute construc-tion, as in E13:

E13 alaLklaA (jeL ) jayíLwoman D.MEDhusband‘the woman’s husband’ (⊂ E51)

As noted in the preceding section, this immediate anaphora is limited to antecedentswhich are semantically definite or generic and is therefore not possible in E12. In pos-sessive attribution, it is grammaticalized as just noted in §2.3.

The adjectival attribute follows its head. It may be adjacent to it, as in E14a.

E14 a. jayíL yakaLaL d-eAL=ju=teAmale robust emerge-PFV=AM=VEN

‘the robust man came’ (FOM)

b. jayíL d-eAL=ju=teA yakaLaLmale emerge-PFV=AM=VEN robust‘there appeared a robust man’ (FOM)

An attribute of a nominal functioning as the absolutive actant of the verb or as the com-plement of a postposition may be displaced to some post-head position, possibly to theend of the entire clause, as in E14b. This is regularly the case on first mention of a refer-ent, as is the case here.

Apart from a plural marker for empathic nouns (for example in E6 and E23), there isno declension and there are no segmental means to signal attribution.

2.5 Determination

Determination is an operation which combines a determiner with a nominal expressionand thereby specifies the latter’s reference (MuA ller & Klinge (eds.) 2008). The structurematching this function is a binary construction in which the determiner either precedesor follows the entire nominal, whether the latter is simple or complex. In Cabecar, deter-miners introduce their noun phrase. There are demonstrative determiners, but noarticles. The categories of (in-)definiteness and (non-)specificity are not coded; nominalexpressions occur freely without any determiner. Different determiners are illustratedby E8 and E15. The same demonstratives that serve as determiners are also pronouns,that is, substitutes for an entire NP. This is illustrated in §2.3.


In many other languages, including English, providing a nominal with a determinerconverts it into a noun phrase, and these two categories differ in their distribution.8 InCabecar, constraints on prenominal constituents of an NP concern more their syntag-matic order than the conditions of their selection; a nominal and a noun phrase sharemost of their distribution.

E15 (JeL ) (diaL ) tipaAL raA sioAL oAL =wa.D.MED D.VIS.low pool COP blue=INTNS

'The/that pool (down there) is very blue.' (FOM)

The parentheses in E15 are meant to indicate that the sentence is possible with nodetermination of its subject, with a single determiner, or with two. On this basis, theterm “noun phrase” will be applied to nominal expressions whether or not providedwith a determiner.

2.6 Subordination and nominalization

Complement clauses are not overtly subordinated and have the structure of an indepen-dent declarative or interrogative clause. Dependent directive clauses are in the infinitive,the addressee figures in the main clause. Dependent declarative and interrogativeclauses will be treated in this order.

E16 shows a declarative clause in absolutive function.

E16 BeALna wa Rogelio seL -r daA DuchíL juL neL -r.all DSP [Rogelio live-D.MID(IPFV) COP ChirripoL ] know-D.MID(IPFV)‘Everybody knows that Rogelio lives in ChirripoL .’ (FOM)

While the complement clause of E16 occupies the absolutive position in front of themain verb, E17 shows a very common alternative:

E17 YíLs te i sh-aL=wa=pa jeL jawaL d-aAL =ju jíLr.1SG ERG 3 say-PFV=TOT=APPP [D.MED healer emerge-IPFV=AM now/today]‘Appropriately I said that the healer would come today.’ (FOM)

The complement clause is shifted to the end of the sentence and is anticipated in theabsolutive position by a cataphoric pronoun, i in E17. This is the process which has beencalled heavy shift or right-dislocation. This construction differs from asyndesis of twoindependent sentences by its intonation. Segmental differences between an extraposedcomplement clause and a corresponding independent clause are limited to interrogativeclauses; see below.

While the absolutive function is the most frequent for complement clauses, othersemantic functions available to propositions are marked by a postposition governing thecomplement clause just as the postposition would govern any NP. E18 illustrates a com-plement clause in ergative function. The complement clause of E19 is marked by theinstrumental required by the main verb.

8 This terminology follows Stockwell et al. 1973, ch. 3: a nominal is a constituent with the distri -bution of a common noun.


E18 I jaL -n-aL =teA kalwaL bata kíL te[3 lower-D.MID-PFV=TEL horse tip SUPER] ERG

parreL pj-aL =wa boL -taAwaA .rib break-PFV=TOT two-CL.ELONG

‘He broke two ribs because the horse threw him over.’ (lit.: ‘his falling down fromthe horse broke two ribs’) (FOM)

E19 S’ waAL batsaAL ijeL te nima kuk-eAL wa.1SG pleased-IPFV [3PS ERG fish catch-IPFV ] INSTR

‘I like for him to fish.’ (FOM)

The complement clause may also depend on a noun, as in E20.

E20 MíLkoA baleAL -n-aL =wa biyoAL sh-eAL ijeL te.[grandmother hide-D.MID-PFV=TOT] news say-IPFV 3.PS ERG

‘He tells the news that grandmother died.’ (FOM)

In this construction, the complement clause precedes its head noun just as the nominalattribute of E12 precedes its head noun.

As noted above, declarative complement clauses are finite and show no signs of nom-inalization. A Cabecar finite clause is less distinct from a non-finite clause than in otherlanguages because the verb lacks person and number as marks of finiteness. Conse-quently, Cabecar clauses lend themselves more easily to subordination without furthermeasures than finite clauses marked for person and number in other languages.

A content (a.k.a. “wh”) interrogative clause is introduced by the proform in theclause-initial focus position. Dependent interrogatives again have the same structure asindependent ones. E21 illustrates a right-dislocated content interrogative clause in abso-lutive function.

E21 Shk-aAL i su-a yíL=baAL tuL -r=keL suL t-suL t=síL raA .walk-IPFV 3 see-SNR [who=EXCL run-D.MID(IPFV)=IPFV2 hop-hop=SPRR COP]‘Let’s go and see who is best in jumping!’ (FOM)

By the same rules as for dependent declarative clauses, a dependent interrogative maybe in situ, as in E22a, or may be right-dislocated, as in E22b.

E22 a. Daniel te maíL baL kaldu chaLk-aL .Daniel ERG [where 2SG standing] ask-PFV

b. Daniel te i chaLk-aL maíL baL kaldu na.Daniel ERG 3 ask-PFV [where 2SG standing INT]‘Daniel asked where you were.’ (FOM)

The interrogative particle na is impossible with the clause in absolutive position of ver-sion (a), but optional in the right-dislocated variant (b). Its presence is then an elementof the asyndesis of two independent clauses.

2.7 Orientation

An expression derived from a verbal predication is oriented if it designates one of theparticipants of the underlying situation, and is non-oriented if it designates the situationitself. Thus, English employer and employee are oriented, while employment is non-ori-


ented. In Cabecar, orientation of deverbal nominalizations occurs in two kinds of mor-phological forms:

a) derivation of agent nouns and nouns oriented towards some other actantb) formation of the stative perfect.

As for (a), the suffix -ä SUBJECT-ORIENTED NOMINALIZER (its nasal allomorph is -a) requires somecomment. This morpheme has two functions: it forms the infinitive (as in E21) andderives agent nouns in the construction [ XV -ä ]N ‘one who Xes’. An intransitive base

yields an agent noun by itself. A transitive base is nominalized with its absolutive actant,jilé in E23.

E23 JileL tju-a-waL jeL=neL wa i juL neL -r daA .[something deal-SNR-PL] D.MED=EMPH DSP 3 know-D.MID(IPFV) COP

‘The buyers are those who know.’ (FOM)

However, a purely syntactic analysis cannot account for the meaning of all agent nouns,since they may be lexicalized, as in yabalá sua (child:PL see:SNR) ‘babysitter’. Other deriva-tions of deverbal nouns are even less regular.

An agent noun only allows for a moderate degree of syntactic complexity, and only tothis extent does it contrast with a relative clause headed by its actor. Like agent nouns inother languages, it categorizes its referent inherently instead of simply identifying it bysome situation that it happens to be involved in. This point is briefly taken up in §3.3.

As for (b) above, the stative perfect is a form oriented towards the absolutive actantof the base. E24 is a construction whose predicate is a stative perfect form of a transitiveverb. E25 shows the same form of an intransitive verb in attributive function.

E24 (ijeL -waL wa) ju raA yoAL -leAL3.PS-PL DSP house COP form-S.PRF

‘the house has been built (by them) / (they) have built the house’ (FOM)

E25 IjeL -waL tsoL chíLchi duL -leAL=wa taAL bi-aA .3.PS-PL EXIST dog die-S.PRF=TOT bury-SNR

‘They are burying the dead dog.’ (GonzaL lez & Obando 2020 s.v. ta� �biä)

The replacement of the ergative by the dispositive postposition and the optional use ofthe copula are symptoms of the non-finite, or rather semi-finite, character of this verbalor deverbal form: it is both a deverbal adjective with passive orientation, for instance inE25, and a conjugation form, for instance in E67 and E78 below.

3 Relative constructions

3.1 Analytic concepts

Much of the terminology used in this paper in describing relative clauses follows that ofLehmann (1984, 2003). We refer again to E26 for illustration.

E26 BaL amíL te chíLchi m-aL =sa Fernando ia su-aL yíLs te.[2SG mother ERG dog put/give-PFV=EGR Fernando DAT] see-PFV 1SG ERG

‘I saw the dog that your mother gave to Fernando.’ (= E1)

A relative construction has a conceptual center towards which the relative clause is ori-ented (§2.7; in E26, this is the dog). The relative clause modifies this center semantically.


If it is overt, it is represented by a domain nominal (DN; Andrews 2007: 208) – chíchi inE26.

A relative clause can be adjoined or embedded. If adjoined, it is a co-constituent ofits main clause. If embedded, it constitutes, together with its DN, a nominal expressioninside the main clause. Following Andrews 2007: 206, a nominal expression containing arelative clause will be called NPmat (matrix NP).

An internal-head relative clause is one which contains its semantic head as an NPoccupying the position corresponding to its syntactic function in the relative clause. Thusin E26, the semantic head chíchi is the absolutive actant of the verb ma� sa. An internal-head relative clause may be adjoined (viz., preposed) or embedded. The embedded vari-ety is called circumnominal (in analogy to pre- and postnominal).9 A circumnominalrelative clause forms the NPmat by itself. Thus, both the relative clause as a whole – likethe bracketed expression in E26 – and its DN constitute nominal expressions which maysimply be called noun phrases (NPs), as is common usage in most of the literature dedi-cated to relative clauses and is justified in the case of Cabecar, as noted in §2.5. More onthis in §3.2.4.

The adjoined relative construction comes in two main varieties: If the relative clauseis preposed, a demonstrative in the main clause anaphorically resumes the relativeclause, as in the sentence Whichever books you put on the list, nobody will read those. ACabecar example appears in E81 below. If the relative clause is postposed, its head is inthe main clause, generally accompanied by a demonstrative in cataphoric function. Thepostposed relative clause does not exist in Cabecar.

In common usage, the DN is called the head of the relative clause. The reader iswarned that, in the case of a head-internal relative clause, the DN is not the head in anystructural sense. The head of a Cabecar relative clause is a semantically-based conceptwhich may or may not materialize as a lexical-nominal expression – a DN – and which isnever marked as a structural head.

Finally, a text-semantic difference between restrictive and non-restrictive relativeclauses may be recalled. The restrictive relative clause is a modifier which forms amore specific concept on the basis of the DN. The clause that it embodies is not assertedand therefore typically uses available information rather than conveying newinformation. For instance, in E26, the speaker does not claim that ‘your mother gave adog to Fernando’; instead, he uses this proposition as a modifier to pin down the referenthe has in mind. Things are different with a non-restrictive relative clause; since this isnot needed to specify a concept, it may assert new information.

The DN nominal may be absent or lexically empty. In E27, there is none.

E27 BaL te i tju-aL yíLs te i sh-aL ska ?2SG ERG 3 deal-PFV [1SG ERG 3 say-PFV] LOC

‘Did you buy it where I told you to ?’ (FOM)

Here, NPmat has the function of a local adjunct in the main clause, marked by the locativepostposition ska. For the semantic interpretation to work, the relative clause has to des-

9 In several recent treatments (for example Basilico 1996), the term “internally headed relativeclause” is restricted to circumnominal relative clauses (de Vries 2002: 20-23 is hesitant aboutthis). This is unfortunate because circumnominal relative clauses are most similar to preposedrelative clauses, to the extent of being indistinguishable from them in many practical cases andgenerally bearing a diachronic relation to them.


ignate a place. In the construction of E27, the relative clause does not contain any traceof this core concept. In such a case, there is nothing for the relative clause to modify; itconstitutes the NPmat by itself. In the English translation of E27, the relative pronounwhere does signal the meaning ‘some place’ inside the relative clause. Both the relativeclause that lacks a DN and the relative clause whose DN is represented by a pronoun aresubsumed under the traditional term free relative.10 At the same time, relative clausesfunctioning as adverbials in the main clause, as in E27, are adverbial relative clauses.They will be dealt with in §3.3.

3.2 Relative clauses with a lexical domain nominal

3.2.1 General structure

The Cabecar relative clause is a circumnominal relative clause, since it contains anominal expression which is understood to be semantically modified by the rest of thisclause, and the clause itself is understood to be oriented towards this nominal compo-nent. The general structure of a Cabecar relative construction with a lexical-nominalhead is as shown in Figure 2.

Figure 2 Circumnominal relative construction

[ (X) [ [… YNP … ]S1 ]NP.i (D.MED)i (Z) ]S2

DN NPmat resumptive

The relative clause S1 depends on one of the constituents of the main clause S2. X and Zmay be anything, but one of them contains the main clause predicate and therefore isnon-zero. There is no grammatical formative to either mark S1 as a subordinate clause –let alone as a relative clause – or to identify its DN Y. The latter is simply a nominal con-stituent of S1. Taken by itself, S1 is, thus, formally indistinguishable from any otherunmarked subordinate clause, including in particular a finite complement clause (§2.6).If S1 precedes or follows S2, S1 may, indeed, be interpreted as an independent declarativeclause if the context permits (§4.2). The relative clause is substantivized implicitly andthus functions as an NP (NP.i) in the main clause., viz. the NPmat.

The paradigmatic relationship between an independent declarative clause and a rela-tive clause is illustrated by E28f.

E28 DulaL gleAL teA kaL lbatioA gloA ts-aL =u=mi busíL ia.boy ERG hat transport-PFV=AM=AND girl DAT

‘The boy took the hat away from the girl.’ (FZH)

E29 KaL lbatioA gloA ts-aL =u=mi dulaL gleAL teA busíL ia[hat transport-PFV=AM=AND boy ERG girl DAT]

(jeL) yul-eAL -geL saL teA .D.MED search-IPFV-IPFV2 1PL ERG

‘We look for the hat that the boy took away from the girl.’ (FZH)

The relative clause in E29 (enclosed in brackets) represents a referent identified by therole it plays in the situation designated by this subordinate clause. The initial position of

10 For Cabecar, it suffices to work with a pronominal DN; the concept “light-headed relativeclause” (Citko 2004) is not needed.


the DN in the relative clause will be discussed in the next subsection. The relative clauseis NPmat and the absolutive actant of the main verb in E29. As anticipated in §2.4, the rela-tive clause is optionally taken up in the main clause by the medial demonstrative jé. Thisis a feature of several examples to follow and will be attended to in §4.

Whether or not the relative clause precedes the main clause, it is prosodically inte-grated into it. There is no intonation break after it, the tone remains high, it is notfollowed by a pause as an independent declarative clause would be. This uninterruptedintonation contour includes the optional resumptive.

Internally, the relative clause may be of any of the clause types introduced in §2.2.E30 is one of the relatively rare occurrences of an ascriptive relative clause.

E30 DíLsaleAL leAL jeL raA bikaLkla duL i seLj -r díLklaAL kjaAL jami.[kingfisher D.MED COP ceremony.master bird] live-D.MID(IPFV) water edge AD

‘The pygmy kingfisher, which is the bird of the master of ceremonies, lives by ariver bank.’ (GonzaL lez & Obando 2020 s.v. dísäle� �le� �)

E33 features an existential, E61 a positional relative clause. Since the full gamut of inter-nal variability of the relative clause is only reached if its predicate is verbal, thediscussion will focus on this clause type.

3.2.2 Selection and syntactic function of the DN

There are no external-head or adnominal relative clauses in Cabecar. A relative clausedoes not differ in its structure from a complement clause, and there is no structuralmeans to identify its DN. The question therefore arises how language users knowwhether a subordinate clause of the structure of S1 in Figure 2 is a relative clause andwhat it is oriented towards. In the following, we present a sequence of steps down a hier-archy of semantic and grammatical constraints which guide the interpretation of such aconstruction. We first summarize this hierarchy in Figure 3. The alternative at the lowestlevel of Figure 3 corresponds to the southern vs. northern Cabecar dialect.

Figure 3 Hierarchy of constraints on relative clause interpretation

1 Orientation

The dependent clause is oriented unless the selection restrictions of the superordinate verb allow a proposition as its actant.

2 DN selection

a Grammatical features of the superordinate predicate

The grammatical number of the DN is identical to the number of the superordinate predicate.

b Selection restrictions of the superordinate predicate

The DN meets the selection restrictions of the superordinate verb

c Syntagmatic position of the DN Syntactic function of the DN

The nominal expression occupying the initial position in the relative clause is the DN.

The nominal expression of the highest position on the hierarchy of syntactic functions (Figure 4) is the DN.


The logic of Figure 3 is that of a filter which constructions enter at the top. Application ofthe constraint at a given level leaves a set of possible interpretations. As long as morethan one interpretation remains available, the construction passes down to the nextlower level. Needless to say, if a sentence passes none of these constraints, it is ungram-matical or senseless. We now consider the levels of Figure 3 one after the other.

Constraint 1 is semantic in nature. Since the orientation to one of the participants inthe case of a relative clause is not coded, it is entirely a matter of context conditions: i fthe superordinate verb selects a propositional actant, as in E31, the subordinate clausecan remain non-oriented and consequently be a complement clause.

E31 AlaLklaA te tebeL líL yakaL kia-r bikeAL ts-aL .woman ERG [knife stuff need-D.MID(IPFV)] think-PFV

‘The woman thought that the iron-thing [i.e. car] was needed.’ (FOM)

A few verbs allow concrete objects and propositions as actants. One of these is the verbsua 'see'. As a consequence, a construction like E32 may remain ambiguous.

E32 I su-aL =raL kalwaL ts-eL biteAL -r=teA i kuta te.3 see-PFV=CLM [bench transport-IPFV come-D.MID(IPFV)=VEN 3 sister.of.male ERG]‘He could see (that) his sister was bringing a bench.’ or ‘He could see his sister (,who was) bringing a bench.’ (historia_12)

As the alternate translations render plausible, the semantic difference between the tworeadings is minimal. Since orientation is a purely semantic process in the Cabecar rela-tive clause, it may or may not happen in constructions like E32 and E1. The same kind ofambiguity exists for headless relative clauses; E69 below is an example.

Moving on to DN selection and deferring the discussion of the determination of theDN to §3.2.3f, we here only note that it may be represented by a lexical-nominal expres-sion, by some pronoun, or it may be zero. In each of these cases, the DN has somesyntactic function in the relative clause (Y in S1 of Figure 2). The following set of exam-ples from the northern dialect illustrates the DN in variants of the absolutive function inthe relative clause. In the non-verbal clause of E33, it is the complement of the existenceverboid.

E33 beALwaLk tsoL ju kjaAL jami kjoL yíL-r taL i[animal EXIST house edge AD] hubbub.make-D.MID(IPFV) much‘the animals that are near the house make much hubbub’ (ser_14)

In E34, the DN is the absolutive actant of an intransitive verb.

E34 BaL te jíLjíL tk-aL=ju yikíL jeL s-aL ?2SG ERG [earthquake cross-PFV=AM yesterday] D.MEDfeel-PFV

‘Did you feel the earthquake that happened yesterday ?’ (FOM)

In E35, the DN is the absolutive actant of a transitive verb.

E35 Yaba te martillo wa kal waAL kuchíL-i kp-aL palaLwa[child ERG hammer INSTR tree stump-SPEC beat-PFV at.one.stroke]

tk-aL=wa=ju i jaL r ka.cross-PFV=TOT=AM 3 cavity LAT

‘The stump that the boy beat with the hammer with one blow penetrated inside.’(GonzaL lez & Obando 2020 s.v. paláá)


In all of the above examples, the order of constituents in the relative clause is unaffectedby the DN status of the absolutive actant. Its interpretation as the concept that the refer-ent in question is based on is the result of applying the constraints #2 of Figure 3.

By constraint 2a, the grammatical number of the DN must be identical to the numberof the superordinate predicate. E36 - E38 feature three NPs in the relative clause all ofwhich satisfy the selection restrictions of the superordinate verb (condition 2b of Figure3), which requires an animate absolutive actant.

E36 JayíL te koAL chi-waL tju-aL busíL ia mineL=julu.[man ERG pig-PL deal-PFV girl DAT] go:PFV=AM:PL

‘The pigs that the man bought for the girl went away.’ (FOM, ⊃ E6)

E37 JayíL-waL te koAL chi tju-aL busíL ia mineL=julu.[man-PL ERG pig deal-PFV girl DAT] go:PFV=AM:PL

‘The men who bought the pig for the girl went away.’ (FOM)

E38 JayíL te koAL chi tju-aL busíL-waL ia mineL=julu.[man ERG pig deal-PFV girl-PL DAT] go:PFV=AM:PL

‘The girls for whom the man bought the pig went away.’ (FOM)

In this series, the one nominal expression of the relative clause which satisfies the num-ber feature of the superordinate verb (condition 2a) is chosen as DN, independently ofits syntactic function (condition 2c).

In E39, constraint 2a of Figure 3 forces no decision. However, only one of the twonominal expressions contained in the relative clause, the ergative actant, meets con-straint 2b, viz. the selection restrictions of the superordinate verb. It is chosen as DNirrespective of its syntactic function.

E39 JayíL te kuaAL tk-aL yikíL mineL=ju.[man ERG corn sow-PFV yesterday] go:PFV=AM

‘The man that planted corn yesterday went away.’ (FOM)

If more than one noun phrase in the subordinate clause passes constraints 2a and b,level 2c of Figure 3 comes into play. The two main dialects of Cabecar differ in thisrespect. We will first briefly review the strategy of the southern dialect, which is rela-tively simple, and then devote more space to the northern strategy.

E28 and E29 above, E40 and E41 below represent the southern dialect. In E40, theDN is the ergative actant, and in E41, the DN is the indirect object of the relative clause,as they are in the independent clause E28.

E40 DulaL gleAL teA kaL lbatioA gloA ts-aL =u=mi busíL ia[boy ERG hat transport-PFV=AM=AND girl DAT]


‘We look for the boy who took away the hat from the girl.’ (FZH)

E41 BusíL ia kaL lbatioA gloA ts-aL =u=mi dulaL gleAL teA[girl DAT hat transport-PFV=AM=AND boy ERG]


‘We look for the girl from whom the boy took away the hat.’ (FZH)


The crucial difference between the three examples E29, E40 and E41 resides in theclause-initial position of the DN. It is demonstrably not the head of a postnominal rela-tive clause, but still an internal head. This is evident from the accompanyingpostposition, which marks its role inside the relative clause independently both of its DNrole in the relative clause and of the syntactic function of NPmat. Since the selectionrestrictions of the main-clause predicate in this case do not determine a particular DNselection in the relative clause, then if word order in relative clauses were free as it is inmain clauses, all of these relative clauses would be ambiguous between the readings ofE29, E40 and E41. The consultant here disambiguates these readings by putting the DNin first position (the same strategy exists in the neighbor language Bribri).

The northern dialect works differently here. In E42, the DN in absolutive function isnot in clause-initial position, nor is the DN in the function of a temporal adverbial in E43.

E42 Yaba te martillo wa kal waAL kuchíL-i kp-aL palaLwa[child ERG hammer INSTR tree stump-SPEC beat-PFV at.one.stroke]

tk-aL=wa=ju i jaL r ka.cross-PFV=TOT=AM 3 cavity LAT

‘The stump that the boy beat with the hammer with one blow penetrated inside.’(= E35)

E43 (JeL ) MaríLa koL naL duaL s ka[D.MED Mary be.born-D.MID-PFV summer PER]

(jeL) kaL juL neL -r s’ waD.MED NEG know-D.MID(IPFV) 1SG DSP

‘I don’t know the year that Mary was born in’ (GonzaL lez & Obando 2020 s.v. kua)

In the northern dialect, the position of the DN in the relative clause does not matter; E29,E40 and E41 are synonymous, with the meaning of E29, as are E44a and b.

E44 a. ChíLchi te wíLshka su-aL (jeL) mineL=ju.[dog ERG cat see-PFV] D.MED go:PFV=AM

b. WíLshka su-aL chíLchi te (jeL ) mineL=ju.[cat see-PFV dog ERG] D.MED go:PFV=AM

Both: ‘The cat that the dog saw went away.’ (FOM)

What matters instead is the hierarchy of Figure 4, which constitutes level 2c in Figure 3for northern Cabecar.

Figure 4 Hierarchy of syntactic functions

absolutive actant

ergative ~ dispositive actant

indirect object

other syntactic function

In E45, all nominal expressions in the relative clause pass the upper constraints of Figure3.


E45 BaL amíL te chíLchi m-aL =sa Fernando ia su-aL yíLs te.[2SG mother ERG dog give-PFV=EGR Fernando DAT] see-PFV 1SG ERG

‘I saw the dog that your mother gave to Fernando.’ (= E1)

By the logic of Figure 4, the DN of E45 must be the absolutive actant, as it is in E44, aswell. The lower positions of Figure 4 only come into play at this stage of the selectionprocess if none of its higher positions meets the constraints above it in the decision treeof Figure 3. Thus, in E46, the ergative actant is chosen because the absolutive does notsatisfy the grammatical features of the superordinate verb, while the benefactive adjunctis lower on the hierarchy of Figure 4.

E46 JayíL-waL te koAL chi tju-aL busíL-waL ia mineL=julu.[man-PL ERG pig deal-PFV girl-PL DAT] go:PFV=AM:PL

‘The men who bought the pig for the girls went away.’ (FOM)

It may be observed at this point that the ordering of the absolutive and the ergativeactants on Figure 4 is one of Cabecar’s few features of syntactic ergativity, complement-ing the generalization made in §2.2 that most behavioral aspects of grammaticalrelations follow the accusative system.

The bottom level of Figure 4 includes complements of the verb governed throughother postpositions and adjuncts like the temporal adjunct of E43 and the benefactiveadjunct of E46. No rules of grammar have been found to hierarchize these.

The position of the dispositive actant in Figure 4 is demonstrated by E47 – E50.These examples are unambiguous; none of the other nominal components of these rela-tive clauses can be DN. Like E39, E47 shows that before the syntactic function hierarchybecomes relevant, the semantic requirements must be met.

E47 AlaLklaA wa tsiruL d-eAL min-eL=ju.[woman DSP cocoa emerge-PFV] go-PFV=AM

‘The woman who brought cocoa left.’ (FOM)

Thus, although tsirú has the highest position on Figure 4, it is not the DN because Figure4 does not even come into play. Like E44 and E45, E48 shows that the absolutive takesprecedence over the dispositive.

E48 AlaLklaA wa chíLchi tsoL min-eL=ju.[woman DSP dog EXIST] go-PFV=AM

‘The dog that the woman has went away.’ (FOM)

Like E46, E49 shows that the dispositive takes precedence over the indirect object.

E49 AlaLklaA wa tsiruL d-eAL yíLs yaba ia jeL min-eL=ju.[woman DSP cocoa emerge-PFV 1SG child DAT] D.MED go-PFV=AM

‘The woman who brought my child cocoa left.’ (FOM)

E50 is the passive version of E46; the dispositive has the same priority over the adjunctas the ergative of the active version.

E50 JayíL-waL wa koAL chi toL -n-aL busíL-waL ia min-eL=julu. [man-PL DSP pig deal-D.MID-PFV girl-PL DAT] go:PFV=AM:PL

‘The men by whom the pig was bought for the girls went away.’ (FOM)

The dispositive occupies the same level as the ergative in Figure 4 because these twonever contrast in one clause.


To complete the picture of the syntactic functions of the DN, E51 shows the DN in thefunction of a possessive attribute.

E51 BaL te alaLklaA jeL jayíL duaL=wa suj -aL ?2SG ERG [woman D.MEDman die:PFV=TOT] see-PFV

‘Did you see the woman whose husband died?’ (GonzaL lez & Obando 2020 s.v. jäyí)

In E51, the DN is taken from the lowest level of Figure 4 although its highest level wouldseem to be available, too. Selection of jayí for DN is disfavored by cultural constraints notaccounted for by Figure 3: since corpses are impure, talking about seeing one would vio-late a taboo.

Finally, it should be noted that since there are no markers of subordination, attribu-tion or DN selection, there is no difference between a relative construction and the kindof attribution seen in E25: the verbal form which was presented there as an adjectivalattribute might as well be the predicate of a relative clause.

3.2.3 Grammatical properties of NPmat

A relative clause may bear the plural suffix, as in E52.

E52 I sh-eAL míLile saL baLkleAL -waL3 say-IPFV [formerly 1PL EXIST.PFV]-PL

mi=keL kaL jaL r yul-aA ,go:IPFV=IPFV2 space/time cavity visit-SNR

‘It is said that our forefathers went to hunt,’ (lit.: ‘They say that the ones of us hav-ing formerly existed went to visit the pit,’) (GonzaL lez & Obando 2020 s.v. po�)

The subordinate clause of E52 is oriented towards the absolutive actant of the existentialverboid and substantivized. The plural suffix marks the NPmat; verboids lack the categoryof number.

We now come to the determination of NPmat. It was said in §2.5 that a determinercombines with a nominal.11 What matters here is the inclusion of modifiers. Suppose thatthe sequential position of the determiner matches its semantic scope, so it obeys a prin-ciple of order iconicity.12 Then a restrictive modifier of the head would be included in thescope, and the determiner would have a peripheral position with respect to the complexnominal. A non-restrictive modifier would not be included in the scope of the deter-miner, so this could separate the modifier from its head.

The sequential order in Cabecar is as follows: In contrast to simpler nominals (§2.5),the nominal constituted by a relative clause cannot combine with a determiner preced-ing it.13 E53 may serve to test this claim.

E53 jeLra jeL siwaL sh-eAL=geL roA koL a.then [D.MED knowledge say-IPFV=IPFV2] COP different‘then the wisdom [song] to be recited is different.’ (chicha_7.4)

11 Stockwell & Schachter & Partee 1973: 123 and 433-440 is among the earliest sources on this.12 On principles of diagrammatic iconicity s. GivoL n 1985, §2.3 and Lehmann 2007, §6.13 De Vries (2002: 136) even has a general principle to exclude this for circumnominal relativeclauses: “As for the external determiner: (i) If D is visible [i.e. if there is such an external deter-miner], it follows the relative clause."


The relative clause is restrictive, so the determiner jé at its beginning has semantic scopeover NPmat. Structurally, however, it is a coconstituent of the DN. This is demonstrated asfollows: If the determiner combined syntagmatically with the relative clause as a whole,it should be insensitive to the category of the first word of this clause. For instance,insert ka� ráá ‘always’ before siwa� in E53. If one does this in the independent declarativeclause corresponding to this relative clause, that is, in E54, the result is fine, since such aclause may be introduced by a temporal adverb.

E54 kaL raL aL siwaL sh-eAL=geLalways knowledge say-IPFV=IPFV2‘one always recites a song’

Using E54 as a relative clause following jé in E53 is ungrammatical precisely because thedeterminer does not combine with a clause (or with an adverb, for that matter). Only anominal expression can be the first constituent of a relative clause preceded by a deter-miner. This is because that determiner relates to that nominal expression, not to therelative clause. An analogous argument applies to E58 below.14

The syntax here is counter-iconic: In the structure, the determiner combines with theDN. Semantically, however, it cannot be determining it because this would force a non-restrictive reading on the construction, while relative clauses like E53 (and E58) areclearly restrictive. Instead, the semantic determination relates to NPmat. This topic will berevisited in §3.2.4.

E55 might appear to be counterevidence to the generalization that a Cabecar relativeclause cannot be the operand of a preceding determiner.

E55 YíLs te jeL i sh-eAL r-aL=mi s-aL .I ERG D.MED[3 say-IPFV emerge-PROG=AND] feel-PFV

‘I heard that, what they have been saying.’ (FOM)

Here, however, there is an obligatory pause after the demonstrative. It is not a deter-miner, but a pronoun cataphorically anticipating the nominal expression to follow.

The last grammatical property of NPmat is its syntactic function. In all of the examplesE29 – E55, this is the function of the absolutive actant of the main verb. In those exam-ples where the relative clause introduces the sentence, the sentence construction almostlooks like asyndetic parataxis. However, it is still unambiguously a subordinative con-struction because NPmat occupies the absolutive slot of the main verb.

The syntactic function born by NPmat is, however, not restricted. In E56 (as in E69below), NPmat is the ergative actant; in E57, it is an instrumental adjunct; in E58, it is apossessive attribute. In E59, NPmat functions as a temporal adjunct; the headless relativeclause of E68 below is a local adjunct.

E56 i jeLjk k-aL=kaL te i duaAL -waL ijaj i sh-aL neL=kaAL i.[3 RFL lift-PFV=ASC] ERG 3 cousin-PL DAT 3 say-PFV D.AUD=MNR

‘having gotten up, he spoke in that way to his cousins.’ (yebule_61)

14 In E43, too, the optional determiner goes with the proper name; it does not relate to the ‘yearthat Mary was born in’.


E57 YíLs nua te yalaL ti-aL j-aL =mi1SG maternal.uncle ERG oak cut-PFV go-PFV=AND

yíLs te baLk peL it-aL (jeL(=neL )) wa.[1SG ERG axe lend-PFV] D.MED=EMPH INSTR

‘My uncle went to fell the oak with the axe I lent him.’ (FOM)

E58 jeL datsíL ji-oAL =geL pa roA batseAL[D.MED fabric put.in-IPFV=IPFV2] body/color COP red‘the color of the clothes he wore was red’ (yer_14.2, following E11)

E59 SaL weAL ikaL -r=keL tuíLna ka tulu kaL d-aAL =kaL kuL na[1.PL harm-D.MID(IPFV)=IPFV2 night] LAT moon NEG emerge-IPFV=ASC be.SBJ

‘During the nights that we were harmed, the moon did not come out’ (ser_12)

As with any other NP, all syntactic functions of NPmat except the absolutive are marked bya postposition. Together with the number marking demonstrated in §3.2.3, these factscorroborate the analysis that the circumnominal relative clause functions as an NP in thematrix, with the peculiarity that it does not combine with a determiner.

3.2.4 Determination of the DN and restrictivity

The following account is structured according to the kinds of determination of the DN;restrictivity of the construction is treated as the dependent variable. There are no con-straints on the internal constituency of the DN. In E60, it is the discontinuous nominaldíkla�� maña� -täwa�� ‘three rivers’.

E60 saL keL i wa díLklaAL tsoL manaL -taAwaAL[1PL space/time:SPEC PER river EXIST three-CL.ELONG]

ki-eAL raA TalíLríL DuchíL JakuíLcall-IPFV COP Telire ChirripoL Pacuare15

‘The three rivers which exist in our lands are called Telire, ChirripoL and Pacuare.’(duchi_1.4)

The scarcity of constraints on the structure of relative clauses includes the determina-tion of the internal DN. If this lacks a determiner of its own, as in E29 - E43 and someother examples, the relative clause may be restrictive. Several languages have beenshown not to allow definite determination of the DN of a restrictive relative clause, espe-cially if the DN is internal (cf. among others Gorbet 1976 for Diegueno, Williamson 1987for Lakhota and Boyle 2016 for Hidatsa). On the basis of such empirical evidence and fol-lowing earlier proposals, a principle according to which the DN of a restrictive relativeclause cannot bear definite or generic determination is postulated in Lehmann (1984, ch.V.2.2). Such a principle presupposes iconicity with regard to the position of the deter-miner (§3.2.3). Now, adnominal relative constructions in which a definite determiner is,in structural terms, a co-constituent of the DN have been known for some time (de Vries2002: 63). Cabecar now falsifies this principle for circumnominal relatives, as well, forwhich it has been maintained in Basilico 1996 and de Vries 2002: 38 et pass., among oth-ers.

15 One use of the copula is to articulate a sentence into topic and comment.


In E56, the DN is a personal pronoun representing an established referent in the uni-verse of discourse. In E61, the DN of the relative clause is both determined by thepossessive pronoun i and represents an established referent.

E61 I dawaL dul kal jula na kaL jeLk dalíL-n-eL .[3 brother.in.law standing tree hand/arm IN] NEG RFL move-D.MID-NEG.PFV

‘His brother-in-law, who was standing on the branch, did not move.’ (yer_29)

The DN of the relative clauses of E30 and E62 has generic reference.

E62 TsibaAL rkaA kaL kia-r jíL yeAL waAL[mountain.cherry NEG want-D.MID(IPFV) what TRL ] fruit

kat-eAL raA beALwaLk te.eat-IPFV COP animal ERG

‘The fruit of the mountain cherry tree, which is good for nothing, is eaten byanimals.’ (GonzaL lez & Obando 2020 s.v. tsiba��rkä)

The DN of the relative clause in E63 is a proper name.

E63 BusuL bulu jeL roA jíLroA waLk keALgoA loA jeLmi tkabeAL waLgeAL[Busubulu D.MED COP animal guardian.spirit and snake boss]‘Busubulu, who is the guardian spirit of the animals and the boss of the snakes,

kaL wa kaLwoAL meL -n-aL jíLroA le kt-oA -gloANEG DSP mandate give-D.MID-PFV [something kill-SNR-FIN]is not going to allow [them] to kill anything.’ (yer_36)

In all of these cases, the relative clause is non-restrictive. Cabecar thus clearly falsifies apossible generalization (Lehmann 1984: 278, de Vries 2002: 35) that circumnominal rel-ative clauses cannot be non-restrictive. However, as seen in §3.2.3, a DN preceded by adeterminer is compatible with a restrictive interpretation. Analogous evidence is pro-vided by the semantically definite pronominal heads of E52, E58, E69f and E76f and bythe quantifier in E64, which structurally accompanies the DN, but semantically relates toNPmat.

E64 Yaba te beALna du ju-aL =mi tk-aL=sa=julu nala eLktaL ka.[child ERG all bird throw-PFV=AND] cross-PFV=EGR=AM:PL path side LAT

‘All of the birds that the child released crossed to the other side of the path.’(FOM)

In such cases, the scope of the determiner is reassigned as relating (counter-iconically)to NPmat instead of the DN. The difference between restrictive and non-restrictive relativeclauses is not structurally marked, not even by intonation (as it is in the postnominal rel -ative construction of modern European languages). In particular, a relative clause whoseDN is a personal pronoun may (E52) or may not (E56) be restrictive. The restrictive ornon-restrictive modification of Y by S1 in Figure 2 is produced only at the level of sen-tence semantics and discourse.

Given these facts, the semantic operations involved in the interpretation of a subordi-native construction as a relative construction may be systematized as follows:(1) On the basis of the structure common to complement clauses and relative clauses,

such an embedded clause is interpreted as a deverbal nominal which may or may notbe oriented.


(2) The embedded clause is interpreted as being oriented towards one of its nominalcomponents, which thus becomes its DN, if the syntactic and semantic properties ofthe context (§3.2.2) force this operation.

(3) If the DN is by itself definite or generic, the resulting relative clause is non-restric-tive. If it is accompanied by a definite or generic determiner, an interpretation of thisconfiguration as iconic yields a non-restrictive relative clause, too.

(4) In the latter case, the scope of the determiner may alternately be reassigned as relat-ing (counter-iconically) to NPmat instead of the DN. The DN itself then comes out asdevoid of a semantic determination just as if there was no determiner in the firstplace. In this case, the relative clause is interpreted as restrictive.

By this account, restrictive interpretation of a relative clause is subject to more condi-tions than non-restrictive interpretation. There is a commonly accepted implicationaluniversal (Lehmann 1984: 279, de Vries 2002: 35) that if a language has non-restrictiverelative clauses, it has restrictive relative clauses. In view of the above, although Cabecardoes not actually falsify this universal, it does throw doubt on it.

3.3 Free and adverbial relative clauses

Cabecar has free relative clauses as introduced in §3.1. However, instead of a binary divi-sion between headed and headless relative clauses, a series of constructions containingpronominal DNs of increasing explicitness is found. We pass through this series startingwith a zero DN. Since there is no mechanism to mark the DN of a relative clause, a zeroDN would only be recoverable in two situations: either it is a clause component whichleaves an identifiable gap as it can only be omitted in certain constructions, including arelative clause; or it is recoverable on a semantic basis. We will first look at the firstalternative. The second alternative applies to adverbial relative clauses, discussed fur-ther below.

A Cabecar NP depends either on a postposition or on a verb. If omitted, it would leavean identifiable gap in either case. However, the complement of a postposition is obliga-tory under all circumstances; there is no postposition stranding. There is, therefore, noquestion of omitting only the NP of a verb dependent marked by a postposition. Amongactant functions, the postpositional phrases functioning as dispositive and indirectobject are optional, so their omission would not leave an identifiable gap. The function ofthe covert head in a headless relative construction would have to be absolutive or erga-tive. However, the absolutive cannot remain covert in principle (E23 and E69). Theergative, too, is inomissible in relative clauses. As an example, consider E65f.

E65 NakeL yakaL kat-eAL kuka-leAL .mantled.howler flesh eat-IPFV roast-S.PRF

‘Howler monkey meat is eaten roasted.’ (Lit.: ‘<they> eat howler monkey meatroasted’) (GonzaL lez & Obando 2020 s.v. nake�)

In an independent active clause in imperfective aspect like E65, the ergative actant isomissible if unidentified. The attempt, however, to convert E65 into a relative clausewhose head in ergative function is zero, as in E66, fails.

E66 YíLs wa nakeL yakaL kat-eAL kuka-leAL suL -leAL .I DSP [mantled.howler flesh eat-IPFV roast-S.PRF] see-S.PRF

‘I have seen that howler monkey meat is eaten roasted.’ (FOM)


E66 only has a complement-clause interpretation as indicated; it cannot mean ‘I haveseen some who eat howler-monkey meat roasted.’16 There are two grammatical solutionsto this expression problem. A partial solution consists in resorting to an agent-noun con-struction (§2.7) instead of a relative clause, as in E67.

E67 YíLs wa nakeL yakaL kat-aA -waL suL -leAL .1SG DSP [mantled.howler flesh eat-SNR-PL] see-S.PRF

‘I have seen eaters of howler monkey meat.’ (FOM)

A close comparison of E67 with E66 reveals that the consultant omitted the secondarypredicate kukale� �, evidently in consonance with the function of the agent noun to catego-rize its referent. The other solution preserves the relative clause, but not the zero head,and is illustrated by E72 below.

The only remaining candidates for a relative clause with zero DN are adverbialclauses like the local clause of E68.

E68 KaL i deAL -neL yíLs kaldu ska.NEG 3 emerge-PFV [1SG standing] LOC

‘He has not arrived at the place where I am (standing).’ (FOM)

The independent declarative clause corresponding paradigmatically to the relativeclause of E68 would comprise a noun of the general meaning ‘place’ and thus have astructure like yís kaldu ka� ska (I standing space/time LOC) ‘I am standing at a place’. Thisnoun would serve as the DN in E68.

Adverbial relative clauses – that is, local, temporal and manner clauses – differ gener-ally from actantial relative clauses in their reduced freedom of choice for the syntacticfunctions of the DN and NPmat: In principle and as shown in §3.2.2f, the two choices aremutually independent in all relative constructions. However, adverbial relative clausesdisplay a strong tendency in favor of parallel functions because of the meaning of their(zero) DN: a place functions typically as a local adverbial, a period or point in time func-tions typically as a temporal adverbial, and likewise for a manner. As a consequence, inadverbial relative clauses, the default is for the (zero or pronominal) DN to have thesame function as the entire clause; and for headless adverbial relative clauses this is nor-mally the only possibility. Consequently, the zero head in E68 poses no interpretationproblem: it must be a local adjunct.

Proceeding on the gamut of increasing explicitness of the DN, the DN in E69f is mini-mally overt.

E69 SaA yuaA te i shaL te saL janeL -waL .[1PL instructor ERG 3 say:PFV] ERG 1PL laughter-CAUS:PFV

‘What our professor said made us laugh.’ (GonzaL lez & Obando 2020 s.v. jañe�wa)

E70 IjeL te i shaL jeL raA chaAL kleAL .[3PS ERG 3 say:PFV] D.MED COP right‘What he said is right.’ (GonzaL lez & Obando 2020 s.v. cha��kle� �)

E69f differ in the presence of the optional resumptive demonstrative in E70. In both, theDN is minimally present in the form of the third person pronoun, taking the position of

16 Generalizing over the preceding set of nonexistent operations, it appears that Cabecar lacks anoperation of gap formation in a clause. This would cover cleft-sentences, which Cabecar lacks, aswell.


the absolutive actant. This may count as a pronominal DN, but a minimal one, as it is anobligatory filler of this position. Observe, incidentally, that these constructions are syn-tactically ambiguous between a relative-clause and a complement-clause reading (cf.§3.2.2). With a complement-clause reading, E69 would translate as ‘(the fact) that ourprofessor said it made us laugh’. As argued in the case of E32 above, orientation heremay depend on the context.

A slightly heavier DN may be provided for E69 as in E71.

E71 SaA yuaA te jileL shaL te saL janeL -waL .[1PL instructor ERG something say:PFV] ERG 1PL laughter-CAUS:PFV

‘Something our professor said made us laugh.’ (FOM)

E71 differs formally from E69 only in the substitution of the personal pronoun by thecorresponding indefinite pronoun. The semantic difference between them is minimal: iin E69 is an expletive element whose semantic definiteness is ignored (§3.2.4), and jilé inE71 is its indefinite counterpart, meaning ‘something, anything’, as in E23. On balance,E69 and E71 are close to synonymous. Cabecar does not appear to possess the indiffer-ent (a.k.a. free choice) relative construction as a specific construction type.

Likewise, the above problem of converting E65 into a relative clause requires anindefinite pronominal DN as in E72.

E72 YíLs wa maneL le te nakeL yakaL kat-eAL kuka-leAL suL -leAL .I DSP [somebody ERG mantled.howler flesh eat-IPFV roast-S.PRF] see-S.PRF

‘I have seen some who eat howler monkey meat roasted.’ (FOM)

The following relative clauses have the shape of a content interrogative clause (§2.6).The interrogative pronoun represents the ergative actant in E73, and the absolutiveactant in E74.

E73 SaL teA yíLroA teA kaL lbatioA gloA ts-aL =u=mi busíL ia yul-eAL -geL .1PL ERG [who ERG hat transport-PFV=AM=AND girl DAT] search-IPFV-IPFV2‘We search the one who took away the hat from the girl.’ (FZH)

E74 JíLroA blaL=wa dulaL gleAL teA busíL yíLga (jeL ) kuj -aL saL teA .[what steal:PFV=TOT boy ERG girl AVERS] D.MED find-PFV 1PL ERG

‘We found what the boy took away from the girl.’ (FZH)

In E75, the DN is represented by a local interrogative adverb; and so the relative clausealso has the function of a local adverbial in the main clause.

E75 KeLgara i d-oAL =geL=ju maL i saL yeALbl-oAL =geL jeLska.always 3 emerge-IPFV=IPFV2=AM [where 1PL hunt-IPFV=IPFV2] there‘He always shows up at the place where we hunt.’ (yer_20.1)

Here, too, the relative clause is resumed by a demonstrative, in this case a local demon-strative adverb. Temporal and manner clauses, too, are formed on the model of adverbialrelative clauses with an interrogative DN.

Finally, the DN may be represented by a demonstrative pronoun, as in E76f.

E76 JeL wa tsiruL deAL jeL raA yíLs jayíL.[D.MED DSP cocoa emerge:PFV] D.MED COP 1SG husband‘He who brought cocoa is my husband.’ (FOM)


E77 neLj jeLj ij bal-aL=waj ijaj i te i sh-aL :[D.AUD RFL hide-PFV=TOT] DAT 3 ERG 3 say-PFV

‘to that one who had hidden himself, they said (it):’ (historia_15)

Observe, again, that despite the definiteness of the DN, the relative clause of E76f isrestrictive (cf. §3.2.4).

4 From independent clause to relative clauseThe structure of a relative clause is not only the same as the structure of a complementclause, but also the same as the structure of an independent declarative clause. This sec-tion will sketch a diachronic progression by which a relative clause may evolve from anindependent declarative clause. This change relates two kinds of construction involvingtwo clauses:(1) asyndetic parataxis of two clauses connected by anaphora(2) matrix clause containing circumnominal relative clause.

The transition from (1) to (2) involves a reanalysis of the first clause of the dyad as a rel-ative clause. As usual, the reanalysis is rendered possible by cases which are structurallyambiguous between (1) and (2). The following three subsections are dedicated first toconstructions which are unambiguously of type (1), followed by ambiguous construc-tions and leading to constructions of type (2). The analytical problems offered bycategory (2) will be highlighted.

4.1 New-referent dyad

The starting point of the development hypothesized is formed by a particular asyndeticparataxis of two clauses which may be called a new-referent dyad and be illustrated byE78.

E78 KeAL matsu raA beAL keALklaA , jeL wa saA weAL ikaL -leAL .KeAL -matsu COP demon primordial D.MEDDSP 1PL harm-S.PRF

‘KeAL -matsu is a primordial demon, he had mistreated us.’ (GonzaL lez & Obando2020 s.v. matsu)

The first of these asyndetically combined clauses contains a referential expression, viz.Ke��-matsu, which introduces a referent into the universe of discourse and gives informa-tion on it. The second clause takes this up anaphorically with another referentialexpression, here jé, and adds some further information to it. This configuration is repre-sented in Figure 5.

Figure 5 New-referent dyad

[ … Xi … ]S1 [ Yi … ]S2

introductory clause resumptive clause

Here X stands for the expression representing referent i, while Y is the expression takingi up anaphorically in S2. Given the canonical configuration for the introduction andimmediate resumption of a referent, by default X is a lexical-nominal expression, while Yis the dedicated anaphoric device for second mentions. In Cabecar, the latter is the


demonstrative of medial deixis jé. Normally, though not obligatorily, Y takes initial posi-tion in S2. E79 is another example instantiating this configuration.

E79 JukaLbata jeLmi yeA leAL keAL tkoA yu-aL i tearrow and long.arrow tip form-PFV 3 ERG

‘The heads of the arrows and long arrows he formed’

dikaL taL goA loA teA leAL wa , jeL tsoL tkeAL -leAL jieL -waL wa.peach.palm piece ripe INSTR D.MEDEXIST plant-S.PRF 3.PS-PL DSP

‘with pieces of ripe peach-palm; these they had planted.’ (yer_03)

Here the introductory clause introduces ‘peach palm’ as a new referent, which again istaken up by jé in the subsequent clause.

Neither of the dyads in E78 and E79 lends itself to a (re-)analysis where S1 is a rela-tive clause: in both of them, S1 gives new information, naturally coded by an independentdeclarative clause, while a restrictive relative clause would instead contribute to theidentification of a referent (§3.1). S2 could only become a postnominal relative clause, onbe�� ke� �klä and diká, respectively. It could not become a circumnominal relative clause in S1,which would then have to be its matrix, because S2 would lack a syntactic function in S1.

Since S1 in Figure 5 is not a constituent of S2, it is resumed in S2 by jé. And since no DNis selected in S1, the antecedent of this demonstrative is not identified by the hierarchy ofsyntactic functions (Figure 4), as it is in relative constructions, but is instead identifiedby a principle of reference tracking which favors textual proximity. This may be seen inE79, where the antecedent of the demonstrative is not selected on the basis of its syntac-tic function; instead, it is the noun phrase to the demonstrative’s left that provides thebest fit with the semantics of S2. Thus, again, the syntagmatic relation of S1 to S2 in theseexamples is unlike the syntagmatic relation born by a relative clause to its matrix.

These examples are only meant to show that not every new-referent dyad lends itselfto a reanalysis of a complex sentence whose introductory clause is a relative clause.

4.2 Ambiguous constructions

Certain clause pairs of the general structure of Figure 5 do allow for a reanalysis suchthat the first clause is a circumnominal relative clause. E80 is an isolated example sen-tence from a dictionary entry.

E80 No ala kíLka kal jaL -n-aL =kaL jeL te i waAL ti-aL=wa.path SUPER tree fall-D.MID-PFV=ASC D.MEDERG 3 obstruct-PFV=TOT

‘Across the path fell a tree, this blocked it completely.’ or: ‘A/the tree that fellacross the path blocked it completely.’ (GonzaL lez & Obando 2020 s.v. tiä)

One may construct a preceding context for E80 so that S1 is an all-new proposition andforms a breath group and intonation unit of its own, with a possible pause after its verb.Then E80 is a paratactic construction as before. In a different context, S1 may be prosodi-cally integrated into S2 as described in §3.2.1 and, instead of asserting something new,may only identify the referent that acts as the ergative in S2 (s. §3.1).17 Then S1 is an inter-nal-head relative clause. It is left-dislocated and resumed by the demonstrative jé in themain clause; so it is a preposed relative clause. The two analyses are paired with the two

17 The online version of this article has a link here to a data set including two passages from a textrecording which have the syntactic structure and the intonation described above.


translations provided for E80. Likewise, the initial clause of E81 introduces a specific ref-erent which gets individuated by what the clause says about it, fulfilling thus the text-semantic condition for its reanalysis as a relative clause.

E81 No aL kaAL i maneL -le-waL tsoL yakeL i-laL , jeL -waL te saL waALyu-eAL=keL .also which-INDF-PL EXIST bad-PL D.MED-PL ERG 1PL cheat-IPFV=IPFV2‘Moreover, some bad people exist, those cheat us.’ or: ‘some existent bad peoplecheat us.’ (duchi_2.6)

Left-dislocation with resumption in the main clause by jé is very frequent in the texts.Typical examples were given in §2.3.18 E63 shows that it even happens inside a relativeclause. Likewise, in the second line of E82, the clause from which the topic (viz., ‘hisbreast, arms and face’) is left-dislocated is a complement clause.

E82 SaL keALklaA -waL te i su-aL jeL te i sh-eAL[1PL ancestor-PL ERG 3 see-PFV] D.MED ERG 3 say-IPFV

i jeL r bata jula waAL ktaAL jeL raA kaL cheL i ta.3 breast arm/hand face D.MED COP space/time worthless PPV

‘Our ancestors saw him and say that his breast, arms and face had thicket onthem.’ or ‘Our ancestors(,) who saw him say that …’ (ser_08)

The first line of E82 exhibits the same structural ambiguity as E80. If the introductoryclause is analyzed as a circumnominal relative clause, this might even be non-restrictive.

Certain factors favor the (re-)analysis of a sequence such as E80 – E82 as involving arelative construction. The first is well-known from the integration of erstwhile left-dislo-cated topics into the following clause structure by suppressing the prosodic break andthe resumptive pronoun. All of this may apply to new-referent dyads of the structure ofFigure 5, as well. Many example sentences are provided by the consultants in alternateversions which differ by the presence/absence of these two features. E83 is an example.

E83 Yawa te du ju-aL =mi (jeL ) tk-aL=sa=ju díLklaAL eLktaL ka.[child ERG bird throw-PFV=AND] D.MEDcross-PFV=EGR=AM river side LAT

‘The bird that the boy threw (i.e., released) crossed to the other side of the river.’(FOM)

Similar examples are E29, E40, E41, E43 and E74.A different factor favoring the reanalysis is referential semantics at the text level. E84

is another example similar to the preceding ones.

E84 seAL rikeAL seAL -r saAL lwíL=síL jeL raA i bata jeLkjeAL[storm.boy feel-D.MID(IPFV) wild=SPRR] D.MED COP 3 tip definitely‘The storm boy who is the wildest is definitely the youngest.’ (ser_06)

Here the resumptive demonstrative is the subject of the ascriptive main clause. Thus, thecriterion for the diagnosis of a subordinate clause which worked for E33, viz., its occupy-ing an obligatory actant position, does not apply: the relative clause could be anindependent clause, and the paratactic sequence could be translated as ‘a storm boy isthe wildest; that is the youngest sibling’. In this particular case, both the intonation con-tour and the immediately preceding context resolve the issue. In the prosodic structure,

18 An analogous structural ambiguity as described here for relative constructions exists withcomplement constructions: the initial clause in E10 may be independent or may be a left-dislo-cated subject complement clause.


there is continuous intonation and no pause between the clauses. The preceding sen-tence says: ‘The storm boys live on the other side of the sea.’ Here, for the interpretationof S1 as an independent sentence to make sense, one would expect the numeral ‘one’ tobe used to single out one element from the set just introduced. Consequently, S1 in E84 ismost plausibly analyzed as a left-dislocated relative clause.

Figure 6 formalizes the reanalysis envisaged.

Figure 6 Reanalysis of paratactic construction as relative construction

Input [ … Xi … ]S1 [ jéi … ]S2

introductory clause resumptive clause

Output [ [ [ … Xi … ]S1 ]Nom.i (jéi) … ]S2

relative clause main clause

As usual, the reanalysis itself is imperceptible. However, it is presupposed by the possi-bility of a construction lacking the resumptive demonstrative, as will be seen in the nextsection.

4.3 Relative construction

As a matter of fact, a paratactic construction as sketched in Figure 5 is the origin of a rel-ative construction in many languages (Lehmann 1984, ch. VI.1.1, Lehmann 2008, §4.1).The preceding section described the crucial context in which the reanalysis from parat-actic construction to relative construction takes place.

Further development leads to the use of an unmarked clause as a circumnominal rel-ative clause in constructions which no longer allow the original analysis as a paratacticconstruction. The decisive step here is to suppress the prosodic break after the introduc-tory clause and with it the resumptive pronoun in the following clause. At this moment,the introductory clause becomes a constituent of the main clause.19 It may then beembedded in a position of the main clause which requires a constituent in a specific syn-tactic function. For example, it suffices to delete the anaphoric demonstrative from E81to obtain a relative construction like E33. E85, too, has the subordinate clause in thefunction of the absolutive actant of the main verb. Putting a sentence boundary after itwould render the following sentence ungrammatical, as it then lacks an absolutiveactant.

E85 Yaba te beALna du ju-aL =mi tk-aL=sa=julu nala eLktaL ka.[child ERG all bird throw-PFV=AND] cross-PFV=EGR=AM:PL path side LAT

‘All of the birds that the child released crossed to the other side of the path.’ (=E64)

The same goes for several of the above examples, including E56f for relative clausesserving as the complement of a postposition. In E57, both the resumptive demonstrativeand the additional emphatic identifier are optional. Whether or not they are added, there

19 De Vries (2002: 38) treats the lack of a resumptive pronoun after a circumnominal relativeclause as an empirical generalization. Later on (p. 146), he seems to recognize that it is actually adiagnostic feature that distinguishes circumnominal from preposed clauses.


is – as long as the sentence syntax is intact – no possibility of interpreting the relativeclause as an independent clause.

Another decisive difference between the function of a clause as a declarative sen-tence and as a relative construction and, thus, as a component of a referential expression,lies in the semantic scope of determination and quantification. As long as the clause is asentence, determiners and quantifiers cooccurring with the nominal that would be theDN can only have scope over it. As soon as the relative clause is a nominal expression,determiners and quantifiers may have semantic scope over this expression (step 4 in§3.2.4). The semantic difference becomes palpable in E85: the sentence does not, ofcourse, imply that the child released all of the birds.

As a last step, the resumptive demonstrative may even follow what is clearly a rela-tive clause. In E86, S1 must be a relative clause because it is embedded in absolutivefunction.

E86 BaL te jíLjíL tk-aL=ju yikíL jeL s-aL ?2SG ERG [earthquake cross-PFV=AM yesterday] D.MEDfeel-PFV

‘Did you feel the earthquake that happened yesterday ?’ (= E34)

In this particular case, no left-dislocation is involved. In E87, the resumptive demonstra-tive even ends the entire sentence.

E87 KeLgara i d-oAL =geL=ju maL i saL yeALbl-oAL =geL jeLska.always 3 emerge-IPFV=IPFV2=AM [where 1PL hunt-IPFV=IPFV2] there‘He always shows up at the place where we hunt.’ (= E75)

In such cases, the sole function of the demonstrative is to mark the final boundary of therelative clause.

5 ConclusionCircumnominal relative clauses are much rarer worldwide than adnominal ones (Dryer2013). They do, however, occur in all continents except Europe and constitute the domi-nant strategy in several languages of America. The construction is known not only fromChibchan languages, with Rama among them, but also from the Yuman languagesMohave, Diegueno and Yavapai, from Navaho, Seri (Sonora, Mexico) and Gaviao (Rondop -nia, Brazil) (Lehmann 1984, ch. III.1.3 and Lehmann 2014).20 The Cabecar relative clauseshares with most of these the lack of any mark of subordination or nominalization.

Moreover, it lacks any structural property that would signal the orientation of theclause and would thus contribute to the identification of its DN. Instead, selection of theDN is governed by a hierarchy of semantic and grammatical constraints. The orientationof the relative clause and the selection of its DN work in the same way whether the nomi-nal in question is overt or zero and whether it is determined or undetermined.Determination of the DN is free, but bears no relation to a restrictive or non-restrictiveinterpretation of the relative clause.

All in all, interpretation of a Cabecar nominalized clause as a relative clause is more amatter of semantics than of grammatical constraints.

20 While circumnominal relative clauses are attested in languages with any basic word order, lan-guages with a dominant circumnominal relative clause tend to be OV languages. See Hawkins2004 on the processing of such structures.


References

Textual sourceschicha: Maricela FernaLndez FernaLndez, ¿Mane�pí sá te dí yue� � rö? – ¿Cómo hacemos la chicha? Oral

text recorded in Shiroles, LimoL n, March 2014.duchi: Rogelio Barquero Reyes, ¿Jíshta� � séjé duchíwák ka� se�ne�wa� ? – ¿Cómo vivimos nosotros los

chirriposeños? Text composed in 2009.historia: Obando MartíLnez, Freddy & GonzaL lez Campos, Guillermo 2015, Historia del clan Kät-

súibawák. Turrialba: Sede del AtlaLntico, Universidad de Costa Rica. Oral text recorded inTurrialba, Cartago, 2014.

ser: GonzaL lez Campos, Guillermo & Obando MartíLnez, Freddy, Se� �rike� � páke� � – Historia de los NiñosHuracanes. Text composed in 2012.

yebule: Freddy Obando MartíLnez, La historia de Ye��bule� �. Oral text recorded in Turrialba, Cartago,February 2016.

yer: FernaLndez, Severiano, “KonoL yeAL ria – El cazador de tepezcuintles”. In: FernaLndez, Severiano &Varas, Valeria 1989, Historias cabécares 1. San JoseL , C.R.: Ed. Universidad de Costa Rica; 9-18.

Secondary literatureAndrews, Avery 2007, "Relative clauses." Shopen, Timothy (ed.), Language typology and linguis-

tic description. Cambridge etc.: Cambridge University Press (2nd rev. ed.); vol. 2: 206-236.Basilico, David 1996, “Head position and internally headed relative clauses”. Language 72: 498–

532.Boyle, John P. 2016, “The syntax and semantics of internally headed relative clauses in Hidatsa”.

Rudin, Catherine & Gordon, Bryan J. (eds), Advances in the study of Siouan languages and lin-guistics. Berlin: Language Science Press; 255–287.

Citko, Barbara 2004, “On headed, headless, and light-headed relatives”. Natural Language &Linguistic Theory 22: 95-126.

Coto-Solano, Rolando & Molina-Munoz, Adriana & GarcíLa Segura, AlíL 2016, "Correlative struc-tures in Bribri". Sadlier-Brown, Emily & Guntly, Erin & Weber, Natalie (eds.), Proceedings ofthe Twentieth Workshop on Structure and Constituency in the Languages of the Americas .(University of British Columbia Working Papers in Linguistics, 43); 27-41.

Dryer, Matthew S. 2013, “Order of relative clause and noun”. Dryer, Matthew S. & Haspelmath,Martin (eds), The world atlas of language structures online. Leipzig: Max Planck Institute forEvolutionary Anthropology (http://wals.info/chapter/90).

GivoL n, Talmy 1976, “Topic, pronoun and grammatical agreement.“ Li, Charles N. (ed.), Subject andtopic. New York etc.: Academic Press; 149-188.

GivoL n, Talmy 1985, “Iconicity, isomorphism and non-arbitrary coding in syntax.” Haiman, John(ed.), Iconicity in syntax. Proceedings of a symposium on iconicity in syntax, Stanford, June 24-6, 1983. Amsterdam & Philadelphia: J. Benjamins (Typological Studies in Language, 6); 187-219.

GonzaL lez Campos, Guillermo 2016, “DeterminacioL n de las frecuencias foneLmicas del idiomacabeLcar mediante un corpus electroL nico” [unpublished manuscript]. Turrialba: Universidadde Costa Rica.

GonzaL lez Campos, Guillermo & Lehmann, Christian 2020, The Cabecar language. To app.GonzaL lez Campos, Guillermo & Obando MartíLnez, Freddy 2020, Ditsa� � duchíwák kte� � chuli�i� i yuäklä

– Diccionario escolar del cabécar de Chirripó. To app.: Turrialba: Universidad de Costa Rica.Gorbet, Larry 1976, A grammar of Diegueño nominals. New York: Garland.Hawkins, John A. 2004, Efficiency and complexity in grammars. Oxford etc.: Oxford University

Press.Jara Murillo, Carla Victoria 2018, Gramática de la lengua bribri. San JoseL : E-Digital.

http://wals.info/chapter/90


Keenan, Edward L. 1985, "Relative clauses". Shopen, Timothy (ed.), Language typology and syn-tactic description. Cambridge etc.: Cambridge UP; Vol. II: 141-170.

Lehmann, Christian 1984, Der Relativsatz. Typologie seiner Strukturen - Theorie seiner Funktio-nen - Kompendium seiner Grammatik. TuA bingen: G. Narr (LUS, 2).

Lehmann, Christian 2003, “Relative clauses”. Frawley, William J. (ed.), International encyclopediaof linguistics. Oxford: Oxford University Press; 3:460-462.

Lehmann, Christian 2007, “Motivation in language. Attempt at a systematization”. Gallmann,Peter & Lehmann, Christian & LuA hr, Rosemarie (eds.), Sprachliche Motivation. Zur Interde-pendenz von Inhalt und Ausdruck. TuA bingen: G. Narr (TuA binger BeitraA ge zur Linguistik, 502);100-135.

Lehmann, Christian 2008, “Information structure and grammaticalization”. Seoane, Elena &Lopez Couso, Maria JoseL (eds.), Theoretical and empirical issues in grammaticalization. Ams-terdam & Philadelphia: J. Benjamins (Typological Studies in Language, 77); 207-229.

Lehmann, Christian 2014, [Review of: Relative clauses in languages of the Americas: A typologicaloverview. Ed. by Bernard Comrie & Zarina Estrada-FernaLndez. Amsterdam: J. Benjamins,2012] Language 90: 534-536.

Margery Pena, Enrique 1989, Diccionario cabécar-español, español-cabécar. San JoseL , C.R.: Ed.Universidad de Costa Rica.

MuA ller, Henrik Høeg & Klinge, Alex (eds.) 2008, Essays on nominal determination. From morphol-ogy to discourse management. Amsterdam & Philadelphia: J. Benjamins (Studies in LanguageCompanion Series, 99).

Stockwell, Robert P. & Schachter, Paul & Partee, Barbara 1973, The major syntactic structures ofEnglish. New York etc.: Holt, Rinehart & Winston.

Villalobos, MaríLa Eugenia 1994, “Bribri, lengua con claLusulas relativas de nuL cleo interno”. Letras29-30: 225-239.

Williamson, Janis Shirley 1987, “An indefiniteness restriction for relative clauses in Lakhota”.Reuland, Eric & ter Meulen, Alice (eds), The representation of (in)definiteness. Cambridge:MIT Press; 168–190.

Wilson, Jack L. 1984, “Relative clauses in Bribri”. Estudios de Lingüística Chibcha 3: 179–199.

Documents

CLIPP Christiani Lehmanni inedita, publicanda, publicata€¦ · 1 We thank the Cabecar speakers named in §2.1 for their cooperation, the members of Martin Haspel-math’s PhD student