Non-Lexical Core-Arguments in Basque, German and Romance: How (and Why) Spanish Syntax is Shifting Towards Clausal Head Marking and Morphological Cross-Reference

Abstract

This article deals with the ways in which non-lexical core arguments can be expressed in various languages. It tries to devise a typological hierarchy for the different types and endeavours to place Romance within this hierarchy. An analysis of Basque verbal markers as cross-reference morphemes introduces the subject with a language radically different from central IE. Using Nichols' (1986 & 1992) typological differentiation between head-marking and dependent-marking languages as its basis, a typological sub-parameter of "clausal head-marking vs. clausal dependent-marking" is suggested which is shown to correspond to two radically different types of clausal co-reference: (1) agreement (concord) and (2) cross-reference. This terminology is then used to describe and explain an ongoing syntactic change in which Spanish object clitics have evolved into obligatory verbal markers closely resembling those of Basque. Their conventional analysis as "agreement markers" is questioned and Spanish is shown to be moving towards a clausal head-marking language in which all core-arguments of the sentence have to be expressed by verbal affixes, while nominal and pronominal argument realisations become mere appositions outside the sentence core. The traditional concept of an emerging new paradigm of "object conjugation" is rejected.

Full text

NON-LEXICAL CORE-ARGUMENTS
IN BASQUE, GERMAN AND ROMANCE
HOW (AND WHY) SPANISH SYNTAX IS SHIFTING TOWARDS CLAUSAL HEAD-
MARKING AND MORPHOLOGICAL CROSS-REFERENCE1
HANS-INGO RADATZ
Catholic University Eichstätt
This article deals with the ways in which non-lexical core arguments can be expressed
in various languages. It tries to devise a typological hierarchy for the different types
and endeavours to place Romance within this hierarchy. An analysis of Basque verbal
markers as cross-reference morphemes introduces the subject with a language radically
different from central IE. Using Nichols’ (1986 & 1992) typological differentiation
between head-marking and dependent-marking languages as its basis, a typological
sub-parameter of “clausal head-marking vs. clausal dependent-marking” is suggested
which is shown to correspond to two radically different types of clausal co-reference:
(1) agreement (concord) and (2) cross-reference. This terminology is then used to
describe and explain an ongoing syntactic change in which Spanish object clitics have
evolved into obligatory verbal markers closely resembling those of Basque. Their
conventional analysis as “agreement markers” is questioned and Spanish is shown to
be moving towards a clausal head-marking language in which all core-arguments of
the sentence have to be expressed by verbal affixes, while nominal and pronominal
argument realisations become mere appositions outside the sentence core. The
traditional concept of an emerging new paradigm of “object conjugation” is rejected.
1. Introduction
Empty categories have traditionally been invoked to account for phenomena
like the optionality of surface subjects in pro-drop languages like Latin, Italian,
or Spanish and it is contended that e.g. in Italian ti amo there is no phonetic
realisation of a referential subject. The descriptive device of empty categories
has, however, not been limited to cases of ‘empty’ subject arguments but is also
used to explain apparently unrealised object arguments in languages like
Brazilian Portuguese (Goldbach 1999) and Basque. The latter has been
1 I’d like to thank Harro Stammerjohann for commenting upon an earlier draft of this paper,
Martin Haase for checking my Basque examples and data, Ulrich Detges and Rolf Kailuweit for
useful comments on my Saarbrücken talk, and an anonymous reviewer for valuable comments
on an earlier draft of this paper. John Cleek (Minneapolis) was kind enough to polish my
English style and punctuation. Many improvements on this text can be credited to them, any
shortcomings, of course, are exclusively due to my own intellectual limitations.

HANS-INGO RADATZ
182
considered to be a ‘null object language’ on the basis that objects need not be
realised on the surface, neither by lexical NPs nor by Pronouns. Goldbach
writes:
Languages like Pashto and Basque, which exhibit agreement with the subject as well
as with the object but which do not license null topics, possess null subjects and null
objects (translated from Goldbach 1999:64).
While this empty category hypothesis has been widely discussed with respect to
the more thoroughly researched IE languages of Europe, I feel, however, that
especially the Basque case calls for further discussion. In the second half of this
paper, the results of this discussion will then be used to shed light also on the
much more well studied languages like e.g. Spanish. Applied to Basque, the
null subject and null object hypothesis has always struck me as particularly
counterintuitive because it doesn’t seem to account correctly for the facts of
this language’s verbal morphology. Basque possesses a complex machinery of
verbal inflection that serves only one purpose: to assure the overt surface
realisation of every single argument in a sentence. Actually, sentences with
effectively unrealised objects are exceedingly rare in Basque and seem to occur
only in connection with nominalisations like the following, where there simply
are no morphological slots on the nominalised element to which non-lexical
objects could possibly be attached:
(1) Zakur-rengana abiatu da
dog-ALATIVE.ANIM.DET.PL proceed aux.1.SG.PRES.ABS
lot-ze-ko.2
tie_up-NOMINALISER-in_order_to
“He went to the dogs to tie (them) up.”
Basque might therefore be described as having a radical dislike for any kind
of non-realisation of object-arguments. However, the general agreement among
generative scholars turns out to be diametrically opposed to this analysis and
makes Basque a language with far-reaching object-drop capabilities. The
question whether Basque either drops its objects regularly or else is reluctant to
ever drop them at all is the starting-point for the following considerations: How
is it possible to see the same state of affairs so differently?
It may prove insightful to set this question in the larger context of the
strategies to be found in various languages for the realisation of non-lexical
subject and object arguments, i.e. I shall take an onomasiological stance on a
universal syntactic problem. In particular, I shall consider the question of
drawing a boundary between the concepts of ‘agreeing with’ and ‘instantiating’
an argument in various language types. There appears to be a continuum of
2 This is a sentence I found in a teach-yourself Basque manual (Beaumont & Lazkano
1998:111).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
183
referential forms used by different languages for the expression of non-lexical
arguments which ranges from morphologically free forms like German or
English personal pronouns, via clitics (most Romance languages) to the
polysynthetic realisations of cross-reference languages like Basque. I shall
argue that the features to be found in certain modern Romance languages
(essentially Spanish, Catalan, Galician, and European Portuguese) were best
explained as a typological shift towards a type quite similar to that of Basque.
As this is a generally much less familiar language than the others to be
discussed later, I feel that it is important to dwell rather more extensively on a
description of those basic facts of Basque syntax I will later cite to analyse the
ongoing typological change within Romance.
2. The Realisation of non-lexical arguments in Basque
Basque is a language isolate showing ergative morphology, i.e. subjects of
intransitive (or rather: unaccusative) verbs as well as direct objects are marked
for absolutive case, whereas subjects of transitive verbs take ergative case
marking. Basque possesses only very few verbs capable of taking synthetic
inflection-affixes: according to the authoritative Basque grammar of the Basque
Language Academy Euskaltzaindia from 1987, out of originally over 60
synthetical verbs in the earliest surviving texts, only 26 are still in modern
usage and even fewer if we exclude those which are limited to the literary
language (cf. Gómez & Sainz (1995:238f.). Most verbal concepts can only be
expressed periphrastically with the aid of an auxiliary.
The small number of synthetic verbs in combination with a morphology of
the agglutinating type, i.e. a verbal morphology with relatively little allomorphy
and amalgamations, allows Basque verbs to co-refer not only with their
subjects, but also with all other arguments, direct and indirect objects alike.
There is even a fourth type of co-reference affix, called ‘allocutive’ referring to
second person non-arguments of the ‘ethical dative’-type, which I shall make
no further mention of as we will only be concerned with the realisation of core-
arguments (cf. Gómez & Sainz 1995:236). Moreover, Basque has consistently
been classified as a pro-drop language in generative accounts, given that
nominal and pronominal subjects appear to be freely elidable. How then do all
these elements interact in the construction of intransitive, transitive, and
ditransitive sentences?

HANS-INGO RADATZ
184
2.1 Basque, one argument (intransitive, Nor3)
By way of exemplification, let us first analyse the intransitive Basque
sentence Ni abiatzen naiz “I set out”, Zu abiatzen zara “You set out” etc. as
follows:
1 sg. (ni-ø)i naizi
2 sg. (zu-ø)i zarai
3 sg. (hura-ø)i dai
1 pl. (gu-ø)i garai
2 pl. (zuek-ø)i zaretei
3 pl. (haiek-ø)i
abia-tzen
dirai
pron-ABS set_out-PTCP.PRS aux be.PRS.IND.ABS\1SG …
“[I / you / he she / we / you / they] set(s) out.”
Figure 1: Basque, one argument (intransitive)
The example shows absolutive case marking on the facultative subject
pronoun (ni, zu, hura etc.), a non-finite verb-form carrying lexical and
aspectual information, and the auxiliary izan “to be”, marked for tense and
mood and co-referenced with the subject pronoun. Utterances like abiatzen da
“he sets out” and abiatzen gara “we set out” without a subject noun phrase or
pronoun nevertheless constitute full sentences. Like the pronouns, nominal 3rd
person subjects as well carry absolutive case which is marked by a ø-ending:4
(2) Jon-ø abiatzen da.
Jon-ABS parting is
“Jon is setting out.”
2.2. Basque, two arguments S-DO (transitive, Nor-Nork)
In Basque transitive utterances with two arguments of the type ‘He sees the
book’, absolutive case marking goes to the direct object whereas the transitive
subject receives ergative case. In this way, subjects of transitive sentences are
morphologically distinct from those of intransitive sentences. Again, the verbal
concept (here ‘to see’) is realised periphrastically; in transitive sentences, the
3 This is the traditional name for the construction in Basque grammar. Nor, nori and nork are
the case-marked forms of the interrogative “who” which are used as the Basque names for
absolutive, dative and ergative as well as for the sentence constructions in which they
participate.
4 The zero-endings marked as ø in the morphological analysis represent slots in the paradigm in
which the absence of an affix is one of a set of possible values within the paradigm; ø therefore
contrasts with the presence of other affixes and marks a zero-ending, not the zero-realisation of
an argument (= Ø). These morphological facts have nothing to do with the descriptive device in
syntactic analysis of positing empty categories like traces or pro and the like.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
185
auxiliary has to be ukan ‘have’, rather than izan ‘be’, as only ukan has the
necessary cross-reference capabilities which allow co-reference not only with
the ergative-subject but also with the absolutive-direct object.
1 sg. (ni-k)i dj-u-ti / ditj-u-ti
2 sg. (zu-k)i dj-u-zui / ditj-u-zui
3 sg. (har-k)i dj-u-øi / ditj-u-øi
1 pl. (gu-k)i dj-u-gui / ditj-u-gui
2 pl. (zue-k)i dj-u-zuei / ditj-u-zuei
3 pl. (haie-k)i
liburu-aj /
liburu-akj
ikus-ten
dj-u-tei / ditj-u-ztei
pron-ERG book-DET.ABS.SG/
book-DET.ABS.PL
see-
PTCP.PRS
ABS.3SG-aux have-PRES.IND.ERG.1SG … /
ABS.3PL-aux have-PRES.IND.ERG.1SG …
“[I / you / he she / we / you / they] see(s) the book(s)”
Figure 2: Basque, two arguments (S + DO-3 Sg)
Somewhat simplifying the morphological situation, one might analyse the
affix du- as co-referencing the singular direct object while ditu- corresponds to
a direct object in the plural. At the same time, du- / ditu- are also marked for
3rd person, as there are special forms for 1st and 2nd person as well:
(3) (Ni-k) ikus-ten zai-tu-t.
I-ERG.SG see-PTCP.PRS ABS.2SG-aux have.PRS.IND-ERG.1SG
“I see you (sg.).”
(Zu-k) ikust-en nau-ø.
you-ERG.SG see-PTCP.PRS ABS.1SG\aux have.PRS.IND-ERG.2SG
“You (sg.) see me”, etc.
Again, the inflected verb group on its own already constitutes a complete
sentence, as all lexical arguments are freely elidable:
(4) Ikust-en di-tu-t.
see-PTCP.PRS ABS.3PL-aux have.PRS.IND-ERG.1SG
“I see them”.
Similar affixes – and similar one-word sentences – also occur with the other
synthetic verbs like e.g. jakin ‘to know’:
(5) a. Arrazoi-a d-aki-t.
reason-DET.ABS.SG ABS.3PL-know.PRS.IND-ERG.1SG
“I know the reason.”

HANS-INGO RADATZ
186
Ba-d-aki-t.
ENUNC-ABS.3PL-know.PRS.IND-ERG.1SG
“I know it.”
b. Arrazoi-ak d-aki-zki-t.
reason-DET.ABS.PL ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1SG
“I know the reasons.”
Ba-d-aki-zki-t.
ENUNC-ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1SG
“I know them.”
c. Arrazoi-a d-aki-gu.
reason-DET.ABS.SG ABS.3SG-know.PRS.IND-ERG.1PL
“We know the reason.”
Ba-d-aki-gu.
ENUNC-ABS.3SG-know.PRS.IND-ERG.1PL
“We know it.”
d. Arrazoi-ak d-aki-zki-gu.
reason-DET.ABS.PL ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1PL
“We know the reasons.”
Ba-d-aki-zki-gu.
ENUNC-ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1PL
“We know them”.
It is this type of structures that has given rise to the description of Basque as a
null object language. Actually then, Basque is not only considered to be pro-
drop but has also been dubbed an ‘object-drop-language’ (Landa & Franco
1996:160). Saltarelli goes even one step further and characterises Basque as a
‘null argument language’ in which any type of lexical argument is always
optional as far as syntax is concerned:
[...] with respect to the issue of empty categories, Basque is a null subject language, as
Italian has been defined. Moreover, Basque is a null object language (direct and
indirect). If we follow Taraldsen’s original hypothesis that null subject languages are a
consequence of the richness of the verb inflectional properties of the language, then
Basque is properly a ‘null argument language’ since its verb inflectional parameter
marks subject (both thematic and non-thematic), direct object and indirect object
(Saltarelli 1988:XIX).
2.3. Basque, three arguments S-IO-DO (transitive, Nor-Nori-Nork)
By adding an indirect object as a further argument, we get the ditransitive
construction of the type “He gives the man the book” which is rendered in
Basque by marking the subject as ergative, the direct object as absolutive, and
the indirect object as dative. Again, all arguments are cross-referentially
represented on the auxiliary ukan:

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
1
87
1 sg. (ni-k)i dj-i-ok-ti
2 sg. (zu-k)i dj-i-ok-zui
3s sg. (har-k)i dj-i-ok-øi
etc.
gizon-arik
liburu-aj
ema-ten
etc.
pron-ERG man-
DET.DAT.SG
book-
DET.ABS.SG
give-
PTCP.PRS
ABS.3SG-AUX
have.PRS.IND-DAT.3SG-ERG.1SG …
‘[I / you / he she / we / you / they] give(s) the man the book’
Figure 3: Basque, three arguments (S-IO-DO)
As can be gathered from the gloss, the auxiliary now has to carry an
absolutive marker referring to the direct object, a dative marker referring to the
indirect object and an ergative marker, securing co-reference with the subject.
Again, the table only gives a glimpse of the actual possibilities and intricacies
of Basque cross-reference morphology. A sentence like (6) below:
(6) (nik)i gizon-arik liburu-aj ema-ten
(pron-ERG) man-DET.DAT.SG book-DET.ABS.SG give-PTCP.PRS
dij-ok-ti
ABS.3SG-AUX have.PRS.IND-DAT.3SG-ERG.1SG
“I give the man the book.”
shows a reflex of plurality on the direct object and on the auxiliary, if the direct
object is in the plural:
(7) (nik)i gizon-arik liburu-akj ema-ten
(pron-ERG) man-DET.DAT.SG book-DET.ABS.PL give-PTCP.PRS
dizkij-ok-ti
ABS.3PL-AUX have.PRS.IND-DAT.3SG-ERG.1SG
“I give the man the books.”
Changing the indirect object to plural as well, we get (8):
(8) (nik)i gizon-eik liburu-akj ema-ten
(pron-ERG) man-DET.DAT.PL book-DET.ABS.PL give-PTCP.PRS
dizkij-ek-ti
ABS.3PL-AUX have.PRS.IND-DAT.3PL-ERG.1SG
“I give the men the books.”
As can be seen in these examples, Basque is markedly different from the
central Indo-European languages in that its finite verbs obligatorily reflect in
their morphology all basic relationships of argumenthood in a sentence, a
syntactic property known as cross-reference in language typology. The status of
these object affixes has been described in various ways in the literature:

HANS-INGO RADATZ
188
– Schwerteck (1984:9), in talking about the incorporating verb forms, calls
the object affixes “pronouns”, “segmental markers” und “pronominal
elements”.5 This terminology reflects the intuition that these affixes
themselves carry out syntactic functions which go beyond mere agreement;
calling them pronouns, however, would endow them with a degree of
morphological independence which they definitely do not possess.
– Franco (1991 and 2000) talks about “verbal agreement morphemes”,
Goldbach (1999:10) about a “system of object agreement”6 and
“inflectional endings on the verb”.7 This is standard generative terminology
which takes into account their stronger morphological integration into the
verb, but sees them as mere agreement devices.
– Saltarelli (1988:XVII) generally uses the unspecific term “marker”; he does,
however, call Basque a “highly inflected language” and seems to place less
weight on the agglutinative character of Basque morphology.
Most authors seem to concur then in considering these markers ‘agreement
affixes’ of some kind and the following analysis by Landa and Franco
represents a generally accepted view concerning the structure of Basque
sentences. As in English or Spanish, verbal markers are seen as only co-
indexing the real (lexical or pronominal) arguments. If these happen to be
absent, this phenomenon is seen as ‘elision’ and the index goes to an empty
category – a trace or a pro:
(9) Jonek aulkia apurtu du eta nik konpondu dut.
Jon-ek aulki-ai apurtu d-u-ø eta ni-k Øi
Jon-ERG chair-ABS break ABS.3SG-AUX have-ERG.3SG and I-ERG
konpondu d-u-t
fix ABS.3SG-AUX have-ERG.1SG
“Jon has broken the chair and I have fixed (it).” (Landa & Franco 1996:161)
5 “Pronomina [...] segmentale Bedeutungsträger [...] pronominale Elemente” in the German
original. As to the historical origin of these polysynthetic verbal complexes of Basque he
writes: “It is by no means new that Basque verbal forms, when retraced to former states in their
history, have been considered to be entire sentences ([Footnote 21:] Examples for this view can
be found e.g. in H. Schuchardt’s Primitiae linguae vasconum)” (translated from Schwerteck
1984:15).
6 “Just as the null-subject possibilities of Italian and Spanish follow from the potential of their
subject inflection, the null-object possibilities of Pashto and Basque result from their system of
object agreement” (translated from Goldbach 1999:65f.). This is, of course, an example for
what Haspelmath (1999:188) has dubbed the “teleological fallacy”: rich agreement morphology
may be an important factor in facilitating the described phenomena but can not in itself be seen
as causing it.
7 “[...] like for example in Basque, where inflectional endings unequivocally indicate the
possible subjects (or objects) of a verb” (translated from Goldbach 1999:10).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
189
According to this theory, the direct object of the coordinated sentence can
be elided easily because the object agreement markers on the verb identify the
null object completely. They co-index the direct object aulkia “chair” with the
empty constituent in the object position of the coordinated sentence. Note,
however, that this is by no means the only possible analysis; after all, both
auxiliaries already contain a marker for 3sg. absolutive and we might envisage
an alternative view in which the observable verbal affixes, rather than a
hypothetical Ø, serve as the locus for co-indexation with the lexical direct
object as in (10):
(10) Jonek aulkiai apurtu di-u-ø
Jon-ERG chair-ABS break ABS.3SG-AUX have-ERG.3SG
eta nik konpondu di-u-t.
and I.ERG fix ABS.3SG-AUX have-ERG.1SG
“Jon has broken the chair and I have fixed (it)”
Why use an empty constituent if there are actual surface elements available
that could be seen either as agreement markers or just as well as pronoun-like
representations of the direct object itself? As a consequence of the non-
obligatory nature of Basque NP-arguments on the one hand and Du Bois’
Preferred Argument Structure constraint (“Avoid more than one lexical core
argument” cf. Du Bois 2003:34ff.) on the other, the great majority of all
arguments would then have to be treated as not occurring on the surface,
triggering an immense proliferation of empty syntactic categories. This is likely
to be an unnecessary violation of any principle of descriptive economy and
therefore a classical candidate for an application of Occam’s razor.
A further relevant observation about this “empty constituent” can be gained
from a comparison with null objects in other languages like e.g. Spanish. A
major difference between Basque and Spanish lies in the fact that in Basque the
only possible alternative to the null object would be either the full lexical noun
phrase aulkia “chair” itself or else an emphatic demonstrative. This is due to
the (highly noticeable!) fact that Basque, unlike Spanish, does not possess a
class of non-demonstrative object pronouns. Nevertheless, the direct object
does not have to be contextually inferred; rather, it is unequivocally specified in
the verbal morphology as to case and number, namely absolutive singular. That
is, the object affixes are specified with the same or even more precision than
the corresponding Romance clitics or the German pronouns. One might ask,
therefore, why Basque verbal affixes are usually not considered to be
realisations of arguments but are seen rather as mere agreement morphology
referring to lexically realised arguments or their respective traces.
These elements are clearly different from such well-established agreement
markers as e.g. subject agreement morphemes in German or English. In these
languages, verbal markers do not only agree with noun phrase subjects but, in

HANS-INGO RADATZ
190
their absence, also with a class of pronominal subjects which in non-pro drop
languages as German are even obligatory:
(11) a. Peter singt. Er singt. *Singt.
b. Peter sings. He sings. *Sings.
This is prototypical ‘agreement’: morphological co-reference with an overt NP
or pronoun. Basque, however, is different. While first and second person
indirect objects might still be overtly realised as case-inflected forms of
emphatic personal pronouns, no such option is available for third person
objects. Basque only has 1st and 2nd person genuine personal pronouns; it is
only for the sake of completeness that grammars tend to list the demonstratives
hura and haiek for the function of emphatic 3rd person pronouns (cf. Zubiri
2000:47). Therefore, if we consider the verbal markers to be mere agreement
morphemes, the least emphatic way of expressing third person pronominal
objects on the surface would be demonstratives. This syntactic fact casts further
doubt on the analysis of Basque verbal affixes as agreement markers, because if
these elements actually were agreement affixes, the question arises: What are
they are supposed to be agreeing with?
Let’s finally turn to a last example in which the auxiliary is not only marked
for cross-reference but also for tense and mood, and which is therefore better
suited to show the full possible complexity of Basque finite verbs and the
agglutinating character of the language (taken from Saltarelli (1988:XVII)):
(12) Guk gizonei liburuak eman diezazkiekegu.
Gu-k gizon-ei liburu-ak ema-n
we-ERG man-DET.DAT.PL book-DET.ABS.PL give-PRTC.PERF
d -i -eza -zki -e -ke -gu
ABS.3PL -DAT.3PL -AUX have.PRS -ABS.3PL -DAT.3PL -MOOD -ERG.1PL
“We can give the books to the men.”
All lexical arguments have to be co-referenced by a marker on the
auxiliary:
– The subject, i.e. the ergative pronoun <guk> by the marker ERG.1PL <-gu>.
– The dative indirect object <gizonei> by the discontinuous dative marker <-
i-> for 3rd person and <-e-> for plural.
– And finally, the absolutive direct object <liburuak> by the discontinuous
absolutive marker <d-> for 3rd person and <-zki-> for plural.
The complete representation of all arguments on the auxiliary allows for any
lexical ergative, absolutive, and dative argument to be optional and the verbal
complex alone constitutes a complete grammatical utterance:

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
191
(13) Eman diezazkiekegu.
Ema-n d -i -eza -zki -e
give-PRTC.PERF ABS.3PL -DAT.3PL -AUX have.PRS -ABS.3PL -DAT.3PL
-ke -gu
-MOOD -ERG.1PL
“We can give them to them.”
The situation in Basque is thus radically different from that in German or, if
slightly (and sometimes even considerably) less, from that of the Romance
languages. Now, there may be the danger of illicitly projecting our intuitions
gained from more familiar Indo-European languages onto the situation in a
non-Indo-European language like Basque, for which reason the following two
questions are by no means trivial:
– Are the ergative, dative and absolutive noun phrases, guk, gizonei and
liburuak, really the subject, indirect, and direct object of the sentence, in the
same sense in which the nominative, dative, and accusative arguments in
the German sentence Wir können den Männern die Bücher geben (“We can
give the books to the men”) would be said to instantiate the subject, indirect
object, and direct object?
– Are the Basque cross-reference affixes really only agreement markers,
comparable to verbal agreement in German Wir können (“We can”)?
An answer to these questions is not only important to Bascologists but could
also aid us in reassessing the well-known facts of the Romance languages,
shedding light on the syntactic status of clitics in those constructions that have
sometimes been called “clitic-doubling” and sometimes “Romance object
conjugation”.
3. The typological parameter head-marking vs. dependent marking
In order to do justice to languages like Basque, Johanna Nichols in her
1986 seminal paper on head-marking vs. dependent-marking languages
introduced a major typological distinction that may well be relevant to the
problem of non-lexical argument realisation in various languages.8 Nichols
draws our attention to an important empirical imbalance in linguistic research
that existed at that time (and most probably persists until today): while head-
marking features are statistically very important and probably even majoritary
among the languages of the world (Nichols 1986:89ff.), this language type had
received only very little if any attention by linguists – including those who
claimed to be working on language universals:
8 In this paper she discusses a sample of over 60 typologically diverse languages, among which
there is – quite surprisingly – not one Romance language! The author developed her theory
further in Nichols (1992).

HANS-INGO RADATZ
192
Despite the efforts of formal grammarians to take a range of languages into
consideration, there is a glaring gap in the typological coverage: the exotic languages
that have so far received significant attention have been almost exclusively dependent-
marking (Japanese, Korean, Finnish, Malayalam, Australian languages) or double-
marking (Turkish, Arabic, Hebrew, Australian languages). Of the head-marking
languages, only Navajo has received significant theoretical attention. In addition,
Relational Grammar has investigated a number of head-marking languages (primarily
from the Algonkian, Salishan, and Wakashan groups); but these languages have not
been used to raise questions of constituency, centricity, syntactic bonds, government
etc. (Nichols 1986:115f.).
Nichols’ typological analysis is based on two universal principles, namely
headedness and overt morphological marking. Headedness is meant to convey
the notion that natural language constituents regularly come as structured into a
head, an obligatory constituent determining the overall syntactic status of the
whole constituent, and one or more facultative dependents which derive their
morphosyntactic accidents from the head. This type of configuration can be
shown to exist not only at the phrase but also at the clause and sentence level:
Constituent Head Dependent
possessed noun possessor
noun modifying adjective
Noun phrase
adposition object of adposition
predicate / verb arguments / adjuncts
Clause auxiliary verb lexical / main verb
Sentence Main-clause predicate subordinate clause
Figure 4: Head-dependent-configurations (cf. Helmbrecht 2001:1425)
Typical dependent-marking structures can be found e.g. in English
possessive constructions like the man’s house, in which the noun house is the
head whereas possession is morphologically marked on the dependent by some
kind of genitive affix. In Hungarian, on the other hand, the same relationship is
expressed by a possessive affix on the head itself (cf. Figure 5). However,
while Nichols’ typology considers head-dependent configurations on all
syntactic levels, I shall from this point on limit myself to phenomena at the
level of sentences, or rather: of non-subordinated or ‘main clauses’. Nichols’
basic claim is that languages throughout the world show a strong tendency to
mark grammatical relations consistently either on the head or on the dependent.
At the same time, however, she concedes that this is only a statistical regularity
and individual languages usually adhere only “predominantly” to one or the
other type while any degree of admixture is always possible.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
193
Dependent Head
English the man’s House
DET man-POSS House
Hungarian az ember ház-a
DET man house-3.SG
Figure 5: Dependent vs. head-marking within the NP
If the phenomenon can occur independently on different levels, it should be
legitimate to extract typological sub-parameters from the general theory. I
therefore suggest a typological parameter clausal head-marking to be derived
and abstracted from the general principle that Nichols advocates for. As to the
distribution of head and dependent on the clause level, Nichols argues
(1986:57) that predicates constitute the head, whereas arguments and adjuncts
can be analysed as dependents.9 Clausal head-marking languages, then, are
those in which all arguments of the sentence are usually realised by
morphological means on their finite verbs or auxiliaries. I contend that the
question of non-lexical argument realisation can be seen along the lines of
clausal head-marking vs. clausal dependent-marking and that Nichols’
parameter may be used to shed new light on the historical development of clitic
systems in Romance in comparison with languages like Basque or German
which represent markedly different types.
Due to the fact that the tendency for consistently marking either on the head
or on the dependent is not absolute, there are actually further types to be found:
in double-marking, syntactic relations are redundantly marked on heads and
dependents alike, whereas in split-marking languages, the situation may be
different, depending on the grammatical subsystem. Basque is actually a good
example for both types: at the sentence level, all syntactic relations are reflected
on the verb (head) but argument noun phrases are also marked for case so that
we get clausal double-marking. Moreover, Basque belongs to the split-type, as
double-marking is limited to the clause level in this language, which is
otherwise largely dependent-marking on the phrasal level.
For an example of a genuine head-marking language we may want to take a
look at Abkhaz. Argument noun phrases in Abkhaz bear no case markers
whatsoever and the entire morphosyntactic information is polysynthetically
realised on the verb, which morphologically cross-references with the lexical
arguments for person, number, and gender. In the absence of morphological
case, the syntactic relations are encoded through the serialisation of verbal
9 In other cases, auxiliaries and their respective lexical (‘main’) verbs may also stand in a
clausal head-dependent relation to each other. I will not dwell further on this interpretation.

HANS-INGO RADATZ
194
affixes. All argument noun phrases are optional and appear only if they are
discourse-pragmatically needed to establish the reference of the verbal affixes.
(14) Pure head-marking language: Abkhaz
a-xàc’a a-p°s a-š°q°’ ø-lc-y-te-yt’
the-man the-woman the-book it-to_her-he-gave-FINITE
“The man gave the woman the book.” (Helmbrecht 2001:1427)
The inflected verb group again constitutes a complete sentence,10 quite
similar to the situation in Basque, as we have already seen:
(15) ø-lc-y-te-yt’
it-to_her-he-gave-FINITE
“He gave it to her.”
Now, Nichols observes that this effect seems to be characteristic of all
languages with head-marking morphology at the clause level. This is hardly
surprising, as these languages code the entire grammatical information about
the clausal arguments like case, number, gender etc. on the verb, leaving noun
phrases with the sole function of providing lexical information about new
discourse participants. Speakers are free to omit lexical or pronominal material
if it is referentially recoverable in the discourse universe. In these languages,
the surface appearance of lexical noun phrases and pronouns is bound
exclusively to discourse motivations and can therefore be completely optional
at the syntactic level. When full noun phrases appear at all, they behave like
appositions with respect to their coreferential verbal affixes and are realised
outside the sentence core.11
In head-marked grammatical relations, the dependent is usually an optional element of
the constituent. For instance, in languages with consistently head-marked clauses, the
verb itself normally constitutes a complete sentence; full NP’s are included only for
emphasis, focus, disambiguation etc. (Nichols 1986:107).
So, in head-marking languages there is a sharing of functions between the
verbal affixes and the full noun phrases which is quite comparable to that
between full verb and auxiliary in verbal periphrases, in that one element, the
auxiliary, specialises in conveying only grammatical information while the
lexical information is left entirely to the other element, the full verb. In the
same vein, head-marking verbal affixes specialise in the purely grammatical
10 “The appositional status of NPs in Abkhaz can be found in many other non-European
languages and may be viewed as a defining feature, among others, of polysynthetic languages”
(Helmbrecht 2001:1427).
11 That lexical noun phrases should be analysed as appositions (or ‘adjuncts’ or ‘dislocated
elements’ respectively) in these cases has also been suggested by generativist authors. Aoun
(1993) proposes such an interpretation for Lebanese Arabic (cf. the critical discussion in
Franco 2000:152ff.).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
195
function while the corresponding appositions carry the lexical information, may
receive contrastive stress and can be freely positioned in the sentence periphery
according to the requirements of discourse pragmatics. The parallelism is,
however, not complete, as in verbal periphrases both elements are obligatory
while in head-marking languages only the verb complex is. In Basque and
Abkhaz sentences, the verbal affixes already seem to constitute the arguments
while the full noun phrases show all the same behaviour of facultative
appositions.
Nichols was by no means the first to propose this analysis; rather it seems
to have been the view of almost any linguist ever to have worked on American
Indian languages:
The dependent in a head-marked constituent stands in a roughly appositive relation to
the head (or, more precisely, to the coreferential marker on the head); the term
‘government’, developed by traditional grammar on the basis of exclusively
dependent-marked relations [...], is not appropriate for head-marked constituents.
Since the appearance of Boas 1911, descriptions of American Indian languages have
insisted that subject and object in these languages are in apposition to the pronominal
markers on the verb, rather than (as in Indo-European) being syntactically governed by
a verb which agrees with them (Nichols 1986:107).
One of the most prominent advocates for this typological distinction is co-
founder of Role and Reference Grammar (RRG) Robert Van Valin, who was
drawn to it by his work on the head-marking language Lakhota. This may be
the reason why RRG is one of the few grammatical theories on whose very
foundations a systematic distinction between head-marking and dependent-
marking was implemented. In a short introductory manifesto on RRG, Van
Valin writes:
In head-marking languages like Lakhota, the bound pronominals on the verb are
considered to be the core arguments; overt NPs are within the clause in apposition to
them (Van Valin 2005:3)
According to this analysis, Basque verbal affixes would no longer have to be
considered agreement markers but rather ‘core arguments’ in themselves; overt
noun phrases, on the other hand, would regularly be seen as appositions, which
would not only account for their unlimited elidability but also for the lack of
syntactic constraints on their position relative to the core-sentence and to each
other.
The fundamental typological difference between these two marking-types
should also be reflected in the terminology that is used to describe the different
type of clausal co-reference that can be observed in head-marking and
dependent-marking languages. Most authors, however, tend to carry over the
notion of ‘agreement’ also to head-marking languages as well, thus obfuscating
the fundamental differences between the two types. This is unnecessary as the

HANS-INGO RADATZ
196
terminological clarification has been around for over seventy years and, what is
more, was published in Bloomfield’s easily accessible, classical monograph
“Language”. Here Bloomfield (1933:191ff.) introduces the term cross-
reference to refer to the type of clausal co-reference which is typically found in
head-marking languages :
Agreement (= concord) Cross-Reference
dependent-marking head-marking
the argument (= dependent) is realised as
a full NP or as a pronoun and is
morphologically marked (dependent-
marking)
there is co-reference with a non-obligatory
full lexical NP which syntactically behaves
like a non-core constituent (apposition)
the verb (head) carries a redundant
agreement marker the argument is realised by an obligatory
affix on the verb (head-marking
example: subject agreement in German
or English example: Basque, Abkhaz; tendencies in
Romance, e.g. French Jean où est-il?
(Bloomfield 1933:193)
Figure 6: Two types of clausal co-reference
Thus, whereas clausal co-reference is morphologically realised as
‘agreement’ in languages with clausal dependent-marking, the corresponding
morphological process in clausal head-marking languages is ‘cross-reference’ –
and therefore radically different. If it is true that in language typology it is not
important what can be but rather what must be expressed in a language, then
Basque definitely has to be described as a language in which all objects have to
be expressed on the surface. And it is as far as can be from a null-subject, null-
object or null-argument language. Rather, null-arguments are impossible in
Basque because the verbal morphology requires the specification of all
arguments. Lexical noun phrases, which by analogy with dependent-marking
languages could be mistaken for subjects and objects, do not belong to the
sentence-core and must be viewed as non-obligatory appositions providing
additional lexical and referential information about the arguments they are co-
referenced with. Rather than being arguments they only partake in the
argumenthood of the verbal affixes.
It should be noted that the typological importance of the parameter ‘clausal
head-marking’ derives largely from the cross-reference morphology on the verb
that comes with it. Whether lexical and pronominal arguments carry additional
case marking, as e.g. in Basque, or at least ‘differential object marking’ (cf.
Bossong 1991), as e.g. in Spanish, is without consequence in this context. On
the clause level we can therefore class head- and double-marking languages
together and oppose them with the dependent-marking languages and their
agreement morphology as the only genuinely different type.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
197
If with cross-reference the traditional terminology of “agreement” vs.
“argument” is misleading and inappropriate, which terms should be used then
in describing the phenomena adduced above? Du Bois (1987), working on the
ergative cross-reference Mayan language Sacapultec, provides us with a
terminology which most adequately captures the syntactic facts in languages
like Sacapultec and Basque. Where dependent-marking languages tend to have
two separate things, i.e. arguments and agreement, head-marking languages
only have ‘mentions’ which Du Bois defines as
a reference item complex consisting of either a bound form alone (a cross-referencing
affix [...]), or an overt free form (full NP or independent pronoun) plus its cross-
referencing bound form within the same clause (Du Bois 1987:813).
A ‘mention’ then can consist of an obligatory cross-reference marker alone,
in case the lexical noun phrase remains un-expressed. If, however, a lexical NP
is present, the NP plus the verbal marker together instantiate one discontinuous
argument realisation, bound together by cross-reference. This notion
acknowledges the fact that in these languages the grammatical and the lexical
aspect of argumenthood are typically split up, only the grammatical element
being obligatory. We will have to see whether the notion of ‘mention’ might
not be better suited to account for certain Romance languages notorious for the
problem of determining the status of clitics and affixes.
In this conception, the verbal affixes in Basque constitute actual instances
of the arguments with an independent external reference. I have to thank an
anonymous reviewer for pointing out to me a number of conventional tests for
argumenthood which, apparently, would all have to yield an affirmative result
in order to substantiate this hypothesis. These tests include the questions:
i. Can these affixes be the answer to a question?
ii. Can these affixes carry the main accent?
iii. Do these affixes allow for ellipsis of the inflected auxiliary?
iv. Do these affixes combine with other DPs or NPs?
v. Are these affixes flexible in their position with respect to each other and the verb?
Quite clearly, the answer to all the above questions is no. This does,
however, not necessarily invalidate these elements’ claim to argumenthood. All
the above tests are diagnostics designed to detect arguments in dependent-
marking languages in which arguments must be realised as free morphemes.
But, as detailed above, it is precisely the nature of cross-reference languages
that they realise arguments not as free morphemes but rather as ‘mentions’
which, by definition, can consist of just a verbal affix. In their mentions we find
a repartition of function between argument-realising verbal affixes and their
facultative appositional counterparts; the affixes fail all the above tests because
the tests rely on those syntactic circumstances which would call for the

HANS-INGO RADATZ
198
otherwise elided appositions. The failure of Basque verbal affixes to comply
with the traditional tests for argumenthood can therefore not be adduced to
disprove their status as arguments in a head-marking language.
4. The other end of the hierarchy: German
At the other end of the hierarchy of clausal head-marking we find languages
like German. Its only head-marking feature on the clausal level is concord, i.e.
the typical Indo-European agreement of verbs with their subjects.12 Actually,
according to Nichols, pure clausal head-marking seems to be the exception
rather than a statistically relevant type, and we may therefore disregard subject
agreement and treat German for all practical purposes as a dependent-marking
language. In German, only full noun phrases and pronouns can function as
arguments:
(16) Ich gebe Günther das Buch.
I.NOM give Günther the.ACC book.
“I give Günther a book.”
(17) Ich gebe es ihm. / Ich gebe ihm das. /
I give it him / I give him that /
Ich gebe dem das. / I give it him.
Ich gebe es dem. / I give him that /
“I give it to him.”
How free or how bound these pronouns may be in each case, i.e. how far
cliticisation has already progressed in German, is not relevant in our context. It
is, however, safe to say that German pronouns are syntactically much closer to
lexical noun phrases than any ‘weak’ or even ‘strong’ pronoun in the Romance
languages (with the possible exception of Brazilian Portuguese, which seems to
be moving away from the rest of Romance in this respect).
In the context of the head-marking morphology in Basque, I asked whether
we wanted to analyse the verbal affixes as actual realisations of or as mere
agreement markers with the subject, direct and indirect object. Irrespective of
the answer to this question, no one can deny that such a question might at least
be raised for Basque. For German, however, it would make no sense at all:
– The subject markers on German verbs are clearly agreement affixes
referring to an independently realised, overt obligatory subject – a full noun
phrase or a subject pronoun.
– Object clitics, pronouns, and noun phrases clearly consitute the realisations
of the respective arguments and are typically not appositional. Noun
12 In this respect, English would be an even better example, as concord has been reduced to a
rather precarious agreement in 3rd person singular verbs. But if English has almost no head-
marking, it also has almost no morphological (case-) marking on the dependents, which makes
it a completely different type.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
199
phrases and pro-forms are in complementary distribution and there can be
no clitic doubling.
(18) a. Ich gebe es ihm.
I give it him
“I give it to him.”
b. *Ich gebe esi ihm dasi.
I give it him that
c. *Ich gebe es ihmi Peteri.
I give it him Peter
d. [*Miri / *Ichi] miri gefällt das.
[to_me / I] to_me appeals that
“This appeals to me / I like this.”
5. The Romance languages – problematic intermediate cases
We have now examined one example of a language with full clausal head-
marking and another one with full dependent-marking morphology. The best
way to characterise the Romance languages would be to locate them on an
abstract scale on different points in between these two more extreme examples.
This intermediate status of Romance has already been noticed by Nichols, who
explicitly singles out the Romance case as an example of ongoing typological
change within the group of the otherwise quite consistently dependent-marking
Indo-European languages:
Indo-European has retained its basic type – dependent-marked with subject inflection
on verbs [...] for languages preserving the inherited morphology – for some 6,000
years, with only a recent trend toward head-marked clauses in the pronominal clisis of
the Romance languages (a process which occurs only after most of the morphology has
been lost) (Nichols 1986:89).
This change is driven by a general Romance tendency to introduce new,
alternative head-marking strategies. Historically, most pronouns were reduced
to pronominal clitics and these have frequently been further amalgamated into
affixes and affix groups which increasingly resemble the cross-reference
morphology of Abkhaz, Lakhota or Basque, especially in those Romance
languages in which obligatory clitic-doubling is gaining ground.
Like most other Indo-European languages, the Romance languages all
possess subject co-reference morphology on their verbs, which has traditionally
been considered to be agreement just as in the case of languages like German.
This view, however, is not undisputed and thus Corbett (1994) finds it
disturbing to speak of agreement in a situation where the non-realisation of the
purported referent of the agreement is the statistically normal, unmarked case:

HANS-INGO RADATZ
200
Another common problem, and a serious one in some theoretical frameworks, is the
existence of so-called ‘pro-drop’ languages. It may be desirable to say of certain
languages, for example, Serbo-Croatian, that verbs agree with their subject in person
and number, as in ja itam ‘I read’, ty itaš ‘you read’. In most circumstances, the
more natural utterances would be itam and itaš. The problem is that the pronominal
agreement controller is normally not present (Corbett 1994:57).
Kailuweit’s (2005) position within the framework of Role And Reference
Grammar is more differentiated, as he considers only part of the cases as
instances of agreement. Kailuweit no longer analyses the subject markers on
Spanish verbs as agreement affixes as has been traditionally done. On his
account, at least as long as no lexical or pronominal subject is present, the
marker not only agrees with the subject argument but rather realises it directly.
Only if a better candidate for subjecthood is present, will this be analysed as the
subject, and the same verbal marker is then seen to function as an agreement
affix:
It is generally assumed that, with the exception of French, Romance languages are so-
called ‘pro-drop’-languages. As illustrated under (1) in a Spanish example, they are
capable of realising the subject-argument through the morphological categories of
Person and Number on the verb – categories which function as agreement markers as
soon as a nominal phrase (NP) appears in the position of the subject.
(1) a. Maríai cant-ói [agreement]
b. Cant-ó [subject realised by means of a verbal suffix]
(translated from Kailuweit 2005:1).
This is intuitively appealing on the one hand, because it no longer leaves
cant-ó (sing-3SG.PST.IND) without a subject as if it were ‘incomplete’ in any respect;
on the other hand it is also problematic because it is unconvincing to treat one
and the same affix alternatingly as an argument or as an agreement marker,
depending on the presence or absence of a more suitable alternative. This
solution has all the air of stopping short halfway. Once we have conceded that
Spanish verbal affixes actually realise their subject arguments in some cases,
we might then also consider the possibility that Spanish has come to be even
more head-marking than that. Like all Romance languages, Spanish has a fully
developed subject co-reference morphology on the verb and freely elidable
lexical or pronominal subjects,13 a feature that has usually been described as
pro-drop or null-subject. This, of course, only makes sense, if one conceives of
Spanish subject co-reference in the same terms as of German subject
agreement. If, however, the situation in Spanish were to be more analogous to
Basque than to German, its subject-markers would have to be seen as an
13 French subject clitics which have traditionally been seen as clitical pronouns are nowadays
analysed as agreement affixes by most linguists. In this case, French would be a null subject
language, just like the rest of her Romance sister languages (cf. Kaiser 1992:115 et passim).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
201
entirely different breed of affixes, namely obligatory cross-reference affixes
which serve to express the subject argument. Du Bois’ concept of ‘mention’
almost imposes itself here, as the facts of this aspect of Spanish syntax
resemble Sacapultec (and Basque) considerably more than those Indo-European
languages on which the traditional analysis is based.
Spanish is therefore on the one hand similar to German in that it is not a
null-subject language; both languages require overt subjects on the surface. The
radical difference, on the other hand, lies in the morphological technique of
subject realisation which is analytic in German and (poly-) synthetic in Spanish
(and in the Romance languages in general). As would be expected for head-
marking languages, lexical and pronominal ‘subjects’ are not integrated into the
sentence core in the Romance languages and can either appear on the periphery
(and only there!) or be elided altogether:
(19) [yoi,] a ese no se=lo=voyi a dar [, yoi] (Spanish)
I to that_one not him=it=go.1SG.PRS.IND to give I
(20) [moii,] je=ne=vaisi pas le=lui=donner [, moii] (French)
I I=not= go.1SG.PRS.IND NEG it=him=give I
(21) [*ich,] dem gebe [ich] es nicht [*,ich] (German)
I to_that_one give.1SG.PRS.IND I it NEG I
“I’m (certainly) not giving it to him!”
While subjects are realised through a head-marking construction in all
Romance languages, the situation with objects is far less uniform across
Romance. The general picture is complex and varies from language to
language; they all seem to have in common, though, that at least certain objects
are realised via head-marking on the verb. Spanish (and Catalan) are probably
the two most advanced languages on the way towards clausal head-marking and
I shall therefore use Spanish to illustrate the point (cf. Figure 7). That table is
admittedly an extreme idealisation of the actual situation and is only intended
to represent the gross overall picture. Thus, the only dependent-marking
solution in the example is only possible as long as the direct object-noun phrase
remains in its canonical position. Once it is raised into a left-dislocated topic
position, the marker on the verb becomes obligatory:
(22) a. Juan conoc-e a María.
John know-s to Mary
“John knows Mary.”
b. A María, Juan la=conoc-e.
to Mary John her.ACC=know-s
“As for Mary, John knows her.”

HANS-INGO RADATZ
202
Head-Marking Dependent-Marking
NP (*)Juan la=conoc-e a María.
John her.ACC=know-s to Mary
“John knows Mary.”
Juan conoc-e a María.
John know-s to Mary
“John knows Mary.”
Juan la=conoc-e a ella.
John her.ACC=know-s to she
“John knows her.”
*Juan conoc-e a ella.
John know-s to she
“John knows her.”
direct objects
Pron
Juan la=conoc-e.
John her.ACC=know-s
“John knows her.”
NP Juan le=habl-a a María.
John her.DAT=speak-s to Mary
“John is talking to Mary.”
*Juan habl-a a María.
John speak-s to Mary
“John is talking to Mary.”
Juan le=habl-a a ella.
John her.DAT=speak-s to she
“John is talking to her.”
*Juan habl-a a ella.
John speak-s to she
“John is talking to her.”
indirect objects
Pron
Juan le=habla.
John her.DAT=speak-s
“John is talking to her.”
Figure 7: Head-marking vs. dependent-marking in Spanish objects
The tendency in Spanish is clearly towards an increase of clausal head-
marking and some varieties are already much further advanced in this than
others. Kaiser (1992) reports:
In some varieties of Spanish, in particular those from the Rio de la Plata region in
Argentina and Uruguay, the object clitic seems to be possible and even obligatory also
in those cases where the co-referent direct object NP carries the semantic feature
[+animate] and is therefore preceded by the preposition a (translated from Kaiser
1992:57).
In these dialects, Juan la conoce a María is completely grammaticised and
we get the full head-marking situation!14
While it is already a noteworthy fact that Romance languages should
exhibit clausal head-marking traits at all, the general typological importance of
the shift depends heavily on the degree to which each language actually makes
use of them. In an empirical study on the related topic of Romance object
14 Van Valin confirms this and acknowledges the existence of Latin American varieties of
Spanish in which clitic-doubling has come to be obligatory (personal communication, 2002).
Jaeggli (1982:14) reports that, in River Plate Spanish, clitic doubling is impossible with a non-
pronominal, inanimate d.o., preferred with animate, specific d.o., highly preferred with a non-
pronominal goal i.o., and obligatory in all other cases.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
203
conjugation, Koch (1993, 1994) has undertaken the task of evaluating the
statistic relevance of ‘clitic-doubling’ in French, Italian, and Spanish, the main
results of which I have conflated into the synoptic table in Fig. 8. It shows a
clear ranking, with French being most reluctant to adopt head-marking, Italian
as a case in which both strategies seem to coexist on all levels, and Spanish
with a clear-cut – and in the case of object pronouns already consummated –
tendency towards clausal head-marking.
Further evidence in favour of a head-marking analysis of Romance objects
comes from the classical examples of segmented sentences which have been
particularly popular in the context of the debate over object conjugation in
Romance and have quite justly (and following Queneau) led Koch (1993) to a
comparison with Chinook and other Amerindian languages:
(23) a. [Il=la=lui=a=donnée], à Jean, son père
he.NOM=her.ACC=him.DAT=has given to John his father
sa moto (Tesnière 1959:175).
POSS motorcycle.F.SG
“His father gave John his motorcycle.”
b. Io in quella casa i piedi [non ce=li=metterò mai] (cf. Koch 1993:172).15
I in that house the feet NEG there=them=put.1SG.FUT never
“I shall never set foot in that house.”
In languages without morphological case, the fundamental syntactic relations
have to be encoded sequentially, with the ensuing consequence that the order of
elements in the sentence becomes increasingly fixed and can no longer be used
for discourse-pragmatic functions. While French exhibits a vast array of
syntactic constructions to achieve this, of which only some exploit head-
marking, these clearly seem to be the preferred structures in Spanish and
Catalan. They provide a perfect solution to the discourse pragmatic problem of
underlining the informational profile of an utterance by detaching lexical from
grammatical information. Syntax proper is condensed into the morphosyntax of
the clitic group, and the noun phrases, which carry the lexical information, are
set free from the sequential restrictions of the core sentence and can now be
used in the sentence periphery, where they can be ordered in any way the
distribution of old and new information may require. Even if French shows
certain head-marking tendencies as well, they have never come to be fully
grammaticised as in Spanish.
15 These traditional examples are obviously and admittedly fabricated by linguists and rather
unlikely to ever appear in spontaneous discourse, as they violate the Preferred Argument
Structure constraint “Avoid more than one lexical core argument” (Du Bois 2003:34ff.). They
are nevertheless grammatically possible utterances in these languages and illustrate their
capability of representing all core arguments as affixes on the verb while converting the
corresponding NPs into appositions outside the sentence core.

HANS-INGO RADATZ
204
French Italian Spanish
[–hum] 95,0
VO = object follows verb;
no clitic
92,0
85,5
[+hum] 89,5
[–hum] 0,7
OV = object precedes verb;
no clitic
1,7
1,8
[+hum] 0,0
[–hum] 0,8
oVO = object follows verb;
pleonastic clitic precedes verb 4,0
4,4
[+hum] 7,9
[–hum] 3,6
realised as an NP
OoV = object precedes verb;
pleonastic clitic precedes verb
3,0
6,4
[+hum] 2,6
VO = object follows verb;
no clitic
25,0
0,0
OV = object precedes verb;
no clitic
4,2
0,0
oVO = object follows verb;
pleonastic clitic precedes verb 25,0
O = Direct object
realised as a Pronoun
OoV = object precedes verb;
pleonastic clitic precedes verb
[no data]
46,0
100,0
VI = object follows verb;
no clitic
95,0
70,1
40,0
IV = object precedes verb;
no clitic
0,0
2,4
0,0
iVI = object follows verb;
pleonastic clitic precedes verb 0,0
14,0
41,8
realised as an NP
IiV = object precedes verb;
pleonastic clitic precedes verb
5,0
8,7
18,2
VI = object follows verb;
no clitic
5,0
13,7
0,0
IV = object precedes verb;
no clitic
0,0
15,7
0,0
iVI = object follows verb;
pleonastic clitic precedes verb 23,5
I = Indirect object
Realised as a Pronoun
IiV = object precedes verb;
pleonastic clitic precedes verb
100,0
45,0
100,0
All numbers are percentages. For details about the corpus used cf. Koch (1993:173). O =
direct object, o = direct object clitic; I = indirect object, i = indirect object clitic; V = verb;
Pron = emphatic, free pronoun; [+/- hum] = object with a human vs. non-human referent.
Figure 8: Percentage of pleonastic object clitics in Romance languages
(data from Koch 1993)

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
205
This is mainly due to the early development of grammaticalised cleft-
constructions in French which constituted an important alternative way of
expressing theme-rheme structures, and while clefting never really caught on in
Spanish, it eventually won out over clausal head-marking in French. The
intermediate position of Italian as seen in the statistics of Fig. 8 is most
probably due to the fact that clefting is more important in Italian than in
Spanish but at the same time less important than in French.
6. Language change: intermediary stages of a change from dependent-
marking towards head-marking
The last 1000 years have seen Spanish pronouns change from free forms
similar to German pronouns towards increasingly bound forms. In a convincing
article from 1990, Rini has tried to reconstruct and date the process in which
Old Spanish clitic pronouns developed into the grammaticalised verbal affixes
of Modern Spanish (Rini 1990). He does this by analysing the history of three
related syntactic phenomena, namely 1. Interpolation, 2. Word order, and 3.
Pronominal duplication.
Old Spanish clitics, though already phonologically dependent, had not yet
developed into bound morphemes (cf. Rivero 1986:775). Their position with
respect to their host was not yet as fixed as it is in Modern Spanish (e.g. él
díxolo, él lo dixo, lo él dixo “he said it” all can be found in the same document)
and the so-called ‘interpolation’ of material between the clitic and its host was
quite frequent:
(24) quien te algo prometiere (Corbacho 145)
who pron.2SG.DAT something should_promise
“whoever should promise you something”
para lo mejor conplir que lo ella non mando (Corbacho 91)16
for it better accomplish that it her not order.1SG.PRS.IND
“in order to accomplish it better that I shall not order her to do it
While clitic-doubling did already occur in Old Spanish, it was far from
being as frequent as today and it was never obligatory. Thus, clitic-doubled
examples like:
(25) E tu as me tollido a mí un capellano (Berceo, 229d)17
and you.SG have me taken from me one priest
“And you have taken away a priest from me.”
16 Arcipreste de Talavera o Corbacho, ed. Gerli, Michael, Madrid: Cátedra, 1979, quoted from
Rini (1990:357).
17 Milagros de Nuestra Señora, ed. García Solalinde, Antonio, Madrid: Espasa-Calpe, 1922,
quoted from Rini (1990:361).

HANS-INGO RADATZ
206
occur freely alongside non-doubled structures in the same text:
(26) A ti lo digo, nuera (Corbacho ch. 38, p. 144)18
to you.SG it I_say daughter-in-law
“I say it to you, daughter-in-law.”
But even where clitic-doubling did occur, as in (25), the construction is
syntactically quite different from the corresponding structures in the modern
language:
In Modern Spanish [... t]he clitic is indispensable, which suggests that it is the basic
pronoun and the tonic is the additional or redundant form, which appears for emphasis.
There is no reason to believe, however, that in Old Spanish the emphatic constructions
were of this same nature as regards the question of which was the basic and which was
the redundant or additional element. In fact, the evidence provided by a diachronic
perspective seems to suggest just the opposite: that the tonic pronoun, not the clitic,
was the basic element of the emphatic constructions, and the clitic was additional or
redundant when the two forms occurred together (Rini 1990:360).
What Rini found out was that, with respect to all three aforementioned
parameters, Spanish seems to have reached the modern type of distribution at
roughly the same time for each one of them, i.e. around the beginning of the
17th century.19 By then, no further examples of interpolation appear in the
texts, pronominal clitics essentially become fixed to the left of finite and to the
right of non-finite verbal forms and clitic-doubling becomes obligatory with
emphatic pronouns. The fact that all three changes reach their conclusion more
or less simultaneously is a strong indication that they have to be seen as three
facets of one and the same grammaticalisation process which at this time must
have reached a new stage: the stage of affixes.
Most interesting, however, is Rini’s observation regarding clitic-doubling.
While in Old Spanish clitic-doubling the free pronouns acted as indispensable
head of the construction and the clitics were just a redundant reinforcement, the
respective roles can be shown to have changed completely in what Rini aptly
dubs a ‘cephalic shift’ (Rini 1990:361). From the beginning of the 17th century
on it is no longer the free pronouns but rather the clitic affixes that constitute
the head of the construction while free pronouns may but need not be added for
emphasis.
Put in the terms of Bybee (2005), the high token frequency of doubling
clitics leads to an increased entrenchment of object clitics in the morpho-
syntactic structure and ultimately “to the grammaticization of the new
18 Quoted from Rini (1990:360).
19 Other authors date the point for completed grammaticalisation much earlier; thus, for Barry,
this state is reached from the thirteenth century onwards (cf. e.g. Barry 1987:219). For further
empirical detail cf. Eberenz (2000:175-208).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
207
construction and the creation of grammatical morphemes and changes in
constituency” (Bybee 2005:6). Object-clitics were first phonologically reduced
in that they lost their capability of being stressed; from now on, stress could
only be placed on the accompanying free pronoun or full NP. As verbs and
clitics were used together ever more frequently, they came to be perceived as a
unit in which the clitics were reinterpreted as obligatory grammatical
morphemes. Because free object pronouns lost their obligatory status at about
the same time, the new verbal markers began to function as cross-reference
morphemes and were now treated as the actual realisations of objects.
In modern Spanish, (1) clitic doubling has become obligatory with any
(emphatic) object pronoun, (2) is regularly used with indirect object NPs, and
sometimes occurs even (3) with direct object NPs (cf. Fig. 8). In all these cases,
the verbal affixes co-occur (and morphologically cross-reference with) their
respective lexical or pronominal appositive phrases. With nominal objects the
extent and obligatoriness of clitic-doubling seems to be in constant flux, and
any researcher adducing examples must provide exact information as to which
dialect or variety of Spanish is envisaged – all clear indications of a still
ongoing process of change.
The cephalic shift created a new type of morphology in Spanish which
would best be described as ‘cross-reference’. Obviously, there remain large
areas in Spanish syntax where cross-reference is not (yet) obligatory or even
possible. On the other hand, there are large areas (mainly with emphatic object
pronouns) where it has already been fully grammaticalised. While Spanish may
still be far from Basque in this respect, it is nevertheless clearly moving in the
same direction.
Traditionally, grammaticalisation theory envisages the typical development
of free (emphatic) pronouns as ‘emphatic personal pronoun > clitic pronoun >
agreement affix’ (cf. e.g. Croft 2000:157). In the case of Spanish we can see
that this is at least not the only possible development and that we are again
faced with the type of shortsighted exclusion of languages with (clausal) head-
marking as deplored by Nichols (1986:115f.; see quote above). While this type
of grammaticalisation may lead to the establishment of a new type of
agreement affixes, it may also, alternatively, create cross-reference affixes thus
turning a language into one with clausal head-marking. For centuries, Spanish
speakers have chosen head-marking strategies in order to resolve discourse-
pragmatic problems of information structure and these preferred but still
optional patterns are now turning into structural patterns of grammar in just the
way Haspelmath (1999:193) envisages it.
As “grammars code best what speakers do most” (Du Bois 1987:811),
Spanish is accepting ever more clausal head-marking as fixed, obligatory
constructions into its syntax and we may be witnessing the emergence of a
more and more consistent head-marking syntax. What has frequently been
described as a change in which clitic pronouns are grammaticalised into

HANS-INGO RADATZ
208
agreement-morphemes must then be seen as an even more radical change, not
into agreement morphemes but into cross-reference affixes.
The difference between French and Spanish in this respect is more of
degree than of principle, as all Romance languages possess at least some of
these head-marking constructions. Clitic-doubled verbal complexes like il la lui
a donnée (or sp. se lo ha dado, cat. la hi ha donat etc., all meaning “he has
given it to him”) are indeed quite analogous to the corresponding structures in
Basque:
(27) Bere aita-ki Jon-ik bere
POSS.3SG father-DET.ERG.SG Jon-DAT.SG POSS.3SG
moto-aj eman dj-i-ok-Øi .
motorcycle-DET.ABS.SG give ABS.3SG-AUXhave.PRS.IND-DAT.3SG-ERG.3SG.
“His father gave John his motor-bike.”
The following examples show the possibilities for expressing differences in
functional sentence perspective and the similarity of the French and the Basque
solutions:
(28) a. Il=la=lui=a donnée, à Jean, son père, sa moto.
he=it=him=has given to John his father his bike
b. Eman dio Joni bere aitak bere motoa.
given he_has_it_to_him John his father his bike
“His father gave John his motor-bike.”
(29) a. À Jean, il=la=lui=a donnée, son père, sa moto.
to John he=it=him=has given his father his bike
b. Joni eman dio bere aitak bere motoa.
to_John given he_has_it_to_him his father his bike
“His father gave John his motor-bike.”
(30) a. Sa moto, il=la=lui=a donnée, à Jean, son père.
his bike he=it=him=has given to John his father
b. Bere motoa eman dio Joni bere aitak.
his bike given he_has_it_to_him to_John his father
“His father gave John his motor-bike.”
The main difference here lies in the fact that this cross-reference marking
with appositional complements is the obligatory normal case in Basque while it
is a stylistically marked exceptional construction, the appearance of which is
bound to a discourse-pragmatic motivation. The French examples are all
instances of segmented phrases, a discourse pragmatic repair strategy which
justifies the apparition of the object clitics in French. With all lexical
arguments in their conventional slots, clitic doubling becomes unacceptable in
French:

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
209
(31) a. Son père [*il=la=lui=] a donné sa moto à Jean.
his father he=it=him= has given his bike to John
b. Bere aitak eman dio bere motoa Joni.
his father given he_has_it_to_him his bike to_John
“His father gave John his motor-bike.”
In Spanish, on the other hand, the corresponding sentence may not be the
preferred one, but is at least grammatically possible:
c. Su padre se=la=ha dado la moto a Juan.
his father him=it=has given the bike to John
On this basis we can tentatively formulate a typological hierarchy, where
the Romance languages can be situated at an intermediate level:
Type example Syntactic consequences
Head-marking
Abkhaz – all arguments realised polysynthetically on the finite verb;
– NPs are facultative appositions without any case marking;
– no deictically unmarked object-pronouns (if at all, only
demonstratives).
Double marking
Basque – all arguments realised polysynthetically on the finite verb (head-
marking);
– NPs are facultative appositions with case marking (additional,
redundant dependent-marking);
– no deictically unmarked object-pronouns (if at all, only
demonstratives).
Split marking
Spanish,
Catalan,
European
Portuguese
mixed;
– Subject head-marking on the verb with appositional subject-NPs
or -pronouns respectively;
Objects are realised
1. as affixes on the verb,
2. as affixes + apposition and
3. as NPs (but not as pronouns!)
Dependent-
marking
German – all arguments realised as NPs pronouns with case marking;
– only head-marking feature is the redundant subject agreement on
the verb.
Figure 9: Head- vs. dependent-marking-hierarchy

HANS-INGO RADATZ
210
7. Conclusion
Syntax, morphology, and phonology are usually so inextricably intertwined
in diachrony that trying to establish exactly where a change originated may
come down to the classical hen-egg problem. As Romance languages lost their
case inflections (morphological phenomenon induced by phonological
processes), syntactic relationships increasingly came to be encoded in the
sequencing of constituents (syntax). As this sequencing became ever more
rigid, new focusing techniques had to arise, as constituents could no longer be
arranged freely to the discourse needs of language users (loss of word stress by
the 16th century may have aggravated this problem in the case of French). The
new techniques have in common that they place a pronominal element (relative
or personal pronoun / clitic) in its canonical place such as to leave sequencing
unaltered in the sentence core. The corresponding lexical NPs are thus freed to
be placed in the periphery where they may be focused according to the
conversational needs. The main solutions along these lines are clefts and
pronominal reprise (or clitic-doubling), and while French and Spanish both use
clefting as well as pronominal reprise techniques, the former are the typical
solution in French, whereas Spanish largely depends on the latter.
As Dufter (this volume) has shown, the success of c’est-clefts in French let
this constructions become so entrenched that their initial motivation as a
focusing device didn’t remain their only function; with the advent of the
‘informative-presupposition cleft sentence’ in the 16th century, French clefts
could now also carry given or thematic information. What Dufter calls clefting
‘beyond necessity’ can analogously also be observed for Spanish clitic-
constructions, the history of which shows a continuous increase of uses that
clearly go ‘beyond necessity’, and what once started out as individual speakers’
discourse strategies has, through the centuries, lost its dependence on the
original motivations and has now been integrated into the very syntax of the
simple sentence in Spanish. The typological drift from a conventional central
IE language with agreement towards the ‘exotic’ type of clausal head-marking
with cross-reference morphology can thus be argued to derive ultimately from
new discourse strategies as a reaction to a syntactic change, namely the
progressive fixation of word order towards rigid SVO (even if Langobardi’s
Inertia Theory would have us expect otherwise).
The cross-reference analysis of Spanish clitics is not a mere terminological
reformulation of the traditional analyses. The corresponding language type has
long been well established and well described, although it has, until recently,
received very little attention within IE linguistics and Romance philology in
particular. If it can be shown that Spanish syntax is in a central aspect moving
away from its IE sister languages like French, English, or German to become
more like Basque, Abkhaz, or Lakhota, this is in itself an important typological
insight. For the syntactic analysis of Spanish, the cross-reference hypothesis
implies that, in many cases, lexical NPs as well as pronouns must no longer be

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
211
seen as the subject and object arguments, but rather as appositional phrases
outside the sentence-core forming ‘mentions’ together with cross-reference
markers on the verb. These mentions are discontinuous realisations of
arguments in which the verbal markers are obligatory (and therefore the heads),
while the lexical NPs or pronouns are syntactically facultative. Spanish does
therefore neither tolerate null-subjects nor null-objects. A further consequence
of this typological drift is that we can expect clitic-doubling to become ever
more obligatory and to be carried over to further new domains.
8. Closing remarks: no “object-conjugation” in Romance (or elsewhere)!
The phenomena discussed here are widely identical with those that have
been invoked in the debate among Romance linguists over a supposed tendency
towards ‘object-conjugation’ in these languages.20 It should be clear from the
above, however, that the notion of object-conjugation is misleading here as it
evokes the notion of conjugation and with it the notion of agreement. As we
have seen, agreement proper is only to be found in pure dependent-marking
languages whereas the phenomena in question represent a typological shift
precisely in the opposite direction, namely towards head-marking. While
allusions to Basque and Hungarian are present in all contributions to the
debate, the syntactic facts of these languages are consistently presented within a
terminological framework developed for the description of dependent-marking
Indo-European languages, treating Basque and Hungarian cross-reference
elements as agreement-markers. This type of ‘agreement’ is then projected onto
the Romance material. We must therefore reject the notion of ‘object-
conjugation’ altogether, not only for the Romance phenomena, but also for the
languages in which it purportedly is most fully developed, namely Basque and
Hungarian. By failing to differentiate between agreement and cross-reference,
the theory of Romance object-conjugation continues to look for the arguments
where they are not to be found – in the noun phrases and strong pronouns – and
perpetuates a strict distinction between a dependent agreement-marker and its
nominal referent. In so doing they miss the very nature of this ongoing process
of change: the emergence of cross-reference structures and clausal head-
marking.
REFERENCES
Ashby, William J. 1977. Clitic inflection in French. Amsterdam: Rodopi.
20 Cf. de la Grasserie (1889:278f.), Lenz (1920:81-3), Heger (1966), Rothe (1966), Llorente &
Mondéjar (1974), Ashby (1977:73-78), Bossong (1980, 1982), Berretta (1989), Schwegler
(1990:101-3), and Koch (1993, 1994).

HANS-INGO RADATZ
212
Aoun, Joseph. 1993. “The syntax of clitic-doubled arguments”. International
Journal of Basque Linguistics and Philology 27. 709-730.
Barry, Anita K. 1987. “Clitic pronoun position in thirtheenth-century Spanish”.
Hispanic Review 55. 213-220.
Beaumont, Jean-Charles & Ramón Lazkano. 1998. Iniciación al euskara –
Euskara baturako hastapenak. Chennevières-sur-Marne: Assimil.
Berretta, Monica. 1989. “Tracce di coniugazione oggettiva in italiano”.
L'italiano tra le lingue romanze, ed. by F. Foreti et al., 125-150. Roma:
Bulzoni.
Bloomfield, Leonard. 1933. Language, Chicago: University of Chicago Press.
Bossong, Georg. 1980. “Aktantenfunktionen im romanischen Verbalsystem”.
Zeitschrift für romanische Philologie 96. 1-22.
----------. 1982. “Historische Sprachwissenschaft und empirische
Universalienforschung”. Romanistisches Jahrbuch 33. 17-51.
----------. 1991. “Differential object marking in Romance and beyond”. New
Analyses in Romance Linguistics, ed. by Dieter Wanner & Douglas A.
Kibbee, 143-170. Amsterdam & Philadelphia: John Benjamins.
Bybee, Joan. 2005. “The Impact of Use on Representation: Grammar Is Usage
and Usage Is Grammar”. LSA 2005 Presidential Address. Online at
http://www.unm.edu/~jbybee/Bybee plenary.PDF.
Corbett, Greville G. 1994. “Agreement”. Encyclopedia of Language and
Linguistics Vol. 1, 54-60. Oxford: Pergamon Press.
Croft, William. 2000. Explaining Language Change. London & New York:
Longman.
Grasserie, Raoul de la. 1889. “De la conjugaison objective”. MSLP 5. 268-300.
Du Bois, John W. 1987. “The discourse basis of ergativity”. Language 63. 805-
855.
----------. 2003. “Argument structure: Grammar in use”. Preferred Argument
Structure, ed. by John W. Du Bois, Lorraine E. Kumpf & William J.
Ashby, 11-60. Amsterdam & Philadelphia: John Benjamins.
Eberenz, Rolf. 2000. “El pronombre personal átono con función duplicadora”.
El español en el otoño de la edad media, ed. by Rolf Eberenz, 175-208.
Madrid: Gredos.
Franco, Jon. 1991. “Spanish object clitics as verbal agreement morphemes”.
MIT Working Papers in Linguistics 14. 99-114.
----------. 2000. “Agreement as a Continuum”. Clitic Phenomena in European
Languages, ed. by Frits Beukema & Marcel den Dikken, 147-189.
Amsterdam & Philadelphia: John Benjamins.
Goldbach, Maria. 1999. Spezifische und arbiträre leere Objekte. Frankfurt am
Main: Vervuert.
Gómez, Ricardo & Koldo Sainz. 1995. “On the Origin of the Finite Forms of
the Basque Verb”. Towards a History of the Basque Language, ed. by José
Ignacio Hualde, Joseba A. Lakarra & Robert Lawrence Trask, 235-274.
Amsterdam & Philadelphia: John Benjamins.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE
213
Haspelmath, Martin. 1999. “Optimality and diachronic adaptation”. Zeitschrift
für Sprachwissenschaft 18. 180-205.
Heger, Klaus. 1966. “La conjugaison objective en français et en espagnol”.
Langages 3. 19-39.
Helmbrecht, Johannes. 2001. “Head-marking vs. dependent-marking
languages”. Language Typology and Language Universals, ed. by Martin
Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible, vol.
2, 1424-1432. Berlin & New York: Walter de Gruyter.
Jaeggli, Osvaldo. 1982. Topics in Romance Syntax. Dordrecht & Cinnaminson:
Foris.
Kailuweit, Rolf. 2005. “‘Pro-drop’, Kongruenz und ‘optimale Klitika’ – Ein
Beschreibungsansatz im Rahmen der Role and Reference Grammar”.
Morphologie und romanistische Sprachwissenschaft, ed. by Carmen
Kelling, Judith Meinschaefer & Katrin Mutz, 87-110. [http://www.ub.uni-
konstanz.de/kops/volltexte/2006/1812/pdf/AP_120.pdf].
Kaiser, Georg. 1992. Die klitischen Personalpronomina im Französischen und
Portugiesischen. Eine synchronische und diachronische Analyse. Frankfurt
am Main: Vervuert.
Koch, Peter. 1993. “Le ‘chinook’ roman face à l’empirie. Y a-t-il une
conjugaison objective en français, en italien et en espagnol et une
conjugaison subjective prédéterminante en français?”. Actes du XXe
Congrès International de Linguistique et Philologie Romanes, Université
de Zurich (6-11 avril 1992), ed. by Gerold Hilty, vol. 3, 169-190. Tübingen
& Basel: Francke.
----------. 1994. “L'italiano va verso una coniugazione oggettiva?”.
Sprachprognostik und das ‘italiano di domani’. Prospettive per una
linguistica ‘prognostica’, ed. by Günter Holtus & Edgar Radtke, 175-194.
Tübingen: Narr.
Landa, Alazne & Jon Franco. 1996. “Two issues in null objects in Basque
Spanish”. Grammatical Theory and Romance Languages, ed. by Karen
Zagona, 159-168. Amsterdam & Philadelphia: John Benjamins.
Lenz, Rodolfo. 1920. La oración y sus partes. Estudios de gramática general y
castellana. Madrid: Revista de Filología Española.
Llorente Maldonado de Guevara, Antonio & Mondéjar, José. 1974. “La
conjugación objetiva en español”. Revista Española de Lingüística 4. 1-60.
Nichols, Johanna. 1986. “Head-marking and dependent-marking grammar”.
Language 62. 56-119.
----------. 1992. Linguistic Diversity in Space and Time, Chicago & London:
The University of Chicago Press.
Rini, Joel. 1990. “Dating the grammaticalization of the Spanish clitic
pronoun”. Zeitschrift für Romanische Philologie 106. 354-370.
Rivero, María Luisa. 1986. “Parameters in the typology of clitics in Romance,
and Old Spanish”. Language 62. 774-807.

HANS-INGO RADATZ
214
Rothe, Wolfgang. 1966. “Romanische Objektkonjugation”. Romanische
Forschungen 78. 530-547.
Saltarelli, Mario. 1988. Basque, London: Croom Helm.
Schwegler, Armin. 1990. Analyticity and Syntheticity. A diachronic perspective
with special reference to Romance languages. Berlin & New York: Mouton
de Gruyter.
Schwerteck, Hans. 1984. Strukturen baskischer Verbformen. Wiesbaden: Franz
Steiner.
Tesnière, Lucien. 1959. Éléments de syntaxe structurale. Paris: Éditions
Klincksieck.
Van Valin, Robert D. Jr. 2005. A summary of Role and reference Grammar.
[http://linguistics.buffalo.edu/research/rrg/RGsummary.pdf].
Zubiri, Ilari. 52000. Gramática didáctica del euskera. Bilbo: Didaktiker.