Specifying a term on a semantically empty base

Concept formation and entity identification need not center around a lexical head notion; something may be conceptualized or identified as whatever plays a certain role in a certain situation. Consequently, the term that a relative clause helps to form may or may not have a lexical core. If it has one, this is the head nominal of the relative clause, which modifies it. If it has none, the relative clause modifies an empty head.

  1. A relative clause which combines with a lexical head bears a specifying relation to it. Depending on the positional type, it may be an adjectival bearing an attributive relation to the head, or its specifying function may emerge from the construction as a whole. S. the section on attribution.
  2. A free relative clause is one whose semantic core is lexically empty. It forms a complex concept by itself.

The operations of (re-)categorization that may be necessary to form a free relative clause depend on the primary categorization of oriented open clauses in the language. Their primary category may be

  1. adjectival: The clause is basically a relative clause. Its use in substantival function requires substantivization.
  2. substantival: The clause is basically an oriented substantival, so no special free relative clause construction is needed. What may require a special operation is the use of such a clause as an attribute to a head.
  3. nominal s.l.: Such a clause serves equally as an adjectival (and consequently as an attributive clause), like #1, and as a substantival, like #2. No recategorization in either direction is necessary. below is likely an example.

These distinctions and recategorization operations can only be verified if a nominalized clause takes different shapes in substantival and adjectival use. Substantivization of an adjectival is often a covert process: in many languages, an adjectival can freely be used as a substantival. Instead of positing a conversion operation here, such cases might be more appropriately analyzed as of the category ‘nominal s.l.’

There are two varieties of the free relative clause:

  1. If the nucleus of a restrictive relative construction is pronominal – no matter whether inside or outside the relative clause –, it is a light-headed relative clause.1
  2. If the relative clause is just an oriented nominal with nothing occupying the relativized position, it is a headless relative clause.

Light-headed relative clause

The head of a light-headed relative clause is a pronoun, typically one of interrogative origin. A frequent manifestation of a pronominal free relative clause is the correlative diptych, as in .

.Etuceyya-t-takk-a-t-ōatuceyyavēṇṭum
Tamil[ whatdo:INF-∅-fit-PTCP ]-∅-PTLthatdo:INFone.must
What is fitting to do, that is to be done.

The following examples are just pronominal free relative clauses, without being correlative. If a relative clause with a relative pronoun lacks a lexical head, the relative pronoun serves as a pronominal head for the complex concept to center around. is an example. The pronominal features – here masculine singular – in which the relative pronoun otherwise agrees with its head now constitute the head and thus become the semantic core of the higher nominal. The coordination in proves that such a light-headed relative clause is syntactically equivalent to an NP.

.Nuncegohacepistulatrisdeludam:
LatinnowI:NOMD1:ABL.SG.Fletter:ABL.SG.Fthree:ACC.PL.Mdeceive:FUT:1.SG
Now I'm going to deceive three people with this letter:
 erumetlenonemetquihancdeditmilitteram.
 master:ACC.SG.Mandpimp:ACC.SG.Mand[ REL:NOM.SG.MD1:ACC.SG.FPRF:give:IND:3.SGI:DATletter:ACC.SG.F ]
 master, the pimp, and the one that gave this letter to me.(Pl. Ps. 690f)

This example has the particular that the syntactic function of the higher nominal in the matrix is different from the function of the relativized position, but is nevertheless not coded.

Again, such a relative clause may be preceded by a determiner.

.eaquaeadeffeminandosanimospertinentimportant
LatinANA.ACC.PL.N[ REL:NOM.PL.Ntoeffeminate:GERUND:ACC.PL.Mmind:ACC.PLpertain:3.PL ]import:3.PL
they import those things which have an effeminating effect on the mind(Caes. BG 1, 1, 3)

In , the two functions in question are the same as in . The introductory pronoun is used to code the function of the higher nominal, thus avoiding eventual problems.

Even languages which do not use a relative pronoun in adnominal relative clauses may use one in free relative clauses. f have the human and the non-human interrogative pronouns in light-head function.

.yaanmáaxil-ik-ech
YucatecEXIST[ whosee-INCMPL-ABS.2.SG ]
(there is) someone (who) sees you(ACC_0491.1)
.hachutsba'xk=umeent-ik
Yucatecverygood[ whatIPFV=SBJ.3do-INCMPL(ABS.3) ]
what he does is good(ACC_0337)

Again, such a relative construction may be determined just like a headed one. f show a definite determiner in front of the relative clause.

.lemáaxyaantáan-ilti'=e'k=unúuk-ikut'aan
YucatecDEM[ whoEXISTfront-ADVRLOC]=TOPIPFV=SBJ.3answer-INCMPLPOSS.3speech
the one who is in front of him will answer him(AAK_016.3)
.t=inwo'lalt-ahleba'xúuchten
YucatecPFV=SBJ.1.SGcause:TRR-CMPLDEM[ whathappen(CMPL.ABS.3.)me ]
I assumed the responsibility of what happened to me(EMB_0403)

The syntagmatic compatibility of a definite determiner with a relative pronoun of interrogative – thus, indefinite – origin is unproblematic here, as they pertain to distinct nominal constituents: The interrogative pronoun represents the core, while the definite determination pertains to the higher nominal. This follows the rule that the core nominal of a restrictive relative clause cannot be semantically definite.

Headless relative clauses

If a relative clause which does not involve a relative pronoun lacks a lexical head, it is a headless relative clause. Omitting hamal from generates an example.

.Bavul-uver-diğ-imhamalnerede?
Turkish[ suitcase-ACCgive-NR-POSS.1 ]porterwhere
Where is the porter whom I gave the suitcase?

Turkish clauses nominalized by the suffix -diğ may be oriented or non-oriented. In the latter case, they are substantive clauses. Rather than assuming a specifically adjectival clause in , it may be more economic to assign the oriented nominalization to the supercategory ‘nominal s.l.’

illustrates how a ‘nominal’ headless relative clause optionally and ceteris paribus combines with a head noun and then forms a prenominal relative clause.2

.a. na=na[pishkatu pay-nin-tu=mika]=tapyan-ta-w
Awa Pit1sg=top [fish buy-caus-impf=nmlz.sg]=acc hit-pst-1
I hit the one who was selling the fish.
b. [na-wasulakwa-t=mika]kwizhanyawalkwa-tɨ-zi
[1sg-acc bite bite-perf.ptcp=nmlz.sg] dog meat steal-pst-2/3
The dog which bit me stole the meat.(Curnow 1997: 286 ap. Dryer 2013)

The situation is less clear-cut if the introductory formative of the headless relative clause differs from the one introducing a headed RC. Several languages including Italian cannot use the naked relative clause as an NP. This is shown in f: mere omission of the head from results in ungrammaticality.

. Lo scolaro [ che non ha mai imparato a lavorare ] l'impara all'università.
Italian The student who has never learnt to work learns it at the university.

One solution, viz. providing the RC with a determiner, is shown by below. The other solution is to use a different introductory formative. This is shown in .

.Chi non ha mai imparato a lavorare l'impara all'università.

The introductory chi here codes a human nucleus. This construction therefore marks the transition between the headless and the light-headed relative clause.

Light-headed vs. headless relative clause

The relative clause may be preceded by pronominal elements which, at first sight, might appear to be its head, as in .

.a. Colui [ che non ha mai imparato a lavorare ] l'impara all'università.
Italian The one who has never learnt to work learns it at the university.
b.Coloro che non hanno mai imparato a lavorare l'imparano all'università.
Those who have never learnt to work learn it at the university.

However, contrary to appearances, colui and its plural coloro are not the head noun. If they were, then since colui is definite, the relative clause would have to be non-restrictive; but it is restrictive. This kind of pronoun represents two grammatical components. The first is a demonstrative feature; the second is the category N, which in this particular case combines with number – a pronominal category. At the semantic level, the relative clause combines with the second component; the pronominal element just serves as a prop for the substantivization of the relative clause. The resulting complex nominal is then determined. Barring etymological analysis, the two components are merged in the determiner colui, although they belong to different levels of semantosyntactic structure. This lack of compositionality is represented in the following diagram:

semantosyntactic structure   NP PredP
╱    ╲
     ╱     Nom
  ╱ ╱       ╲
Det   N          RC VP
coding  ╲  ╱          △
expression structure colui [ che non ha mai imparato a lavorare ] l'impara all'università.

‘Formal’ heads

At the end of this section, we consider a variety of relative construction which does the service of a free relative in languages which have neither a headless nor a light-headed relative clause.

.Mary=obut-tahito=waBillda
Jap[ Mary=ACChit-PRT ]person=TOPBillCOP
The one who hit Mary is Bill.
.kinooit-takoto=owasure-ta
Jap[ yesterdaysay-PRT ]thing=ACCforget-PRT
I forgot what I said yesterday.

English translations of such constructions do not require a head. The Japanese construction uses a noun in head function which is lexically as empty as the pronouns appearing in the translations and to be discussed further below. Such nouns are called ‘formal nouns’ in the Japanese descriptive tradition. Japanese has no other free relative clause; this restrictive relative construction does the service. The Japanese relative construction (s. the section on positional types) is not marked by any means. Moreover, the orientation of the relative clause is not signalled, either. Under such conditions, a headless relative clause would not be recognizable as such.

Free relative clauses are an integral part of pseudo-cleft-sentences.


1 This definition differs significantly from the one provided in Citko 2004 and is instead closer to the definitions provided in Touratier 1980, Lehmann 1984 and Gutiérrez Bravo 2012. In Citko 2004, anything that precedes the relative clause in a postnominal relative construction is its head. However, one cannot maintain both that la in Span. he visto a la que me presentaste ‘I have seen the [female] one that you introduced to me’ is a light head and that the head of a restrictive relative clause is what is restricted by it. In the present account, instead, light head means ‘light nucleus’. Pronominal elements preceding the relative clause that cannot be part of the nucleus of a restrictive construction – as la in the example; cf. also f – do not count as or contribute to a light head and instead determine the higher nominal.

2 This solution may also generalize to relative constructions which involve a relative pronoun, like the Latin one: If the difference between a headed and a headless relative clause reduces to presence vs. absence of a lexical head, it may be artificial to consider the relative clause as an adjectival if it is headed, and as substantivized if not.