Professional Documents
Culture Documents
2012 Comrie & Estrada-Fernández Relative Clauses in Languages of The Americas - A Typological Overview-John Benjamins
2012 Comrie & Estrada-Fernández Relative Clauses in Languages of The Americas - A Typological Overview-John Benjamins
Editors
Spike Gildea Fernando Zúñiga
University of Oregon University of Zurich
Editorial Board
Balthasar Bickel John Haiman Marianne Mithun
Zurich St Paul Santa Barbara
Bernard Comrie Martin Haspelmath Doris L. Payne
Leipzig / Santa Barbara Leipzig Eugene, OR
Denis Creissels Bernd Heine Franz Plank
Lyon Köln Konstanz
William Croft Paul J. Hopper Dan I. Slobin
Albuquerque Pittsburgh Berkeley
Nicholas Evans Andrej A. Kibrik Sandra A. Thompson
Canberra Moscow Santa Barbara
Carol Genetti František Lichtenberk
Santa Barbara Auckland
Volume 102
Relative Clauses in Languages of the Americas. A typological overview
Edited by Bernard Comrie and Zarina Estrada-Fernández
Relative Clauses in Languages
of the Americas
A typological overview
Edited by
Bernard Comrie
Max Planck Institute for Evolutionary Anthropology
Zarina Estrada-Fernández
University of Sonora
For a number of years the Seminar on Linguistic Complexity held at the University
of Sonora, Hermosillo, Mexico, has brought together a number of general linguists
and specialists in particular languages to discuss issues relating to the identifi-
cation, analysis, and genesis of linguistic complexity. Given the emphasis of the
University of Sonora’s Department of Linguistics on indigenous languages of the
Americas, and especially of Mexico, this has been the main areal thrust of the proj-
ect. A major emphasis in discussions of complexity within the project has been on
subordination and related types of clause combining, as one aspect of syntactic
complexity going beyond the structure of the simple sentence. One type of sub-
ordinate clause that has given rise to particular interest, both within the Sonora
project and well beyond it, is the structure and diachronic development of relative
clauses, and it is to this particular area that the contributions in the present volume
are devoted, authored by contributors to the Seminar and associated scientists.
The first three articles are concerned with general questions of relative clause
structure and diachrony, approached from the viewpoint of a cross-linguistic
perspective, paying due attention to the differences among languages while also
trying to draw conclusions of general validity.
Givón’s chapter provides a clear illustration of this methodology, with an
emphasis on the diachronic issue of the genesis of relative clauses. The author
shows how, on the one hand, relative clauses can develop from increasingly mor-
phosyntactically explicit indication of the links between erstwhile paratactically
juxtaposed clauses, thus directly addressing the issue of the genesis not only of rel-
ative clauses but also of linguistic complexity. The other source considered, namely
nominalizations, draws attention to an important but often neglected alternative
source of relative clauses, though interestingly one that is perhaps less obviously
a case of increasing complexity, since sentential nominalizations are themselves
complex structures, absent as distinct structures from the grammars of many
languages; such relative clauses may originally have been noun phrases in apposi-
tion to their notional head, giving rise to headed relative clauses, or indeed single
noun phrases, giving rise to headless relatives.
Kuteva and Comrie again take up the issue of the development of complexity,
this time on the basis of a particular set of languages, namely pidgins and creoles.
Bernard Comrie & Zarina Estrada-Fernández
They show that in terms of the number of overt morphosyntactic markers of rela-
tive clause structure, pidgins and creoles typically just have a single marker, in a
sense the functionally optimal solution, while other languages frequently use more
than one marker (with the current record-holder having five), and also frequently
use none. They argue that at least in this respect pidgin and creole languages,
with short histories, in general reflect a less complex structure, while languages
with longer histories have had time to develop more complex structures. Interest-
ingly, pidgins and creoles differ in this respect from other kinds of intense contact
between languages, where familiarity with two languages often leads to complex
structures that combine features of the two input languages; this suggests that pid-
gins and creoles may indeed reflect situations where the creators of the pidgin/
creole had only restricted access to the superstrate language.
Finally, Van Valin argues that any general theory of the syntax-semantics
interface must be able to answer to the full range of cross-linguistic variation in
relative clause constructions. Many existing theories are heavily rooted in the
often typologically rare and areally restricted properties of relative clauses in the
major European languages, and few theories deal elegantly with internally-headed
relative clauses, although these are not infrequent cross-linguistically and seem
particularly thick on the ground in indigenous languages of the Americas. (More
generally, it seems that languages of the Americas have a greater aversion, whether
in their grammars or in discourse, to externally-headed relative clauses than do
languages of most other parts of the world. Indigenous languages of the Ameri-
cas have a high incidence of internally-headed and headless relative clauses.) Van
Valin argues further that Role and Reference Grammar provides just the kind of
theory of the syntax-semantics interface that is adequate for the analysis of such
cross-linguistic variation.
The other ten chapters deal with individual languages of the Americas, rang-
ing roughly from north to south from Tuscarora (Iroquoian) through Northern
Paiute, Pima Bajo, Yaqui (all Uto-Aztecan), Seri (?Hokan), Yucatec Maya (Mayan)
to Hup (Nadahup), Gavião (Tupian), and Toba (Guaycuruan). (The fact that nine
languages are covered in ten chapters stems from the fact that two chapters are
devoted to Yaqui, from rather different perspectives.) The following discussion,
though organized overtly by language, is also organized conceptually by topic.
Álvarez González’s study of Yaqui emphasizes the nature of relative clauses in
that language in their relation to nominalizations, arguing indeed that so-called
relative clauses in Yaqui are nominalizations. (An alternative formulation might
be that they have the function of relative clauses, but the structure of nominaliza-
tions.) Different kinds of nominalizations are used depending, for instance, on the
position relativized, thus giving rise to an intricate interaction of nominal versus
verbal properties and other aspects of relative clause structure.
Introduction
T. Givón
Linguistics Department University of Oregon
and White Cloud Ranch Ignacio, Colorado
1. Introduction
1.1 Background
In a companion paper (Givón 2009b) I suggested that the diachronic rise of
complex verb phrases proceeds through the following general steps, in order:
(1) General diachronic trend of complex-VP formation:
a. Parataxis: The two clauses are packed under separate intonation
contours.
* I am indebted to Tania Kuteva for proofing the German data, and to Matt Shibatani for
comments on the history of Japanese REL-clauses. The materials presented here can also be
found in Givón (2009a).
T. Givón
I further suggested that this general trend overrides the considerable typological
variation found in the diachrony of complex VPs, so that both major typological
pathways, as in (2) below, still follow the same general trend in (1).
(2) Two main pathways to clause union:
a. The clause-chaining pathway
b. The nominalized v-comp pathway
Of these, method (i) is quite reliable, but the historical records often skip crucial
intermediate stages and variants. They are, typically, edited written records,
whereas diachronic change takes place, overwhelmingly, in the spoken language.
More to the point, for many languages such records do not exist. Method (ii) is the
sweetest for elucidating the detailed mechanisms of change. And it is sweeter yet
1. Lexis within the condensed complex clause can lead, eventually to the rise of new lexical
words. In the case of complex VPs (main verb plus a COMP clause), the product is a new
lexical verb. In the case of complex NPs (head noun plus a REL-clause), the product is a
new lexical noun.
Toward a diachronic typology of relative clause
when combined with (i). But you have to catch the language at the right stage,
which is largely a game of chance. Method (iii) is bold, speculative and theory
dependent (Givón 2000), and thus should be practiced with care. However, often
it is the only method available. In this paper I have attempted to avail myself of all
three, relying more heavily, of necessity, on a mix of (ii) and (iii).
c. Embedded rel-clause:
n ye cε min ye [Ø] ye muru san.
I past man rel see past knife buy
‘The man that I saw bought the knife’.
(Hist.: ‘I saw that man [,] [he] bought the knife’.)
One REL-clause type, which Carlson (1994; pp. 513–514) calls “semi-embedding”,
represents the beginning of a syntactic, embedded, pattern:
In (7a), a chunk of the main clause (‘Then they…’) is given before the
pre-posed REL-clause. That chunk is then recapitulated in the full main clause
in (7b), where the co-referent noun is marked with an emphatic resumptive
pronoun.
In light of what we have seen in Bambara and Supyire, we can now
re-interpret the Japanese zero-marking relativization strategy. Japanese is a well-
documented clause-chaining language, although perhaps not the most t ypical
one. It is also a zero-anaphora language. The condensed s yntactic v ersions of the
embedded REL-clauses are the ones normally cited. What is u sually not men-
tioned is that the mere insertion of an intonation break would render the com-
plex s yntactic construction a viable paratactic one, albeit somewhat ungainly
with the backward pronominalization (though one could easily interpret it as
an apositive paratactic construction). Thus (Katsue Akiba, i.p.c.):
T. Givón
Like Bambara, Japanese has a rather meager marking machinery for relativiza-
tion, utilizing whatever anaphoric pronouns and demonstratives are available
from clause chaining. The only change between the paratactic and syntactic
version of the construction is intonational: (i) Two clausal intonation con-
tours merge into one; and (ii) (in Bambara) the erstwhile stressed demonstra-
tive ‘that’ becomes de-stressed/lower-tone in the embedded REL-clause, and
thus now part of the new REL-clause morphology. This is very much in line
with Mithun’s (2006) observation about embedded clauses in Mohawk, where a
merged intonation contour is the only reliable indication of embedding.2 And
2. In a sweeping paper, Everett (2005) has claimed that ‘his’ Amazonian language, Pirahã,
has no embedded clauses. In support he cites Pirahã clause-chaining constructions that
Toward a diachronic typology of relative clause
The conjunction may be dropped from the first clause, yielding a more emphatic
focus:
(10) ku-iš pa-apreez-zi, nu apaa-aš-pat gín
rel-nom be-impure-3sg conj that-one-nom-prt shekel/acc
ku.babbar paa-i
silver give-3sg
‘Whoever is impure, that very man shall give (three?) silver shekels’.
‘function as’ REL-clause, very much like Bambara, but are not embedded. Everett suggests
that all such clauses are separated by an intonation break from their main clause. As further
support for his claim of non-embedding, he cites other clause-chaining serial-verb languages
(Pawley 1987; Matisoff 1969). At face value, this seems to be an early stage of grammaticaliza-
tion (Givón 1991b, 2009b; Mithun 2006, 2007). Only a text-distribution study of intonation
contours would tell whether Pirahã has already advanced beyond the earliest paratactic stage
(like Bambara) or not.
T. Givón
To drive home how typical a clause-chaining pattern this is, consider the following
example from Chuave (Gorokan, East Highlands, Papua-New Guinea), a clause-
chaining serial-verb language par excellence. In this language, all presuppositional
clauses, including restrictive REL-clauses, are nominalized, and could only come
at the chain-initial position (Thurman 1978):
What Robert (2006) argues about the presence vs. absence of the conjunction in
Hittite, and its connection to the diachronic evolution of Hittite REL-clauses, is
worth citing:
Robert notes that there is a strong association between the presence of a conjunction
in the main (‘resumptive’) clause and the presence of an explicit anaphoric
(‘resumptive’) pronoun there. The Hittite scribes either no intonation-break
Toward a diachronic typology of relative clause
(comma) marking or else marked them at the end of every clause (as Robert does).
Still, it is fairly clear that what Robert describes is a drift from an old paratactic
clause-chaining pattern, with the main (‘resumptive’) clause marked by both a con-
junction and an anaphoric (‘resumptive’) pronoun, to a later syntactic-embedded
pattern, where both the conjunction and pronoun are dispensed with. And I think
it is a safe bet that the intonation contours merged in the process.
One may as well note, lastly, that the clause-chaining source for REL-clauses
is universal, and can be found – with a discerning eye for informal oral
discourse – in just about any spoken language. As an example, consider the
following exchange between a mother and her 2 yr. 9 months-old daughter. At
this early age, the child produces not a single bona-fide adult-like R EL-clause
(Diessel 2005), and her mother produces virtually none either during her
conversations with her daughter. But the paratactic precursor is already there,
often spread across two-person turns, as in (14) (Nina, CHILDES data-base; see
Givón 2009a):
c. Dative:
Ich kenne den Mann, dem hat Martin das
I know the/acc man, THAT/dat has Martin the/acc
Buch gegeben
book given
‘I know the man, the one that Martin gave the book to’.
(Hist.: ‘I know the man. THAT one Martin gave the book to’).
By removing the intonation break (and thus in written language the conservative
comma), de-stressing the demonstrative pronoun, and a minor adjustment to a
non-contrastive word-order, the set of non-restrictive REL-clauses in (17) can be
turned into restrictive ones. Respectively (ignoring the written German conven-
tion about commas, a relic of the non-restrictive pattern):
As in German, the demonstrative loses its stress when the paratactic n on-restrictive
REL-clause (19c) is condensed into the syntactic restrictive REL-clause (19d). So
while the source of the parenthetical clause is different, the condensation pattern –
the essence of this pathway, from parenthetical non-restrictive to restrictive – is
the same.
The naturalness of selecting the clause-type to be used as the parenthetical
(non-restrictive) portion of the paratactic construction is, roughly, that it must
topicalize the preceding co-referent (‘head’) noun. The Y-movement used in
German is certainly such a construction (Givón 2001, Chapter 15). The head-
less REL-clause of Hebrew carries the same topicalizing function (as do all
REL-clauses).
The use of stressed demonstratives is almost entirely predicted from the con-
flation of two necessary attributes of such constructions:
The stressed demonstrative is rather well suited for this function (Linde 1979),
so it is not an accident that it is distributed so widely across the typological
chasm, in the clause-chaining and verb-serializing Bambara and Hittite, as in
the more embedding and nominalizing German and Hebrew. The only lan-
guages that are less-likely to show this feature are zero-anaphora languages like
Japanese.
Toward a diachronic typology of relative clause
e. Non-restrictive rel-clause:obj:
tupuych ’uru (‘uru) mamachi
rock/s the/s (that/o) woman-gen
tuka’napu-vwan wacu-ka-n(a)...
table/o-on put-perf-nom
‘the rock, (that one) that the woman put on the table...’
(Hist.: ‘the rock, (that) of the woman’s putting on the table,...’)
f. Restrictive rel-clause: Indirect obj
tuka’napu ’uru pu-vwan mamach-i tupuychi-ci wacu-ka-n(a)...
table/s the/s rel-on woman-gen rock/o put-ant-nom
‘the table on which the woman put a rock...’
(Hist.: ‘the table of the woman’s putting the rock on’...’)
g. Non-restrictive rel-clause: Indirect obj
tuka’napu ’uru, (‘uru) pu-vwan mamach-i
table/o the/s, (that/o) rel-on woman-gen
tupuychi-ci wacu-ka-n(a)...
rock/o put-ant-nom
‘the table, (that one) on which the woman put a rock,...’
(Hist.: ‘the table, (that) of the woman’s putting the rock on...’)
Of the two nominalizing suffixes on the verb, -tu, the subject nominalized is
still used synchronically to mark lexical subject nominalizations. In combi-
nation with the old passive/perfect marker -ka-, it can also be used to mark
direct-object
(technically subject-of-passive) nominalization. Thus (Givón
1980, 1988):
(21) a. Main clause:
ta’wach ’u pѳ’ѳ-mi
man/s the/s sing-hab
‘The man writes’.
b. Subject nominalization:
’ú ta’wach pѳ’ѳ-mi-t(u) ’ura’-’ay
that/s man/s write-hab-nom be-pers
‘That man is a writer’.
c. Object (subject-of-passive) nominalization:
‘ích’-ara pѳ’ѳ-kwa-tu ’ura-’ay
this/s-be write-pass-nom be-pers
‘This is a book’.
The Ute REL-clause data fit our scenario of parataxis-to-syntaxis rather snugly.
And the non-restrictive REL-clauses still function synchronically as nominalized
clauses, as in:
Toward a diachronic typology of relative clause
(22) a. Subject:
’ú tupyuchi tuka’napu-vwan wacu-ka-t(u)
that/s rock/o table/o-on put-perf-nom
tuguvu-n ’ura-’ay
friend-my be-pers
‘That one who put the rock on the table is my friend’.
(Hist.: ‘That [putter-of-the-rock-on-the-table] is my friend’.
b. Object:
’uru mamachi tuka’napu-vwan wacu-ka-n(a)
that/o woman/gen table-o-on put-perf-nom
tupuych ’ura-’ay
rock/pred be-pers
‘What the woman put on the table is a rock’.
(Hist.: ‘That [the-woman’s-putting-on-the-table] is a rock’.)
c. Indirect object:
’uru pu-vwan mamachi tupuychi wacu-ka-n(a)...
that/o rel-on woman/gen rock/o put-ant-nom
‘That (thing) on which the woman put a rock is a table’
(Hist.: ‘That [the-woman’s-putting-the-rock-on] is a table’).
b. Place derivation:
yod-sa ‘place of residence’
live-place
Much like ‘free-dom’ and ‘child-hood’ in English, these noun suffixes (‘nominal
classifiers’) originated as the head nouns in noun compounds. The use of these
suffixes in Tibetan relativization can be seen in (Mazaudon 1978; DeLancey
1988):
What the Tibetan data above suggest, I think, is that there is no binding
correlation between the nominalization case-recoverability strategy and the
non-restrictive paratactic pathway to embedded REL-clauses. Tibetan is a rather
classical clause-chaining SOV language. What is more, like related languages
in the Bodic region, and like many other clause-chaining languages, chain-
medial clauses in Tibetan are typically nominalized (i.e. non-finite; Givón 2001,
Chapter 18). A clause-chaining source of restrictive REL-clauses is thus very likely
here. What is more, the pre-nominal position of Tibetan REL-clauses makes the
non-restrictive pathway to embedded REL-clauses much less plausible, given that
non-restrictive REL-clauses are parenthetical after-thought devices that most com-
monly follow the head noun – regardless of word-order type.
Toward a diachronic typology of relative clause
In light of the discussion above, I would like to examine briefly the pre-nominal
restrictive REL-clauses of Mandarin Chinese. In earlier discussion, I have tended
to interpret their pre-nominal position in this otherwise – rigid SVO and clause-
chaining (V-serializing) language as a relic of earlier SOV syntax. Restrictive
REL-clauses in Mandarin are marked with the clause-final nominalizer suffix -de
(Li & Thomson 1981):
Cleft constructions are said to have a REL-clause tucked under the same into-
nation contour, following a contrasted (stressed) noun (Schachter 1971). But in
many languages the data exist to suggest that this syntactic construction is a con-
densation of an earlier paratactic one, where the REL-clause was packaged under
a separate intonation contour. What is more, in some languages the same can be
shown for WH-questions. As an illustration of both patterns, consider Kihungan
(Bantu, Takizala 1972; Givón 2001, Chapter 15):
c. Non-restrictive rel-clause:
kit, ki a-swiim-in Kipes...
chair dem 3sg/rel-buy-past K.
‘the chair, the one that Kipes bought...’
d. Syntactic cleft:
kwe kít ki a-swiim-in Kipes
be chair dem 3sg/rel-buy-past K.
‘It’s a CHAIR that Kipes bought’.
e. Paratactic (non-restrictive) cleft:
kwe kít, (kiim) ki a-swiim-in Kipes
be chair thing dem 3sg/rel-buy-past K.
‘It’s a CHAIR, (the thing) that Kipes bought’.
f. Syntactic WH-question:
(kwe) kí (ki) a-swiim-in Kipes?
(be) what (dem) 3sg/rel-buy-past K.
‘What did Kipes buy?’
(Lit.: ‘(It’s) WHAT (that) Kipes bought?’)
g. Paratactic (non-restrictive) WH-question:
kwe kí, (kiim) ki a-swiim-in Kipes?
be what (thing) dem 3sg/rel-buy-LAST K.
‘It’s WHAT, (the thing) that Kipes bought?’
Of the 7–8 major relativization strategies considered earlier (Givón 2001), I have
suggested here that at least four:
fit into either one of the two paratactic mega-pathways that give rise to
embedded REL clauses: The clause-chaining pathway (i, ii), or the non-restrictive
(parenthetical-clause) pathway (iii, iv). One more type, the Philippine verb-coding
T. Givón
strategy (v), has probably risen from a nominalized source to begin with, and may
thus p arallel the case of Ute. Both languages have post-nominal REL-clauses, which
are more compatible with the non-restrictive pathway. Another type, the Hebrew
resumptive-anaphoric pronoun strategy (vi), has a long history that leads back to
a nominalized source (Givón 1991a). The use of simple anaphoric pronouns in
Hebrew relativization, combined with the post-nominal position of REL-clauses,
are both compatible with the non-restrictive (parenthetical) paratactic source.3
There is obviously a lot more to be done here, and more corroborative evi-
dence to be gathered and collated. The correlation between the non-restrictive
REL-clause source and word-order needs to be further studied. But both major
pathways that emerge out of the typological data seem to follow the parataxis-to-
syntaxis route.
A final point concerns some cognitive correlates of the two developmental
steps I have posited at the start:
The processing speed of lexical words (28c) is ca. 250 msecs/word, relying heavily
on automated spreading activation of semantic networks. The processing speed
of single syntactic clauses (28b) is ca. 1–2 secs/clause. And the processing speed
of two chained clauses (28a) is at least double. The level of semantic complex-
ity may not vary all that much from (28a) to (28b) to (28c), but the processing
speed surely does. These two steps of condensation involve increased processing
speed and automaticity. Whether this is the primary driving motivation for such
3. In Biblical Hebrew, the later finite relativization pattern with the generalized
REL-subordinator ’asher was preceded by an earlier layer of nominalized REL-clauses
(Givón 1991a). The etymology of ’asher may go back to ’athar ‘place’ (Hetzron, in personal
communication). If so, there may have been a spreading of the pattern from a nominalized
locative REL-clause to the entire case paradigm, a phenomenon also attested in spoken Greek
(pou ‘where’), spoken Southern German (wo ‘where’) and Krio (we ‘where’).
Toward a diachronic typology of relative clause
Abbreviations
References
4. Another study (Givón 2008) seems to suggest that the primary motivation for the rise
of V-complement constructions is communicative rather than cognitive. The subsequent
condensation into hierarchic structure, and the presumed increase in automaticity, may be a
secondary development.
T. Givón
Givón, T. (ed.). 1979a. Discourse and Syntax. New York NY: Academic Press.
Givón, T. 1979b. On Understanding Grammar. New York NY: Academic Press.
Givón, T. 1988. Tale of two passives: Internal reconstruction in Ute. In Passive and Voice [Typ-
logical Studies in Language 16], M. Shibatani (ed.), 417–439. Amsterdam: John Benjamins.
Givón, T. 1990. Syntax: A Functional-Typological Introduction, Vol. 2, Amsterdam: John
Benjamins.
Givón, T. 1991a. The evolution of dependent clauses in Biblical Hebrew. In Approaches to Gram-
maticalization. Focus on Types of Grammatical Markers [Typological Studies in Language
19], E. Traugott & B. Heine (eds), 257–310. Amsterdam: John Benjamins.
Givón, T. 1991b. Some substantive issues concerning verb serialization: Grammatical vs. cogni-
tive packaging. In Serial Verbs: Grammatical, Comparative and Cognitive Approaches, C.
Lefebvre (ed.), 137–211. Amsterdam: John Benjamins.
Givón, T. 1994. Nominalized clauses in Ute: The diachronic seesaw of finite and non-finite struc-
ture. In II Encuentro de Linguística en el Noroeste, Hermosillo: Universidad de Sonora.
Givón, T. 2000. Internal reconstruction: As method, as theory. In Reconstructing grammar. Com-
parative linguistics and grammaticalization [Typological Studies in Language 43], S. Gildea
(ed.),107–60. Amsterdam: John Benjamins.
Givón, T. 2001. Syntax, Vol. 2. Amsterdam: John Benjamins.
Givón, T. 2009a. The Genesis of Syntactic Complexity, Amsterdam: John Benjamins.
Givón, T. 2009b. Multiple routes to clause-union: The diachrony of syntactic complexity. In
Syntactic Complexity. Diachrony, acquisition, neuro-cognition, evolution [Typological Stud-
ies in Language 85], T. Givón & M. Shibatani (eds), 81–118, A msterdam: John Benjamins.
Heine, B. & Kuteva, T. 2007. The Genesis of Grammar, Oxford: OUP.
Justus-Raman, C. 1973. The Old Hittite Relative Construction, Ph.D. dissertation, University of
Texas at Austin.
Justus, C. 1976. Relativization and topicalization in Hittite. In Subject and Topic, C. Li (ed.),
215–45. New York NY: Academic Press.
Li, C.N. & Thompson, S.A. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley
CA: University of California Press.
Linde, C. 1979. Focus of attention and the choice of pronoun in discourse. In Syntax and Seman-
tics, Vol. 12: Discourse and Syntax, T. Givón (ed.), 337–354. New York NY: Academic Press.
Matisoff, J. 1969. Verbs concatenations in Lahu. Acta Linguistica Hafriensia, 12(2): 169–220.
Mazaudon, M. 1978. La formation des propositions relatives en tibétain. Bulletin de la Societé de
Linguistique de Paris 73(1).
Mithun, M. 2006. Structural parameters of clause integration: Elusive complementation, Semi-
nario sobre Complejidad Sintáctica. Ms, University of Sonora, Hermosillo.
Mithun, M. 2007. Threads in the tapestry of syntax: Complementation and Mohawk. Ms, Uni-
versity of California at Santa Barbara.
Pawley, A. 1987. Encoding events in Kalam and English: Different logics for reporting events. In
R. Tomlin (ed.), 329–360.
Robert, P. 2006. Clause boundaries in Old Hittite relative sentences. Transaction of the Philologi-
cal Society 104(1): 17–83.
Schachter, P. 1971. Focus and relativization. Language 47(1): 19–46.
Shibatani, M. 1972. Three reasons for not deriving ‘kill’ from ‘cause to die’. In J. Kimball (ed.),
125–137.
Toward a diachronic typology of relative clause
Takizala, A. 1972. Focus and relativization in Kihungan. Studies in African Linguistics 3(2):
259–87.
Thurman, R. 1978. Interclausal Relations in Chuave. MA Thesis, University of California,
Los Angeles.
Watters, D. 1998. The Kham Language of West-Central Nepal (Takale Dialect) Ph.D. disserta-
tion, University of Oregon at Eugene.
The evolution of language and elaborateness
of grammar
The case of relative clauses in creole languages*
Starting from the assumption that creole languages present an opportunity for
testing hypotheses on the evolution of complexity in language, we examine
the number of markers used to construct relative clauses, more specifically
in relativization on subjects. On the basis of a sample of 52 creole languages,
we show that such languages more often than not have simply marked relative
clause constructions, encoded by no more than one relativization marker.
This typological result stands out as particularly significant if we view it against
the background of non-creole languages, for which we have been able to identify
cases with up to five relativization markers.
1. Introduction
* We wish to thank the following colleagues for generously sharing with us their k nowledge
on creoles and – in many cases – also unpublished data from their fieldwork on these creoles:
Philip Baker, Angela Bartens, Alan Baxter, Clancy Clements, Hubert Devonish, Joseph
Farquharson, Stephanie Hackert, Tjerk Haggemeijer, Philippe Maurer, John McWhorter,
Susanne Michaelis, Bettina Migge, Susanne Mühleisen, Paula Prescod, Edgar Schneider,
Eeva Sippola, Hein van der Voort, Don Winford. Special thanks go also to Bernd Heine and
Alexandra Aikhenvald for providing an insightful analysis of a number of linguistic facts. The
first author is greatly indebted to SOAS, University of London as well as the Alexander von
Humboldt Foundation for their generous financial support.
Tania Kuteva & Bernard Comrie
complexities of all or many presently attested languages are unlikely to have been
present in early human language. Heine and Kuteva’s (2002: 394, 2007) findings
about the grammaticalization processes in the languages of the world lend support
to this standpoint. In other words, one way of thinking about the evolution of
language involves the assumption that language in prehistoric time had simple,
functionally motivated grammatical structures.
The complexification of grammar may mean more complexity being added
to a less complex grammar with respect to contents, i.e. additional categories,
additional variants of the same grammatical category, etc. It may also involve
more complexity with respect to form. In the present study, we are interested in
this latter kind of complexity, which we refer to as elaborateness of grammatical
marking (cf. also Comrie & Kuteva 2005; Kuteva & Comrie 2005). As Comrie and
Kuteva (2005) point out, the assumption about a simple-before-complex cline of
language evolution does not mean an exclusive complexification process at each
and every point of language development: once elaboration sets in, there is also
the possibility of simplification. But given that all languages with m
illennia of
history have some degree of elaboration somewhere in their grammars, the forces
leading to elaboration have overall had an edge over those leading to greater
simplicity.
The assumption of simple-before-complex grammar, however, remains just
what it is – a speculative assumption only – unless we can buttress its plausibility
by means of observable linguistic data. Creole languages present an opportunity
for testing hypotheses on language evolution since – in the case of some of these
languages at least – we are in a position to observe how grammar is being created,
a situation which has led some analysts of language to view the process of creolo-
genesis as a “language laboratory” (Hagège 1993), cf. also the image of greenhouse
in Plag (1994: 19).1
In a recent paper, McWhorter 2001 investigates precisely the “language
laboratory” of creologenesis, and propounds the idea that creole languages
have the world’s simplest grammars. The reactions to McWhorter’s standpoint
have ranged from sympathetic to highly critical (see Issue 5, 2/3 of Linguistic
Typology).
In the present study, our goal is to test the prediction that Comrie’s (1992)
and Heine and Kuteva’s (2002, 2007) as well as McWhorter’s (2001) approach
1. Plag (1994: 19) makes a parallel between creolization and “a sort of greenhouse effect: just
as plants grow faster and in higher numbers in a greenhouse than in their natural environ-
ment, in creolization language change proceeds faster and more drastically than in natural
change”.
The evolution of language and elaborateness of grammar
2. Encoding the relative clause construction in the languages of the world
Here the relative clause precedes the head noun and it contains no element
co-referential to the relativized noun.
At the other end of the scale, there are languages with no less than five mor-
phosyntactic segments serving as markers of the relativization strategy. Ngemba
(Bantoid, Niger-Congo, spoken in Cameroon), for instance, marks relative clauses
by means of:
The complementizer -bah is optional, the other four relativization markers are
obligatory, however.
The evolution of language and elaborateness of grammar
For the purposes of this study, we have examined all the creoles for which we
could find a grammatical description of the way they mark relativization upon the
subject. Our sample consists of 52 languages, see Appendix 1. Even though this is
basically a convenience sample, it has a relatively good coverage of all creoles since
it represents almost all groups of creoles according to lexifier:
English-based
French-based
Portuguese-based
Dutch-based
Apart from only two sample languages – Louisiana Creole and Tok Pisin –
which appear to have zero relativization marking more often than not, all the cre-
oles from our sample turn out to explicitly mark the relative clause construction.
Twelve of our sample languages have zero marking as an additional option. On
the whole, however, zero marking cannot be regarded as characteristic of the way
creoles encode the relative clause construction. For Gullah, for instance, Mufwene
(1986: 15) states that sentences with zero marking – exemplified in (3) below – are
a “tiny minority”:
(3) Gullah (Mufwene 1986: 15)
Uh think a man see he wife dress up an bloom up will try get de way too.
‘I think a man [who] sees his wife dressed up and blossomed will try to get
in the/her way too.’
In a more generalizing way, Bruyn (1995: 157) points out that “zero marking
may be viewed as the strategy that can more easily be adopted by an expand-
ing pidgin or an emerging creole”. The detailed investigation of relativization in
Hawaiian Creole also supports this observation. Thus Peet (1978: 96–99) shows
that a zero-marked relative clause construction for relativizing on subjects pre-
ceded historically two other, one-marker relative clause constructions (one with
a personal pronoun functioning as a relative clause marker, and another with an
interrogative pronoun used as a relative pronoun). Furthermore, in the case of Tok
Pisin, which is one of the two languages in our sample where zero relativization
marking seems to be the preferred option, it stands to reason that the explanation
Tania Kuteva & Bernard Comrie
of this fact has to do with the specific status of Tok Pisin, namely a creole which
for a number of speakers is a pidgin. In other words, the more creole-like a variety
is – as c ontrasted to a pidgin-like variety – the less likely it is for that variety to have
zero relativization marking.
The results of our investigation are thus compatible with Hancock’s 1986
survey, cited in Mufwene (1986: 19), of 33 Anglophone Creoles, only seven of
which have been identified as having zero marking.
The main result of our examination of creoles has to do with the precise
number of the explicit markers of relativization: our study shows that creoles
regularly employ no more and no less than precisely one relativization marker:
(4) Capeverdean (Hutchinson 1986: 81)
kel omi ki bai
the man rel go
‘the man that went’
(5) Guyanese Creole English (Hubert Devonish p.c.)
Di maan wa kom a di Instithuut laas nait bin a wan
the man who came to the Institute last night ant be a
Kriyool Stodiiz profesa.
creole studies professor
‘The man who came to the Institute last night was a professor of
creole studies.’
Note that some creoles may make use of more than one way to mark
r elativization. Tok Pisin, for instance, may use either ia or we, or husat, or third
person pronoun, or zero marking for the relative clause construction. Yet, in each
particular case, if the relative clause construction is marked at all, this is done by
means of no more than one marker.
Bislama is a creole which at first sight does not conform to the generalization
“one relative clause: one relativization marker”. In this creole, it seems that the
relative clause construction is marked by more than one element. In addition to
the invariable relativizer we (deriving from where), there is also the third person
pronoun used after it in the relative clause:
(6) Bislama (Tryon 1987: 118)
Em i save man ia we em i ded.
(he pred.marker know man here rel he pred.marker died)
‘He knew the man who died.’
the pronoun is not something which is necessary for the relative clause only (as
contrasted to the corresponding main clause). Moreover, the relative clause may
just as well be formed without the subject pronoun:
(7) Bislama (Tryon 1987: 118)
Em i save man ia we i ded.
(he pred.marker know man here rel pred.marker died)
‘He knew the man who died.’
Besides, Tryon (1987: 119) presents the form ia as some sort of marker of rela-
tivization: “Ia, then, is very widely used in Bislama following and modifying the
object of the main clause to signal the introduction of a following relative clause.”
From the examples given, however, it appears that this form functions as a definite
article rather than as a relativization marker, which is the reason why we have not
treated it as a relativizer here:
(8) Bislama (Tryon 1987: 119)
Mi luk man ia. em i stap brekem windo.
(I saw man here he pred.marker stop brake window)
‘I saw the man. He was breaking the window.’
Likewise, Tok Pisin has also been pointed out as a variety where double
marking is possible. Thus Sebba (1997: 114) analyses the demonstrative ya ‘here’
as a morpheme which “may occur twice: once after the head noun and once at the
end of the relative clause; both the first and second ya are optional, but the second
one rarely occurs without the first” (Sebba 1997: 114):
(9) Tok Pisin (Wurm et al. 1979, cited in Sebba 1997: 114)
meri ya [i-stap long hul ya] em i-hangre.
woman rel sm-stay in hole rel she sm-hungry
‘The woman who stayed in the hole was hungry.’
Just as in the case of Bislama, however, the example illustrating the use of ya speaks
in favour of a treatment of this morpheme – when used immediately after the head
noun – as a definiteness marker rather than a relativization marker.
One more of our sample languages, Tayo, seems – at first sight at least – to
exhibit a double marking; this, however, is only the first impression. It turns out
that the second element used after the relativization marker sa – see (10) below – is
the subject index, which is obligatory in both relative clauses and in independent
main clauses, see (11) (Corne 1994: 287–289):
Following Hinskens and van Rossem 1995, we treat the occurrence of sender as a
case of an agreement marker on the relative pronoun die. Accordingly, for us, this
second way of marking the relative clause construction, die sender, is also a case of
a single relative clause marker.
What is of much more interest to us here is the third kind of marking, which
involves not only the relative pronoun die but also a copy of the personal pronoun
serving as the antecedent, e.g. mi…die mi…(lit. ‘I/me…who/I…’):
Examples like these testify to double marking of the relative clause construction.
Note, however, that:
ii. The texts analysed by Hinskens & van Rossem 1995 are highly representative –
quantitatively – but they are limited to one lect only, “Religious Negerhollands”,
which is different from either “Low” or “High” Negerhollands referring to the
2. Hein van der Voort (p.c.) also points out to us that this construction is encountered in
18th century Herrnhut translations of religious texts and that it is obviously modelled on High
German.
Tania Kuteva & Bernard Comrie
language as it was used by the Slaves and by the white people respectively, that
is, the Negerhollands used by the Moravian Brethren in their translations is
somewhat artificial (Hinskens & van Rossem 1995: 76).
This latter fact then might be the reason why double marking of the relative clause
construction can be observed in the liturgical Negerhollands texts discussed here,
which does not necessarily mean that this particular marking was also typical of
ordinary, that is, stylistically unmarked discourse.3
Finally, Palenquero is the only creole in our sample which clearly involves double
marking as one possibility for the encoding of the relative clause construction,
whereby this type of encoding does not seem to be stylistically marked. Schwegler
and Green (2007) describe it as a creole having three relativization markers:
(i) lo ke ‘that which; s/he who’, (ii) ke ‘who, which, that’, (iii) i ‘who, which, that’:
(16) Palenquero (Schwegler & Green 2007: 285)
moná lo ke sabé fecha e tat’ éle ta aí.
child rel know date of father 3sg cop there
‘The child who knows the father’s birthday is over there.’
(17) Palenquero (Schwegler & Green 2007: 301)
a-ten jende ke bae monte má nu.
EXIST people rel go field more neg
‘There are people that no longer go (work in) the field.’
(18) Palenquero (Schwegler & Green 2007: 301)
ao ese ma mujé i t′ aí.
all dem pl woman rel cop there
‘All these women that are there.’
The first of these is the only possible – and stylistically neutral – example of
elaborate relativization marking (that is, consisting of two component elements)
that we have come across in our sample languages. Notice, however, that here we
may still be dealing with a single relativization marker for the following reason: the
two morphemes lo ke occur always together and nothing can be inserted between
them, which would support a hypothesis of them being regarded as a single “fixed”
form (with thanks to John McWhorter p.c.).
3. The parallel we can make to eighteenth-century Sranan can be very instructive here. Thus
Bruyn (1995: 152n), cited in Hinskens and van Rossem (1995: 71), presents data that include
eighteenth-century Sranan Bible translations – again by Moravian Brethren – which also
testify to double marking of the relative clause construction in liturgical Sranan (une di une
‘you, who you’). This, however, does not make Sranan a creole with a doubly marked relative
clause construction overall.
The evolution of language and elaborateness of grammar
The present finding – that is, the consistency with which creoles mark the r elative
clause construction by no more than one marker – is rather puzzling in the light
of the regularities observed in recent research on language contact. There have
been two major types of situation identified according to the type of impact
social network structure and stability have on linguistic structure. T rudgill
(2004: 437–438) describes these as a contact-and-simplification situation on the
one hand, and a contact-and-complexification situation on the other. The former
involves cases where “simplification may occur in high-contact languages as a
result of pidginization, which is what occurs in those situations involving adult and
therefore imperfect language acquisition on the part of speakers who have passed
the critical threshold…” In other words, language contact causes loss (of phono-
logical material, grammatical structures), and Trudgill has proposed an account
for this loss in a series of works starting with Trudgill (1983; see also Trudgill 1996;
Trudgill 2001). Trudgill (2001: 372) is particularly explicit on this point: “Just as
complexity increases through time, and survives as the result of the amazing
language-learning abilities of the human child, so complexity disappears as a
result of the lousy language-learning abilities of the human adult. Adult language
contact means adult language learning; and adult language l earning means simpli-
fication, most obviously manifested in a loss of redundancy and irregularity and
an increase in transparency.”
The contact-and-complexification situation, on the other hand, involves cases
where increased complexification may occur in languages as a result of borrowing,
whereby long-term contact and childhood bilingualism are necessary accompany-
ing factors. What is relevant to the present discussion is the conclusion Trudgill
(2004: 43–438) arrives at: “high-contact languages may demonstrate more redun-
dancy if child language contact is involved.” We ourselves have not investigated
the amount of child language contact involved; nevertheless, we are in a position
to say that Trudgill’s conclusion is in agreement with the conclusions of other
recent works on language contact such as Heath (1978) and Nichols (1992), and
much more recently, Aikhenvald (2002), Heine and Kuteva (2005), Kuteva (2008),
who also show that language contact may well bring about diversification and
complexification of grammar.
Tania Kuteva & Bernard Comrie
5. Discussion
In the previous section we argued that a high degree of language contact may lead
to increase in elaborateness of marking of grammatical categories. As a reaction
to the innate, “bioprogram” position articulated in Bickerton (1981), it has by now
become a common practice to treat creoles as the result of an extremely high degree
Tania Kuteva & Bernard Comrie
Thus one may be tempted to ascribe the zero marking for the relative clause
construction in Gullah to the substrate, which would rule out the credibility of a
functional motivation here. As Mufwene (1986: 15) convincingly argues, however,
this situation is hardly the result of substrate influence since in serialization in
African languages, all the verbs in the construction share – typically – subject,
tense, and aspect, which – at least with respect to the subject – is not true of the
Gullah example above. If the role of substrate serialization is to be ruled out, what
can then be a plausible account?
Our response to this is that at the stage expanding pidgin – emerging creole
(in those cases where a creole emerges out of a pidgin), a pidgin/creole language
variety does not have a lot of morphosyntax in the first place. In fact, even at the
stage of an established creole, the amount of morphosyntax – especially as far as
relative clause subordination is concerned – is not that impressive, which explains
The evolution of language and elaborateness of grammar
why creolists observe that even though a relativization marker may be available
in the system, relative clauses are relatively rare in spontaneous discourse. What
happens is that the initial lack of marking characteristic of pidgins is followed by
evolving simple marking which – at a later stage – may well be expected to become
elaborarate.
This is supported by the observations on Tok Pisin, the historical docu-
mentation for which is better than for most other pidgins/creoles. Thus Sebba
(1997: 113–114) points out: “In this case we can actually see the development
of strategies for making relative clauses where only the most rudimentary ones
existed in early stages. In early examples… we find that relative clauses were not
marked by overt means at all:
(23) South Seas Pidgin (Sebba 1997: 113–114)
You savez two white men [stop Maputi] he got house.
‘You know two white men who live at Maputi who have houses.’
Transcriptions by Hall (1943) of the speech of New Guineans who were born
around the beginning of the twentieth century show the same:
(24) Tok Pisin (Sebba 1997: 113–114)
Spos yu lukim man [i-kisim poisen] yu i-ken tok.
if you see man sm-get magic you sm-can talk
‘If you had seen the man who made black magic, then you might talk.’
More recently, from the 1950s onwards, there have appeared several markers
of the relative clause construction: ia/ya, we, husat, each of which can be used
optionally.”
Even if we are to summarize the diachronic development of marking for the
relative clause construction in expanding pidgins/emerging creoles as lack of
marking – simple marking – elaborate marking, this is in no way problematic for our
analysis of the evolution of grammatical structure in creoles, and by extrapolation,
in human language, as simple-before-complex.
Abbreviations
Appendix 1
Creoles
American Earlier Black English 1/0 (Schneider 1989: 213–217, p.c.)
Angolar 1/0 (Lorenzino 2007)
Annobón 1 (Holm 1989: 284)
Bajan 1/0 (Donald Winford p.c.)
Bahamian Creole 1 (Stephanie Hackert p.c.)
Batavia 1 (Philippe Maurer p.c.)
Berbice Dutch 1 (Kouwenberg 1995: 236)
Belizean Creole 1 (Greene 1999: 92)
Bislama 1 (Tryon 1987: 118, 152)
Capeverdian 1 (Hutchinson 1986: 81)
Cavite Chabacano 1 (Eeva Sippola p.c.)
Daman Creole Portuguese 1 Clancy Clements p.c.)
Dominican 1 (Chapuis 2007)
Eastern Maroon Creoles 1 (Bettina Migge p.c.)
Fa d’Ambu (1) (Post 1996: 197, 202)
Guadeloupe Creole French 1 (Holm 1989: 368)
Gullah 1/0 (Mufwene 1986)
Guyanese Creole English 1 (Hubert Devonish, p.c.)
Jamaican Creole 1 (Joseph Farquharson, p.c.)
Haitian 1 (Muysken & Veenstra 1995: 162;
Holm 1989: 386)
Hawaiian English 1/0 (Peet 1978: 96)
Korlai Creole Portuguese 1 (Clements 1996: 181–192;
Clancy Clements p.c.)
Krio 1 (Nylander 1984: 131)
Kristang 1/0 (Baxter 1988: 108–113)
Kriyol 1 (Kihm 1994: 177)
Lingala 1/(0) (Comrie & Kuteva 2005: 206)
Pointe Coupee Parish
Louisiana Creole 0/1 (Klinger 2003: 225–226)
Martinique Creole French 1 (Holm 1989: 368)
Morisyen 1 (Adone 2001: 64)
Nicaraguan Creole English 1/0 (Angela Bartens, p.c.)
Negerhollands 1/(2) (Hinskens & van Rossem 1995: 68–69; 74)
Ndyuka 1 (Huttar & Huttar 1994: 90–91)
The evolution of language and elaborateness of grammar
Appendix 2
References
Alsagoff, L. & Lick, H.C. 1998. The relative clause in colloquial Singapore English, World
Englishes 17/2: 127–138.
Azeb, A. 2001. The Maale language. Ph.D. dissertation, University of Leiden.
Baxter, A. 1988. A Grammar of Kristang (Malacca Creole Portuguese), Pacific Linguistics, Series
B-No.95. Canberra: A.N.U. Printing Service.
Bell, A. 2000. Maori and Pakeha English. New Zealand English [Varieties of English around the
World G25], A. Bell & K. Kuiper (eds.), 249–278. Amsterdam: John Benjamins.
Bickerton, D. 1981. Roots of Language. Ann Arbor: Karoma.
Bruyn, A. 1995a. Grammaticalization in creoles: The development of determiners and relative
clauses in Sranan. Ph.D. dissertation, University of Amsterdam.
Bruyn, A. 1995b. Relative clauses in early Sranan. In The early stages of creolization, J. Arends
(ed.), 149–202. Amsterdam: John Benjamins.
Byrne, Francis X. 1988. Deixis as a non-complementizer strategy for creole subordination
marking, Linguistics 26(3): 335–64.
Carrington, L.D. 1984. St. Lucian Creole. A descriptive analysis of its phonology and m orpho-syntax.
Hamburg: Helmut Buske Verlag.
Chapuis, D. 2007. Kwéyòl, or Dominican (Creole French). In Comparative creole syntax: P arallel
outlines of 18 creole grammars, J. Holm & P.L. Patrick (eds.), 83–100. London: Battlebridge
Publications.
Chumbow, B. 1977. Relatives as determiners: A case from Ngemba. In Language and linguistic
problems in Africa. Proceedings of the VII conference on African linguistics, P.F.A. Kotey & H.
Der-Houssikian (eds.), 283–302. South Carolina: Hornbeam Press.
Clements, C. 1996. Genesis of a Language. Amsterdam: John Benjamins.
Comrie, B. 1992. Before complexity. In The Evolution of human language. Proceedings of the
Workshop on the evolution of human languages, August 1989, Santa Fe, New Mexico,
J. Hawkins & M. Gell-Mann (eds.), 193–211. SantaFe, NM: Addison-Wesley.
Comrie, B. & Kuteva, T. 2005. The evolution of grammatical structures and “functional need”
explanations. In Language Origins: Perspectives on Evolution, M. Tallerman (ed.), 185–207.
Oxford: Oxford University Press.
Corne, C. 1994. Relativization and thematization in Tayo and the implications for creole genesis.
Journal of Pidgin and Creole Languages 9(2): 283–304.
Corne, C. 1995. Nana k nana, nana k napa: The paratactic and hypotactic relative clauses of
Reunion Creole, Journal of Pidgin and Creole Languages 10(1): 57–76.
Crowley, T. 1990. Beach-la-Mar to Bislama. Oxford: Clarendon Press.
Dolan, T.P. (ed.). 1999. A dictionary of Hiberno-English: The Irish use of English, Dublin: Gill &
Macmillan.
Fought, C. 2003. Chicano English in context. Palgrave: Macmillan.
Greene, L. 1999. A grammar of Belizean Creole. Compilations from two existing United States
Dialects. NewYork/Washington, D.C./Baltimore/Boston/Bern/Frankfurt am Main/Berlin/
Vienna/Paris: Peter Lang.
Gumperz, J.J. & Wilson, R. 1971. Convergence and creolization: A case from the Indo-Aryan/
Dravidian border in India. In D. Hymes. 151–67.
Hagège, C. 1993. The language builder. Amsterdam: John Benjamins.
Hall, R.A., Jr. 1943. Melanesian pidgin english; grammar, texts, vocabulary. Baltimore: Linguistic
Society of America.
Hancock, I.F. 1986. The domestic hypothesis, diffusion, and componentiality: An account
of Atlantic Anglophone Creole origins. In Substrata versus universals in creole genesis,
P. Muysken & N. Smith (eds.), 71–102. Amsterdam: John Benjamins.
The evolution of language and elaborateness of grammar
Hansen, K., Carls, U. & Lucko, P. 1996. Die Differenzierung des Englischen in nationalen
Varianten. Berlin: Eric Schmidt Verlag.
Hartford B.S. 1996. The relationship of New Englishes and linguistic theory: A cognitive-
based grammar of Nepali English, In South Asian English: Structure, use, and users,
R.J. Baumgardner (ed.), 88–103, Urbana and Chicago: University of Illinois Press.
Heath, J. 1978. Linguistic diffusion in Arnhem Land. (Australian Aboriginal Studies Research
and Regional Studies, 13). Canberra: Australian Institute of Aboriginal Studies.
Heine, B. & Kuteva, T. 2002. On the evolution of grammatical forms. In The transition to
language, A. Wray (ed.), 376–398. Oxford: Oxford University Press.
Heine, B. & Kuteva, T. 2003. On contact-induced grammaticalization, Studies in Language 27(3):
529–72.
Heine, B. & Kuteva, T. 2005. Language contact and grammatical change. Cambridge: Cambridge
University Press.
Heine, B. & Kuteva, T. 2007. The genesis of grammar: A reconstruction. Oxford: Oxford University
Press.
Hinskens, F. & van Rossem, C. 1995. The Negerhollands word sender in eighteenth-century
manuscripts. In The Early Stages of Creolization, J. Arends (ed.), 63–88. Amsterdam: John
Benjamins.
Hoekstra, J. 2002. Relativization in Frisian. In Relativization on the North Sea Littoral, P. Poussa
(ed.). LINCOM EUROPA.
Holm, J.A. 1986. The spread of English in the Caribbean area. In Focus on the Caribbean, 8, 1:
1–22.
Holm, J.A. 1989. Pidgins and Creoles: References survey, (Vol. 2). Cambridge: Cambridge
University Press.
Hutchinson, J. 1986. Aspects of complementary development in creole relativization. In
Developments in linguistics and semiotics. Language teaching and learning. C ommunication
across cultures, 81–92. Georgetown University Round Table on Language and Linguistics.
Huttar, G.L. & Huttar, M.L. 1994. Ndyuka. London: Routledge.
Hymes, D. 1971. Pidginization and creolization of languages: Proceedings of a conference held
at the University of the West Indies, Mona, Jamaica, April 1968. Cambridge: Cambridge
University Press.
Klinger, T. 2003. If I could turn my tongue like that. Baton Rouge: Louisiana State University.
Kouwenberg, S. 1995. Berbice Dutch. In Pidgins and creoles: An introduction (Creole Language
Library, 15), J. Arends, P. Muysken & N. Smith, 233–243. Amsterdam: John Benjamins.
Kuteva, T. 2008. On the “frills” of grammaticalization. In Rethinking grammaticalization: New per-
spectives for the twenty-first century (Typological Studies in Language), M.J. López-Couso &
E. Seoane (eds.) in collaboration with T. Fanego. 189–219. Amsterdam: John Benjamins.
Kuteva, T. & B. Comrie. 2005. The typology of relative clause formation in African languages. In
African Studies. E. Voeltz (ed.), 209–208. Amsterdam: John Benjamins.
Lorenzino, G.A. 2007. Angolar (Creole Portuguese) or Lunga Ngola. In J. Holm & P.L. Patrick
(eds.), 1–23.
McWhorter, J. 2001. The world’s simplest grammars are creole grammars. Linguistic Typology
5: 125–166.
Mesthrie, R. 1991. Syntactic variation in South African Indian English. In English around
the world. Sociolinguistic perspectives, J. Cheshire (ed.), 462–473. Cambridge/New York:
Cambridge University Press.
Migge, B. 2003. Creole formation as language contact. The case of the Suriname creoles.
Amsterdam: John Benjamins.
Tania Kuteva & Bernard Comrie
Mufwene, S. 1986. Restrictive relativization in Gullah. Journal of Pidgin and Creole Languages
1(1): 1–31.
Mufwene, S. 1990. Transfer and the substrate hypothesis in creolistics. Studies in Second
Language Acquisition 12: 1–23.
Muysken, P. & Veenstra, T. 1995. Haitian. In Pidgins and creoles: An introduction (Creole Language
Library, 15), J. Arends, P. Muysken & N. Smith, 153–164. Amsterdam: John Benjamins.
Myers-Scotton, C. 1997. “‘Matrix language recognition’ and ‘morpheme sorting’ as possible
structural strategies in pidgin/creole formation. In The structure and status of pidgins and
creoles, A.K. Spears & D. Winford (eds.), 151–174. Amsterdam: John Benjamins.
Nadkarni, M.V. 1975. Bilingualism and syntactic change in Konkani. Language 51: 672–683.
Nichols, J. 1992. Linguistic diversity in space and time. Chicago: University of Chicago Press.
Nylander, D.K. 1984. WH-phrases, relative clauses and case assignment in Krio. York Papers in
Linguistics 11, 241–252.
Peet, W. Jr. 1978. Relativization in a Creole Continuum. Ph.D. dissertation, Ann Arbor,
Michigan: UMI.
Plag, I. 1994. Creolization and grammatical change: A comparison. In Creolization and l anguage
change, (Linguistische Arbeiten), A. Dany & I. Plag (eds). 3–23. Tübingen: Niemeyer.
Post, M. 1996. Aspect marking in Fad’Ambu. The elements xa and sa and their respective aspec-
tual values. From Contact to Creol and Beyond, Philip Baker, 189–206. London: Battlebridge
Publications.
Prescod, P. 2004. Une description grammaticale du syntagme nominal dans le créole anglophone
de St-Vincent-et-les-Grenadines. Ph.D. thesis. Université Paris III.
Romaine, S. 1992. The evolution of complexity in a creole language: acquisition of relative
clauses in Tok Pisin. Studies in Language 16-1: 139–182.
Schneider, E. 1989. American Earlier Black English: Morphological and syntactic variables.
Tuscaloosa: University of Alabama Press.
Schwegler, A. & Green, K. 2007. Palenquero (Creole Spanish). In Comparative creole syntax:
Parallel outlines of 18 creole grammars, J. Holm, & P.L. Patrick (eds.), 273–306. London:
Battlebridge Publications.
Sebba, M. 1997. Contact languages. Macmillan Press, Basingstoke.
Slobin, D. 2002. Language evolution, acquisition and diachrony: Probing the parallels. In The
evolution of language out of pre-language, T. Givón & B.F. Malle (eds.), 374–393. A
msterdam:
John Benjamins.
Thomason, S.G. & Kaufmann, T. 1998. Language contact, creolization, and genetic linguistics.
Berkeley, CA: University of California Press.
Trudgill, P. 1983. On dialect. Social and geographical perspectives. Oxford: Basil Blackwell.
Trudgill, P. 1996. Dual source pidgins and reverse creoles: Northern perspectives on language
contact. In Language Contact in the Arctic: northern Pidgins and contact languages. Ernst H.
Jahr and Ingvild Broch (eds.), 5–14. Berlin: Mouton.
Trudgill, P. 2004. The impact of language contact and social structure on linguistic structure:
Focus on the dialects of Modern Greek. In Dialect grammar from across-linguistic perspec-
tive, (Trends in Linguistics), B. Kortmann (ed.), 435–452. Berlin: Mouton de Gruyter.
Tryon, D.T. 1987. Bislama: An introduction to the national language of Vanuatu. Canberra: Dept.
of Linguistics, Research School of Pacific Studies, Australian National University.
Van Ness, S. 1990. Changes in an obsolescing language: Pennsylvania German in West Virginia
(Vol. 336), Gunter NarrVerlag.
Wurm, S., Laycock, D.C., Mühlhäusler, P. & Dutton, T. 1979. Handbook of New Guinea Pidgin.
Canberra: Pacific Linguistics.
Some issues in the linking between syntax
and semantics in relative clauses
1. Introduction1
1. I would like to thank Ranko Matasović and Dejan Matić for comments on an earlier draft.
Robert D. Van Valin, Jr.
The head noun of the relative clauses in both of these sentences is bestya ‘horse’.
In the externally-headed relative clause in (1a), the head noun appears after the
relative clause and is case-marked for its matrix clause function, namely, that of
subject. It does not occur in the relative clause at all, and consequently there is no
direct marking of its function in the embedded clause at all. Conversely, in the
internally-headed relative clause in (1b), the head noun appears inside it and is
case marked for its function in it, namely, that of direct object. There is no direct
indication of its matrix clause function. Thus, the hearer faces the problem of
determining the function of the head noun within the relative clause in externally-
headed relative clauses and the problem of determining the function of the head
noun within the matrix clause in internally-headed relative clauses.
Standard analyses of relative clauses are grounded in the analysis of English-
type externally-headed constructions involving a relative pronoun, e.g. The horse
which the man bought was a good horse, despite their typological rarity, and such
accounts typically involve the mechanisms used for handling long-distance
dependencies, i.e. movement rules (or the equivalent, e.g. slash categories) and
phonologically null elements (e.g. traces, empty WH-operator). With respect to
internally-headed relative clauses, generative analyses (e.g. Cole 1987; Basilico
1996) posit a null external head, so that they are structurally similar to externally-
headed relative clauses; furthermore, they posit covert movement of the head
noun, usually to the same position occupied by the head noun in externally-
headed relative clauses.
In this paper the issue of the determination of the function of the head noun in
the clause in which it does not appear will be carried out in a theory which eschews
all of these theoretical mechanisms, namely, Role and Reference G rammar [RRG]
(Van Valin 2005; Van Valin & LaPolla 1997). RRG is a a p arallel a rchitecture theory
2. Abbreviations: acc - accusative, anti - antipassive, atv - active voice, clm - clause-linkage
marker, dat - dative, det - determiner, ehrc - externally-headed relative clause, erg - ergative,
evid - evidential, ihrc - internally-headed relative clause, ls - logical structure, nom - nomi-
native, pass - passive, past - past tense, perf - perfect, prcs - precore slot, prfv - perfective,
psa - privileged syntactic argument, refl - reflexive, rel - relative pronoun/marker, rp -
reference phrase.
Some issues in the linking between syntax and semantics in relative clauses
syntactic representation
Discourse-pragmatics
Linking
algorithm
semantic representation
3. See Van Valin (2006) for discussion of how the RRG linking system fits into models of
sentence processing.
Robert D. Van Valin, Jr.
2. The RRG analysis of clause structure and the linking algorithm
RRG features a non-endocentric syntax; that is, the major phrasal categories
are not projections of lexical heads. The head of the clause is the nucleus, which
contains the predicate, which may be a verb, a combination of verbs, a nominal
phrase, an adjective phrase, or a prepositional phrase. Argument expressions
are analyzed as ‘reference phrases’ [RP] (Van Valin 2008), which are typi-
cally headed by a nominal expression but need not be in many languages. The
approach to clause structure is called ‘the layered structure of the clause’, with
a nucleus, a core containing the nucleus and the arguments of the predicate, a
clause, which contains the core and optionally a pre-core slot [PrCS], which is
the position in which WH-elements and relative pronouns occur in languages
like English and German; there are potentially adjuncts modifying each of these
layers, and such adjunct modifiers occur in a periphery modifying the particu-
lar layer involved. In Figure 2 the layered structure of What did Robin show to
Pat in the library yesterday? is given. Grammatical categories like tense, aspect,
modality and illocutionary force, termed ‘operators’ in RRG, are represented in
a separate projection of the clause which is not given here; the auxiliary verb
did would be attached to the operator projection, since its function is to express
tense and illocutionary force.
SENTENCE
CLAUSE
RP NUC PP
PRED
RP V PP ADV
The be¢ in (2a) indicates that this is an attributive construction; it is not a reflex
of English be, and it would occur in the semantic representation of attributive
predications in languages which lack a copula. This is illustrated in Figure 3.
In Lakhota the stative verb hą́ska ‘tall’ occurs directly inflected for its subject,
and there is no copula or other element corresponding to English be; yet it and its
English translation have the same semantic representation.
Two additional components essential to the linking system are the seman-
tic macroroles, actor and undergoer, and the notion of ‘privileged syntactic argu-
ment’ [PSA], which replaces the notion of ‘subject’ in RRG. There are substantial
differences between PSA and subject, but for the purposes of this paper, they will
be taken to be roughly equivalent.4 In (2a), Kim would be the undergoer of the
stative predicate tall, while in (2b) Kim would be the actor of the activity verb sing.
In (2c) I is the actor and the window is the undergoer, and likewise in (2d) Kim
is the actor and the window is the undergoer. The relationship between seman-
tic macroroles and PSA can be summarized as follows: in an accusative language
like English or German, the actor is the default choice for PSA in a core with
a transitive verb, with the undergoer the non-default choice requiring a special
construction, namely the passive.
4. See Van Valin (2009) for detailed discussion of the differences between subject and PSA.
Robert D. Van Valin, Jr.
The linking from syntax to semantics in (2d) is given in Figure 4; the numbers
refer to the steps in (3).
The parser outputs a labeled tree structure, step (3a). Because English is a
language with a voice system, an important first step in (3b) is to identify the voice
of the verb, because it signals the semantic role of the PSA (‘subject’), which is the
first RP in the core in English.
In this instance the voice is active, meaning that the PSA, Kim, is an actor. The
immediately post-nuclear RP, the window, must therefore be an undergoer. The
next step, (3c), is to retrieve the LS for smash from the lexicon and assign mac-
roroles, if possible. In this case it is straightforward: the x argument would be the
actor and the y argument the undergoer. In the final step, (3d), the results of the
second and third steps are matched up: Kim is an actor, the actor is the x argument
in the verb’s LS, and therefore Kim is the x argument. The same reasoning applies
to the other argument, yielding the conclusion that the window is the y argument.
The Completeness Constraint, which states that all referring expressions in the
5. See Van Valin (2005: 136–49) for a detailed discussion of semantics-to-syntax linking in
simple sentences.
Some issues in the linking between syntax and semantics in relative clauses
SENTENCE
PARSER CLAUSE
3a
CORE
RP NUC RP
PRED
V
3b
Voice? – Active Kim smashed the window
∴ PSA = Actor
Actor Undergoer
3d
Actor Undergoer
3c
LEXICON [do¢ (x, Ø)] CAUSE [INGR smashed¢ (y)]
Figure 4. Example of linking from syntax to semantics in a simple English clause
syntax must be linked to something in the semantics (and vice versa), is satisfied.
When it comes to relative clauses, the linking is more complex, but it follows these
basic principles, in addition to some construction-specific rules.
ronoun does not signal the function of the head noun inside the relative clause,
p
and the complementizer is likewise invariant. The only clue is that the comple-
mentizer that may be omitted when the head noun is not the subject of the relative
clause, as in (4b, c). Thus, English presents a good example of a language in which
there is no formal identification of the function of the head noun within the rela-
tive clause.
It should be noted that there are languages with EHRCs in which the f unction
of the head noun in the embedded clause is clearly signaled grammatically. In
German and a few other languages,6 there is a relative pronoun whose case
indicates the function of the head noun inside the relative clause. This is illustrated
in (5).
(5) a. Ich sah den Mann, [dem Maria das Buch gegeben hat].
I saw the.acc man, rel.dat m. the book given has
‘I saw the man [to] whom Maria gave the book.’
b. Der Mann, [den Maria sah], ist Spion.
the.nom man, rel.acc m. saw is Spy
‘The man who Maria saw is a spy.’
The head noun Mann ‘man’ is case-marked for its matrix clause function (accusative
for direct object in (5a) and nominative for subject in (5b)), and the relative pro-
noun is case-marked for the function of the head noun inside the relative clause
(dative for indirect object in (5a) and accusative for direct object in (5b)). Hence
there is no problem in principle with ascertaining the function of the head noun
within the relative clause.
This can also be achieved by strictly syntactic means. Since the 1970’s it has
been recognized that some languages have strong restrictions on the possible func-
tion of the head noun within the relative clause (see Keenan & Comrie 1977), and
the strongest restriction is that the head noun can only serve as the subject of the
relative clause. This is exemplified in the following Malagasy data (Keenan 1976).
(6) a. Na-hita ny vehivavy
[(izay) nan-asa ny
prfv.atv-see det woman clm prfv.atv-wash det
zaza] Rakoto.
child Rakoto
‘Rakoto saw the woman that washed the child.’
*‘Rakoto saw the woman that the child washed.’
6. Indo-European languages provide the majority of languages with relative pronouns;
outside of Indo-European relative pronouns, especially case-marked relative pronouns, are
rare. (R. Matasović, personal communication).
Some issues in the linking between syntax and semantics in relative clauses
In the first two Malagasy examples, the head noun, which precedes the rela-
tive clause, can only be interpreted as the the subject of the relative clause; since
Malagasy is an accusative language and the voice of the verb is active, it is inter-
preted as the actor. In (6c) it is interpreted as the undergoer, because the voice of
the verb in the relative clause is passive. Because of this syntactic restriction the
function of the head noun within the relative clause is always unambiguous and
immediately recoverable.7
The three are central aspects to the RRG analysis of EHRCs are (1) the syn-
tactic representation, (2) the semantic representation, and (3) the construction-
specific linking rules. The syntactic structures assigned to I saw the window (that)/
which Kim smashed are given in Figure 5.
Within the layered structure of the RP restrictive modifiers such as adjec-
tives and restrictive relative clauses are modifiers at the nuclearR level and occur
in the periphery modifying the RP nucleus. There is no empty RP-slot for the
head noun inside the core of the EHRCs; this is consistent with the point made in
§1 that RRG does not allow phonologically null elements in syntactic representa-
tions. In the first example the relative clause is marked by the complementizer
that, which functions as a clause-linkage marker, and in the second the relative
pronoun which appears in the PrCS. It is possible to omit that, yielding a structure
lacking a clause-linkage marker or a PrCS.
The second aspect is the semantic representation of the sentence. Like adjec-
tives, relative clauses express attributes of the head noun, e.g. the tall man vs. the
man who is tall, and accordingly the semantic representation of the relative clause
is represented as filling the same slot in an attributive predication that an adjective
does (see (2a), Figure 3) i.e. in be¢ (x, [pred¢]), the LS of the relative clause occurs
in the pred¢ slot. Hence the LSs for the EHRCs in Figure 5 are given in (7)
7. See Van Valin (2005: 260–65) for detailed discussion of the linking in Malagasy relative
clauses.
Robert D. Van Valin, Jr.
SENTENCE SENTENCE
CLAUSE CLAUSE
CORE CORE
RP RP RP RP
NUC NUC
CORER CORER
PRED PRED
NUCR PERIPHERYR NUCR PERIPHERYR
V V
(CLM )CLAUSE CLAUSE
V V
I saw the window (that) Kim smashed I saw the window which Kim smashed
The attributive LS fills the second argument slot of see¢, and the argument which
is shared between the matrix and embedded clauses is indicated by dashed under-
lining. It is also co-indexed with one of the argument positions in the LS of the
embedded predicate, and this co-indexed position may be filled by a relative
pronoun.8 The clause-linkage marker (complementizer) that is not represented
in the LS of the relative clause; it would be a property of the syntactic template for
EHRCs.
The third aspect is the construction-specific linking rules. In addition to the
general syntax-to-semantics linking principles in (3), the rules in (8) apply to the
linking of EHRCs.
8. If the head noun functions as an adjunct PP, the adjunct PP is represented by a LS and the
co-indexing would be between the head noun and one of its argument positions.
Some issues in the linking between syntax and semantics in relative clauses
The interaction of the rules in (3) and (8) is illustrated in Figure 6, which depicts
the syntax-to-semantics linking in the EHRC in (7a).
SENTENCE
PARSER
3a CLAUSE
CORE
RP RP
NUC
CORER
PRED
NUCR PERIPHERYR
V
CLM CLAUSE
CORE
N
RP NUC
PRED
3b
V 3b
Main verb:
Voice?–Active Relative clause verb:
∴PSA = Actor Voice?–Active
I saw the window that Kim smashed ∴PSA = Actor
3d Actor Undergoer
3c
[do¢ (y, Ø)] CAUSE [INGR smashed¢ (z)] LEXICON
(8a)
Actor Undergoer be¢ (xi, [[do¢ (Kim, Ø)] CAUSE [INGR smashed¢ (zi)]]) be¢ (x, [pred¢])
3c
LEXICON see¢ (v, w) Coindexing (8b)
RC LS substitution (8c)
[see¢ (1sg, [be¢ windowi, [do¢ (Kim, Ø)] CAUSE [INGR smashed¢ (zi)]]
SENTENCE
PARSER
3a CLAUSE
CORE
RP RP
NUC
CORER
PRED
NUCR PERIPHERYR
V
CLAUSE
PrCS CORE
N
RP RP NUC
3b PRED
3b
Main verb:
Voice?–Active V Relative clause verb:
∴PSA = Actor Voice?–Active
I saw the window which Kim smashed ∴PSA = Actor
3d Actor Undergoer
3c
[do¢ (y, Ø)] CAUSE [INGR smashed¢ (z)] LEXICON
(8a)
Actor Undergoer be¢ (xi, [[do¢ (Kim, Ø)] CAUSE [INGR smashed¢ (whichi)]]) be¢ (x, [pred¢])
3c
LEXICON see¢ (v, w) Coindexing (8b)
RC LS substitution (8c)
[see¢ (1sg, [be¢ windowi, [do¢ (Kim, Ø)] CAUSE [INGR smashed¢ (whichi)]]
The linking principles in (3) apply to the relative clause just as they do to the
main clause. When there is no relative pronoun, then the linking in the EHRC will
be incomplete at this point, because the relative clause is missing an argument.
After the linking within the main and subordinate clauses has been carried out
following (3), the rules in (8) come into play. The first step, (8a), is the creation of
the derived LS for the relative clause, an example of which was given in (7a′). The
second step, (8b), involves the co-indexing of the first argument of the attributive
LS with an argument position in the embedded LS. If there is no relative pronoun
(which is the usual case cross-linguistically), then the head noun is co-indexed
with a variable in the LS; this variable will remain lexically unfilled. This is the case
in (7a). The final step, (8c), involves substituting the attributive LS for the head
Some issues in the linking between syntax and semantics in relative clauses
noun in the matrix LS and replacing the variable in the first argument position in
the attributive LS with the head noun. The head noun is the noun in whose nuclear
periphery the relative clause occurs. Because the lexically unfilled variable, z in
Figure 6, is co-indexed with the head noun window, the Completeness Constraint
is satisfied. The presence or absence of the complementizer that does not affect the
linking.
The linking in EHRCs with relative pronouns differs in two respects; first,
the relative pronoun in the PrCS must be linked to an argument position in the
semantic representation, following step (3d), and second, the co-indexing in step
(8b) necessarily involves the relative pronoun rather than a lexically unfilled vari-
able. This is illustrated in Figure 7. During the first phase of the linking, following
(3), the relative pronoun gets linked to the undergoer of the LS, due to its being the
only unlinked argument position in the EHRC’s LS, and this is required in order to
satisfy the Completeness Constraint.
Thus, the general syntax-to-semantics linking algorithm, summarized in (3),
together with the EHRC-specific linking rules, provides an account of how the
head noun is correctly interpreted within the EHRC, and this is achieved without
syntactic transformations or phonologically-null entities.
Unlike in EHRCs the function of the head noun within an internally-headed rela-
tive clause [IHRC] is readily recoverable, since the head noun occurs inside the
IHRC; the issue, as noted in §1, is determining the main clause function of the
head noun. An example of an IHRC from Bambara (Bird 1968) is given in (9a)
along with its LS in (9b).
(9) a. [Ne ye so min ye] tye ye san
1sg past horse rel see man past buy
‘The man bought the horse that I saw.’ Bambara (Bird 1968)
b. [do¢ (tye, Ø)] CAUSE [BECOME have¢ (tye, [be¢ (xi, [see¢ (ne, soi)])])]
Bambara is an SAuxOV language, and inside the IHRC the head noun is marked
by the relativizer min. The LS has the same embedded attributive LS as in EHRCs,
but it differs from the ones in (7a′, b′) in that it is the ‘external variable’, i.e. the
argument of the attributive LS, which is left lexically unfilled. The structure of
(9) is given in Figure 8; the past tense markers ye would be linked to the operator
projection, which is not given here. Because the matrix undergoer argument so
‘horse’ is part of a relative clause, it precedes the actor ‘man’ in order to avoid a
center embedding.
Robert D. Van Valin, Jr.
SENTENCE
CLAUSE
CORE
RP RP NUC
CLAUSE
CORE
RP RP NUC PRED
PRED
V V
The construction-specific linking rules for IHRCs are given in (10); they are
very similar to the rules for EHRCs.
(10) Rules governing linking from syntax to semantics in IHRCs
a. Retrieve from the lexicon an attributive LS and substitute the LS of
the verb in the relative clause for the second argument.
b. Co-index the first argument in the attributive LS with the argument
in the relative clause LS identified as the head noun.
c. Insert the attributive LS into the open argument position in the
matrix LS.
There are two differences between the principles in (8) and those in (10). First, in
(10b) there is no option involving a relative pronoun, since IHRCs never involve
relative pronouns, and second, in (10c) there is no replacement of the external
variable by the head noun, since it is already present in the LS of the relative clause.
The linking from syntax to semantics for (9a) would go as in Figure 9.
The parser outputs a labelled tree structure, (3a). Step (3b) is executed
with respect to the relative clause as well as the main clause; because Bambara
has no passive construction and no case marking on RPs, the semantic roles are
determined primarily by word order and adpositional marking. In the relative
Some issues in the linking between syntax and semantics in relative clauses
SENTENCE
PARSER CLAUSE
3a
CORE
RP RP NUC
CLAUSE
CORE
RP RP NUC PRED
PRED
V V
Actor Undergoer
3c Actor
LEXICON see¢ (y, z)
Undergoer 3c
(10a)
[do¢ (v, Ø)] CAUSE [BECOME have¢ (v, w)] LEXICON
[be¢ (x, [pred¢])] [be¢ (xi, [see¢ (ne, soi)])]
[do¢ (tye, Ø)] CAUSE [BECOME have¢ (tye, [be¢ (xi, [see¢ (ne, soi)])])]
clause, ne ‘I’ is the ‘subject’ and actor, as it is RP immediately before the tense
auxiliary, and so ‘horse’ is the undergoer, as it follows the auxiliary and precedes
the verb; it is marked by the relativizer min, indicating that it is the head noun. In
the main clause, tye ‘man’ is the RP immediately before the tense auxiliary, hence
it is the main clause ‘subject’ (actor). In step (3c) the LSs for ye ‘see’ and san ‘buy’
are retrieved, and macroroles are assigned. In the next step, (3d), the information
from steps (3b) and (3c) are matched up, yielding the linking of ne ‘I’ to the y
argument of the LS for ye ‘see’ and so ‘horse’ to the z argument position in the
Robert D. Van Valin, Jr.
r elative clause, and the linking of tye ‘man’ to the v argument of the LS of san ‘buy’;
there is nothing at this point to link to the w argument in this LS.
At this point the rules in (10) come into play. Following (10a) an attributive LS
is retrieved from the lexicon, and the LS of the IHRC, in this case see¢ (ne, so), is
inserted as the predicate¢ in the attributive LS. The next step, (10b), is to co-index
the lexically unfilled x variable with the head noun so ‘horse’ (its head noun sta-
tus is indicated by the double underlining). The final step, (10c), is to insert the
attributive LS into the unlinked argument position in the matrix LS, satisfying the
Completeness Constraint and yielding the LS in (9b).
While there is no major problem in terms of identifying the function of the
head noun within an IHRC, it is not always as clear which RP should be inter-
preted as the head noun, as it is in Bambara with its relativizer min. There is, for
example, no corresponding marker in the Quechua example in (1b). Languages
with IHRCs have developed different strategies for indicating the head within the
IHRC. In Lakhota, for example, the head noun must be indefinite (Williamson
1987), but this is only distinctive if there is only one indefinite RP in the clause; if
both RPs are indefinite, then the result is ambiguity.
5. Conclusion
The purpose of this paper has been to investigate the linking between syntax
and semantics in relative clauses, both EHRCs and IHRCs, within a monostra-
tal syntactic theory that disallows phonological null elements in syntactic repre-
sentations, Role and Reference Grammar. It has been necessary to augment the
general syntax-to-semantics linking algorithm with construction-specific link-
ing rules, given in (8) and (10). This is very much in the spirit of RRG, in which
construction-specific linking properties interact with general linking properties
(cf. Van Valin 2005, §5.1.1). The rules in (8) were presented as being for EHRCs,
and those in (10) for IHRCs, but a close examination of them and their interaction
with the general linking principles suggests that there is but a single set of rules,
with the differences following from the different properties of EHRCs and IHRCs.
The proposed unified relative clause linking rules are given in (11).
(11) Rules governing linking from syntax to semantics in relative clauses
a. Retrieve from the lexicon an attributive LS and substitute the LS of the
verb in the relative clause for the second argument.
b. Co-index the first argument in the attributive LS with the argument in
the relative clause LS identified with the head noun.
c. Insert the attributive LS into the head noun’s argument position in the
matrix LS, with the head noun incorporated into the attributive LS.
Some issues in the linking between syntax and semantics in relative clauses
(11a) is the same as both (8a) and (10a). (11b) specifies the co-indexing of the
attributive LS argument with an argument within the relative clause LS, the argu-
ment “identified with” the head noun. In an EHRC with a relative pronoun, this
would mean co-indexing with the relative pronoun; in an EHRC with no relative
pronoun, this would mean co-indexing with the unlinked argument in the relative
clause LS; it is unlinked because there is no RP in the relative clause corresponding
to the head noun, hence there is nothing to link to it. Finally, in the case of IHRCs,
it is the head noun itself which is co-indexed with the attributive LS argument. The
final rule, (11c), specifies that the attributive LS is merged into the matrix LS in the
position of the head noun argument; it furthermore states that the head noun must
be incorporated into the attributive LS. How can this be accomplished? In an EHRC
with a relative pronoun, as in Figure 7, all argument positions in the r elative LS are
lexically filled, but the first argument position is not, and therefore the head noun
can fill that argument slot. In an EHRC without a relative pronoun, as in Figure 6,
the head noun is not a constituent of the relative clause and therefore cannot be
inserted into the unlinked argument position in it; it can, however, fill the argument
position in the attributive LS, which is co-indexed with the unlinked argument
position. Finally, the head noun fills an argument position in the relative clause LS
in an IHRC, and consequently this requirement is met by definition. Thus, despite
the formal differences between EHRCs and IHRCs, the rules in (3) and (11) can
account for the linking from syntax to semantics in both types of relative clauses,
without invoking movement (overt or covert) or phonologically null elements.
References
Basilico, D. 1996. Head position and internally headed relative clauses. Language 72: 498–532.
Bird, C. 1968. Relative clauses in Bambara. Journal of West African Languages 5: 35–47.
Cole, P. 1987. The structure of internally headed relative clauses. NLLT 5: 277–302.
Jackendoff, R. 2002. Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford:
OUP.
Keenan, E.L. 1976. Remarkable subjects in Malagasy. In Subject and Topic, C. Li (ed.), 247–301.
New York NY: Academic Press.
Keenan, E.L. & Comrie, B. 1977. Noun phrase accessibility and universal gramar. Linguistic
Inquiry 8: 63–99.
Van Valin, R. 2005. Exploring the Syntax-Semantics Interface. Cambridge: CUP.
Van Valin, R. 2006. Semantic macroroles and sentence processing. In Semantic Role Universals
and Argument Linking: Theoretical, Typological and Psycho-/neurolinguistic Perspectives, I.
Bornkessel, M.Schlesewsky, B. Comrie & A. Friederici (eds), 263–302. Berlin: Mouton de
Gruyter.
Van Valin, R. 2008. RPs and the nature of lexical and syntactic categories in Role and Reference
Grammar. In Investigations of the Syntax-Semantic-Pragmatics Interface, R.D. Van Valin, Jr.
(ed.), 161–178. Amsterdam: John Benjamins.
Robert D. Van Valin, Jr.
Van Valin, R. 2009. Privileged syntactic arguments, pivots and controllers. In Studies in Role and
Reference Grammar, L. Guerrero, S. Ibáñez Cerda & V. Belloro (eds), 45–68. México City:
UNAM.
Van Valin, R. & Lapolla, R. 1997. Syntax: Structure, Meaning and Function. Cambridge: CUP.
Williamson, J. 1987. An indefiniteness restriction on relative clauses in Lakhota. In The Repre-
sentation of (In)definites, E. Reuland & A. ter Meulen (eds), 168–190. Cambridge MA: The
MIT Press.
part ii
Uto-Aztecan
Relative clauses and nominalizations
in Yaqui
This paper focuses on the different types of relative clauses (RCls) in Yaqui,
a Uto-Aztecan language spoken in northwestern Mexico, demonstrating,
on the basis of the nominalization properties exhibited by these types of
construction, that they are in fact better analyzed as noun phrases headed
by a nominalized verb. After reviewing the main features used in the literature
to define RCls and presenting the major relativization types identified
cross-linguistically, the different constructions that exist in Yaqui to relativize
subject, object, indirect object, oblique and locative complements, are
described. As these constructions show different nominalization properties,
the nominal (non-finite) and verbal (finite) characteristics associated with
the Yaqui RCls are explored in greater detail in the next section. Based on this
survey of finiteness in Yaqui RCls, relativization in Yaqui is then considered as
a nominalization process in which a finite verbal clause is adjusted to a noun
phrase in order to be used as a modifier of the head noun. The last section
provides a discussion about the connection between relativization
and nominalization, and the characteristics of two different types of
nominalization: lexical and clausal. Finally, the notion of referentialization,
the act of referring to an entity, is proposed to refer to the nominalization
function and to explain how clausal nominalization can be the basis of
relativization in a large number of languages around the world, as is the case
in Yaqui.
1. Introduction
This paper studies the different types of relative clauses (RCls) existing in Yaqui,
a Uto-Aztecan language spoken in the northwest of Mexico. The main goal is to
demonstrate, on the basis of the nominalization properties exhibited by these
Albert Álvarez González
types of construction, that they are in fact better analyzed as noun phrases headed
by a nominalized verb.1
In Section 2, I will review the main features used in the literature to define RCls
and I will present the different relativization types identified cross-linguistically
according to three criteria: (a) the position of the clause in relation to the head;
(b) the mode of expression of the relativized noun phrase within the RCl; and
(c) the kind of grammatical relations that can be relativized. These criteria will
be used in Section 3 for the presentation of the Yaqui RCls. I will then describe
the different constructions existing in Yaqui to relativize subject, object, indirect
object, oblique and locative complements. In addition to the positional types of
RCls and to the relativization strategies, special attention will be paid to the types
of markers involved in relativization, as well as the presence or absence of agree-
ment between the head noun and the RCl.
As these constructions show different nominalization properties, in Section 4
I will explore in greater detail the nominal (non-finite) and verbal (finite) charac-
teristics associated with the Yaqui RCls. This survey of finiteness in Yaqui RCls will
lead us to consider that relativization in Yaqui is done via a nominalization process
in which a finite verbal clause is adjusted to a noun phrase in order to be used as a
modifier of the head noun. As part of this adjustment process, nominalization and
finiteness will thus appear to be a matter of degree.
Based on recent work by Shibatani (2009) concerning the connection between
relativization and nominalization in different languages around the globe, I will
discuss in Section 4 whether it is appropriate to speak of relativization regarding
the Yaqui constructions under consideration. The discussion here will also allow
us to further define the notion of nominalization, by identifying two different
types – lexical and clausal nominalizations – (Koptjevskaja-Tamm 2006; Genetti
et al. 2008; Shibatani 2009), and by pointing out the main contrasts between them.
In the final observations, the notion of referentialization will be proposed to dis-
tinguish between the transcategorial process of nominalization and the function
associated with this process (the act of referring to an entity). This functional
notion will finally be used to explain how clausal nominalization can be the basis
of relativization in a large number of languages around the world, as is the case in
Yaqui.
1. The data included in this paper are the result of direct elicitation or taken from texts in-
cluded in Estrada et al. (2004) and in Estrada and Alvarez (2008). I deeply thank Melquiades
Bejípone Cruz (a Yaqui native speaker) for his kind and generous help during the research
for this article.
Relative clauses and nominalizations in Yaqui
2.1 Definition
Traditionally, RCls are believed to represent a type of clause combinations. They
are considered to be a multi-verb construction that is part of the different construc-
tion types of interclausal connectivity, such as serial verbs, complement clauses,
adverbial clauses, clause chains and coordination (Payne 1997: 306). Syntactically,
RCls would be dependent on the matrix clause, exhibiting a type of subordina-
tion. Andrews (2007: 206) points out this relation of dependency and embedding
between two clauses when he defines a RCl as “a subordinate clause which delimits
the reference of a noun phrase by specifying the role of the referent of that noun
phrase in the situation described by the RCl.” This definition insists more on the
function of the RCl than the definition provided, for example, by Comrie and
Kuteva (2005), which considers a RCl as “a clause narrowing the potential refer-
ence of a referring expression by restricting the reference to those referents of which
a particular proposition is true.” So, a RCl would function as a nominal modifier
(Keenan 1985) by restricting the semantic domain covered by a syntactic constitu-
ent (typically a noun termed the “head noun” or the “relativized noun”). Because
the function of an RCl would be to act as a type of noun modifier,2 RCls are also
referred to as adjective clauses or as a type of adjectival subordination.
Interestingly, this same function is shared by other constructions in which the
restriction is not introduced by a typical clause (for example, by non-finite con-
structions such as participial or nominalized constructions). In Section 4, I will
discuss the relation between relativization and nominalization in further detail.
As part of this discussion, the defining features of RCls presented above will be
subjected to critical appraisals in view of the nominalization approach.
2. Givón (2001: 175) considers RCls as “clause-size modifiers embedded in the noun phrase.”
Albert Álvarez González
a particular syntactic function within the main clause, it always has a coreferent
within the RCl. Depending on the mechanisms by which each language expresses
the syntactic-semantic role of the head noun in a RCl, we can distinguish the fol-
lowing relativization strategies (Comrie & Kuteva 2005):
The further to the right a syntactic function is placed on the hierarchy, the more
difficult it is to be relativized. So, if a language can relativize a given position on the
hierarchy, all positions to the left will also be relativizable. Additionally, Keenan
and Comrie (1977) point out that all languages with RCls allow relativization of
subjects, the easiest grammatical relation to relativize.
3. This community settled in the USA from its original homeland in the south of the
neighboring state of Sonora, in Mexico, at the beginning of the twentieth century, fleeing
persecution by the Mexican dictator, Porfirio Diaz.
Relative clauses and nominalizations in Yaqui
Examples (1) and (2) show us that the basic order in the transitive construc-
tion is SOV. In noun phrases, nominative case is unmarked, whereas accusative
case is marked by the suffix -ta, with the exception of plural objects, because
there is an incompatibility in Yaqui between the accusative marker and the plural
marker.
4. Abbreviations: acc – accusative, advz – adverbializer, caus – causative, com – comitative,
comp – completive, dat – dative, dem – demonstrative, des – desiderative, det – determiner,
dir – directional, fut – future, gen – genitive, impf – imperfective, inst – instrumental,
loc – locative, neg – negation, nom – nominative, nmlz – nominalizer, obl – oblique, pas –
passive, perf – perfective, pl – plural, poss – possessive, red – reduplication, refx – reflexive,
rel – relativizer, res – resultative, sg – singular, vbz – verbalizer.
Albert Álvarez González
When the head noun is the object of the main clause, the RCl is marked by the
suffix -ta as in (5), which illustrates the case-agreement between the head noun
and the RCl. The presence of the suffix -me indicates that the head noun chu’u is
the subject of the RCl.
(5) Joan uka chu’u-ta [Maria-ta ke’e-ka-m]-ta me’a-k
John det.acc dog-acc Mary-acc bite-perf-rel-acc kill-perf
‘John killed the dog that bit Mary.’
Recall that in (4) the subject head noun and the RCl did not exhibit number agree-
ment; the same pattern is presented in (6), whereby there is a lack of agreement
between the object head noun and the postnominal RCl. The RCl here contains an
intransitive verb with a locative adjunct.
(6) U-me wakas-im ba’a-m [tina-po ayuka-me] lu’uta
det-pl cow-pl water-pl tub-loc exist-rel finish
‘The cows are finishing the water that is in the tub.’
In Yaqui, the indirect object can be marked by a dative suffix that is, in fact, the
combination of the accusative suffix -ta and the directional suffix -u. The subject
relativization of a head noun functioning as an indirect object of the main clause
results, once again, in case-agreement between the head noun and the RCl, as
shown in (7).
Relative clauses and nominalizations in Yaqui
When the ditransitive construction involves an indirect object in the plural, its
plurality prevents the presence of the dative marker, as is usual in Yaqui, and no
number-agreement appears then between the head noun and the subject-RCl. A
relevant example is (8).
(8) U yoeme u-me usi-m [yeewe-me] kaka-m maka-k
det man det-pl child-pl play-rel candy-pl give-perf
‘The man gave candies to the children who are playing.’
The pluralization of the subject head noun implies the presence of number-
agreement between the head noun and the RCl, as shown in (10).
(10) U-me bisikleeta-m [in jinu-ka-’u]-m sikili
det-pl bicycle-pl 1sg.gen buy-perf-rel-pl red
‘The bicycles that I bought are red.’
If the head noun is the direct object of the main clause, as in (11), no case agree-
ment occurs between the head noun and the object RCl, in contrast to cases
involving subject RCls.
(11) Joan inika bachi-ta [em jinu-ka-’u] bwa’a-ka
John dem.acc corn-acc 2sg.gen buy-perf-rel beat-perf
‘John ate this corn that you bought.’
Number agreement between a plural head noun and a direct object RCl still occurs
if we have a head noun as direct object of the main clause. The subject of the RCl
Albert Álvarez González
remains in genitive form and its object remains in accusative. The order of ele-
ments within the RCl corresponds to the basic SOV word order. Consider (12).
(12) mache’eta-m [em nee reuwa-ka-’u]-m ne jippue
machete-pl 2sg.gen 1sg.acc lend-perf-rel-pl 1sg.nom have
‘I have the machetes that you lent me.’
Both the absence of case agreement, and the presence of number agreement
between the head noun and the direct object RCl, are exemplified, again in (13)
and (14), in which a head noun functions as the indirect object of the main clause.
(13) U jamut yoem-ta-u [nim ke’e-ka-’u]
det woman man-acc-dir 1sg.gen bite-perf-rel
rebo’osam jinu-k
mantilla buy-perf
‘The woman bought the mantilla from the man I bit.’
(14) U jamut yoeme-me-u [nim ke’e-ka-’u]-m
det woman man-pl-dir 1sg.gen bite-perf-rel-pl
rebo’osam jinu-k
mantilla buy-perf
‘The woman bought the mantilla from the men I bit.’
If the subject head noun is marked as plural, it will trigger double number-
agreement with two elements within the RCl: the third person plural dative
pronoun (ameu), and the plural marker -m suffixed to the verb of the RCl.
(16) U-me yoeme-m [em rebo’osam ameu
det-pl man-pl 2sg.gen mantilla 3pl.dat
Relative clauses and nominalizations in Yaqui
jinu-ka-’u]-m naamuk-ia-me
buy-perf-rel-pl get_drunk-res-pl
‘The men from whom you bought the mantilla are drunk.’
The indirect object relativization of a direct object head noun is seen in (17)
involving a singular head noun, and in (18) involving a plural head noun. These
constructions show the absence of case agreement (17) and, once again, the trig-
gering of the double number agreement (e.g. the plural dative pronoun and the
plural suffix) by the object head noun (18).
(17) U chu’u yoem-ta [em rebo’osam au
det dog man-acc 2sg.gen mantilla 3sg.dat
jinu-ka-’u]
ke’e-ka
buy-perf-rel bite-perf
‘The dog bit the man from whom you bought the mantilla.’
The same characteristics can be seen in the indirect object relativization of an indi-
rect object head noun, illustrated in (19) and (20). The postnominal RCl lacks
case-agreement with the head noun, and exhibits a double number agreement.
(19) Juan yoem-ta-u [em rebo’osam au jinu-ka-’u]
John man-acc-dir 2sg.gen mantilla 3sg.dat buy-perf-rel
karo-ta nenka-k
car-acc sell-perf
‘John sold the car to the man from whom you bought the mantilla.’
just seen, the postnominal RCl contains a resumptive pronoun that indicates the
case role of the anaphorically referenced head noun in the RCl. Similarly to the
indirect object relativization cases, the subject of the RCl in oblique relativization
appears to be in genitive form, and the RCl verb is marked by the suffix -’u.
(21) U baabu [itom pu’ato-m a-e
det mud 1pl.gen plate-pl 3sg.acc-inst.sg
joa-’u] ba’a-yejte-la
make-rel water-sit_down-res
‘The mud we are making the plates with is liquid.’
the direct object head noun (marked by the suffix -ta), nor does number a greement
occur between the plural direct object head noun and the oblique RCl.
The oblique relativization with an indirect object head noun doesn’t involve any
change in relation to the examples above. The postnominal RCl has an overt case-
marked reference (the oblique pronoun), the subject of the RCl is in the genitive
form, and there isn’t case or number agreement between the head noun and
the RCl verb. The following examples illustrate the relativization of comitative
complements.
The locative relativization of a head noun functioning as direct object of the main
clause doesn’t involve case agreement between the head noun and the RCl (33), or
number agreement with a plural head noun (34).
chu’u-ta jinu-k
dog-acc buy-perf
‘I bought a dog for the house where I was born.’
i. If we consider the positional type, Yaqui RCls are external-head RCls and they
are always postnominal, i.e. the RCl always occurs after the head.5
ii. Concerning the expression type, Yaqui RCls use two different strategies that
are contingent on the relativized syntactic function: the gap strategy is used
in subject, direct object and locative relativization, whereas the pronoun
retention strategy is used in the indirect object and oblique relativization.
iii. Regarding the type of markers involved, Yaqui RCls exhibit the use of three
different suffixes: the suffix -me is used in subject relativization, the suffix
-’u occurs in non-subject relativization except for locative (direct object,
indirect object and oblique), whereas the suffix -’Vpo is used in locative
relativization.
iv. Considering the relativized grammatical relations, Yaqui does not exhibit
possessor relativization, which is in accordance with Keenan and Comrie’s
(1977) Noun Phrase Accessibility Hierarchy, which identifies the possessor
relation as the most difficult grammatical relation to relativize, consequently
occupying the lowest position in the hierarchy.
v. As for the inheritance capacities of the head noun properties for the RCl,
Yaqui exhibits differences that depend on the syntactic function of the head
noun within the RCl: case agreement is present in subject relativization
only, whereas number agreement only appears in direct and indirect object
relativization.
5. However, Yaqui RCls are not always embedded within the main clause. They can also be
adjoined to the main clause. In this case, the RCl that is not a constituent of the main clause
appears to be postposed, especially with large RCls (i.e., if the RCl contains an object). There is
no clear evidence as yet that this switch of position involves a distinction between a restrictive
and a non-restrictive function.
Albert Álvarez González
6. The locative complement is the easiest case role to relativize in Yaqui.
Relative clauses and nominalizations in Yaqui
At first glance, the second rule seems to be more problematic, although that
isn’t really the case. This rule states that “if a language has agreement at a given
position on the hierarchy of syntactic functions, it will have agreement for all
HIGHER positions” (Lehmann 1986: 674). This is true for case agreement since
the only syntactic function involving case agreement is the one in the highest posi-
tion, i.e. the subject function. The other functions don’t exhibit case agreement.
As for number agreement, the observance of the rule would imply the presence
of number agreement in the cases of subject relativization seen, as number agree-
ment is observed in both direct and indirect object relativization, whereas this
agreement type is absent in cases of oblique relativization. In fact, this apparent
violation of the rule can be easily explained if we consider the impossibility, in
Yaqui, to combine the agentive/subject relativization suffix -me with the plural
suffix -m(e). The examples in (37) show that these two homophonous suffixes are
mutually exclusive.7
(37) a. U ne-nenka-me
det red-sell-nmlz
‘the seller’
b. U-me ne-nenka-me
det-pl red-sell-nmlz
‘the sellers’
c. *U-me ne-nenka-me-m(e)
In (37), the suffix -me is glossed as NMLZ instead of REL, which may seem
contradictory as compared with the gloss given to this suffix in all other previ-
ous examples. This choice is not arbitrary, however. In fact, it is clear that Yaqui
RCls show some nominalization properties. Nonetheless, nominalization signs of
a different nature may be associated with the three suffixes involved in Yaqui RCls
(e.g. -me, -’u and -’Vpo).8 Next I explore this situation in detail, first describing
7. This case represents an obvious example of morphological haplology, a fairly common
phenomenon of inflection in which an affix or clitic is absent when the adjacent part of the
stem is homophonous to it (Stemberger 1981).
8. In their study of RCls in Yaqui, Martínez Fabián and Lagendoen (1996) focused on the
suffixes -me and -’u. They concluded, based on a generative analysis, that the suffix -me is a
nominalization marker whereas the suffix -’u is a relativization marker. The present study
proposes that both suffixes (and also -’Vpo) are nominalization markers although they exhibit
different signs of nominalization.
Albert Álvarez González
the nominal characteristics associated with each construction, and then shifting
attention to their verbal characteristics.
In other words, I next survey the finiteness properties of Yaqui RCls. Based on
the scalar approach of finiteness proposed by Givón (1990: 852–891, 2001: 24–37),
I mainly consider the following morphological/syntactic features, which serve well
to illustrate the differences between finite and non-finite clauses, thus determining
the degree of finiteness exhibited by a given clause:
i. Tense-aspect-modality (TAM)
ii. Pronominal agreement
iii. Nominalizing affixes
iv. Subject/Object case marking
v. Articles, determiners
vi. Types of modifiers (adverbs, adjectives)
d. ko’okoe-me
be_sick-nmlz
‘sick person’
e. majta-wa-me
teach-pas-nmlz
‘student’
This agentive suffix, with an origin probably related to the word yoeme ‘person,
man’, can be used more idiosyncratically (in accordance with its derivational sta-
tus), as is the case with some animal names (39), or with the sequence -wa ‘passive’
+ -me, which is more productively associated, nowadays, with an action/result
meaning (compare examples in (40) with (38e)).
(39) a. bujte-me
produce_ears of corn-nmlz
‘whale’
b. totte-me
fold-nmlz
‘snail’
(40) a. allee-wa-me
be_happy-pas-nmlz
‘happiness’
b. etbwa-wa-me
steal-pas-nmlz
‘theft’
Based on its derivational behavior, the suffix -me appears to be, in fact, a nominal-
ization marker, i.e. a morpheme that turns verbs into nouns. It is predominantly
used in the lexicon to create agentive deverbal nouns. We can then propose that
verbs in subject RCls carry a nominalizing suffix.
Case agreement is another nominal property exhibited by the relativized con-
struction involving the suffix -me. As seen earlier (see (5) and (7) above), sub-
ject RCls exhibit case-agreement with singular object head nouns. Therefore, the
whole clause suffixed by -me acts as a noun phrase, taking a nominal case marker
(the object suffix -ta).
i. The subject within the RCl is marked with genitive case, which is a nomi-
nal inflectional case that mainly marks the relationship between two nomi-
nals. The subject is interpreted as the possessor of the action denoted by the
verb that is suffixed by the marker -’u. The further genitive case marking on
the construction is indication that the verb suffixed by -’u has been in fact
nominalized.
ii. The direct and indirect object RCls exhibit number agreement with the head
noun. This is a nominal property, since Yaqui verbs are not inflected to show
number agreement, but only nominals and postnominal adjectives are. Addi-
tionally, the marker used to indicate number agreement, the suffix -(i)m, is the
same plural marker used with nominals (postnominal adjectives in Yaqui may
also use reduplication to mark plurality).
As in cases involving the non-subject RCl marker -’u, the use of the locative suf-
fix -po as a locative oblique-RCl marker triggers a morphologically conditioned
case of epenthesis, although this is not exactly the same type of insertion as the
Relative clauses and nominalizations in Yaqui
one described earlier (cf. 3.1.2). In (31) to (36), a vocalic re-articulation may
be observed after the glottal stop. Nonetheless, the condition that triggers both
morphophonological processes is identical, since epenthesis of an “echo-vowel”
(Dedrick & Casad 1999: 28–29) only occurs when the suffix is attached to verbs,
not nominals.
Additionally, the Yaqui lexicon includes some words created as a result of the
suffixation of this marker to nominals or verbs. This process, not a very produc-
tive one, is illustrated in (43). Notice some idiosyncratic uses associated with it, in
(43b–43c).
(43) a. te’o-po
God-loc
‘church’
b. bakeo-m-po
cow_boy-pl-loc
‘cowboy game’
c. e’-eusi-wa-m-po
red-hide-pas-pl-loc
‘hide-and-seek game’
The suffix -po may also be used as a headless RCl with idiosyncratic meanings as
shown in (44). Its use in the creation of toponyms is a further illustration of its
lexicogenetic function, as in (45).
(44) Junuka’a yeu sim-su-ka-’apo.
dem.acc out go-comp-perf-loc
‘After this happening …’
(45) wata-ba’am-po
willow-water-loc
‘Huatabampo (a village in Sonora)’
Finally, as in cases involving non-subject RCls marked by -’u, the subject within
the locative oblique RCl is marked with genitive case, an indication of the nomi-
nalization process undergone by the verb.
In sum, the origins of the three suffixes discussed here, their participation in
derivational processes as nominal suffixes (except for the suffix -’u), the case or
number agreement with the head noun, as well as the genitive case marking of the
RCl subject (except for -me) are clear evidence of nominalization. One more sign
of nominalization may be added to these: the fact that RCls may combine with
elements of nominal modalities such as determiners or demonstratives. The exam-
ples in (46) illustrate RCls exhibiting demonstrative pronouns as the head noun,
whereas the examples in (47) show headless RCls introduced by a determiner.
Albert Álvarez González
It is even possible to find the same type of construction lacking any introducing
element, as in the cleft constructions in (48), or in the free RCl in (49).
(48) a. Wa-me yabe-m tea-ka-me Joan
dem-pl key-pl find-perf-nmlz John
‘The one who found those keys is John.’
b. Joan-ta tea-ka-’u yabe-m-tu-kan
John-gen find-perf-nmlz key-pl-vbz-impf
‘What John found were the keys.’
(49) In yaa-bae-’u ne kopta-k
1sg.gen do-des-nmlz 1sg.nom forget-perf
‘I forgot what I was going to do.’
If we compare (49) with (50), we can see that the expression suffixed by -’u is inter-
changeable with a simple noun, which shows that both expressions share the same
general nominal function.
(50) In kabuji ne kopta-k10
1sg.gen drum 1sg.nom forget-perf
‘I forgot my drum.’
9. A tapanco is an altar used for offerings during the Yaqui celebration of All Souls’ Day.
10. Examples (49) and (50) show that if the direct object noun phrase is possessed, the
accusative marking is not allowed as can be seen with the ungrammaticality of the following
example * In kabuji-ta ne kopta-k.
Relative clauses and nominalizations in Yaqui
The case marking exhibited by the accusative pronoun enchi may be associated
to finiteness because it expresses a purely syntactic relation between a verbal
predicate (the verb within the RCl) and a noun phrase. This same argument
is valid regarding the resumptive pronouns we may find in indirect object and
oblique relativization. This pronoun indicates the case role of the anaphori-
cally referenced head noun in the RCl and, importantly, its formation always
involves an accusative pronoun and a relational suffix/postposition, such as
a directional suffix (for dative marking), an instrumental suffix, a comitative
suffix, etc.
A third feature indicating finiteness in Yaqui RCls is the use of adverbs rather
than adjectives as RCl modifiers. The ungrammaticality of adjectival modification,
and the grammaticality of adverbial modification are shown in (52) with a subject
RCl, and in (53) with a direct object RCl. They are evidence that the RCl retains
part of the finiteness of the verbal base since adjectives are nominal modifiers,
while adverbs are verbal modifiers.
The previous section on the nominalization properties exhibited by the Yaqui RCls
logically leads us (i) to question the notion of relativization as applied to the con-
structions presented in Section 3 and (ii) to further think about the concept of
nominalization, in an attempt to understand the connection between these two
different processes.
5.1 Relativization
As we have already seen (see Section 2), relativization implies that an expression
with a clause/sentence-type internal structure is used to restrict the set of potential
referents corresponding to a noun phrase. It is considered crucial for relativization
purposes that the relativized construction reproduces the construction of an inde-
pendent clause/sentence. If we consider the nominalization properties exhibited
by Yaqui RCls (see Section 4.1), it is obvious that this condition is only partially
fulfilled in this language. So, identifying Yaqui RCls as ordinary sentences seems
to be an error. In fact, the structures referred to as relative clauses in Section 3
should be more accurately termed as nominalized entities rather than full clauses
or sentences.
Relative clauses and nominalizations in Yaqui
11. Comrie and Thompson follow the same analysis when they explain why “it is not difficult
to understand how a nominalization can function as a relative clause: the nominalization and
the noun with which it is in construction can be thought of as two juxtaposed nominal ele-
ments [NOM] [NOM], the modifying relationship between them being inferred by language-
users (rather than being specified by the grammar, as it is in languages with specific relative
clause morphology), just as the modifying relationship is inferred in a noun-noun compound
such as tree-house, in which the two nominal elements simply happen to be single nouns.”
(Comrie & Thompson 1985: 394).
Albert Álvarez González
subject RCl, whereas in (56) I show a patient nominalization used as an object RCl.
In (57), a locative nominalization is used as a locative oblique RCl.
The main reason for associating these nominalization types with relativization
patterns is that these nominalized entities do display some formal features indi-
cating finiteness, such as TAM markers, in the verbal stem. However, because the
process of nominalization has canceled the sentential predicative function in favor
of an adnominal modifier function, we cannot consider the resulting construction
as a sentence or full clause. It can be considered, at best, a nominalized clause with
no predicative function, on the basis of the similarities between the internal syntax
of these nominalizations and a full clause with a finite verb and the full array of its
arguments realized (Shibatani 2009: 195).
5.2 Nominalization
At this point, it is necessary to develop a little further the notion of nominaliza-
tion, and to make an important distinction between two different types: lexical
Relative clauses and nominalizations in Yaqui
15. As a good example of syntactic iconicity, this functional change from referential to modi-
fying function is expected to also imply a prosodic integration, from two separate intonation
contours in the coreferential construction (two referring expressions with the same referents)
to a single intonation contour in the modifying construction (one referring expression).
Relative clauses and nominalizations in Yaqui
Back to Yaqui, the suffixes -me and -’Vpo are used in both types of nomi-
nalization (regarding lexical nominalizations, see (38–40) for suffix -me, and
(43–45) for suffix -’Vpo), while the suffix -’u is only used in cases of clausal
nominalization.
Yaqui RCls are not, strictly speaking, clauses. In fact, they are expressed syntacti-
cally as noun phrases headed by a nominalized verb (nominalized by the suffixes
-me, -’u or -’Vpo depending on the case role exhibited by the element being rela-
tivized). As Shibatani (2009) argues, Yaqui is part of a large number of languages
in which the RCls are nominalized entities, illustrating the strong connection that
exists between clausal nominalization and relativization. We could even consider
that, in languages like Yaqui, there is no such thing as relativization. If we con-
tinue using this term in describing such cases, it is probably due to a general Euro-
centric bias among linguists. But why is relativization, in certain languages like
Yaqui, indistinct from nominalization? Or, in other words, why can nominaliza-
tion be the basis for relativization? Let us recapitulate the explanation proposed
here, which roughly agrees with Shibatani’s.
Clausal nominalization is used to convert a finite verbal clause into a noun
phrase. So, from a semantic/functional perspective, it is used to turn events/
states into entities characterized in terms of these events/states. In this sense,
nominalization is not only a category-changing operation, but it is also, obviously,
Albert Álvarez González
References
Andrews, A. D. 2007. Relative clauses. In Language Typology and Syntactic Description, Vol. II.,
2nd edn, Timothy Shopen (ed.), 206–236. Cambridge: CUP.
Comrie, B. 1981. Language Universals and Linguistic Typology. Chicago: CUP.
Comrie, B. & Kuteva, T. 2005. Relativization strategy. In World Atlas of Language Structures
(WALS), M. Haspelmath, M.S. Dryer, D. Gil & B. Comrie (eds), 494–497. Oxford: OUP.
Comrie, B. & Thompson, S. 1985. Lexical nominalization. In Language Typology and Syntactic
Description, Vol. III, Timothy Shopen (ed.), 349–398. Cambridge: CUP.
Croft, W. 1994. Semantic universals in classifier systems. Word 45: 145–171.
Relative clauses and nominalizations in Yaqui
Dedrick, J.M. & Casad, E.H. 1999. Sonora Yaqui Language Structures. Tucson AZ: University of
Arizona Press.
Di Sciullo, A.M. & Williams, E. 1987. On the Definition of Word. Cambridge MA: The MIT Press.
Estrada Fernández, Z., Buitimea Valenzuela, C., Gurrola Camacho, A.E., Castillo Celaya,
M.E. & Carlón Flores, A. 2004. Diccionario Yaqui-Español y Textos: Obra de Preservación
Lingüística. Hermosillo: Universidad de Sonora y Plaza Valdés.
Estrada Fernández, Z. & Alvarez Gonzalez, A. 2008. Parlons Yaqui. Paris: L’Harmattan.
Genetti, C., Bartee, E., Coupe, A., Hildebrandt, K. & Lin, Y.J. 2008. Syntactic aspects of
nominalization in five Tibeto-Burman languages of the Himalayan área. Linguistics of the
Tibeto-Burman Area 31(2): 97–143.
Givón, T. 1990. Syntax. A Functional-Typological Introduction, Vol. II. Amsterdam: John
Benjamins.
Givón, T. 2001. Syntax: An Introduction, Vol. II. Amsterdam: John Benjamins.
Haspelmath, M. 2002. Understanding Morphology. London: Arnold.
Keenan, E.L. 1985. Relative clauses. In Language Typology and Syntactic Description, Vol. II,
T. Shopen (ed.), 141–170. Cambridge: CUP.
Keenan, E.L. & Comrie, B. 1977. Noun phrase accessibility and universal grammar. Linguistic
Inquiry 8: 63–99.
Koptjevskaja-Tamm, M. 2006. Nominalization. In Encyclopedia of Language & Linguistics, Vol.8,
2nd edn, Keith Brown (ed.), 642–659. Oxford: Elsevier.
Lehmann, Christian. 1986. On the typology of relative clauses. Linguistics 24: 663–680.
Malchukov, A.L. 2004. Nominalization/Verbalization: Constraining a Typology of Transcategorial
Operations. Munich: Lincom.
Malchukov, A.L. 2006. Constraining nominalization: Function/form competition. Linguistics
44(5): 973–1009.
Martínez Fabián, C. & Lagendoen, T.D. 1996. Sobre las llamadas cláusulas relativas en yaqui. In
Memorias del Tercer Encuentro de Lingüística en el Noroeste, Tomo 1: Lenguas Indígenas,
Vol.2, Z. Estrada Fernández, M. Figueroa Esteva & G. López Cruz (eds.), 443–464.
Hermosillo: Editorial Unison.
Payne, T.E. 1997. Describing Morphosyntax. A Guide for Field Linguists. Cambridge: CUP.
Searle, J.R. 1969. Speech Acts: An Essay in the Philosophy of Language. Cambridge: CUP.
Shibatani, M. 2009. Elements of complex structures, where recursion isn’t it. The case of rela-
tivization. In Syntactic Complexity. Diachrony, Acquisition, Neuro-cognition, Evolution
[Typological Studies in Language 85], T. Givón & M. Shibatani (eds), 163–198. A
msterdam:
John Benjamins.
Stemberger, J.P. 1981. Morphological Haplology. Language 57(4): 791–817.
On relative clauses and related
constructions in Yaqui
Lilián Guerrero
IIFL-Universidad Nacional Autónoma de México
This paper examines the form and function of relative clauses in Yaqui. Two
major types of relative clauses are identified, subject relatives marked by -me
and non-subject relatives marked by -’u. Additionally, there are three structure
types which closely resemble relative units: ‘non-restrictive’ relative clauses, the
nominalized complement of certain matrix predicates, and the complement
of a seem-like verb. Based on the assumption that nominalization is a gradient
phenomenon (Comrie & Thompson 2007; Koptjevskaja-Tamm 1993), it
is proposed that Yaqui relative clauses establish different points inside the
continuum, i.e. from clauses genuinely modifying a referential element to
different degrees of clausal nominalization governed by a main predicate.
Functionally, it is shown that true relatives introduce or further establish new
information into discourse, whereas non-modifying nominal clauses serve as
either appositive units (Keenan 1985; Carlson 1977), or as a core argument.
1. Introduction1
1. This research has been supported by a CONACyT grant (No. 83529). The author also
thanks two anonymous readers for their insightful comments, Jesús Villalpando and the book
editors for his meticulous revision on the last version of the paper.
Lilián Guerrero
Rel-clauses have been the topic of many grammatical studies and not without
reason. Firstly, relativization is a helpful mechanism to distinguish grammatical
relations, e.g. subject from object, direct object from indirect object, and so
on (Keenan & Comrie 1977, 1979). And secondly, relativization is also a pow-
erful strategy to derive new elements such as nouns, adjectives and participles
(Comrie & Thompson 2007). There is a third property characterizing a Rel-clause,
the fact that a referential expression simultaneously plays a syntactic-semantic role
in the main unit as well as in the dependent unit. In the construction the puppy
[we adopted last year] is enormous, the head noun puppy functions as the object in
the event described by the Rel-clause (we adopted the puppy last year) and has the
role of subject in the main unit (the puppy is enormous). Conversely, in we adopted
the puppy who approached to us with her tail wagging, the head noun has the role
of the dependent subject (the puppy approached to us with her tail wagging), and
it functions as the main object (we adopted the puppy). Furthermore, languages
may exhibit different constraints with respect to which syntactic functions can be
subject of relativization and – according to the well-known Accessibility Hierarchy
(Keenan & Comrie 1977, 1979; Keenan 1985), they may also use different strate-
gies depending on function of the head noun within the Rel-clause.
Finally, inside the domain of subordination, Rel-clause have a special status
since the dependency between the two units is not structural (i.e. complementa-
tion), or in terms of clausal modification (i.e. adverbial clauses). Instead, the event
described in the dependent unit narrows the potential reference of a referring
expression (Comrie & Kuteva 2005), delimits the reference of a NP by specify-
ing the role of the referent of that NP in the situation described by the Rel-clause
(Andrews 2007: 206). Thus, in the previous example, the relative clause we adopted
last year, restricts the potential reference of the referring expression puppy, to only
referents of which the proposition the puppy is enormous is true.
This paper examines the form and function of Rel-clauses in Yaqui. In this
Uto-Aztecan language, relatives are mainly expressed via nominalization.2 Two
2. Yaqui belongs to the Sonoran group of the Southern branch of the Uto-Aztecan family.
The language is spoken mainly in Mexico, by more than 15,000 people living along the Yaqui
River in the Central West part of Sonora, and by an estimated of 6,000 speakers across the
US-Mexican border, in Pascua, Arizona. There are several grammatical studies on Yaqui;
among the most significant are Crumrine (1961), Johnson (1962), Lindenfeld (1973), E scalante
(1990), Jelinek and Escalante (2000), all based on the Arizona dialect; Dedrick and Casad
(1999), Félix Armendáriz (2000), Hernández Doode (2002), Guerrero and Van Valin (2004),
Silva Encinas (2004), Álvarez (2006), Martínez (1996), Guerrero (2004, and further work), as
well as several articles in Estrada Fernández et al. (2008) and Estrada Fernández et al. (2007),
all of them based on the Sonoran dialect.
On relative clauses and related constructions in Yaqui
types of clauses are identified, subject relatives marked by -me (1b), and non-subject
relatives marked by -’u (1c). Like any other attributive modifier, Rel-clauses are
structurally optional (1a). In the examples, the dependent clause appears within
brackets while co-referential arguments are co-indexed.3
(1) a. Aapo siika.
3sg.nom go.sg.pfv
‘He/she left.’
b. U o’ou-Ø [enchi bicha-ka-me] siika.
det man-nom 2sg.acc see-pfv-clm go.sg.pfv
‘The mani who __i saw you, left’
c. U-me o’ou-im [em bicha-ka-’u] saja-k.
det-pl man-pl 2sg.gen see-pfv-clm go.pl-pfv
‘The meni you saw__i, left.’
Along with modifying Rel-clauses, Yaqui has a variety of clauses that structurally
resemble relatives in various ways, but they do not fit the definition of a Rel-clause
narrowing the potential reference of a referring expression. Instead, the non-
restrictive relative clause in (2a) provides additional information or makes a com-
ment about a participant without delimiting its reference (Carlson 1977; Keenan
1985: 168–9). The ‘pseudo-relative’ clause (Lambrecht 1981; van der Auwera 1985)
in (2b) functions as a clausal argument of perception complement-taking verbs.
And, finally, the dependent unit marked by -me in (2c) functions as a core argu-
ment of a seem-like verb, resulting in some sort of ‘raising’ construction.
(2) a. [In sai-tu-ka-’u] aman tawa-ne.
1sg.gen brother-vblz-pfv-clm there stay-pot
‘The one who was my brother, he will remain over there.’
(H’ life story: 115)
b. Nim achai [jaibu enchi siika-m-ta] te’a-k.
1sg.gen father already 2sg.acc go.sg.pfv-clm-acc find-pfv
‘My father found out that you already left.’ (Guerrero 2006a: 142)
c. Ivan-Ø [ka tua Torim-meu wee-pea-m-ta] bena.
Ivan-nom neg true Torim-pl.dir go-desid-clm-acc seem
‘It seems Ivan doesn’t want to go to Torim.’ (Guerrero 2004: 266)
3. Abbreviations: acc: accusative, appl: applicative, clm: clause linkage marker, desid:
desiderative, dem: demonstrative, det: determiners, dir: directional, gen: genitive, ins:
instrumental, ints: intensifier, loc: locative, neg: negation, nom: nominative, pastc: past
continuative, pfv: perfective, pl: plural, pot: potential, obl: oblique, prs: present, ref: referen-
tial, sg: singular, vblz: verbalizer.
Lilián Guerrero
The pronominal system also keeps track of the major syntactic functions (Table 1).
Yaqui pronouns have been traditionally divided into independent (“full”) and
dependent (“reduced”) forms. While full pronouns are expected to behave as lexi-
cal elements in terms of their distribution, reduced nominative pronouns have
been considered “second position” clitics, while reduced accusatives (available
only for third person) cliticize to the verb. A third set of reduced forms occur as
objects of postpositions.
When the head noun functions as a non-subject argument inside the Rel-clause
(e.g. dependent object or oblique), the dependent unit takes the clause linkage
marker -’u. Thus, the head noun mesa ‘table’ is the object of kokta ‘break’ in (6a);
4. The nominalizing strategy where subject- and non-subject relatives are overtly marked
on the verb is also observed in Ute, O’odham, Shoshoni, Luiseño, Hopi, Cupeño inside the
Uto-Aztecan family. Comrie and Kuteva (2005) also cite similar patterns in Berber, Turkish,
Kambera, Tukang Besi, Lhasa Tibetan and several other languages.
On relative clauses and related constructions in Yaqui
u jamut ‘the woman’ is the oblique argument of waate ‘miss’ in (6b), while ‘the
woman’ has the role of recipient of bittua ‘send’ in (6c). Notice also that, when the
dependent subject is overtly expressed, it occur as a genitive pronoun (1st person,
2nd person), as in (6a-b), an accusative pronoun (3rd person), or an accusative
nominal, as in (6c); the rest of the core arguments are marked the same way as they
appear in independent clauses.
(6) a. Mesa-Øi [em __i kokta-ka-’u] sikili-tu-kan.
table-nom 2sg.gen break-pfv-clm red-vblz-pasc
‘The table that you broke was red.’
b. Jamut-ta-ui [nim __i etejo-ka-’u] ne waate-Ø.
woman-acc-dir 1sg.gen chat-pfv-clm 1sg.nom miss-prs
‘I miss the woman with whom I chatted.’
c. U jamut-Øi [Joan-ta ili usi-ta a-ui
det woman-nom Juan-acc little child-acc 3sg-dir
bittua-ka-’u] siika.
send-pfv-clm go.sg.pfv
‘The woman to whom Juan sent the child, left.’
Although less common, the locative postposition -po may also derive locative
Rel-clauses as in (7), but it is still unclear whether these relatives restrict the iden-
tity of a head noun or delimit the event within a time/location frame in discourse.
Because of this and because there are very few examples of locative Rel-clauses
in the corpus, they are not included here. There is no data on relatives modifying
genitives or objects of comparison.
(7) a. Poso-poi [__i kuchu’m ane’e-po] a wo’ota-ne.
pond-loc fish.pl exist-loc 3sg.acc throw-pot
‘(The fish’s skin) it is thrown in the pondi wherei the fishes are.’
(Johnson 1:2)
b. Inimi jiba aane,
there always exist
[__i junu bwe’u mako’ochin-ta weye-ka’a-po]
dem big guamuchil-acc be.stand-pfv-loc
[__i ne senu-k bicha-k sestul ta’a-po].
1sg.nom one-acc see-pfv one day-loc
‘They have always existed therei, in the place wherei the big
guamuchil three stands up; in the place wherei one day I saw one.’
(Grandfather: 7–9)
Yaqui closely follows the cross-linguistic tendencies found by Comrie and Kuteva
(2005: 495–501) with respect to different strategies of relativization: s ubject relatives
Lilián Guerrero
make use of the ‘gap’ strategy only, as seen in (5a) and (5b), while non-subject
relatives may use the ‘gap’ strategy, as in (6a) and (6b), as well as the ‘pronoun-
retention’ strategy, as in (6c). Moreover, since the Rel-clause tends to follow the
head noun in Yaqui, it is sometimes difficult to determine whether the head is
inside or outside the dependent unit. Evidence that the head noun is e xternal to
the Rel-clause in the sentences in (5) and (6) comes from case-marking: the head
noun is marked as nominative in (5a), (6a) and (6b), as accusative in (5b) and
oblique in (6c), reflecting its syntactic status with respect to the main predicate.
Internally-headed relatives are infrequent in Yaqui. The head noun jamut ‘woman’
is the subject of the non-verbal predicative clause in (8), and the hearer of the
speech act verb nooka ‘talk to’ inside the Rel-clause, and so it is marked as an
oblique argument by the dependent verb. Notice also that the head noun appears
in the canonical position for non-subject arguments inside the linked unit, e.g.
preceding the verb.
(8) [Kajlos-ta jamut-ta-u nooka-ka-’u] Maria-tu-kan.
Carlos-acc woman-acc-dir talk-pfv-clm María-vblz-pastc
‘The woman to whom Carlos talked was María.’
Indeed, Rel-clause as a whole agrees in case and number with its head noun.
This agreement pattern is more systematic for -me clauses, especially for
accusative case (e.g. (5a), (10a)), and oblique cases (e.g. (10b–c) below). For -’u
Rel-clauses, accusative agreement is atypical but number agreement is not, as
shown in (9b). In (9c), the Rel-clause takes the instrumental plural postposition
-mea.
(9) a. Ju’u yoeme-Ø chu’u-tai me’a-k
det man-nom dog-acc kill-pfv
[a __i kiki-su-ka-m-ta].
3sg.acc bite-finish-pfv-clm-acc
‘The man killed the dog that bit him.’
b. Kaa mache’eta-mi ne jippue-Ø
neg machete-pl 1sg.nom have-prs
[em ne __i reuwa-ka-’u-m].
2sg.gen 1sg.acc lend-pfv-clm-pl
‘I don’t have the knives that you lent me.’
c. Kuchi’i-mi ne maka-’e
knife-pl 1sg.acc give-imp
[wakaj-ta em a-meai chukchukta-’u-m-mea].
meat-acc 2sg.gen 3sg-ins.pl red.cut-clm-pl-ins
‘Give me the knives that you chop the meat with!’
On relative clauses and related constructions in Yaqui
5. The so-called ‘participial’ relatives (Buelna 1891: 16, 50) in Yaqui retain the past perfec-
tive suffix -ka, and occasionally the passive suffix. However, compared to the productivity
of the suffix -(ka)me as deriving adjectives and participles in Ralámuli (Tarahumara; Islas
2010), Warihío (Félix 2005), Cora (Vázquez 2002) and Huichol (Iturrioz & Gómez 1993), this
strategy is very limited in Yaqui (see Guerrero 2009, 2010 for comparison).
Lilián Guerrero
Once derived, these forms behave as any other nominal with respect to case and
plural marking, co-occurrence with other adjectives, and word order, i.e. exam-
ples of decategorization (from verbs) and re-categorization (as nominal) processes
(Malchukov 2004, 2006; Lehmann 1988).
Leaving aside the cases of lexical nominalization, Yaqui Rel-clauses show a mixture
of nominal and verbal characteristics. The degree of syntactic nominalization of
Yaqui relatives is outlined in the next section, together with a thorough discussion
regarding their nature and function in discourse.
It has been claimed that external Rel-clauses often appear in two different forms,
commonly called ‘reduced’ and ‘unreduced’ (Andrews 2007: 211). The former are
less like full clauses, e.g. they have a reduced tense-mood marking, r estrictions
on the coding of co-referent elements and, sometimes, features of adjectival or
nominal morphology on the verb. Further, ‘reduced’ relatives usually appear
in the canonical position of adjectival modifiers, whereas ‘unreduced’ relatives
may appear in a different position. Generally, this distinction correlates with
different degrees of nominalization (Comrie & Thompson 2007; Koptjevskaja-
Tamm 1993).
On relative clauses and related constructions in Yaqui
There is also a correlation between reduced and unreduced clauses and their
position in the construction in Yaqui. In the corpus under study, 88% of the 137
Rel-clauses marked by -me immediately follows the head noun, and only 12%
are extraposed to the right; in contrast, 43% of object and oblique relatives are
extraposed, leaving an empty slot inside the Rel-clause. What this distribution
shows is that subject relatives prefer to be embedded within the main clause, and
be closer to their head noun in a similar way than other nominal modifiers, i.e. a
higher degree of nominalization compared to object and oblique relatives.
The distribution of Yaqui Rel-clauses with respect to the syntactic functions
of the head noun is shown in Table 2 (around 150 examples from direct elicitation
where excluded here). Although relativization has access to all direct and oblique
core arguments, the syntactic role which is most commonly modified in Yaqui is
the intransitive subject (S). Notice that there is a significant difference between the
frequency of relatives modifying head nouns in S role, with respect to head nouns
functioning as a transitive agent (A) and a transitive object (O).
6. I am grateful to Carlos Silva, Rolando Félix and Cresencio Buitimea for letting me use
their oral narratives. For this paper, I have analyzed 14 oral texts (2494 clauses in total). The
complete references for the oral texts are included as part of the reference section; the English
translations and the clause numbering for the texts are mine.
7. Martínez and Langendoen (1996) previously suggested that the occurrence of S-relatives
in Yaqui highlights its nominal function by introducing new referents together with its more
relevant attributes; at the time, the authors based their analysis on directly elicitated data only.
On relative clauses and related constructions in Yaqui
Oral texts 63 22 5 9
Dictionary 38 14 4 6
101 (67%) 26 (17%) 9 (6%) 15 (10%)
The following passage shows object relatives. In (16a) the head noun namurokoa
‘mud’ functions as the object in the main clause and has the subject role in the
dependent unit. The same is true of wa’a yoawa ‘the sacred animal’ in (16b); here,
the head noun together with the Rel-clause is extraposed to the right, so there is a
resumptive pronoun inside the main clause.
(16) a. senu bea namurokoa-ta soota-k [junum
one then mud-acc raise-pfv there
ba’a-po yuka-m-ta].
water-loc be.inside-clm-acc
‘Then, one raised the mud that was laying there in the water.’
(Grandfather: 55)
b. nien ket wa-me’e itom yo’owa-m ai ju’uneya
thus too dem-pl 1pl.acc ancestor-pl 3sg.acc know
wa’a yoawa-tai [potcho’oku ane-m-ta].
dem.acc animal-acc mountain exist-clm-acc
‘And our ancestors knew about him too, about that animal living
in the mountain.’ (Saint: 16)
The following narrative starts by setting out the location of the story; the first
two clauses express spatial locations by means of oblique relatives marked by -po
in (17b) and (17c); soon after there is an object relative introducing the topic of
the story (17g). Once all the settings and new protagonists are laid out, the story
continues by introducing the definite noun of the thing that was moving under the
water, the ‘female lizard’.
(17) a. Sestul ta’a-po te batwe-po nau rejte-n,
one day-loc 1pl.nom river-loc together walk.pl-pastc
b. [susu’e= te kate-ka‘a-po]
little.hill=2pl.nom be.seated.pl-pfv-loc
c. [bau ba’a-ta tatawa’a-po]
close water-acc RED.stay-prs-loc
On relative clauses and related constructions in Yaqui
8. Maybe, there is also a S-relative in (17d) encoding some sort of temporal location yuku
mecha-m ‘being the season of rain’; in Yaqui copulas are Ø in the present and -tu when
no-present. Otherwise, it is hard to explain the plural marking on mecha-m and the singular
supletive verb sime.
Lilián Guerrero
Although more data is necessary, two possible explanations can be suggested. First,
the clause linkage marker -’u is extending its domain to other syntactic functions,
such as it also has access to subject NPs, i.e. -’u as a general and multi-functional
subordinator. Second, it is possible that Yaqui is making a distinction between
restrictive and non-restrictive relatives. Accordingly, the information coded in a
relative unit may be either essential to understanding who the designated entity is
(i.e. restrictive), or neither essential nor defining, but merely specifying in further
detail some information about that noun (i.e. non-restrictive). Although there is
a referentially shared argument in both types, non-restrictive clauses specify the
head noun in a way similar to appositive nouns (e.g. Garibaldi and Bartola, our
kittens), whereas restrictive clauses does not plainly specify their heads, but rather
restrict their meaning in a direct way (de Vries 2002: 71). For several authors,
non-restrictive clauses are not true Rel-clauses since they merely made a com-
ment about a referential entity without delimiting its reference (Keenan 1985: 168;
Carlson 1977; Lehmann 1984).
What is of interest here is that Yaqui S-relatives marked by -’u neither delimit
nor modify the head noun, but they provide given information made salient in a
different way by focusing on some properties of that entity. Take as an example
the clause in (18a), which may be paraphrased as the one, that is, my Compadre
Timo, or Timo, my Compadre. Although the referent is already identifiable, it may
be relatively less accessible in discourse and so requires a more substantial lexical
realization, i.e. tracking old topics. Therefore, it may be the case that S-relatives
marked by -’u are somewhere between the characterization of subject’s attributes
expressed by -me, and the most active events expressed by O-relatives marked by
-’u.
In the next section, modifying Rel-clauses are formally distinguished from
dependent clauses governed by a main predicate.
The second strategy involves fully syntactic complements; here, the linked unit
overtly expresses all its core arguments, the verb is marked for the relevant tense-
aspect-mood operators, its position with respect to the main clause is variable,
and the dependent unit is generally marked by -’u, and less frequently by the loca-
tive postposition -po. When the matrix subject and the dependent subject are non
co-referential (20a) and (20b), the dependent subject must be accusative; when
they are identical (20c), there must be an anaphoric genitive pronoun. In addition,
the position of the dependent unit varies; it can be embedded, as in (20a), or can
be extraposed to the right, as in (20b) and (20c). Occasionally, when the subordi-
nated unit is clause-final, there may be a co-referential pronoun inside the main
unit, as in (20b).
(20) a. Peo-Ø [kaba’i-m enchi jinu-ka-’u] suale-n.
Pedro-nom horse-pl 2sg.acc buy-pfv-clm believe-pastc
‘Pedro believed that you had bought the horses.’
b. Peo-Ø ai suale-Ø [kaba’i-m enchi jinu-ka-’u].
Pedro-nom 3sg.acc believe-prs horse-pl 2sg.acc buy-pfv-clm
‘Pedro believed that you had bought the horses.’
Lilián Guerrero
The last two strategies are limited to a few complement-taking predicates and
their distribution is mainly based on the subject’s identity: the nominalized
marked by -m(e) and followed by the accusative -ta, as in (21a), demands dif-
ferent subjects, while the participial-like unit marked by -kai in (21b) requires
identical subjects. In the former, the dependent subject must be accusative, and
the verb can be unmarked or be marked by the perfective -ka; in the latter, the
co-referential subject must be omitted and the verb must be fully unmarked.
Notice that the nominalized complement type is structurally similar to subject
relatives.
(21) a. Aurelia-Ø [enchi laaben-ta pona-ka-m-ta] jikka-k.
Aurelia-nom 2sg.acc violin-acc play-pfv-clm-acc hear-pfv
‘Aurelia heard you play the violin.’
b. Maria-Ø bo’obicha-Ø [sim-bae-kai].
María-nom hope-prs go-desid-clm
‘María hopes to leave.’
Among the four complement types, the most common ones are the co-lexical
(19) and syntactic (20a) structures. Psych-action, jussives, propositional attitude,
knowledge, indirect perception, and speech act verbs, all take the -’u complement
(e.g. ‘imperfect nominals’) and some of them also take the morphological struc-
ture as an alternative. In contrast, the nominalized type is limited to direct per-
ception (21a), and two mental predicates where the content of the complement
encodes the perception of an event, such as te’a ‘find, discover’ and teenku ‘dream,
imagine’. The predicate bena, a ‘seem-like’ predicate, also takes this structure as a
complement (22).
indicate clausal complement of ‘ea ‘think about’ in (23c).9 Then, out of 48 examples
from discourse corpora, bena functions as a main predicate in 9 instances; it intro-
duces nominal arguments in 19 examples; it marks adverbial clauses in 8, and com-
plement units only in 4; finally, there are 8 instances where bena acts as a matrix
predicate (23d). The discussion focuses only on the last function.10
(23) a. U ili jamut-Ø ankeles-ta bena.
the little woman-nom angel-acc seem
‘The girl seems an angel.’
b. Ame-t chachai-ne bake’o-ta wakas-im nama’a benasi
3pl-loc RED.yell-pot cowboy-acc cow-pl guide seem.like
‘He yells to them, like a cowboy guiding his cows.’ (Saint: 7)
c. Wa’ame o’owi-m [kaa tuisi wakabaki-ta
these.pl man-pl neg good wakabaki-acc
bwase-ka-benasia] ’ea-Ø.
cook-pfv-clm think-prs
‘These men have the feeling that the wakabaki was not well cooked.’
d. Into au take tua lutu’uria-ta benasi,
and 3.refl shake really true-acc clm
wante-ka-m-ta benasi.
run.sg-pfv-clm-acc seem-ints
‘(“I won, I won” said the turtle), and she shook herself as if was true, as
if she had really run.’ (Turtle: 52–4)
9. ’ea is the only verb that takes this marker, and it is systematically followed by the particle
si. Historically, it is then unclear whether bena was a verb or a member of the be-postpositions,
e.g. beas/beasi ‘around’, bepa ‘over’, betuk ‘under’, betana ‘from’, betchi’ibo ‘for’.
10. There is a typology of apposite (non-restrictive) Rel-clauses in discourse where one
of the types, the continuative type, enables a movement within narrative time, by depicting
two successive extra-linguistic events (Look 2007: 339). That may be the function of apposite
clauses marked by benasi in (23b).
Lilián Guerrero
(ii) The coding of the dependent subject. True Rel-clauses and complement units
differ with respect to coding of the dependent subject. In this language, only
the 1st and 2nd person singular pronouns are formally distinguished from the
genitive and accusative functions, whereas the rest are the same (Table 1). In the
Rel-clause in (25a) and (25b), the pronominal subjects must be genitive; in con-
trast, in the (25c) complement unit, the pronominal subject must be accusative;
any other combination results in ungrammaticality. When nominal, it is marked
by the accusative -ta in both cases, i.e. as in Ute (Givón 1980).
(25) a. [Em/*enchi bwika-’u] ne yi’i-ne.
2sg.gen/acc sing-clm 1sg.nom dance-pot
‘I will dance whatever you sing.’
b. Aurelia-Ø bicha-k [ tajo’o-ta nim/*ne baksia-ka-’u].
Aurelia-nom see-pfv cloth-acc 1sg.gen/acc wash-pfv-clm
‘Aurelia saw the clothes that I washed.’
On relative clauses and related constructions in Yaqui
Possessive subjects for Rel-clauses are very common cross-linguistically, i.e. the
possessive-accusative nominalized type in Koptjevskaja-Tamm (1993: 110–128),
while possessive subjects for complementation are less frequent (Cristofaro
2003: 130–1). As for Yaqui, the strong tendency for marking genitive subjects for
modifying relative clauses and accusative subjects for all other non-independent
clauses (e.g. complements and most adverbials) provides another piece of evi-
dence for the nominalization continuum, with the modifying Rel-clauses at the
top. Indeed, there is an additional argument in support of an attributive/possessive
function of relatives: sentential nominalization acting as core arguments marked
by -me requires the main and the dependent subject to be different, while more
than the half of S/A-relatives involve coreferential subjects.
Hence, Yaqui nominalized clauses are used with predicates that (i) disal-
low same-subject constructions, e.g. causative or jussive verbs, or (ii) predicates
that allow both same-subject and different-subject interpretations, e.g. some
psych-action and mental verbs as in (26a). All the same, the alternative control
construction (26b) is never marked by -me or -’u, but by the clause linkage marker
-kai. Contrast the pair of examples below.
(26) a. Nepo [Peo-ta enchi kuna-ka-m-ta] teenku-k.
1sg.nom Pedro-acc 2sg.acc marry-pfv-clm-acc dream-pfv
‘I dreamed of Pedro marrying you!’
b. Tuuka beako Lupe-Ø teenku-k [Peo-ta kuna-kai].
yesterday night Lupe-nom dream-pfv Pedro-acc marry-clm
‘Last night, Lupe dreamt of (herself) marrying Pedro.’
The situation for bena-clauses is completely different. On the one hand, there is
only one subject participant and it notionally belongs to the dependent verb; the
Lilián Guerrero
clause in (27a) is ruled out because each verbal unit has its own subject. On the
other, the notional dependent subject does not occur in its canonical position, but
it serves as the matrix subject and so it must be marked nominative (27b), i.e. a
‘raising’ or matrix-coding-as subject construction.
(27) a. *ne [enchi ka tua Suichi-u
1sg.nom 2sg.acc neg ints Vicam-dir
wee-pea-m-ta] bena.
go.sg-desid-clm-acc seem
‘It seems to me that you are not going to Vicam.’
b. empo [tua Torim-meu wee-pea-m-ta] bena.
2sg.nom ints Torim-dir.pl go.sg-desid-clm-acc seem
‘You seem to want to go to Torim.’
c. *[enchi/em tua Torim-meu wee-pea-m-ta] bena.
2sg.acc/gen ints Torim-dir.pl go.sg-desid-clm-acc seem
‘It seems that you want to go to Torim.’
For a predicate like seem in English, there are two possible structures. In the
clause it seems that my cats really enjoy the garden, the NP my cats is the subject
of the embedded clause, whereas in the sentence my cats seem to enjoy the garden,
the NP is the subject of the matrix predicate, and the dependent verb appears in
its infinitive form. Yaqui only shows the last option, as shown by the ungram-
maticality of (26c). In the matrix-coding construction, there is no change in the
semantic role of the NP; what changes is its syntactic function with respect to the
main verb.
(iii) Argument coding. True Rel-clauses and complement units differ with
respect to missing participants. We have seen that inside the Rel-Clause, only
the dependent subject must be marked by genitive or accusative case, but the
rest of the arguments are coded the same way as in simple clause, i.e. there is
no re-arrangement of case marking. Also, although Rel-clauses usually follow the
head noun, we saw cases where the relative can be extraposed to final position
and, most of the time, the head noun remains as a matrix core argument. As a
result, there is one verbal slot left empty in the dependent unit, i.e. externally-
headed (28a). This phenomenon can be also seen as some sort of noun extraction
or ‘raising’. In contrast, all the slots required by the dependent verb in a comple-
ment unit must be overtly expressed (28b); the clause in (28c) is ruled out since
the dependent subject serves as an argument of the matrix core, something that is
fine for relatives. What it is possible for complementation but not for modification
is to copy the dependent subject as a main core argument, as in (lit.) Aurelia saw
me that I washed the clothes in (25c) above and repeated below, i.e. perception by
means of first-hand evidential (Guerrero 2006b: 148).
On relative clauses and related constructions in Yaqui
(iv) Finiteness. Although most relatives encode states generally unmarked for
tense or marked as past/perfective, it is possible for any Rel-clause to express a
future-oriented event (31a), take some modal markers as in (29) above, and be
independently negated (31b).
(31) a. bea [nee mujtitua-ne-me] yaja-k,
then 1sg.acc be.crossed-pot-clm arrive.pl-pfv
ju-me’e bikenyom.11
det-pl viqueño.pl
‘And then, the ones who will made the sign of the cross on me,
will arrive, the Viqueños.’ (Maejto: 62)
b. ju-me [kaa tu’i-m kakare-ka-me ]
det-pl neg good-pl red.house-pfv-clm
into kaa kimu-k che’e tu’ii-ne.
and neg enter.pl-pfv more good-pot
‘Those who do not have their houses clean, do not enter, it
won’t be good.’ (Star: 25)
11. The word ‘Viqueño’ is the family name of the inhabitants of Vicam town.
On relative clauses and related constructions in Yaqui
The bena construction is special with respect to TAM operators. Not only is the
dependent verb strongly unmarked (33a), the same as subject relatives, but
the matrix verb itself appears unmarked too (33b). Also, negation has scope over
the dependent unit only.
(33) a. Kia ne ju’ubwa eje-ta yeu
ints sg.nom just palo fierro-acc out
tomte-m-ta bena ¡ketun kaa momoli!
born-clm-acc seem still neg mature
‘I seem to have just born, like a palo fierro, not mature!’
(Experience: 15)
b. Lili-Ø [kaa tajo’o-ta baksia-su-ka-m-ta] bena.
Lili-nom neg cloth-acc wash-finish-pfv-clm-acc seem
‘Lili seems to have not washed the clothes.’
5. Discussion
As Givón has pointed out, there is good evidence to argue for nominalization as
the major diachronic pathway for all subordinated clauses in the Uto-Aztecan
family (Givón 2006, 2007, and earlier studies). Yaqui relatives are a clear example
Lilián Guerrero
such a state of affairs. As a result, the dependent unit may be construed as a prop-
erty attributed to the entity bringing it about, as the mental conceptualization of
the event as a whole, or something between.
This paper has examined the internal syntax of Yaqui relative clauses, it has explored
the distribution of subject and non-subject relatives in discourse, and it has intro-
duced the major differences between modifying nominal units and sentential nom-
inalized arguments. Structurally, Rel-clauses in Yaqui show a mixture of nominal
and verbal characteristics, such as nominalization exists as a continuum. Although
further research focusing on the discourse functions of Rel-clause remains to be
undertaken, this analysis provides ample examples of what is or can be in-between
the two extremes of the continuum of nominalization. On the one hand, the more
nominalized a Rel-clause is, the less accessible to various positions for relativiza-
tion and the more restrictive is its occurrence as a complement unit. On the other
hand, whereas certain Rel-clauses may encode some background information in
texts, it is also true that Rel-clauses marked by -me introduces new participants
into discourse, and non-restrictive or appositive clauses marked by -’u may be
re-introducing some participants by using a proper name followed by a comment.
References
Comrie, B. 2005. Further thoughts on the Japanese-type relative clauses, and its implications for
the general typology of relative clauses. Paper presented at ‘On the Typology, Acquisition &
Processing of Relative Clauses’. Leipzig: MIP.
Comrie, B. & Keenan, E.L. 1979. Noun phrase accessibility revisited. Language 55: 649–664.
Comrie, B. & Kuteva, T. 2005. Relativization in subjects and relativizations in obliques. In The
World Atlas of Language Structures, M. Haspelmath, M.S. Dryer, D. Gil & B. Comrie (eds),
494–497. Oxford: OUP.
Comrie, B. & Thompson, S.A. 2007. Lexical nominalization. In Language Typology and Syntactic
Description, Vol. 3, T. Shopen (ed.), 334–381. Cambridge: CUP.
Cristofaro, S. 2003. Subordination. Studies in Typology and Linguistic Theory. Oxford: OUP.
Crumrine, L.S. 1961. The phonology of Arizona Yaqui, with Texts. Anthropological Papers of the
University of Arizona 5. Tucson AZ: The University of Arizona Press.
Dedrick J.M. & Casad, E.H. 1999. Sonora Yaqui Language Structures. Tucson AZ: The University
of Arizona Press.
de Vries, M. 2002. The Syntax of Relativization. Ph.D. dissertation, University of Amsterdam.
Dixon, R.M.W. 2006. Complement clause and complementation strategies in typological
perspective. In Complementation. A Cross-Linguistic Typology, R.M.W. Dixon & A.Y.
Aikhenvald (eds), 1–48. Oxford: OUP.
Dryer, M. 2005. Order of relative clause and noun. In The World Atlas of Language Structures. M.
Haspelmath, M.S. Dryer, D. Gil, and B. Comrie (eds.), 366–367. Oxford: OUP.
Escalante, F. 1990. Voice and Argument Structure in Yaqui. Ph.D. dissertation, The University
of Arizona.
Estrada Fernández, Z. et al. 2004. Diccionario yaqui-español y textos: Obra de preservación
lingüística. México: Plaza y Valdés.
Estrada Fernández, Z. et al. 2008. Mecanismos de voz y formación de palabra. México: Editorial
Plaza y Valdes.
Estrada Fernández, Z., Wichmann S., Chamoreau, C. & Álvarez González, A. 2008. Studies in
Voice and Transitivity. Munich: Lincom.
Félix Armendáriz, R. 2000. Las relaciones gramaticales en yaqui: Un análisis en el marco de la
Gramática del Rol y Referencia. MA thesis. University of Sonora.
Félix Armendáriz, R. 2005. A Grammar of River Warihío. Ph.D. dissertation. Rice University.
Fox, B. 1987. The noun phrase accessibility hierarchy reinterpreted: Subject primacy or the abso-
lutive hypothesis? Language 63(4): 856–70.
Fox, B. & Thompson, S.A. 1990. A discourse explanation of the grammar of relative clauses in
English conversation. Language 66(2): 297–316.
Guasti, M.T. 1989. Romance infinitive complements of perception verbs. MIT Working papers
in Linguistics 11: 31–45.
Givón, T. 1980. The binding hierarchy and the typology of complements. Studies in Language
4: 333–77.
Givón, T. 2001. Syntax, Vol. 2. Amsterdam: John Benjamins.
Givón, T. 2006. Multiple routes to clause-union: The diachrony of syntactic complexity. Paper
presented at I Seminario Sobre Complejidad Sintáctica. Hermosillo: Universidad de Sonora.
Givón, T. 2007. Toward a diachronic typology of relative clauses. Paper presented at II Semi-
nario de Complejidad Sintáctica. Hermosillo: Universidad de Sonora.
Gómez, P. & Iturrioz, J.L. 1993. La escala de estructuras atributivas en huichol [Languages of the
World 6]. Munich: Lincom.
On relative clauses and related constructions in Yaqui
Guerrero, L. 2004. The Syntax-Semantic Interface in Yaqui Complex Sentences, a Role and
Reference Grammar Analysis. Ph.D. dissertation, University at Buffalo.
Guerrero, L. 2005a. Unión de cláusulas en textos Yaquis. In The Second Conference on Indigenous
Languages of Latino America. Austin TX: The University of Texas. 〈http://www.ailla.utexas.
org/site/cilla2_toc.html〉
Guerrero, L. 2005b. Yaqui relative clauses. In Proceedings of the 8th Workshop on American
Indigenous Languages [Santa Barbara Papers in Linguistics 16], C. Jany & L. Harper (eds),
15–27. Santa Barbara CA: University of California.
Guerrero, L. 2006a. The Structure and Function on Yaqui Complementation [Studies in Native
American Linguistics 54]. Munich: Lincom.
Guerrero, L. 2006b. Verbos de percepción en yaqui. In Memorias del VIII Encuentro Internacio-
nal de Lingüística en el Noroeste, Z. Estrada Fernández (ed.), 161–182. Hermosillo: Univer-
sidad de Sonora.
Guerrero, L. 2009. La formación de conceptos de propiedad en lenguas yuto-aztecas. Paper pre-
sented at X Congreso Nacional de Lingüística de la AMLA. México.
Guerrero, L. & Van Valin Jr., R.D. 2004. Yaqui and the analysis of primary object languages.
International Journal of American Linguistics 70(3): 290–319.
Hernández Doode, G. 2002. Marcadores discursivos: Una exploración en la narrativa yaqui. MA
thesis, Universidad de Sonora.
Islas, B. 2010. Caracterización morfo-sintáctica y semánticas de los conceptos de propiedad en
ralámuli de choguita. BA thesis, ENAH.
Jelinek, E. & Escalante, F. 2000. Unaccusative and unergative verbs in Yaqui. In Uto-Aztecan:
Structural, Temporal and Geographic Perspectives, E. Casad & T. Willett (eds), 171–182
México: Universidad de Sonora.
Johnson, J. 1962. El idioma Yaqui. México: INAH.
Keenan, E. 1985. Relative clauses. In Language Typology and Syntactic Description: Complex
Constructions, T. Shopen (ed.). Cambridge: CUP.
Keenan, E.L. & Comrie, B. 1977. Noun phrase accessibility and Universal Grammar. Linguistic
Inquiry 8: 63–99.
Koenig, J.P. & Lambrecht, K. 1999. French relative clauses as secondary predicates. In Empirical
issues in Formal Syntax and Semantics 2, F. Corbin, C. Dobrovie-Sorin & J.M. Marandin
(eds), 191–214. The Hague: Thesus.
Koptjevskaja-Tamm, M. 1993. Nominalizations. London: Routledge.
Lambrecht, K. 1981. Topic, Antitopic, and Verb Agreement in Non-Standard French [Pragmatics
and Beyond 2: 6]. Amsterdam: John Benjamins.
Lehmann, C. 1984. Der Relativisatz. Tübingen: Günter Narr.
Lehmann, C. 1988. Towards a typology of clause linkage. In Clause Combining in Grammar and
Discourse [Typological Studies in Language 18], J. Haiman & S. Thompson (eds), 181–225.
Amsterdam: John Benjamins.
Lindenfeld, J. 1973. Yaqui Syntax. Berkeley CA: University of California Press.
Look, R. 2007. Appositive relative clauses and their functions in discourse. Journal of Pragmatics
39: 336–362.
Malchukov, A. 2004. Nominalization/Verbalization: Constraining a Typology of Transcategorial
Operations. Munich: Lincom.
Malchukov, A. 2006. Constraining nominalization: Function/form competition. Linguistics
44(5): 973–1009.
Lilián Guerrero
Martínez, C. & Langendoen, T. 1996. Sobre las llamadas cláusulas relativas en yaqui. In Memo-
rias del III Encuentro de Lingüística en el Noroeste, Z. Estrada Fernández, M. Figueroa & G.
López Cruz (eds), 443–462. Hermosillo: Universidad de Sonora.
Miller, J. 1989. Sentential Complementation in Huaraz Quechua: A Typological-functional
Study. MA thesis, The University of Texas at Arlington.
Reh, M. 1985. Die-Krongo-Sprache. Berlin: Reimer.
Rude, N. 1996. Objetos dobles y relaciones gramaticales: El caso del yaqui. In Memorias del III
Encuentro de Lingüística en el Noroeste, Z. Estrada Fernández, M. Figueroa & G. López
Cruz (eds), 491–522. México: Universidad de Sonora.
Schüle, S. 2000. Perception Verb Complements in Akatek, a Mayan Language. Ph.D. disserta-
tion, Universitat Tubingen.
Shibatani, M. 2009. Elements of complex structures, where recursion isn’t it. The case of
relativization. In Syntactic Complexity. Diachrony, Acquisition, Neuro-cognition, Evolution,
T. Givón & M. Shibatani (eds), 163–198. Amsterdam: John Benjamins.
Silva Encinas, M.C. 2004. La secuencia temporal en el discurso narrativo Yaqui. Tesis de
Maestría. Hermosillo: Universidad de Sonora.
van der Auwera, J. 1985. The predicative relatives of French perception verbs. In Predicates
and Terms in Functional Grammar, A.M. Bolkestein, C. de Groot & J.L. Mackenzie (eds),
219–234. Dordrecht: Foris.
Vázquez, V. 2002. Cláusulas relativas en cora meseño. In Del cora al maya yucateco, P. Levy (ed.),
269–348. México: UNAM.
Vendler, Z. 1967. Facts and events. In Linguistics in Philosophy, 12–146. Ithaca NY: Cornell Uni-
versity Press.
Vendler, Z. 1970. Say what you think. In Studies in Thought and Language. J.L. Cowan (ed.),
79–97. Tucson AZ: They University of Arizona press.
Weber, D.J. 1989. A Grammar of Huallaga (Huánuco) Quechua. Berkeley CA: University of Cali-
fornia Press.
From demonstrative to relative marker
to clause linker
Relative clause formation in Pima Bajo
Zarina Estrada-Fernández
Universidad de Sonora
1. Introduction
been provided except for Hale’s 2002 discussion of the origin of the relative marker
in Pima Bajo.
The aim of this paper is threefold: (i) to show that RCs in Pima Bajo are encoded
via a relative marker which is distinct from the relativization strategies observed
in other Uto-Aztecan languages from the same geographic area, e.g. O’odham,
Opata, Yaqui, and Tarahumara; (ii) to show that the relative marker in Pima Bajo
follows the expected predictions (Heine & Kuteva 2002) about the grammaticali
zation of demonstratives as a source for such markers, and (iii) to show that Pima
Bajo has only subject and object RCs; unexpected patterns observed in oblique
constructions fulfill the semantic expectations of RCs since this language doesn’t
allow this type of RCs. Our analysis findings show that within certain oblique
constructions the relative marker is being reanalyzed as a clause linker or clausal
connective (Heine & Kuteva 2002).
The analysis of RCs in Pima Bajo posits some challenges, first to linguis-
tic theory, since the language shows certain unexpected patterns in the case
of oblique constructions that seem to be functionally equivalent to RCs, and
second, to linguists interested in language documentation, since the language, in
comparison with other Uto-Aztecan languages from the same region, is chang-
ing rapidly.
The paper is organized as follows: Section 2 provides some basic grammatical
information about Pima Bajo; Section 3 provides background on RCs, in p articular
the process of nominalization, and describes subject and object RCs, where word
order restrictions as well as the degree of nominalization of RCs are also addressed;
Section 4 deals with the historical development of the relative marker in Pima
Bajo; and finally, Section 5 sheds new light on the oblique RCs recently observed
in the language.
Pima Bajo is an agglutinative language with a very small number of suffixes, i.e.
there are no more than three suffixes in a verb root. Other relevant grammatical
properties are: (i) the language shows no case marking on nouns, only on deter-
miners, e.g. (1a–b); (ii) the unmarked or preferred word order is APV, which it
is quite flexible depending on the context; (iii) lexical or pronominal arguments
can be freely omitted if they are predictable from the discourse; and (iv) personal
pronouns only distinguish between subject and non-subject, i.e. a nominative-
accusative alignment, e.g. (2a–c) and Chart 1:
Zarina Estrada-Fernández
A P V
(1) a. ig gogis ik lii oob kik-im.1
det.sbj dog det.nsbj dim person bite-cont2
‘The dog is biting the child.’
A P V
b. ig kil ik supon hug.
det.sbj man det.nsbj chicken eat.pfv
‘The man ate the chicken.’
A P V
(2) a. takav aan Peier nohog-id.
yesterday 1sg.sbj Peter help-pfv-appl
‘Yesterday I helped Peter.’
A R T V
b. aap lii oob tai bih as kait.
2sg.sbj dim person fire bring.pfv rep say.unmk
‘“You brought fire to the child,” he replied.’
A V
c. aap in-a’as-tar.
2sg.sbj 1sg.nsbj-laugh.pfv-caus
‘You make me laugh.’
Basic declarative sentences in Pima Bajo are usually marked by a restricted set
of tense-aspect-mode (TAM) suffixes: -im ‘continuous’ or ‘progressive’ in (3a),
root truncation for ‘perfective’ in (3b), -va ‘completive’ in (3c), -tad ‘remote’ or
‘imperfective past’ in (3d), -a/-hag ‘future/prospective’ in (3e), -ia ‘probable
future’3 in (3f), and an unmarked form for ‘present/habitual/imperfective’ in (3g).
1. I want to thank to all the Pima Bajo speakers who for more than twenty years have helped
me to understand the language, in particular to Cleotilde Galaviz who is the one who has
helped me most during the past years. The RCs data come from my own field notes. Very few
of them come from texts, since these kinds of constructions are scarce in those materials.
2. a - agent, acc - accusative, ant - anterior, appl - applicative, asp - aspect, cmpl - completive,
cond - conditional, conj - conjunction, cont - continuous, dat - dative, dem - demonstra-
tive, det - determiner, dim - diminutive, dir - directional, dur - durative, dv - deverbalizer,
e - iqui, fut - future, gen - genitive, hab - habitual, imp - imperative, impf - imperfective, int -
intensifier, intr - interrogative, irr - irrealis, loc - locative, m - modal, neg - negative, nom -
nominative, nmlz - nominalizer, nsbj - non-subject, obj - object, obl - oblique, p - patient,
pfv - perfective, pl - plural, poss - possessive, prtc - participial, pst - past, r - recipient, refx -
reflexive, reit - reiterative, rel - relative, rem - remote, sbj - subject, sg - singular, st - stative,
sub - subordinator, t - theme, unmrk - unmarked.
3. This suffix provides a major degree of certainty to the clause in comparison to the plain
future.
From demonstrative to relative marker to clause linker
Non-finite, or atelic, sentences make use of another set of suffixes; the imperative
suffixes, -in ‘imperative.sg.’ and -ivar ‘imperative.pl’; both of these suffixes often
require the irrealis suffix -an.
b. ho’og-in tuuk-an.
side-imp inside-irr
“Take it inside!”
The use of all the suffixes on the first group, except for the ‘completive’ -va, the
‘future’ -a/-hag and the ‘probable future’ -ia, as I will show in the following section,
occurs in not fully nominalized RCs in Pima Bajo, i.e. clauses that are a half way to
being totally nominalized.
All RCs in Pima Bajo which are clauses that form a constituent with their head
nouns are embedded clauses, since they all appear within the main clause.
From a morphosyntactic point of view, RCs may be considered to be nominal-
ized. Nominalization of a clause is a well-known process in clause combining
that is observed when a clause starts functioning as an argument of a verb or
as a nominal element or noun phrase, as in RCs. Nominalization and non-
finiteness are s yntactic-related processes; both terms are used to name distinct
syntactic properties shown in a clause when it degrades, desententializes or
adjusts to function as a noun phrase or as an argument of another clause (cf.
Givón 2009: 63–73; Bisang 2001: 1400–1402). Givón’s (2009: 66) definition of
nominalization is the following:
(5) Nominalization is the process via which a finite verbal clause – either in
its entirety or only the subject-less verb phrase – is converted into a noun
phrase.
Other authors, like Lehmann (1984), have pointed out the following aspects as
those that show the “increasing nominalization” status of a subordinate clause:
(7) a. agglutination of the subordinator to the verb
b. limited choice of verbal categories on the verb (non-finite markings)
c. appearance of the subject in the genitive
d. condensation of the clause to its verbal center
4. Instances of external correlative clauses also appear in the language, and as I will show
later, they appear in an adjunct of peripheral position.
5. Linguistic change that Heine and Kuteva (2002) have previously discussed as a alternative
route for the grammaticalization or relative markers.
6. Troike (forthcoming) observes for Coahuilteco, an extinct isolated language spoken in
Texas, that “is the only SOV language in North America with externally-headed relative clauses
following the Noun, and followed in turn by an independent (not encliticized) D
emonstrative
(N-RC-Dem), a structure apparently unique in North America.”
Zarina Estrada-Fernández
There are three main types of RCs that appear in the language; all of them
show the relative marker -kig on the verb and one or two other strategies for
relativization (Comrie 1998; Kuteva & Comrie 2006): (i) the externally-headed
RC with no subordinator element at the beginning of the clause; (ii) the externally-
headed RC with a personal pronoun at the beginning of the clause, as in (8c), and
(iii) the headless RC, where a determiner appears at the beginning of the clause,
as in (8d). For (8a–b) a distinction is made between subject RCs, as in (8a), where
the strategy of omission (gap)7 applies for the argument which is coreferent with
its head noun, and the object RCs, as in (8b), where a non-subject or genitive
pronoun encodes the notional subject within the RC:
(8) Externally-headed RC with no subordinate relative pronoun:
a. hig a’an [ gii-kig ] vig.
det.nom feather fall.pfv-rel red
‘The feather that fell is red.’
b. ig okosi [ in=niir-kig ] ig gi’id.
det.sbj woman 1sg.nsbj=see.pfv-rel det.sbj big
‘The woman I saw is big.’
Headless RC:
d. [ig da-kig] gii.
det.sbj jump.pfv-rel fall.pfv
‘The one that jumped, fell.’
At first glance, the distinction between the encoding of subject and object
RCs is an important one in terms of strategies of relativization (Comrie 1981,
1998; C omrie & Kuteva 2005; Kuteva & Comrie 2006), since a distinct strategy
characterizes a different function of the head noun within the RC. In subject
relatives, i.e. clauses where the subject is identical to the head noun, as is observed
in (8a), the omission, i.e. gap, of the subject of the RC (zero anaphora) condenses
the clause into its verbal center (i.e. Lehmann 1984, cf. 7). Object RCs, in contrast,
i.e. clauses with no semantic identity between the head noun and the subject of the
RC, also omit the relativized argument, but the clause encodes the notional subject
by means of a non-subject (genitive) pronoun. This property clearly indicates the
7. Comrie and Horie (1995: 66) define a gap as the constituent that is omitted when
relativized.
From demonstrative to relative marker to clause linker
non-finiteness or nominalized status of the RC; e.g. in= ‘1SG.NSBJ’, as in (8b) and
(9a), and am= ‘2SG.NSBJ’, as in (9b):
(9) a. okosi [ in=niir-kig ] ni’i-im.
woman 1sg.nsbj=see.pfv-rel sing-impf
‘The woman that I saw is singing.’
b. gogos [ in=niar-kig ] si’ lii.
dog 1sg.nsbj=buy.pfv-rel int small
‘The dog that you bought is small.’
No other finite or non-finite markers are allowed to modify the verb within the
RCs in Pima Bajo, not even other nominalizing suffixes, i.e. -dam, which is found
in subordinated clauses of purpose, as in (12a), nor the stative suffix -ka, as in
(12b), nor the irrealis suffix -an, example (12c):
Zarina Estrada-Fernández
Finally, RCs like those provided from (8–12) fall within a single intonation con-
tour. This property, according to Givón (2009: 62 & 277) is good evidence of the
closeness that exists between the head noun and its modifier. Adjoined or correla-
tive clauses (cf. Andrews 2007), which have two distinct intonation contours, are
usually ordered after the main clause, and the clause does not immediately follow
the head noun, example in (13):
(13) aan takav gogis nii [tiskil kii-kig].
1sg.sbj yesterday dog see.pfv pig bite.pfv-rel
‘Yesterday I saw a dog (the one) that the pig bit.’
To sum up, the structural strategies observed in subject and object RCs from Pima
Bajo are: (a) all RCs are postnominal; (b) all show a relative marker -kig on the
verb; (c) the clause is not fully nominalized, since the verb within the RC accepts
a limited choice of finite TAM suffixes, as well as non-finite morphology; and (d)
in object RCs, a non-subject pronoun (genitive) appears, i.e. RCs where the sub-
ject of this construction is not identical or coreferent to the head noun.8 This set
of strategies is also useful to characterize RCs from Pima Bajo as distinct from
complement clauses (Comrie & Horie 1995). Examples in (14) show these differ-
ences; both clauses are equi-subject since their subject is omitted or gapped, but
the clauses have two different connectives: the suffix -kig in the RC in (14a) and the
subordinator suffix -it, in (14b), the clauses differ in that the complement clause
bears the suffix -an ‘irrealis’ which is not accepted in RCs:
(14) a. kil [tikpaan-im-kig.]
man work-cont-rel
‘The man who is working.’
b. kil maat-it [diid-an.]
man know-sub smoke-irr
‘The man knows how to smoke.’
8. The non-subject pronoun can be, according to our description of the basic properties of
Pima Bajo, considered to be a genitive pronoun.
From demonstrative to relative marker to clause linker
The relativization of oblique arguments is not possible in Pima Bajo. This situation
supports the Accessibility Hierarchy proposed by Keenan and Comrie (1977). The
data will be addressed in Section 5.
4. The comparative scenario and the origin of the relativizer -kig
Both subject and object RCs in (15a–b) bear a participial (PRTC) or nominalizing
suffix -k(a) at the end of the verb within the RC. Immediately to the right of this
nominalizing suffix a determiner hig functions as subject of an attributive non-
verbal predication. Paratactic attributive clauses – with two distinct intonational
contours – as in (16), are linked to those in (15) and crucially show that the rela-
tive marker -kig from Pima Bajo originated from a demonstrative. Example (16a)
shows an independent emphatic determiner higai, following the attributive clause;
in (16b), in contrast, the subject of the clause is a first person pronoun, aan ‘I’;
Zarina Estrada-Fernández
both clauses in (16), however, bear the stative or participial suffix modifying the
dependent verb:
(16) a. [ naksil gogis ki’i-k] higai aan mia.
scorpion dog bite-prtc det.emph 1sg.sbj kill.pfv
‘To that one I killed. The scorpion that bit the dog.’
b. [ naksil gogis ki’i-k.] aan a=mia.
scorpion dog bite-prtc 1sg.sbj 3sg.obj=kill.pfv
‘I killed it. The scorpion that bit the dog.’
The diachronic origin of the relative marker -kig is traceable in Névome, an histor-
ical variety of Pima Bajo now extinct. Loaysa (Smith 1862), author of the Arte de
la lengua nevome, mentions that RCs in this language have no relative pronouns,
but only verbal participles. Among these, the author mentions -cama, -dama, and
-cugai. Constructions in (17) are taken from Smith (1862: 31) to illustrate RCs
from Nevome:
(17) a. T’-oga [ tidamacatum’-ami da-cama9] s’-cuga
our-father heaven-loc sit-nmlz st-good
m’-tu-na mu-tuguiga.
r-great-cond your-name
‘our Father, who sits in Heaven, may your name be regarded well’
b. [ governaro tu-tuanu-cugai ] si-bu[h]ogurhida-raga.
governor rei-order-nmlz int-obey-worthy
‘What the governor orders should be obeyed.’
c. [ n’-ohana’-cugai ] s’-amurhida-mut’-api posa pim
my-write-nmlz st-know-want-2sg.sbj but neg
‘an’-t’-io m’-agui.
1sg-pfv-fut you-tell.pfv
‘you want to know what I will write, but I will not tell you.’
d. [ n’-usi’-cugai ] si-gugu an’-igui.
my-plant-prtc st-need 1sg.sbj-e
‘I need what I planted.’
9. A clear example showing that -cama was a derivational suffix in Nevome is the following:
Paparh hipuidag-cama
bad.pl heart-nmlz
‘The people with bad hearts’
From demonstrative to relative marker to clause linker
The nominalized status of the clause is marked by the combination of two suffixes;
the stative suffix -ca and a derivational suffix -me, as in (20a). This combination
of suffixes appears to be frequent in complement clauses, as in (20b-c), which also
can be case marked with the accusative -ta:
(20) a. Idaqui temata [ theopachi hio-ca-me-ta] üida.
dem bread church paint-st-impf-acc get.pl.imp
‘Get some bread to the one that is painting the church.’
b. [Eme=ne hio-ca-me-ta] erà.
2sg.acc=1sg.nom write-st-impf-acc think
‘I think that you write.’
c. [Eme=ne cai gua-ca-me-ta] eràve.
2sg.acc=1sg.nom neg eat-st-impf-acc think
‘I thought that you didn’t eat.’
10. In most Uto-Aztecan languages which are basically SOV languages, RCs are ordered
after the Head Noun.
Zarina Estrada-Fernández
A look at RCs from Yaqui and Tarahumara confirms that the nominalizing mark-
ers are the main relativization strategy. The examples in (21) show that RCs in
Yaqui are marked with two distinct combinations of suffixes: -ka-me, for subject
relative, in (21a) or -ka-’u, for object relative, in (21b):
(21) a. U yoeme [ aman weye-ka-me ] ripti.
det.nom man.nom loc stand-pfv-nmlz11 blind.st
‘The man who stood there is blind.’ (Guerrero 2005: 4)
b. U bisikleeta [ in jinu-ka-’u ] sikili.
det.nom bike.nom 1sg.gen buy-pfv-nmlz red
‘The bike that I bought is red.’ (Guerrero 2005: 5)
RCs from Tarahumara also show a cognate set of those markers, where the
participle -ka shows a voiced velar consonant: -ga-me in a stative/perfective
nominalization,12 e.g. (22a), or -(a)me in a non-stative relativization, (22b–c); all
are examples from Burgess (1984: 131):
(22) a. ‘larigá me’á-me ka-rá-če [ yá
thus kill-nmlz be-quot-emph before
múčí-ga-me ] ralámuli-ka…
be:pl-st-nmlz people-emph
‘That is the way people who lived before were killed…’
b. yé rió [ marsiál ani-li-ame ] imé
this man Marcial say-pass-nmlz agave
me’čí-ame ka-rá-tu.
cut-prtc be-quot-nonspec
‘They say this man called Marcial was a cutter of agave (maguey).’
c. ačí né bilé rió [ rono-či o’kó-me.]
now I a man leg-loc hurt-nmlz
‘I know a man whose leg hurt.’
11. For sake of comparison, I have changed some of Guerrero’s (2005) glosses, e.g. the suf-
fixes -ka-me were glossed by her as ‘perfective-complementizer’.
12. The suffix -ga also appears in Burgess (1984: 23) glossed as perfective; e.g. ’lige ma
muku-ga then now dead-pfv ‘there was dead’.
From demonstrative to relative marker to clause linker
I shall consider here two kinds of empirical data from Pima Bajo. The first group
supports the analysis that this language lacks oblique RCs. The second group pro-
vides evidence for another grammaticalization pathway related to the relative
marker -kig. Initially, I will demonstrate that the attribution of oblique arguments
is only possible by means of paratactic constructions, i.e. with no relative marker
or pronoun (i.e. which also functions as an interrogative pronoun). I will finish this
section with the presentation of empirical data that counts as evidence to demon-
strate that during the diachronic change from paratactic to embedded RCs, the rel-
ative marker gained the possibility of functioning as a clause linker or subordinator.
Oblique constructions in Pima Bajo support the Accessibility Hierarchy from
Keenan and Comrie (1977) in that this language doesn’t allow RCs in this type
of arguments; in other words, the patterns observed for subject and object RCs
do not apply for the relativization of oblique arguments. Paratactic constructions,
where the attributive clause does not form a constituent with its head and without
any relative marker, are provided for a locative argument in (25), or a secondary
object, i.e. a theme participant in a ditransitive construction, in (26):
(25) Peier kor-tam dai [ aki-vui kor-tam. ]
Pete fence-loc jump.pfv river-dir fence-loc
‘Pete jumped the fence, the one which is by the river.’
(26) Marii Peier kav maa [ takav aan niar. ]
Maria Peter horse give.pfv yesterday 1sg.sbj buy.pfv
‘Maria gave Peter the horse that I bought yesterday.’
13. The grammaticalization of a demonstrative into a relative clause marker starting from
paratactic constructions as is the case of the relative marker -kig in Pima Bajo also sup-
ports the proposal given by Givón (2009), where in a process of change, new structures “may
reacquire finite properties,” but also show some nominalized ones. In Pima Bajo, the relative
clauses accept only a restricted set of finite TAM suffixes in the verb root.
Zarina Estrada-Fernández
14. The shirt’s red sleeves used to characterize a group of Mexican soldiers fighting against
the Apache at the beginning of the 20th century.
15. Postpositions in some Uto-Aztecan languages such as Pima Bajo and Yaqui can option-
ally be attached to the noun they modify. This may be a prosodic feature that announces their
change from postpositions to case markers.
From demonstrative to relative marker to clause linker
Constructions in (29–31) have three special features. First, the marker previ-
ously known as a relative marker, -kig, appears in all the examples; this time not
directly attached to the verb, but to the postpositional phrase, where two of the
postpositions are clearly identified: -vui ‘to’, and -vuika ‘for’. Second, two of the
clauses following the suffix -kig are clearly dependent clauses, since both contain a
notional subject encoded as a non-subject pronoun. Third, two of the clauses have
a suffix -ta- which is clearly an old Uto-Aztecan case marker (Langacker 1977),
which is actually non-productive in Pima Bajo.
Even though the examples may be seen as awkward, the analysis of other con-
structions such as those in (32) where the suffix -kig attaches to a relative (i.e.
grammaticalized from the interrogative aita’a)16 pronoun demonstrates that this
marker has been reanalyzed to function as a clause linker or connective.
(32) a. in-maak-in himak boteii aita-kig sudag nukad!
1sg.sbj-give-imp one bottle inter-lin water have.prs
‘Give me the bottle that contains water!’
b. ig kil mua gogos aita-kig in-kiik.
det.sbj man kill.pfv dog inter-lin 1sg.nsbj-bite.pfv
‘The man killed the dog that bit me.’
c. ig gogos aita-kig ig kil mua ko’ok-ad.
det.sbj dog inter-lin det.sbj man kill.pfv sick-rem
‘The dog that the man killed was sick.’
In this paper I have addressed the question of RCs in Pima Bajo. In my analy-
sis I have shown the applicability of the Accessibility Hierarchy in the language
since only subject and object RCs are permitted. The fact that Pima Bajo doesn’t
16. The interrogative aita’a grammaticalizes from hai’ta ‘thing’. This interrogative pronoun
usually questions a theme participant, as for example:
We have also focused on the historical development of the relative marker from a
determiner. This strategy appeared to have a quite distinct diachronic origin from
other nominalization strategies of other Uto-Aztecan languages from the same
geographical area. The process of suffixation of the determiner higai follows a pro-
cess of analogy, which is also observed with other subordinators in the language,
as for example the subordinator ko, which usually appears in verbal complements,
example (34a), but it can be also suffixed to the interrogative aita- to form another
clause linker, as in (34b):
(34) a. Peier maat ko kav tiis-ab.
Peter know.prs sub horse ride.prs-dir
‘Peter knows how to ride a horse.’
b. ig kil mua gogos aita-ko kiik-im-tad.
det man kill.pfv dog inter-sub bit-impf-rem
‘The man killed the dog that was biting him.’
My analysis also supports Comrie & Kuteva’s (2005) proposal, i.e. languages may
employ different morphosyntactic relativizing strategies for the distinct functions
of the head noun within the RCs. For Pima Bajo, a two-way proposal proved to be
necessary: first, for the subject RCs, where the main relativization strategies were a
verbal relative marker -kig and omission of the coreferent element, and second, for
object RCs, where the main relativization strategies were the use of the verbal rela-
tive marker -kig, and the notional subject obligatorily encoded as a non-subject
pronoun.
Furthermore, non-conventional patterns observed in functional equivalent
constructions to oblique RCs support an alternative grammaticalization pathway
from paratactic constructions to embedded ones. This pathway is observed when
the determiner -kig functions as a clause linker:
(35) Marii Peier kav maa takav-kig aan niar.
Maria Peter horse give.pfv yesterday-lin 1sg.sbj buy.pfv
‘Yesterday, Maria gave Peter the horse that I bought.’
Final remarks: our findings have been able to show that typological properties are
not to be seen as a closed or packaged set of grammatical features. Languages change
over time and across boundaries. As consequences of such changes, languages
which are considered to be part of a single family show no rigid and consistent
From demonstrative to relative marker to clause linker
References
Andrews, A.D. 2007. 4. Relative clauses. In Language Typology and Syntactic Description, Vol. 3:
Complex Constructions, T. Shopen (ed.), 206–236. Cambridge: CUP.
Bisang. 2001. 100. Finite vs non-finite languages. In Language Typology and Language Univer-
sals. An International Handbook, Vol. 2. M. Haspelmath, E. König, W. Oesterreicher & W.
Raible (eds), 1400–1430. Berlin: Walter de Gruyter.
Burgess, D. 1984. Western Tarahumara. In Studies in Uto-Aztecan Grammar, Vol. 4: Southern
Uto-Aztecan Grammatical Sketches [Summer Institute of Linguistics Publications in Lin-
guistics 56], R.W. Langacker (ed.), 1–49. Dallas TX: The Summer Institute of Linguistics
and The University of Texas at Arlington.
Casad, E.H. 1984. Cora. In Studies in Uto-Aztecan Grammar, Vol. 4: Southern Uto-Aztecan Gram-
matical Sketches [Summer Institute of Linguistics Publications in Linguistics 56], R.W. Lan-
gacker (ed.), 153–140. Dallas TX: Summer Institute of Linguistics and the University of
Texas at Arlington.
Comrie, B. 1981. Relative clauses. In Language Universals and Linguistic Typology. Syntax and
Morphology, 131–157. Oxford: Blackwell.
Comrie, B. 1996. The unity of noun modifying clauses in Asian languages. In Pan-Asiatic
Linguistics: Proceedings of the Fourth International Symposium on Languages and Linguis-
tics, January 8–10, 1996, Vol. 3, 1077–1088.
Comrie, B. 1998. Rethinking the typology of relative clauses. Language Design 1: 59–86.
Comrie, B. & Horie, K. 1995. Complement clauses versus relative clauses: some Khmer evidence.
In Discourse Grammar and Typology. Papers in Honor of John W.M. Verhaar, W. Abraham,
T. Givón & S.A. Thompson (eds), 65–75. Amsterdam: John Benjamins.
Comrie, B. & Keenan, E.L. 1979. Data on the noun phrase accessibility hierarchy. Language
55(2): 333–351.
Comrie, B. & Kuteva, T. 2005. 122–123. Relativization strategies. In The World Atlas of Language
Structures, M. Haspelmath, M.S. Dryer, D. Gil & B. Comrie (eds), 494–501. Oxford: OUP.
Dedrick, J.M. & E.H. Casad (ed.). 1999. Yaqui Language Structures. Tucson AZ: University of
Arizona Press.
Downing, B.T. 1978. Some universals of relative clauses structure. In Universals of Human Lan-
guage, Vol. 4: Syntax, J. Greenberg (ed.), 375–418. Stanford CA: Stanford University Press.
Estrada Fernández, Z. 1991. Arguments and Clausal Relations. Ph.D. dissertation, The Univer-
sity of Arizona.
Estrada Fernández, Z. 1996. Pima Bajo. A Grammatical Sketch of Pima Bajo [Languages of the
World/Materials 71]. Munich: Lincom.
Estrada Fernández, Z. 2008. Cláusulas relativas en pima bajo. Revista LIAMES (Línguas Indíge-
nas Americanas) 8: 69–83.
Givón, T. 2001. Syntax. An Introduction, Vol. II. Amsterdam: John Benjamins.
Zarina Estrada-Fernández
Tim Thornes
Boise State University
The present paper explores the grammatical formation and synchronic variation
of relative clauses in the Northern Paiute (Western Numic; Uto-Aztecan) language,
as determined by their functional and grammatical connections to nominalization.
We find support for several hypotheses from the literature regarding the
development of syntactic complexity along a paratactic-to-syntactic pathway. An
approach that seeks functional explanations for diachronic developments helps to
make sense of the data, particularly in connecting nominalization to relativation
not as one of several available strategies for relative clause formation, but as part of
the same complex functional and grammatical domain.
1. Introduction1
1. Institutional support for this work has been provided by NSF grant #0418453 and is
hereby gratefully acknowleged. I would also like to thank Tom Givón, the editors, and two
anonymous reviewers for comments and suggestions on earlier drafts, which have no doubt
improved the quality of this paper.
Tim Thornes
been proposed, for example, for some Tibeto-Burman languages (Noonan 1997;
DeLancey 1986) as well as some Turkic languages (Greg S. Anderson in personal
communication). Relative clauses formed via nominalization can be viewed as
morphosyntactically complex noun phrases, just as simple nominalizations are
morphologically complex nouns. Their role in narrowing (in the case of headless
relative clauses, establishing) reference, or in characterizing nominal attributes,
typically through a genitive (possessive) relationship, can then be better articu-
lated. These features are part of the synchronic description of relative clauses
undertaken here for Northern Paiute.
As with most attempts at synchronic description, however, explanation of
the patterns we find can often best be sought through an exploration of probable
historical processes. The relationship between diachronic explanation and syn-
chronic description has been a storied one in the field of linguistics, at least since
the “Saussurean prohibition against mixing synchrony and diachrony, bolstered by
the Chomskyan argument that the language-learning child must construct their
grammar without reference to anything but synchronic facts” (Evans & Dench
2006: 19). Synchronic facts are not mono-dimensional objects, however, and so
language-internal variation has a important role to play not only in helping us
understand how we acquire language, but also, for our present purposes, how we
describe and explain these facts.
As Givón (2007, and within the present volume), points out, the study of the
synchronic variation of related constructions is but one method for syntactic recon-
struction, others being either the study of historical records or the application of
internal reconstruction methodologies based on morphological or s yntactic “relics”.
Northern Paiute, as with most languages, lacks a lengthy written h istory necessary
2
for clarifying the diachronic pathways presented here. Internal reconstruction
I owe an incredible debt of gratitude to my language teachers Rena Adams Beers and Ruth
Hoodie Lewis. I must also acknowledge the late Irwin Weiser (1909–1996), Maude Washington
Stanley (1913–2000), Myrtle Louie Peck (1934–2006), Nepa Kennedy (1918–2010), and
Justine Louie Brown (1918–2011) for their monumental patience in sharing their language
with me. I would also like to thank Lloyd Louie, Patricia Miller, Shirley Tufti, Ken Barney,
Phyllis Harrington Miller, and Yolanda Manning for their generous assistance in developing a
better understanding of Northern Paiute.
The beauty and interest inherent in the language rest clearly with these and the many
generations of speakers that preceded them, whereas the responsibility for any errors of
interpretation, sloppy analyses, or other problems the reader finds here rests solely with me.
2. I should point out that written documentation does reach back a little more than a century,
and the detailed comparison of these and later records may eventually help us to understand
the possible impact of language attrition.
Functional underpinnings of diachrony in relative clause formation
plays a minor role in this study, since one of the main goals of this paper is to
explore the extent to which the Northern Paiute data fit Givón’s proposed “mega-
pathways” of diachronic development. As part of that endeavor, we will need to
consider not only the possible historical source(s) of relative clause structures, but
also their place in a complex and dynamic continuum of development that includes
clausal subordination more generally. To accomplish this, a look at synchronic vari-
ation in related constructions is the main method applied here.
Although relative clause formation in Numic is fairly uniform in morpho-
syntactic profile, little work has explored how detransitive voice constructions
interact with relativization in these languages, what relationships there may be
between relative clauses and other subordinate clause types, or historical syntactic
developments more generally. As far as the interaction of voice and relativiza-
tion is concerned, the focus will be on understanding a genre of headless rela-
tive clause, including their function in establishing and controlling reference and
information flow in discourse through participant selection, focus, and suppres-
sion. An expansion of the topic belongs more properly in a more detailed study
of voice operations in the language than can be managed within the confines of
the present work.
Regarding the relationship between relativization and subordination in gen-
eral, a growing body of recent work suggests that adverbial clauses may arise his-
torically from relative clauses through various means. As Epps (2007, and within
the present volume) suggests for Hup, a Nadahup language of Amazonia, adver-
bial clauses can develop via the reanalysis of headless relative clauses of a particu-
lar type. The suggestion that nominalized relative clauses could have arisen from
reinterpreted verb complements is also found, for example, in Heine and Kuteva
(2007). Again, data presented here suggests possible corroboration with these
findings, particularly with respect to the use of the Northern Paiute participle suf-
fix to mark a wide range of subordinate clause types. This study represents an early
first step in understanding the nature of these historical relationships.
In terms of broad historical trends, a major goal of the present paper is to
explore how well the Northern Paiute data fit a diachronic typology of relative
clauses, such as that presented by Givón (2007 and 2009, Chapter 5, and within the
present volume), and whether what is brought to bear here can shed further light
on pathways leading to syntactic complexity. The Northern Paiute data appear to
fit well within the process of expansion (versus integration) described in Heine and
Kuteva (2007: 216–224) and Heine (2008) – that is, clausal subordination (in this
case, relative clause formation) arising from “the reinterpretation of a thing-like
(nominal) participant as a propositional (clausal) participant (Heine 2008: 1)”.
A look at the formal and functional connections between non-subject relative
clauses and other subordinate clause types is thereby undertaken.
Tim Thornes
3. Abbreviations used in the examples include 1, 2, 3, 4 for personhood of pronominal form;
apl - applicative; aps - anti-passive (unspecified patient/object); caus - causative; cisl -
cislocative (i.e. “motion toward”); com - comitative; cont - continuous; dem - demonstrative;
denom - denominalizer; disjunct - disjunctive; dl - dual; dur - durative (usu. intervocalic
glottalization or medial consonant fortition/gemination); emph - emphatic; excl - exclusive;
fut - future; hab - habitual; inch - inchoative; incl - inclusive; int - intensifier; intr - intran-
sitive; ip/ - instrumental prefix (with simplified gloss); logo - logophoric (reflexive) possessor;
mm - middle marker; mod - modal; neg - negation; nmr - nominalizer; nom - nominative
case; obl - oblique/non-nominative case; obv - obviative; pfv - perfective; pl - plural; pnc -
punctual aspect; ptcp - participle; re - reduplication; restr - restrictive (relative) pronoun;
rndm - random motion; sg - singular; simil - similitive; spl - suppletive form; stat - stative;
subj - subjunctive; tns - tense; tr - transitive: trnsl - translocative (i.e. “motion away”); a dash
‘-’ indicates a morpheme boundary and an equals sign ‘=’ indicates a clitic boundary.
Tim Thornes
Particularly important for the data presented in this paper is the syncretism that
exists among the Northern Paiute pronominal proclitics of Table 3. These forms
function both as direct objects of verbs and possessors of nouns. Only the last form
on the chart, the proclitic marking logophoric (reflexive) possession, is uniquely
attached to nominals, and so its appearance in relative clauses makes the role of
nominalization in relativization processes clear.4
4. A historical connection between the logophoric (reflexive) possessor proclitic and the
antipassive (unspecified patient) verbal prefix is suggested both in Thornes (2003: 177–178)
and Langacker (1977: 46). Langacker suggests a broad functional connection, whereas Thornes
suggests a functional split in the history of *tɨ-.
5. I am using the term ‘logophoric’ as an extension of the traditional sense where there are
two separate pronominal forms to distinguish cases of possible (but potentially ambiguous)
coreferentiality with respect to a third person argument. In English, “The boy saw his mother”
is ambiguous as to whether or not “his” refers to the boy. Classic logophoricity serves to distin-
guish the referential properties of subsequent clausal subjects, particularly in reported speech.
For example, “The mani said that hei/j didn’t see her” is ambiguous, whereas some languages
would have a logophoric form for hei versus hej.
Functional underpinnings of diachrony in relative clause formation
The coreferentiality requirements are not exclusively local, but may indicate pos-
session by an ongoing topic, whether or not it is the grammatical subject. In the
following example, there has been no overt mention of the coreferent of the logo-
phoric proclitic (Wolf) for several clauses, and it is his younger brother (Coyote)
who is the subject of the second clause.
(5) ooʔno pisa miʔi tɨɨkwiʔi-na.
dem good QUOT tell-ptcp
‘“That’s good”, (Wolf) was saying.’
yaisi tɨ= kwaŋa mia.
then logo= younger.brother go.sg
‘Then his younger brother went’. (Marsden 1923: ‘The Cave Myth’)
6. See Toosarvandani (2010) for an insightful discussion of nominalization patterns across
Numic.
7. The range of functions associated with this suffix renders it possible to adopt the term
quasi-converb based on Nedjalkov’s (1995) typology. The key feature of a converb is in the
marking of adverbial clauses. Toosarvandani (2010) refers to this as the “patient and event
nominalizer”.
Tim Thornes
(teach-NMR) ‘teacher’), the suffix appears at first blush to be a typical agent nomi-
nalizer. However, we see it with passivized verb forms, as in (6), in order to make
reference to a semantic patient, and so subject nominalizer is more precise. With
morphosyntactically more complex examples like the following, it is possible to
interpret such forms as headless relative clauses:
(6) [nɨɨtaʔnidɨ]
na-wɨtaʔni-dɨ
mm-gather.by.whacking-nmr
‘buckberries’ (literally, “what is gathered by whacking”)
(7) taba-tsiboi-kwa katɨ-dɨ
sun-emerge-loc sit-nmr
‘the President’ (literally, “(the one) who sits to the east
(sun-emerging side)”)
Of interest with regard to both (8) and (9) is their interpretability either as head-
less relative clauses like ‘what I dream’ and ‘what I shade’ or as nominalizations
like ‘my dreaming’ or ‘my shading’. This is only an issue with translation, however,
although it reveals rather clearly a functional connection between relative clause
formation and nominalization. As we will see further below, notional subjects of
relative clauses are typically treated syntactically as possessors.
Comrie and Kuteva (2005) define a relative clause as a “clause narrowing the
potential reference of a referring expression by restricting the reference to those
referents of which a particular proposition is true (1)”. For the purpose of rela-
tive clause typologies, Northern Paiute relative clauses typically follow the head,
or domain, noun – the “referring expression”. In other words, they are postnom-
inal, and, since the head noun maintains the morphosyntactic properties of a
main clause argument, we can consider them externally-headed. The typological
Functional underpinnings of diachrony in relative clause formation
relevance of this feature is thrown into question by the fact that an overt refer-
ring expression (i.e. a head noun) is not a requirement, since there are countless
recorded instances in Northern Paiute natural speech of headless relative clauses.
A key function of such constructions is still clearly to establish or narrow refer-
ence, however.
Since relative clauses are most typically considered constituents of the noun
phrase, an important property of Northern Paiute relative clauses is that they may,
with or without the head, be case-marked for the syntactic role of the complex
noun phrase in the main clause. This property is most typically independent of
the actual role of the head noun within the relative clause proper. In this paper, the
role of the head noun inside the relative clause is described as notional. In an Eng-
lish example like the linguist you met finally got a real job, the head noun linguist
and its associated noun phrase constitutes the main clause subject, but is in fact
the notional object of the relative clause. In terms of relative clause type, the linguist
you met is considered an object relative clause.
As has been widely discussed in the typological literature, languages may dif-
fer in the range of relative clause types that are allowed. Another way to put it is
that only certain notional arguments are accessible to relativization. In Northern
Paiute, there appear to be few restrictions on the role a head noun may serve in
the relative clause – that is, on the issue of accessibility (Keenan & Comrie 1977).
That said, I will focus on the properties of three of the most common relative
clause types – subject, object, and oblique. These vary in two basic ways: (1) which
nominalizer is used to mark the verb of the relative clause and (2) whether or not
there is a relative pronoun present, marked for the role the head noun plays in
the relative clause. I will describe these three types in turn, beginning with their
canonical formal properties and proceeding to a discussion of how these proper-
ties are modified under extraposition, under which conditions the relative clause
appears in appositional relationship to the would-be head noun. It is under these
latter conditions that we see potential evidence for what Givón (2007, 2009) cites
as reflecting historically earlier, paratactic patterns in the diachronic development
of relative clauses.
suffix appear in bold in the following examples, and the relative (nominalized)
clause appears in square brackets.8
(10) nɨ ka= tɨhɨkya [oʔo wɨnɨ-dɨ] punni
I obl= deer dem stand.sg-nmr see.dur
‘I see the deer (that is) standing out there.’
(11) umɨ [kai u= pidzabi-dɨ] ɨmɨ-nɔ tuʔi na-koiwɨnai-ʔyakwi
they neg 3= like-nmr them-with try mm-fight.against-hab
‘... those that didn’t like it would try to fight with them.’
(NK: ‘Boarding School’)
Relative clauses may appear extraposed, particularly when the head noun is the
object of the main clause, often under a separate intonation contour and with a
presumptive pronominal appearing as a proclitic on the main clause verb:
(12) paana kai mɨ= punni, ka= [mɨ=aapo tɨ-tɨha-ga-dɨ]
however neg pl=see obl=pl=apple re-steal-TRNSL-nmr
‘…but (he) didn’t see them, (those) who went to steal the apples…’
(NK: ‘Boarding School’)
Note that the extraposed relative clause in (12) is treated syntactically as a noun
phrase, including case-marking as the syntactic object (oblique) in the main
clause. Although coreferential with the pronominal proclitic, one can question
whether or not the proclitic constitutes a proper head. The structure in (12)
does, however, share properties with other forms of extraposition. Extraposed,
post-verbal direct objects in general require a presumptive pronominal proclitic
on the verb.
(13) uu u= patsa-tabɨa, ka= kutsu
just.so 3= kill.sg-appear obl= cow
‘That’s how (Porcupine) killed her, the cow.’ (NK: ‘Porcupine and Coyote’)
8. I interpret these not as internally-headed based upon the fact that, although the head
noun of a subject relative clause appears in the position for subject of the relative clause
verb, its syntactic case role is clearly marked by the proclitic. I also model this analysis based
upon the non-subject relative clause types described in the following sections, and consider
left-headedness the default pattern for all headed relative clauses in Northern Paiute.
Functional underpinnings of diachrony in relative clause formation
Note that the pronominal proclitic referring here to the notional subject of the
relative clause can readily be interpreted as the possessor of a nominalized verb
(recall (8) and (9) above). When the notional subject appears as a full lexical noun
or noun phrase, the logophoric (reflexive) possessor proclitic appears on the rela-
tive clause verb, coreferring to the head of the relative clause.
(15) su= tɨpi [naatsi tɨ= wɨnai-hu-na]
nom= rock boy logo= throw-pnc-ptcp
‘The rock the boy threw…’
more robust nominal case-marking also supports the possessor analysis for these
constructions.
As has been noted in other Numic languages (cf. Bunte’s 1986) discussion
of Southern Paiute (southern Numic) subordinate clauses) the perfective suffix
-pɨ (pfv), in lieu of -na, does occur, but appears to be optional for speakers of
the dialects I have worked with (and very rare in narrative). The temporal setting
for the relative clause event in these cases very clearly precedes that of the main
clause with the perfective. Dayley’s (1989) description of Tümpisa Shoshone (cen-
tral Numic) demonstrates a sharper functional distinction between the patterns of
use for cognate suffixes along identical lines. Compare:
(16) a. su= miidɨ [i= tɨmɨ-na] sɨda ʔmani-pɨ
nom= meat 1= buy-ptcp bad become-pfv
‘The meat I bought spoiled.’
b. su= miidɨ [i= tɨmɨ-pɨ] sɨda ʔmani-pɨ
nom= meat 1= buy-pfv bad become-pfv
‘The meat I bought (a while back) spoiled.’
These formal parallels are worth noting, since they demonstrate some finite prop-
erties with regard to relative clauses, but are otherwise set aside for our present
purposes. The use of the participle –na is far more common in Northern P aiute,
rendering the temporal relationship between relative clause and main clause
underspecified or, more accurately, context dependent. The participle is the sub-
ordinating suffix of choice not only for non-subject relative clauses, but also in
a wide range of functions, including verbal complements and adverbial clauses.9
The role of the participle in marking clauses that carry paraphrased or otherwise
backgrounded information in narrative is described in some detail in Thornes
(2003: 466–472).
As with subject relative clauses, object relative clauses may also appear
extraposed, as in (17), particularly when the head noun is the object of the main
clause, as with the subject relative clause of (12) above. The following example
is from direct elicitation and illustrates either (1) a relative clause separated,
or extraposed, from the head noun it modifies or (2) a headless relative clause
in apposition to the would-be head noun whose reference it serves to narrow.
In any case, I would argue that the properties of nominalization allow for the
flexibility we see. We may not, in fact, be observing the tight syntactic bond
between head noun and relative clause in a single complex noun phrase, but
9. As pointed out by one reviewer, this property of the participle suffix is widespread
throughout the Uto-Aztecan family.
Functional underpinnings of diachrony in relative clause formation
rather two separate noun phrases in apposition, one a simple noun, the other a
nominalized clause.
(17) mɨ= naʔatsi ka= tɨpi mayɨ-u, ka= [tsiaʔa ti= wɨnai-hu-na]
pl= boys obl= rock find-pnc obl= girl logo= throw-pnc-ptcp
‘The boys found the rock that the girl threw.’
Again, such extraposition also typically occurs under a separate intonational con-
tour. Notice that the extraposed clause is case-marked oblique for its syntactic role
in the main clause, but is otherwise syntactically independent of it, and as such
could be interpreted as headless. The example could then read, “The boys found
the rock, (the one that) the girl threw.” The case-sensitive determiner proclitic ka=
is a historically reduced form of a demonstrative, as I mentioned earlier, in line
with Givón’s (2007) diachronic paratactic-to-syntactic typology.
The (notional) head noun itself may also appear extraposed (as a kind of par-
enthetical afterthought), here under a clearly separate intonational contour and
syntactically separate from both the relative clause and the main clause.
(18) nɨ u= mayɨ, [mɨ= tɨ-watsikɨ-na], ka= kii
I 3= find pl= aps-lose-ptcp obl= key
‘I found it, what they lost, the key.’
There are several features of interest in (18). Prior to the first comma is a com-
plete sentence. Both the extraposed noun phrase ‘the key’ and the nominalized
(structurally, headless relative) clause ‘what they lost’ refer to the direct object of
the main clause. Both are in apposition to the pronominal object proclitic. The
common thread here is that if we consider relative clause formation to be one
function of nominalization, we can interpret differences in degree of syntactic
integration as a function of the appositional relationships holding between the
different noun phrase types exemplified above: (1) the pronominal object proclitic
on the main clause verb, (2) the possessed, nominalized clause (what we can inter-
pret as a headless relative clause), and (3) a lexical noun, case-marked for its role
in the main clause. The variation in syntactic integration then represents a kind of
iconic reflection of the different functional relationships holding between nouns
and nominal attributes. That is, tight syntactic integration correlates with restric-
tive modification, loosely integrated structures correlate with nominal apposition.
The loosely integrated structures we have explored thus far would correspond with
Givón’s postulation of a non-restrictive parenthetical pathway in the diachonic
development of postnominal relative clauses in Northern Paiute and elsewhere in
Numic, and Uto-Aztecan more generally.
Modeling embedded relative clauses on possessor-possessum relationships
also makes both syntactic and semantic sense of the structures we find, a point to
Tim Thornes
which we will return as we look at the prevalence of headless relative clauses and
their interaction with detransitive voice more closely in Section 3. We turn first to
oblique relative clause formation.
One also finds the postnominal oblique relative clause occurring under a separate
intonational contour from the notional head, as a non-restrictive relative clause.
Syntactically, all that’s left in these stand alone nominalized clauses is the merg-
ing of the intonation contour with the phrase that contains a head noun, since it
appears in the usual object position before the verb of the higher clause.
(20) yaisi hima, uuni-ku =tiaʔ, [pɨ-kwai
then what dem-obl =thusly restr-loc
ni= hani-kwɨ-na] uuni-ku ni= himi-na
1.pl.excl= do/wear-fut-ptcp that.kind-obl us= give.pl-ptcp
‘…and (they) gave us those sorts of things, what we were to wear…’
(NK: ‘Boarding School’)
The oblique relative clause construction is commonly used to form what may read-
ily qualify as headless relative clauses. In these cases, there is no referring expres-
sion to which the nominalized clause stands in apposition – it is itself a referring
expression in the form of a (nominalized) oblique relative clause and may alone
occupy the syntactic position required by a verb, as in (21).
(21) u -su [pɨ-kuba u= katɨ-čai-na,] yaisi oo-tu patsa-u
3-nom restr-SUPRA 3= sit.sg-hab-ptcp then dem-ALL kill.sg-pnc
‘He killed the one he was riding (literally, “upon whom his sitting”) there.’
(NK: ‘Porcupine and Coyote’)
Functional underpinnings of diachrony in relative clause formation
One factor that comes into play within the context of relative clause typology
relates to the role of voice operations in the formation of relative clauses. From a
strictly syntactic perspective, the “promotion” of a syntactic object to subject sta-
tus, as in a passive, may serve to provide the necessary syntactic context for relative
clause formation, as in well-known cases where, for example, direct objects are
not otherwise “accessible” to relativization. But what of languages, like Northern
Paiute, where accessibility is considerably less restrictive? What roles do detransi-
tive voice operations play in the formation of different relative clause types via
nominalization?
Following a short introduction to the detransitive voice morphology of
Northern Paiute, a look at data demonstrating the use of such morphology in con-
cert with nominalization to form headless relative clauses is in order. Although
a more extensive discourse study is needed, it is clear that the various combina-
tions of detransitivizing and nominalizing morphology serve clearly distinct
communicative foci. As nominal attributes, the frequency of such clauses gives
Tim Thornes
Reflexive:
(23) a. i= pɨta ni pa-kia-wɨnɨ
1= arm I ip/water-give-cont.sg
‘I’m washing my arms’
b. nɨ na-pa-kia-wɨnɨ
I mm-ip/water-give-cont.sg
‘I’m bathing’ (i.e. washing myself)
Passive:
(24) a. oʔo uu ka u=patsa tabɨʔa;
dem thusly KA 3=kill.sg appear
b. u-su na-patsa tabɨʔa
3-nom mm-kill.sg appear
‘That’s what killed her, I guess; she was killed, apparently.’
(NK: ‘Bear and Deer’)
As with typical passive constructions, the use of na- indicates the topicality of
the patient of a transitive event and the corresponding pragmatic and syntactic
demotion of the agent (obligatorily absent). Valence is reduced, leaving one core
argument, the patient subject, in the clause. It is the passive function that will be
the focus here in its interaction with nominalizing morphology in forming relative
clauses.
Also referred to as the “unspecified argument” prefix (Snapp & Anderson
1982; Langacker 1977), the antipassive prefix tɨ- is required on most transitive
Functional underpinnings of diachrony in relative clause formation
verbs when the patient-object is absent, as we see by comparing (25a) and (25b),
generally due to the fact that it is low in topicality, unknown, or unimportant.
(25) a. nɨ ka= tuku kuhani
I obl= meat cook
‘I’m cooking the meat.’
b. nɨ tɨ-kuhani
I aps-cook
‘I’m cooking.’
In (26a), the benefactive object is unspecified, but the overt patient object sup-
presses the occurrence of the antipassive prefix. In (26b), we find the co-occurrence
of both the antipassive prefix and an object proclitic referring to a benefactive.
Therefore, the prefix appears only to be sensitive to unspecified patients rather than
syntactic objects. Another feature of the antipassive construction is that the focus
could be construed as directed to the action denoted by the verb, not to its effect.
These nominalized clause types are referring expressions whose notional agents,
as in a passive construction, are pragmatically unimportant or detopicalized. Note
that, like other extraposed relative clauses, that of (27) falls under a separate into-
national contour, but remains marked by the determiner proclitic as the would-be
nominative argument of the main clause verb.
The tɨ- Verb -dɨ combination. The head/domain noun is still the notional
subject/semantic agent (A > S) and the patient/object is unspecified, implicit, or
generic.
(28) uuni-ku tɨ-ma-yakwi-na,
that.kind-obl aps-ip/hand-carry-ptcp
su= [tɨ-woitsami-dɨ
ka= tokano]
nom= aps-watch-nmr obl= night
‘That’s what (he) was carrying, that night watchman.’
(NK: ‘Boarding School Days’)
This combination typically serves to name a person by a habitual action (as with
occupations, as in tɨ-gwɨhɨ-dɨ (cf. gwɨhɨ ‘grab’) ‘police officer’ (literally, “the one
who grabs”), but often, as in (28) with the additional clarifying features of a clause,
thereby creating a complex noun phrase. Again, the entire nominalized clause is
case-marked for its role in the main clause.
The na- V -na combination. In the case of this combination, the head/domain
noun must function either as the notional secondary object or an oblique with
respect to the relative clause. The syntactically promoted subject of passive may
appear within the relative clause, but is not available to serve as the head noun
(unlike with the na- V -dɨ combination of example (27)):
Here the interrogative pronoun appears to serve the role of head, although techni-
cally it does not refer to anything beyond the referent established by the restrictive
relative clause that follows it.
Functional underpinnings of diachrony in relative clause formation
Relative clauses of this type are quite commonly used to refer generically to the
result(s) of some transitive action. The logophoric possessor (LOGO) proclitic
corefers to the notional subject – the agent – of the relative clause verb. The coref-
erence requirements associated with it extend to the current topic, last mentioned
several clauses back.
One does often find this construction preceded by a form of the demonstra-
tive translating as ‘the/that kind’ with a case suffix to indicate its non-subject rela-
tion – a kind of empty-headed relative, also known as a fact-S construction:
(31) yaisi himma =bina, uuni-ku [tɨ-ma-kwɨhɨ-na]
then what =be that.kind-obl aps-ip/hand-grab-ptcp
‘Then whatever it was, the kind (of thing he was) holding ...’
(NK: ‘Boarding School Days’)
Note that unniku is not a typical head in the sense that, properly speaking, it plays
no syntactic role with respect to the relativc clause verb, which has been detransi-
tivized with the antipassive prefix.
A clear head or domain noun is only available in the context of an oblique
relative clause whose patient/object is unspecified:
(32) su= “barrel” [pɨ-kwai-ku mɨ= tɨ-woisa-na]
nom= barrel restr-loc-ESS pl= aps-wash.clothes-ptcp
‘the barrel that they did their washing in..’ (MS: ‘Autobiography’)
What we have been referring to as the participle suffix carries a range of functions,
including nominalization and non-subject relative clause formation, as we have
seen, as well as marking the verbs of complement clauses and adverbial clauses
coding temporal simultaneity.10 Note the following example of an embedded ver-
bal complement [in square brackets].
The subject of the complement clause verb in (33) appears in the telltale pro-
clitic position, as in a non-subject relative clause, in a possessor relationship to
the nominalized verb form. Crucial to our understanding of these construc-
tions as nominalizations is just this possessor-possessum relationship between
the notional subject and the nominalized verb or verb phrase. With transitive
verbal complements, as in (34), both notional subject and object appear in oblique
(non-nominative) case forms.
But what of the adverbial clause function? Here, the functional or syntactic rela-
tionships appear to be more like they would be in a finite main clause. The fol-
lowing is a typical example of the participle marking verbs in a clause-chaining
sequence coding event simultaneity in narrative:
10. It is perhaps more properly called a converb suffix since it serves to mark “a non-finite
verb form whose main function is to mark adverbial subordination” (Haspelmath, 1995: 3).
Functional underpinnings of diachrony in relative clause formation
In these cases, the adverbial clauses marked with the participle are otherwise syn-
tactically identical to fully finite clauses in the expression of their arguments. That
is, aside from the marking on their verbs as subordinate clauses, the rest of the
finite properties remain intact. Indeed, we have already seen examples to dem-
onstrate that there do not appear to be clear restrictions on the co-occurrence of
tense-aspect marking with the participle (cf. examples (15), (17), (20), and (21)).
In examples like (35), however, there is a tendency for such marking to occur on
just one verb in the chain – namely, that of the (often final) main clause.
5. Summary
References
1. Introduction
1. The Encuesta Complementaria de Pueblos Indígenas (INDEC 2006) estimates 69,462 Tobas
living in Argentina; this estimate was done following the self-definition criteria either as a
member or as a first generation descendant of this group. The II Censo N acional I ndígena de
Población y Viviendas 2002 (DGEEC 2003) estimates 1,500 Toba-Qom (Guaycuruan) living
in Paraguay. Despite the information given by Gordon and Grimes (2005), who e stimates
that 147 Tobas live in southern Bolivia, data from the 2001 population census (Instituto
Nacional de Estadística 2002), as well as many other sociolinguistic reports (López 2006),
do not mention the existence of a Toba population (i.e. organized as ‘Toba People’) today
in Bolivia.
Clauses as noun modifiers in Toba (Guaycuruan)
maʒe x (with x
demonstratives
as prefixes)
Demonstratives ra x x x
(dem)
na x x
so x x
ʒe x x
ka x
ñi x x
Demonstratives- ra-m x x
topicalizer
na-m x x
(dem-top)
so-m x x x
ʒe-m x x
ka-m x x
ñi-m x
Toba is a polysynthetic language that has a flexible word order (mainly VS, SVO,
and OVS when O is pronominal), a verb/noun opposition, and head marking
in the possessive and argument-predicate relations. It also shows an alienable/
inalienable possession distinction, no case marking nor an adjectival word class. In
this language, gender (masculine-feminine), number (singular-plural), c ollective,
and distributive are encoded on nouns, and deictic features are expressed
through demonstratives (Censabella 2001, 2002; Carpio 2004, 2007a–b; Carpio &
Censabella 2010).
In the verbs, there is an active/middle opposition expressed by dependent
pronominal markers for all persons (Table 2). A split alignment of verbal p erson
marking is observed in the active voice: nominative-accusative for the speech
act participants and tripartite with an additional split in the encoding of S for
the non‑speech act participants (third person). The use of the person markers
that express the S argument is motivated by the semantic aspectual properties of
the verbs. In a clause, to be grammatically acceptable, the P argument, whether
required by an applicative morpheme or not, must be overtly expressed by a
noun phrase. The S/A argument need not be expressed by a noun phrase because
María Belén Carpio & Marisa Censabella
the personal markers on their own guarantee its recognition (Censabella 2006;
Carpio 2007a).
1sg s- ñ-
2sg aw- an-
3sg i-; r-; Ø-; t-; w- n-
1pl s-…q ñ-…q
2pl qaw-…i qan-…i
3pl i-…ʔ; r-…ʔ; Ø-…ʔ; t-…ʔ; n-…ʔ
w-…ʔ
The Toba language lacks verbal tense, and there is an obligatory perfective/
imperfective opposition in all verbs. It has ‘directional’ and ‘locative-movement
oriented’ classes of verbal suffixes. The valence-changing operations are middle,
antipassive, non-promotional passive, causative (four different kinds of deriva-
tional causative constructions), and applicative (Censabella 2008, 2009a, 2010;
González 2008, 2009).2
2. The phonemic inventory of Toba is shown below (Censabella 2002: 53). The bracketed
graphemes are the ones used in this paper. The opposition e/i is neutralized after pre-palatals;
so the morpheme we have studied here could be pronounced as maʒe or maʒi depending on
the speaker’s dialect.
plosive p t tʃ[ch] k q ʔ
voiceless frict. s ʃ [sh] h
voiced frict. ʒ ɣ [g] ʁ [ɢ]
lateral l ʎ
glide w j [y]
tap ɾ [r][d]
nasal m n ñ
front Back
close i o
open e a
Clauses as noun modifiers in Toba (Guaycuruan)
3. Corpus
The corpus is composed of nine narrative texts3 told by eight different male and
female elder adults. They are speakers of the three most mutually-intelligible
dialects spoken in the Chaco province: lañaɢashek, noʔolɢranaq and rapigemleʔk.
Six of the narrative texts have been recorded on magnetophonic tapes at the
speakers’ houses; the other three are published Toba texts transcribed by Toba
teachers. In order to avoid word-by-word translations from Spanish, we do not
study relative clauses and related constructions via elicitation techniques.
Pronouns, proper names or lexical nouns may function as heads of noun phrases.
As lexical nouns refer to types of entities rather than unique referents, they may
require some modifiers to avoid misunderstandings. Thus, as Givón (2001: 1)
points out, noun modifiers are used to further specify or narrow down the domain
of reference of their head nouns, and they show different degrees of syntactic
3. a. Historical Narrative; speaker Domingo López, noʔolɢranaq dialect; age 67; October
1994; length: 16 minutes, aprox. 90 sentences. Magnetophonic recording.
b. Life story; speaker Anastasio Peñaloza, lañaɢashek dialect; age 70; October 1994;
length: 32 minutes, aprox. 214 sentences. Magnetophonic recording.
c. Story of the Anthropophagic Woman; speaker Pablo Rojas, rapigemleʔk dialect; age 59;
November 1992; length 22 minutes, aprox. 175 sentences. Magnetophonic recording.
d. Story of the Rhea Hunter; speaker Pablo Rojas, rapigemleʔk dialect; age 61; October
1994; length 40 minutes, aprox. 288 sentences. Magnetophonic recording.
e. The collapse of La Cangayé; speaker Montiel Romero, lañaɢashek dialect; age 68;
November 1987; length 3 minutes, aprox. 16 sentences. Magnetophonic recording.
f. Pear Story; speaker Ruperta Pérez, rapigemleʔk dialect; age 46; July 2005; length 4
minutes, aprox. 29 sentences. Magnetophonic recording.
g. So Pinyo’olec. Published in Toba and Spanish by the Nate’elpi Nsoquiaxanaxanapi
(Keeping Mothers) from Pampa del Indio in January 2006; possibly lañaɢashek dialect;
length 52 sentences.
h. Nedec ‘Book’ N0 5 (extract), told by Cabito Leiva in Juan José Castelli (Chaco province),
possibly rapigemleʔk dialect; recorded by O. Sánchez in December 1977; length: 50
sentences. Text transcribed by O. Sánchez (2008).
i. Nedec ‘Book’ N0 6 (extract), related by Petoxoi in Miraflores (Chaco province), possibly
rapigemleʔk dialect; recorded by O. Sánchez in December 1977; length: 50 sentences.
Text transcribed by O. Sánchez (2008).
N.B.: sentences include more than a clause. In some texts, one sentence may even include
several clauses linked by coordinators.
María Belén Carpio & Marisa Censabella
We describe the use of the subordinators: maʒe, specific to relative clauses, and
ra [da], so, na, ʒe, ñi, ka (dem-type), and ram, som, nam, ʒem, kam (dem-top
type), which introduce relative clauses and noun complements. Regarding RelCls,
we demonstrate that the subordinators that signal the beginning of RelCls do
not permit the recovery of the syntactic function of the head noun within the
dependent clause; on the contrary, pragmatic factors related to the kind of infor-
mation encoded in the dependent clause condition their use. Nevertheless, the
syntactic function of the head noun is inferred by subtraction of the missing argu-
ment in the RelCl. It is worth remembering that, in main clauses, the P argument,
whether required by an applicative or not, must be overtly expressed by a noun
phrase. But the S/A arguments need not be expressed by a noun phrase because
they are obligatorily indexed on the verb. The non-occurrence as a noun phrase
of the argument co-referent with the head noun is the clue to recover its syntactic
function within the RelCl.
María Belén Carpio & Marisa Censabella
4. In the examples, we underline head nouns and indicate dependent clauses between square
brackets, and the subordinator morpheme in bold face. Abbreviations: 1 - first person; 2 - second
person; 3 - third person; 3u - unspecified third person; all - allative; antipass - antipassive;
appl - applicative; ben - benefactive; coll - collective; compl - complementizer; cont - contin-
uous aspect; coord - coordinator; dl - dual; dem - demonstrative; dem.pron - demonstrative
pronoun; dim - diminutive; dir - directional; dist - distributive; emph - emphatic; ex.pres -
existential presentative; f - feminine gender; loc - locative; m - middle voice; m.adv - modal
adverb; mal - malefactive; masc - masculine; neg - negative; nom - n ominalizer; np.pass - non
promotional passive; obj - object; pres.pron - presentative pronoun; pl - plural; poss - posses-
sive; prog - progressive aspect; sg - singular; t.adv - temporal adverb; t - transitive.
Clauses as noun modifiers in Toba (Guaycuruan)
(2) A-rel
chaʔaʒi woʔo so shiyaɢawa [maʒi yi-ʔaɢat-tak
coord ex.pres dem person rel 3t-say-prog
da woʔo ka i-choɢoden]
compl ex.pres compl 3t-have.pity
‘… because there is a person who is saying that he has pity.’
O. Sánchez; Nedec 5 - Cabito Leiva: 69
(4) Possessor-rel
qataq qoʔoʎaq ñaq qa-y-kaa-tak
coord t.adv t. adv np.pass-3t-chase-prog
na Qom-pi [maʒi l-maʔ neʔena ʔalwa]
dem Toba-coll rel 3poss-home dem.pron land
‘… and, in those days, someone is still chasing the Tobas whose home
is this land.’ O. Sánchez, Nedec 5 - Cabito Leiva: 69
(5) Possessed-rel
qaq nachida da w-aygi somaʒi
coord pres.pron compl 3-be.inside 3sg
sowaɢat neʔena l-ayipi qataq
coord dem-pron 3poss-people coord
na ʔalwa [maʒi l-maʔ
na Qom]
dem land rel 3poss-home dem Tobas
‘And this is (the place) where he was because of his people and the land
which is the home of the Tobas.’ O. Sánchez; Nedec 6 - Petoxoi: 73
As in the case of the maʒe relativizer, the syntactic function of the head noun
within the dependent clause is not recoverable through the dem type subordi-
nator, but no relativized possessors were observed in the corpus and only one
dubious example of a relativized possessed was attested introduced by them. The
use of the dem type subordinators is pragmatically motivated because they occur
when new information relevant to the continuity or further development of the
discourse is encoded in the dependent clause. The syntactic functions of the head
noun within a relative clause introduced by the dem type subordinators, inferred
by subtraction, are: S (6), A (7), P (8), and applied P (9).
(6) S-rel
woʔo na qom [ra t-a-ygi na wataɢanaq]…
ex.pres dem Toba
rel 3-go-appl:loc dem military service
‘There are Tobas who go to the military service…’ DL#77
(7) A-rel
nache a-ñaʔañi yaɢayn-ole nsoɢoy [ra
coord f-dem.pron old.woman-dim Nsogoi rel
i-kitchikchi-gi ʒeʔeʒe norek]…
3t-blow-appl:loc dem.pron fire…
‘And this old woman Nsogoi, who blows inside the fire…’ PR-N#26
(8) P/T-rel
qalaɢaʒe ra-m yataqteʔ y-alamaɢato
coord dem-top m.adv 1poss-belongings
y-enaɢat [ra ayim qo-y-ana-gi]
1poss-name rel 1sg np.pass-3t-give-appl:loc
‘But this is true, my heritage, the name that they have given me.’ AP#84
(9) applied P-rel
qaq nachiso seʔeso yaʔaɢayki-pi so-m qomiʔ
coord pres.pron dem.pron old.man-coll rel-top pers.pron
y-amaɢ-aw-ʔa-lo so gobernador [da se-yaɢan-aɢ-aw-ʔa
3t-send-dir-all-pl dem gobernador rel 1pl-call-1pl-dir-appl:all
neʔena
ʔalwa]
dem.pron tierra
‘And this is it, those old men that send us to the Gobernador from whom we
claim this land.’ O. Sánchez, Nedec 5 - Cabito Leiva: 70
and the spatial relation or visibility between the deictic center and the entity
referred to by the noun:
In the singular, demonstratives agree in gender with the modified noun as shown
in (15a–b), while in the plural, they agree in number with the head noun by the
addition of the suffix -wa, without gender marking (15c):
b. na i-wal r-keʔe
dem 1poss-grandchild 3-eat
‘My grandchild (masc) eats.’
c. na-wa i-waʔaʎi r-keʔe-ʔ
dem-dl 1poss-grandchild.pl 3-eat-pl
‘My grandchildren (fem/masc) eat.’
(19) P-rel
qayka ka ʎaʔ [ka-m qa-y-aɢat]
neg.ex.pres compl another one rel-top np.pass-3t-tell
‘There is not anybody else to be told.’ DL#29
(20) S-rel
so yaʎe-pi [so-m w-aʔaw-chigiñi] qom
dem man-coll rel-top 3-go.first-prog.dir Toba
‘Those men who are from the beginning are Toba.’6DL#1
6. Toba people’s customs are the main topic of this text.
7. The suffix -m with a demonstrative used in a noun phrase composed of two nouns (gener-
ally one alienable and the other inalienable) encodes the relation of possession:
6. Conclusions
singular form, and the dem-top type subordinators also modify highly topical
nouns.
The description of the subordinators in RelCls and noun complements pro-
vided in this paper constitutes a contribution to further typological and gram-
maticalization studies in Toba and other languages of the Gran Chaco region.
References
Carpio, M.B. 2004. La categoría de colectivo en la lengua toba (flia. Guaycurú, Argentina). Actas
del XXIV Encuentro de Geohistoria Regional. Instituto de Investigaciones Geohistóricas
(CONICET). Resistencia. 114–121.
Carpio, M.B. 2007a. Sistemas de alineación en toba (familia guaycurú, Argentina). MA thesis,
Universidad de Sonora, México.
Carpio, M.B. 2007b. Coherencia referencial en toba (flia. guaycurú, Argentina). Seminario de
ComPLejidad Sintáctica. Maestría en Lingüística. División de Humanidades y Bellas Artes.
Departamento de Letras y Lingüística. Universidad de Sonora (México). Hermosillo.
15–16 November, 2007.
Carpio, M.B. 2009. Accesibilidad referencial en toba (flia. guaycurú, Argentina). Revista Estudios
en Ciencias Humanas. Estudios y monografías de los postgrados de la Facultad de Humani-
dades. Resistencia (Chaco, Argentina): Universidad Nacional del Nordeste. Facultad de
Humanidades. 〈http://hum.unne.edu.ar/revistas/postgrado/revista7.htm〉.
Carpio, M.B. & Censabella, M. 2010. Género y número en toba (flia. guaycurú, Argentina).
Estudios de lenguas amerindias 2. Contribuciones al estudio de las lenguas originarias de
América, 57–83. Sonora, México: Editorial UniSon.
Censabella, M. 2001. Sincronía dinámica de los determinantes demostrativos en un corpus nar-
rativo en lengua toba, 1–12. Actas Cuartas Jornadas de Etnolingüística. Rosario: Universi-
dad Nacional de Rosario, Escuela de Antropología.
Censabella, M. 2002. Descripción funcional de un corpus en lengua toba (familia Guaycurú,
Argentina). Sistema fonológico, clases sintácticas y derivación. Aspectos de sincronía
dinámi+ca. Ph.D. dissertation, Universidad Nacional de Córdoba.
Censabella, M. 2006. Relaciones gramaticales en lengua toba (flia. Guaycurú, Argentina). In
Memorias del VIII Encuentro Internacional de Lingüística en el Noroeste, Z. Estrada Fernán-
dez (ed), 81–103. Hermosillo: Universidad de Sonora.
Censabella, M. 2008. Derivación causativa en lengua toba. In Studies in Voice and Transitivity,
Z. Estrada Fernández, S. Wichman, C. Chamoreau & A. Álvarez González (eds), 105–125.
Munich: Lincom.
Censabella, M. 2009a. Aplicativos en toba: Significado, función y procesos de gramaticalización.
53rd Congreso Internacional de Americanistas, México D.F.
Censabella, M. 2009b. Chapter Chaco ampliado and Argentina en el Chaco. In Atlas
sociolingüístico de los pueblos indígenas de América Latina, Tomo I, Inge Sichra (ed. and
comp.), 143–167. Cochabamba: UNICEF/FUNDPROEIB Andes.
Censabella, M. 2010. Beneficiaries and recipients in Toba (Guaycurú). In Benefactives and
Malefactives. Typological Perspectives and Case Studies [Typological Studies in Language
92], F. Zúñiga & S. Kittilä (eds), 185–201. Amsterdam: John Benjamins.
María Belén Carpio & Marisa Censabella
Patience Epps
University of Texas at Austin
* I am very grateful to my Hup hosts, friends, and language teachers; to the Instituto
Socioambiental, FOIRN, and the Museu Goeldi in Brazil; and to Fulbright-Hays, the National
Science Foundation (BCS0111550), and the Max Planck Institute for Evolutionary Anthro-
pology, Leipzig for supporting this research. Many thanks to Orin Gensler for enlightening
discussion of the Hup material, and particularly for the initial insight that the traditional
headed vs. headless distinction does not apply well to Hup. I would also like to thank two
anonymous reviewers and Stephanie Ramos Bierge for their very helpful comments on this
paper. All shortcomings are of course my own. Finally, thanks to Zarina Estrada for the
invitation to participate in this volume and in the ‘Seminarios de Complejidad Sintáctica’
in Hermosillo, Mexico (2006–2007).
Patience Epps
1. Introduction
1. In these definitions and in the choice of the term ‘domain nominal’ to refer to the noun
phrase modified by the relative clause, I follow Andrews (2007) and other typologically based
discussions of relative clauses. I consider only ‘restrictive’ relative clauses here.
2. Examples include the following, from French (i) and Polish (ii) (Citko 2004: 96–97):
and in other respects. In Polish and other European languages, such ‘light’ heads
are typically related to wh-pronouns, and include demonstratives, determiners,
indefinites (e.g. ‘somebody’), and other elements.
Yet, as this paper argues, even a three-category approach may not adequately
characterize the property of a relative clause to appear with or without the noun
whose reference it delimits, and may risk obscuring interesting variation along
this parameter both cross-linguistically and within a given language. In particu-
lar, for languages in which relative clauses are nominalizations and can thus be
understood as occurring in an appositional relationship with the domain nominal
(i.e. head) to form a noun phrase (cf. Andrews 2007: 232; Comrie & Thompson
2007: 338), a relative clause may associate with a range of nominal elements that
are more or less like full nouns in their own right. This is most notably the case
when the domain ‘nominal’ has undergone processes of grammaticalization that
result in a partly lexical, partly grammatical identity.
The following discussion illustrates this claim by considering data from Hup,
a language of the northwest Amazon. In Hup, relative clause constructions can be
classified as headed, headless, or as falling somewhere on a continuum between
the two. In these intermediate cases, the lexical vs. grammatical identity of the
domain nominal corresponds to differing degrees of grammaticalization, from full
noun to ‘bound noun’ to classifier to (arguably) a marker of number agreement.
The Hup data therefore suggest that the ability of a relative clause to appear with a
domain nominal should not be viewed as a dichotomy between headed or head-
less, nor even as involving a third, clearly defined intermediate category of ‘light’
heads. Rather, for languages like Hup, the property of ‘headedness’ in relative
clauses may be best represented as a gradient phenomenon, and this approach is,
arguably, descriptively richer and typologically more accurate than the alternative.
3. Information on Hup (Hupda, Jupde) was obtained via fieldwork on the Rio Tiquié, Brazil,
2000–2004. The family name ‘Nadahup’ is preferred because ‘Makú’ is used to refer to several
unrelated language groups in Amazonia and is thus prone to confusion, and because ‘Makú’
is widely recognized in the Vaupés region as an offensive ethnic slur. ‘Nadahup’ combines
elements of the names of the four established languages that make up the family (Nadëb, Dâw,
Yuhup, Hup).
Patience Epps
4. Nasalization is thus (in general) a property of the entire morpheme in Hup, but it is
indicated orthographically here on the segmental level in the interest of reader-friendliness:
Where at least one consonant or vowel in a syllable is nasal, all other segments (other than
those transparent to nasalization) should be assumed to be nasal even if unmarked. High tone
is indicated by v́, rising tone by v̌.
5. The following abbreviations are used: assoc.pl - associative plural; decl - declarative; dep -
dependent; dim - diminutive; dir - directional; dist - distributive; dst.cntr - distant contrast;
dynm - dynamic; emph - emphasis; fact - factitive; fem - feminine; flr - filler; foc - focus;
ideo - ideophone; inch - inchoative; int - interrogative; interj - interjection; itg - intangible
(demonstrative); msc - masculine; neg - negative; neg:ex - negative existence; obj - object;
obl - oblique; perf - perfective; pl - plural; poss - possessive; prx.cntr - proximate contrast;
rep - reported evidential; resp - respect; sg - singular; tag -interactive tag; tel - telic.
6. Case-marking in Hup falls on the final element of the noun phrase. The fact that the
case suffix attaches to the domain nominal and not to the relative clause as well is evidence
Between headed and headless relative clauses
that relative clause + domain nominal form a single syntactic constituent (whereas headless
relative clauses are directly case-marked; see below).
Patience Epps
In addition to the headed and headless relative clauses outlined above, Hup has at
least three further categories of relative clauses, as defined by the type of domain
nominal with which they associate. As I argue here, these elements have the status
of domain nominals, but are distinct from the full nominal elements described in
§2.1 above, and are likewise distinct from each other, ranging from more to less
noun-like.
7. However, its function in other contexts in Hup is less clearly like that of the -Vp suffix; see
Epps (2008: 688, 834).
Between headed and headless relative clauses
[mɔ̌h g’íg-ip]=ʔĩh
tinamou shoot.arrow-dep=msc
‘So, it’s said, there was a (respected) man standing there listening, [a man
who was out shooting tinamou birds (Tinamus sp.)]’.
For inanimate entities, bound nouns commonly found with relative clauses include
plant-part terms (12), among others (13):
Relative clauses associated with bound nouns as domain nominals are arguably
less fully headed than are those associated with full noun phrases, and can there-
fore not be assimilated to the category of fully headed relative clauses in Hup. This
perspective could perhaps be challenged by the suggestion that the relative clause
construction in Hup is by definition a bound construction, such that even a noun
that can appear independently elsewhere is best considered ‘bound’ when it func-
tions as a domain nominal (compare the discussion of ‘repeater’ classifiers in Hup’s
Tariana and Tukanoan neighbors, among other languages of the northwest Ama-
zon; Aikhenvald 2000: 361, Seifart 2005). Nevertheless, there is no escaping the fact
that the bound noun is a distinct class of domain nominal that lacks a syntactically
independent identity (and has a relatively generic meaning). A bound noun that is
associated with a relative clause cannot therefore be identical to any independent
co-referential nominal constituent in the discourse, since another instance of the
bound noun will necessarily be associated with a distinct nominal element. For
example, the masculine animate bound noun =ʔĩh in (8) above, which attaches
to the relative clause (hɔ̌p̃ kǝ́k-əp=ʔĩh [fish pull-dep=msc] ‘man who was fishing’)
refers anaphorically to the nominal ʔayǔp=ʔĩh [one=msc] ‘a man’ (an alternative
antecedent could be tiyǐʔ ‘man’, but =ʔĩh simply cannot occur by itself). In contrast,
domain nominals that are full noun phrases may be identical to co-referential con-
stituents, as is the case for ‘clothes’ in (2) above.
Relative clauses involving bound nouns also cannot be identified as cases in
which the domain nominal of is null (i.e. dropped, but underlyingly present –
a possibility mentioned in Citko 2004), with which the bound noun associates,
yielding the structure [Relative]-Ø=Bound noun. First, except in certain limited
cases (primarily involving generic reference) Hup classifying nouns cannot appear
without a preceding nominal element; they must always be directly bound to
Between headed and headless relative clauses
something, if only the default third person singular pronoun tɨh (e.g. tɨh=típ ‘egg’).
This suggests that the relative clause itself provides the relevant nominal host for
the bound noun. This argument is supported by the fact that the insertion of a
candidate nominal element between the relative clause and the bound noun is rare
and frequently unacceptable. For example, the default element tɨh is ungrammati-
cal in this position:
(14) *ʔãh dó-op=tɨh=tip
1sg take-dep=3sg=egg
Intended meaning: ‘the egg I took’
Relative clauses taking bound nouns likewise cannot be considered a type of head-
less relative. In the majority of cases, the bound noun is quite clearly functioning
as the head of the noun phrase, rather than as a modifier or inflectional element
(regardless of the type of nominal it associates with – numeral, demonstrative,
relative clause, or nominalized verb root). As such, the bound noun carries the
primary referential load, with the preceding nominal element restricting refer-
ence, and the bound noun encodes features of animacy that determine whether
the noun phrase will receive case-marking if it is functioning as the object of the
main clause (see Epps 2007, 2008: 232ff).
In summary, relative clauses associated with bound nouns in Hup should be
classified as headed, but the status of these heads is ‘lighter’ than is that of full
nominal elements.
‘thing that does V’, ‘thing for doing V’ (e.g. hĭʔ=teg [write=stick/long.thing]
‘pencil’), and they regularly occur together with certain nouns, particularly loans
(e.g. dúc=tat [light (Pt. luz) =fruit/round] ‘lightbulb’; ‘ball’ in (17) below). Like
their b ound-noun counterparts, they are also common with demonstratives,
pronouns, and numerals, and frequently appear with relative clauses. As such, they
commonly refer anaphorically to a full noun phrase in the preceding discourse (as
in 17), but a co-referent noun phrase is not necessarily required (as in 18):
(17) nup bóda=tat-ʔěʔ, [núp d’ɔh-yǽt-æp]=tat
this ball=fruit-perf this rot-lie-dep=fruit
‘This was a ball, [this rotting round thing lying here].’
(18) [ʔin cák-ap]=teg
1pl climb-dep=thing
‘ladder’ (lit. ‘thing we climb up’)
Relative clauses involving classifying nouns are structurally and functionally very
similar to those involving bound nouns, in keeping with the close historical rela-
tionship between these two categories. Relatives like those in examples (17–18)
may, therefore, be considered headed, i.e. having a classifying noun as a domain
nominal, for the same reasons as those given above for bound-noun relatives.
There is also no evidence to support a null head, inflected by a classifier: the clas-
sifying noun typically carries a referential load, there is often no reasonable can-
didate for a null head (as in 18, for example), such candidate nominals do not
normally appear between the relative and the classifying noun, and in many cases
they would in fact yield a questionable or ungrammatical utterance.
However, because classifying nouns are distinct from bound nouns and have
a more grammatical identity, relative clauses with which they associate are argu-
ably not as fully headed as are those that appear with bound nouns. While clas-
sifying nouns in Hup often clearly have a syntactic identity as nouns in their
own right, they may also appear more like dependent morphological elements
associated with noun phrases – that is, their identity ranges from more lexical
(i.e. like the head of a noun phrase) to more grammatical (i.e. like a marker of
agreement), with variation both across particular classifying nouns and across
constructions in which they appear. At their most lexical, classifying nouns are
obligatory heads of agent or instrument nominalizations, contributing crucially
to the meaning of the construction (e.g. b’ǒy=g’æt [study=leaf] ‘school book’; as
a nominalized verb root b’ǒy alone means ‘activity of studying’). At their most
grammatical, classifying nouns are optional elements associated with other
nouns (frequently, but not exclusively, Portuguese loanwords), and which make a
relatively minor semantic contribution; e.g. bóda or bóda=tat [ball=fruit] ‘ball’
(from Portuguese bola, see also examples above). A classifying noun may have a
Patience Epps
relatively grammatical identity of this kind even when it follows a verb root or
phrase, since verbs in Hup may be nominalized simply by the omission of any
verbal inflectional morphology (typically accompanied by the addition of rising
tone); e.g. tegd’uh hɔ̌k̃ (=teg) [tree cut(=thing)] ‘chainsaw’ (note that this bare
verb root form of nominalization normally yields an activity interpretation, as
in the case of ‘studying’ above, but occasionally an agent/instrument interpreta-
tion is also possible). In their semantics, some classifying nouns have been only
marginally extended beyond their bound-noun meanings (e.g. =g’æt leaf > book,
paper), while others have developed more abstract meanings (e.g. =teg ‘tree,
stick’ > ‘shaft’ > ‘thing in general’).
The range of lexical to grammatical features exhibited by Hup classifying
nouns reflects their identity as historically intermediate between lexical and
grammatical elements (see Grinevald 2000 for discussion of this intermediate
nature of classifiers generally).8 While classifying nouns in Hup still retain a
primarily nominal identity, they are not all undergoing exactly the same set of
changes at the same time (as the different degrees of semantic abstraction indi-
cate). If these processes of grammaticalization continue, at least some of Hup’s
classifying nouns may develop a productive agreement function, as is exhibited
by certain classifiers in other Amazonian languages, such as Miraña (Bora, see
Seifart 2005). As such, the status of particular classifiers on Hup relative clauses
may become that of inflectional agreement, rendering the relative clause itself
effectively headless.
In sum, the classifiers in examples like (17–18) may be understood as syntacti-
cally marginal domain nominals, which are co-referential with some other nomi-
nal entity in the preceding discourse or in the pragmatic context, but which have
also developed some features of grammatical elements. Relative clauses associated
with classifiers in Hup thus have features of both headed and headless relative
clauses; they are best understood as intermediate between the two, and are closer
to the headless end of the spectrum than are the relative clauses having bound
nouns as domain nominals.
8. Relevant to this question is whether the relative clause or the classifier (to the extent
that it can be considered a lexical element at all) is in fact the ‘head’ of the noun phrase as a
whole (where ‘head’ refers to the element of the noun phrase that is syntactically and semanti-
cally more central, often assumed to be an obligatory property of all phrasal constituents, as
opposed to ‘head’ as another name for the domain nominal associated with a relative clause;
cf. Fraser et al. 1993: 1–2). The development of the classifier construction in Hup has arguably
involved an initiated historical shift (through reanalysis) of a compounded or bound noun
from head to dependent, and thence to a more grammatical element (see Epps 2007).
Between headed and headless relative clauses
̃
(20) [ʔíp pǎ =mǽh]=d’ǝh hid ʔɔ̃h=yíʔ-íy
father neg:ex=dim=pl 3pl sleep-tel-dynm
‘[The little fatherless ones] fell asleep’.9
Various features of Hup =d’ǝh suggest that its status is as much like that of a clas-
sifying or bound noun (as in fact it was considered by Moore & Franklin 1979) as
it is that of an inflectional morpheme. Like the masculine/feminine class terms in
Hup, =d’ǝh is for the most part reserved for animate entities, and thus contributes
semantic content other than simply number (compare also its ‘collective’ use in
Yuhup [Ospina 2002] and in some Hup contexts). Its nominal-like identity is in
fact clearest in relative clauses, where =d’ǝh acts as the suppletive plural form of
the masculine/animate classifier or bound noun, and at the same time replaces the
Dependent suffix -Vp:
9. The ‘Negative existence’ morpheme pã is considered a predicative particle, akin (though
not identical) to a verb.
Patience Epps
̃
(21) a. hɔ̌p kǝ́ k-ǝp=ʔĩh
fish pull-dep=msc
‘man who is fishing’
b. hɔ̌p kǝ́ ̃ k=d’ǝh
fish pull=pl
‘men/animate entities who are fishing; fishermen’
Where other domain nominals are present, however (whether classifying nouns,
bound nouns, or full noun phrases), =d’ǝh does not take their place and its role is
more clearly that of number marking:
̃
(22) a. hɔ̌p kǝ́ k-ǝp=ʔãý
fish pull-dep=fem
‘woman who is fishing’
b. hɔ̌p kə́ ̃ ́
k-əp=ʔãy=d’ǝh
fish pull-dep=fem=pl
‘women who are fishing’
In examples like (19–20 and 21b) above, in contrast to (22b), Plural =d’ǝh arguably
functions as a kind of domain nominal, heading the relative clause just as do the
classifying nouns discussed above. Its similarity to classifiers may also be seen in
its behavior with numerals, which (with the exception of ‘one’) cannot constitute
noun phrases on their own, but obligatorily co-occur with some other nominal
entity: a full lexical noun, a classifier (as in 16 above), or – if the referents are
animate – the Plural marker =d’ǝh (Example 23):
(23) kaʔǎp=d’ǝh (*kaʔap) ʔín-íh, ́
tãʔãy=d’ǝh
two=pl 1pl-decl woman=pl
‘We are two, (us) women’.
In their structure (i.e. a verb root inflected with a nominal suffix, lacking the
Dependent marker -Vp) these d’ǝh-marked relative clauses bear some resem-
blance to the headless relatives taking nominal case-marking morphology
(see 5–7 above). However, differences between these two clause types indi-
cate that the plural-marked type should not be classified as a headless relative
clause on a par with the case-marked variant. First, case markers may appear
on relative clauses only in combination with the ‘Filler’ suffix -Vw-, which (as
noted above) may effectively act as a marker of subordination.10 Additionally,
10. Compare also the behavior of case markers with numerals, which can appear as nominal
arguments if combined with the Plural form =d’ǝh, and which between the noun and the case
marker only take case-marking if another nominal element intervenes *kaʔǎp-át [two-obl];
kaʔǎp=d’ǝh-ét [two=pl-obl] ‘with two [animate entities]’.
Between headed and headless relative clauses
as noted above, the plural marker on relative clauses (when there is no inter-
vening nominal element) carries information about animacy and gender
(masculine or gender-neutral), as well as number; case markers, on the other
hand, indicate only the nominal status of the relative clause and its grammati-
cal relation within the main clause. Moreover, while objects in Hup normally
inflect for object case only if animate, this does not apply to ‘headless’ relative
clauses, which receive object-marking regardless of animacy; in contrast, P lural
marking on relative clauses (as on other nominals) is heavily dependent on
animacy.
While the status of the Plural marker =d’ǝh with relative clauses is much like
that of a classifying noun in Hup, the lack of the Dependent suffix -Vp (which
comes between the verb and the associated domain nominal in other relative
clause types) raises the question of whether these Plural-marked relatives should
be considered relative clauses at all, or rather a form of gerund. In fact, the clas-
sifier-like status of d’ǝh suggests that the Plural-marked clauses discussed here
could be assimilated to the distinct type of agent/instrument nominalizations
mentioned in §3.2 above, which involve a verb root followed directly by a noun,
usually a classifying noun (e.g. hĭʔ=teg [write=stick/thing] ‘writing thing =
pencil’; compare the relative clause híʔ-ip=teg [write-dep=stick/thing] ‘thing
for writing’). Semantically, the agent/instrument nominalization indicates an
inherent quality of the referent, whereas the relative clause variant refers to a
quality that is more temporary. (As noted above, verbs may also be derived as
nominals – usually relating to an activity – by appearing as bare roots with a
rising tonal pattern; accordingly, the classifying (or other) nominal element in
the agent/instrument nominalization discussed here is not itself required for
nominalization.)
However, there are several indications that the Plural-marked clause is not
simply a subtype of the agent/instrument nominalization, but shares important
features with relative clauses. First, as illustrated in (21) above, the =d’ǝh-marked
clause is effectively the only option when pluralizing the masculine animate bound/
classifying noun =ʔĩh (i.e. the form -Vp =ʔĩh=d’ǝh is not attested). In keeping with
this fact, the =d’ǝh-marked clause carries no information about the temporary vs.
inherent nature of the association, whereas the agent/instrument nominalization
refers only to relatively inherent qualities:
(24) a. mɔ̌y hæ̌p=ʔĩh
house sweep=msc
‘(male) house-sweeper’
b. mɔ̌y hǽp-æp=ʔĩh
house sweep-dep=msc
‘man who is sweeping the house’
Patience Epps
c. mɔ̌y hǽp=d’ǝh
house sweep=pl
‘house-sweepers’, ‘men/people who are sweeping the house’11
Even more importantly, the Plural-marked clause behaves syntactically like a rela-
tive clause, rather than like an agent/instrument nominalization: it does in fact
involve a full clause, and a variety of arguments of the verb may be relativized. The
agent/instrument nominalization, on the other hand, can only involve a (nominal-
ized) verb phrase; it may thus include an object of the verb (as in ‘house-sweeper’
in 24 above), but not a subject, and the agent may appear only externally, as a pos-
sessor (Examples 25a–b; see Epps 2008: 839). The associated classifying noun may
refer only to the agent of the nominalized verb or to an associated instrument or
location, but not to an object.
(25) a. *tɨh g’ét=mɔy
3sg stand=house
Intended meaning: ‘the house where she always stays’ (referring to
someone who rarely goes out)
b. tɨnĭh g’ét=mɔy
3sg.poss stand=house
‘her staying-house’; ‘the house where she always stays’
In Hup relative clauses, on the other hand, a subject nominal that is not
co-referential with the domain nominal typically appears as an argument of the
verb within the relative, and the arguments that may be relativized are not limited
to agents and associated instrument/locations. Crucially, this is true for both the
singular variant (26a) and the plural variant (26b), supporting the analysis of the
Plural-marked clause as a relative.
(26) a. [tɨh g’ét-ep]=mɔy
3sg stand-dep=house
‘the house where she stays/lives’
b. papad-næn-yǽt-ǽy=mah yɨ-d’ǝ̌h-ǝ́h,
moan-come-lie-dynm=rep that.itg-pl-decl
[hid g’ig-póg]=d’ǝh
3pl shoot.arrow-emph1=pl
‘They were lying around moaning, those [whom they’d shot]!’
The preceding discussion has established that the Plural-marked clause in Hup
is best understood as a type of relative clause, and that it has characteristics
11. Whether or not the tonal patterns on verbs inflected with additional segmental material
carry meaningful information requires further investigation. Tonal contours in this environ-
ment are indistinct, and consultants do not appear to attend to them.
Between headed and headless relative clauses
of a headed relative, particularly with respect to the features that Plural =d’ǝh
shares with classifying and bound nouns in this language. However, its status
as headed is nonetheless more marginal in comparison to these other types of
relative clause. As noted above, in many contexts the contribution of =d’ǝh is
primarily one of number-marking: It occurs generally as a marker of number
on noun phrases referring to animates, and likewise on the domain nominals
of relative clauses when these are not the masculine/animate bound noun =ʔĩh,
as illustrated in (22b) above. As a marker of plural number, =d’ǝh even occurs
occasionally (though very rarely) with nouns referring to inanimates, suggesting
that neither animacy nor gender is an essential semantic component. Moreover,
whereas the set of classifying nouns represents a grammaticalized subset of the
bound nouns in Hup (e.g. masculine/animate < ‘man’; shaft/thing < ‘tree, stick’),
Plural =d’ǝh has no current alternative use as a comparably referential nominal
element, on a par with a bound noun. The behavior of =d’ǝh is therefore more like
that of a grammatical element, in comparison to the behavior of the bound and
even the classifying nouns in Hup. Finally, as discussed above, the plural-marked
relative clause is structurally similar to the case-marked headless relative clause,
in that in both cases the presence of nominal morphology makes the presence
of the Dependent marker unnecessary, or at least has altered its form. We may
thus conclude that the plural-marked relative clause in Hup is closer to a headless
relative clause than a headed one, but that it is still best characterized as a distinct
type that falls somewhere between the two.
ominative headless relative clause (see Example 4 above): the verb is inflected
n
with the Dependent marker -Vp, and an argument is obligatorily gapped (in con-
verbs, the null argument is almost always the subject, which is co-referential with
the subject of the associated main clause).
In addition to their identical structures, evidence that the headless relative is the
source of the converbal clause in Hup comes from their near-complementary
distribution: Nominative, singular headless relatives are extremely rare, in contrast
to headless relatives in object or oblique roles, whereas the Dependent-marked
verbal construction is quite common as a converbal clause. Furthermore, occa-
sional ambiguity occurs, by which a converbal and a headless relative interpre-
tation are both possible (compare the similar ambiguity found in the ‘adjoined
relative’ clause in Australian languages; see Hale 1976 and Epps 2009 for further
discussion). An example of this ambiguity can be seen in the first -Vp-marked
clause in (29); the second is unambiguously a converb:
Plural-marked relative clauses (also in nominative case within the main clause)
have experienced a similar diachronic transition from relative to converb, but cru-
cially they have been less fully affected. Only a very few Plural-marked clauses
function clearly as converbs (Example 30), whereas many are ambiguous between
converb and relative (31), and examples functioning unambiguously as plural
relative clauses are not particularly rare (as in 19–20 above).
4. Conclusion
The Hup case suggests that the traditional division of relative clauses into two dis-
crete categories, headed and headless, is overly simplistic and does not do justice
to the complex range of relative clause variants that languages like Hup may have.
Even expanding the two-category approach by adding a third, intermediate set
of ‘light-headed’ relatives appears to be similarly inadequate. The data discussed
Patience Epps
here indicate that the ability of relative clauses to appear with or without a domain
nominal – their ‘headedness’ – may be best understood as a gradient phenome-
non, based on the degree to which the element appearing in the role of the domain
nominal may be understood as a lexical or a grammatical entity. This is likely to be
particularly relevant for languages in which relative clauses are nominalizations,
as in Hup.
Virtually by definition, processes of grammaticalization do not tend to involve
abrupt transitions by which a given linguistic element changes in a single step
from a lexical entity to a grammatical one; rather, a grammaticalizing element
typically goes through intermediate stages in which it has both lexical and gram-
matical features, and may have more of one or the other depending on how far
along it is in the process. Therefore, if an element of this kind associates with a
relative clause – itself a nominalization – it will likely have some of the lexical
characteristics expected of a domain nominal, but also some of the grammatical
characteristics expected of an element of nominal morphology (e.g. a classifier or
marker of collectivity/number) that associates with the nominalization that is the
relative clause. Accordingly, the relative clause construction as a whole will share
features of both headed and headless relatives, resulting in an intermediate iden-
tity. As in Hup, where a language has more than one such relative variant, and/or
different instances of the same construction will vary with respect to the lexical or
grammatical identity of the domain ‘nominal’, this intermediate region between
headed and headless is best viewed as a continuum. This perspective guarantees us
a greater descriptive richness and a more sophisticated typological understanding
of the phenomenon of relative clauses.
References
Aikhenvald, A.Y. 2000. Classifiers: A Typology of Noun Categorization Devices. Oxford: OUP.
Aikhenvald, A.Y. 2003. A Grammar of Tariana. Cambridge: CUP.
Andrews, A. 2007. Relative clauses. In Language Typology and Syntactic Description, Vol. 2, T.
Shopen (ed.), 206–263. Cambridge: CUP.
Citko, B. 2004. On headed, headless, and light-headed relatives. Natural Language and Linguistic
Theory 22: 95–126.
Comrie, B. & Thompson, S.A. 2007. Lexical nominalization. In Language Typology and Syntactic
Description, Vol. 3, T. Shopen (ed.), 334–381. Cambridge: CUP.
Corbett, G.G., Fraser, N.M. & McGlashan, S. (eds). 1993. Heads in Grammatical Theory.
Cambridge: CUP.
Curnow, T.J. 1997. A Grammar of Awa Pit (Cuaiquer): An Indigenous Language of South-western
Colombia. Ph.D. dissertation, Australian National University.
Dryer, M. 2004. Noun phrases without nouns. In Grounding and Headedness in the Noun Phrase,
J.-C. Verstraete (ed.). Special issue of Functions of Language 11(1): 43–76.
Between headed and headless relative clauses
Epps, P. 2007. Birth of a noun classification system: The case of Hup. In Language Endangerment
and Endangered Languages: Linguistic and Anthropological Studies with Special Emphasis on
the Languages and Cultures of the Andean-Amazonian Border Area [Indigenous Languages
of Latin America], Leo Wetzels (ed.), 107–128. Leiden: Research School of Asian, African,
and Amerindian Studies (CNWS).
Epps, P. 2008. A Grammar of Hup [Mouton Grammar Library 43]. Berlin: Mouton de Gruyter.
Epps, P. 2009. Escape from the noun phrase: From relative clause to converb and beyond in an
Amazonian language. Diachronica 26(3): 287–318.
Fraser, N.M., Corbett, G.G. & McGlashan, S. 1993. Introduction. In Corbett et al. 1–10.
Gildea, S. 1998. On Reconstructing Grammar: Comparative Cariban Morphosyntax [Oxford
Studies in Anthropological Linguistics 18]. Oxford: OUP.
Grinevald, C. 2000. A morphosyntactic typology of classifiers. In Systems of Nominal Classifica-
tion, Gunter Senft (ed.), 50–92. Cambridge: CUP.
Hale, K. 1976. The adjoined relative clause in Australia. In Grammatical Categories in Australian
Languages [Linguistic Series 22], R.M.W. Dixon (ed.), 78–105. Canberra: Australian Insti-
tute of Aboriginal Studies.
Haspelmath, M. 1995. The converb as a cross-linguistically valid category. In Converbs in
Crosslinguistic Perspective, M. Haspelmath & E. König (eds), 1–55. Berlin: Mouton de
Gruyter.
Moore, B. & Franklin, G.L. 1979. Análise preliminar de locuções nominais da língua Hupda. In
Breves Notícias da Língua Maku-Hupda [Ensaios Linguísticos 6], B. Moore & G.L. Franklin,
9–31. Brasília: SIL.
Ospina Bozzi, A.M. 2002. Les structures élémentaires du Yuhup Maku, langue de l’Amazonie
Colombienne: Morphologie et syntaxe. Ph.D. dissertation, Université Paris 7 – Denis
Diderot.
Seifart, F. 2005. The Structure and Use of Shape-based Noun Classes in Miraña (North West
Amazon). Ph.D. disserstation, Radboud University Nijmegen.
Zwicky, A. 1985. Heads. Journal of Linguistics 21: 1–29.
Relative clauses in Seri
Stephen A. Marlett
SIL International & University of North Dakota
This article presents the basic facts about relative clauses in Seri, including their
morphology (as nominalizations), syntax (head-internal), and distribution
in discourse (relatively infrequent). The heads of intransitive relatives may
be formally marked as definite when the content of the relative is being
emphasized. Ambiguity is avoided in transitive relatives by the omission of
the definite article on the head. While some kinds of recursion of relative clauses
are possible, others are not. Throughout this presentation, relative clauses are
distinguished from formally identical but pragmatically distinct constructions
in the language.
1. Introduction
The topic of relative clauses (RCs) in the Seri language is both simple and com-
plex.1 Some of the complexity comes from identifying exactly what are RCs in the
language based on general,1cross-linguistically viable definitions. The d
efinition
1. I have done periodic fieldwork on the language for more than thirty years and also
benefited greatly from previous work done by Edward and Mary B. Moser. Our work has
been carried out under the auspices of the Summer Institute of Linguistics. Recent work
has been supported in part by grants from the National Science Foundation (BCS-0110676,
for a dictionary and texts project) and a Documenting Endangered Languages grant from
the National Endowment for the Humanities (FN-50007–06), for a reference grammar
(Marlett, in preparation). Unless otherwise noted, the examples in this paper are from direct
elicitation.
The text examples Topete, Conejo_Cuernos, Conejo_Puma, Gigante_Comelon, and
Hombre_Barril are from Moser and Marlett (2006); they were oral texts that were transcribed
and slightly edited by competent Seri writers. The examples from Cmaam_Ipca_Quiho
and Alo_Quicös, texts written by Lorenzo Herrera Casanova, are part of Montaño Herrera,
Moreno Herrera and Marlett (2007), as is the text Siete_Filos by René Montaño Herrera, and
the Hant_Quixoaa text by Francisco Xavier Moreno Herrera. The Siete_Filos text will also
Stephen A. Marlett
appear as Montaño Herrera (in press). The Xtiip example is from an unpublished text written
by René Montaño Herrera. The Consejos_Perro is from an unpublished exhortative oral text
by Roberto Herrera Marcos, recorded and transcribed by Edward W. Moser sometime shortly
prior to 1968 and included in a concordance project directed by Joseph E. Grimes and executed
at the University of Oklahoma Computer Laboratory under the Project for Computer Support
for Linguistic Research, partially supported by a National Science Foundation Grant GS-1605.
Some of these same data are found also in Moser and Marlett (2005).
I thank René Montaño Herrera, Francisco Xavier Moreno Herrera, and Lorenzo Herrera
Casanova for their kind and invaluable help in the research for this article. I also thank Zarina
Estrada Fernández, Carolyn O’Meara, Rudolph Troike, Cindy Williams and two anonymous
reviewers for very helpful suggestions.
Relative clauses in Seri
The contrast between finite verbs and nominalized verbs is illustrated by the
very simple examples in (1)–(4), respectively, which are explained below in a bit
more detail.2
(1) hapxa quij cö-i-m-aasitim. (finite verb)
cottontail the.cm 3.io-3.subj:3.obj-px-deceive
‘The cottontail rabbit deceived him/her.’
(2) hapxa cö-c-aasitim quij (nominalized verb)
cottontail 3.io-sn+tr-deceive the.cm
‘the cottontail rabbit that deceived him/her’, ‘the cottontail that is deceiving
him/her’, ‘the cottontail that was deceiving him/her’
(3) hapxa quij ih-mii-ho. (finite verb)
cottontail the.cm 1s.subj.tr-px-see
‘I saw the cottontail rabbit.’
(4) hapxa h-oco-ho quij (nominalized verb)
cottontail 1.pos-on-see the.cm
‘the cottontail rabbit that I saw’
2. The italicized representation corresponds to the practical transcription used in the Seri
community, and the use of “c” and “qu” for /k/ follows the same conventions as in Spanish. The
technical transcription (given within diagonals in some places) is a broad transcription using
the conventions of the International Phonetic Association (Handbook 1999). The same symbols
are used for showing the proposed abstract underlying forms, given within curly braces.
Many forms have phonological complexities due to epenthesis, deletion, fusion and ablaut
that make morpheme breaks difficult to show in linear format. Nevertheless, for lack of space
to do more and despite the drawbacks, such morpheme breaks have been given here. A plus
sign (+) joins two morpheme glosses that end up being shown by only one string of phonemes
in the surface form. A diagonal (/) joins two morpheme glosses that are jointly represented
by a single formative in the underlying form (apart from any other morphonological changes
that may also apply).
Abbreviations: ap – adjective phrase, aux – auxiliary, aux.o – auxiliary (for irrealis
non-subject relatives), aux.s – auxiliary (for irrealis subject relatives), aw – away, caus –
causative, cm – compact, dcl – declarative, dem – demonstrative, dn.i – oblique and indirect
object-oriented deverbal noun, dn.s – subject-oriented deverbal noun, dn.o – direct object-
oriented deverbal noun, do – direct object, ds – different subject, dt – distal, fl – flexible,
foc – focus, have – have, hz – horizontal, impf – imperfective, intns – intensifier, io –
indirect object/oblique, ir.dp – dependent irrealis, ir.id – independent irrealis, lc – location,
md – medial, neg – negative, np – noun phrase, obj – object, on – object nominalizer,
pl – plural, pon – proposition/oblique nominalizer, pos – possessor, pro – pronoun, pv –
passive, px – proximal, rl – realis, sf – surface form, sn – subject nominalizer, subj – subject,
tr – transitive, twd – toward, uf – underlying form, us – unspecified subject, vt – vertical,
1 – first person, 1s – first person singular, 2 – second person, 2p – second person plural,
2s – second person singular, 3 – third person.
Stephen A. Marlett
Examples (1) and (3) have verbs in the proximal realis finite form. Other realis
forms could have been used as well, with slightly different meaning, of course,
such as distal ihyooho and emphatic ihxooho. Finite forms include overt mark-
ing of the subject person (as a prefix), or indicate third person by the lack of such
marking except when both subject and direct object are third person, in which
case an overt marker /i-/ appears, as in (1). This prefix /i-/ is distinguished phono-
logically from the epenthetic /i/ that appears in some syllables, as at the beginning
of the verb in (3), since the epenthetic vowel does not appear when the preceding
word ends in a vowel.
Realis deverbal forms are formed with nominalizing prefixes. Irrealis deverbal
forms all use the independent irrealis prefix {si-} and one of them is slightly differ-
ent in structure. The forms are discussed in detail in Section 2 since understanding
the morphology is crucial to understanding the RCs. The construction itself is evi-
dently somewhere intermediate on a scale of deverbalization and nominalization
in the senses described in Malchukov (2004).
For two of the three types of realis deverbal nouns, the “subject” is cross-
referenced by possessor morphology, as illustrated in example (4) where first per-
son possessor on the deverbal noun is parallel to the ‘my’ in ‘my seeing it’.
The structure of the finite verb ihmiiho /iʔmiiʔo/ in example (3) is shown in (5).
(5) Direct Object Subject Mood Root Aspect
Inflection Inflection
UF (null) ʔ- mi- aʔo (null)
1s.subj.tr px see
‘I see/saw him/her/it/them’
The nominalized forms in (2) and (4) have no alternatives unless changed to irrea-
lis or unless negation is added. RCs therefore are much less expressive than finite
clauses since they lack options that are available in finite clauses. The morphological
structures of the deverbal nouns in (2) and (4) are shown in (6) and (7), respectively.
(6) Indirect Object Subject Transitive Root Aspect
Agreement Nominalizer Marker
UF ko- k- i- aasitim (null)
3.io sn tr deceive
cöcaasitim /k wkaasitim/ ‘who deceives/deceived him/her’
(7) Subject as Object Root Aspect
Possessor Nominalizer
UF
ʔi- oko- aʔo (null)
1.pos on see
hocoho /ʔokoʔo/ ‘what I see/saw’
Relative clauses in Seri
Any attempt to put a finite verb form such as cöimaasitim (the finite verb in (1)) or
ihmiiho (the finite verb in (3)) – or any other finite verb form – into a noun phrase
is ungrammatical.3
As is expected with languages that use a nominalization strategy (Andrews
2007: 263), Seri has neither relative pronouns nor complementizers.
The definitions of RC cited above exclude so-called non-restrictive relatives. It
is not at all clear that anything like a non-restrictive relative occurs in Seri.4 Other
data are also excluded by these definitions; these are discussed in Section 4.
The syntactic properties of RCs in Seri are discussed in Section 3. It is shown
that heads may be nouns, personal pronouns, names, and null (3.1). It is claimed
that all of the RCs are head-internal (3.2) and that the head is responsible for cer-
tain properties of the entire nominal (3.3). The head may be formally definite or
indefinite, with some interesting distributional facts that require detailed discus-
sion (3.4). It is possible to have recursion of RCs of a certain type (3.5), including
center-embedding. The RC may have a predicate nominal and no verb (3.6), and it
may have a finite clause that is subordinate to the deverbal noun (3.7).
The distribution of RCs in texts is rather limited, and the frequency of RCs is
quite low (Section 4.2). Subject relatives are much more common than object and
oblique relatives. In part this is true because almost all descriptive predicates are
verbs in Seri rather than adjectives and are presented with the same nominaliza-
tion morphology as other verbs when they modify nouns. In addition, since the
use of a postpositional phrase in a determiner phrase (DP) is generally restricted
3. Occasionally there is homophony of forms that could cause confusion on this point, but
it is clearly true generally. See also the discussion of irrealis forms in section 2 and the discus-
sion in section 3.7 of the inclusion of a finite clause that is subordinate to the nominalized
clause.
. Only one example of what might be interpreted as a non-restrictive relative has been
found, although I have almost certainly missed others. The example I have is found in the
opening sentence of the text; the RC has the same form as the RCs presented in this study.
My understanding is that the phrase tahac iti quiij is not identifying a particular creator out
of many, but rather simply indicating that the Creator person was there at the (unidentified)
location in which this legend takes place.
Stephen A. Marlett
to fixed expressions and names, the necessary way to talk about ‘the people at
Hastoosxöl’, as in Topete’s narrative (Topete 79), is to use the place name modi-
fied by a subject relative that translates more literally as ‘the people who were at
Hastoosxöl’. A DP such as ‘the man in the brown hat’ must be rendered as some-
thing like ‘the man who is wearing the brown hat’. Similarly, locative adverbs do
not occur in DPs, and so a phrase such as ‘that girl over there’ must be rendered
with an RC that includes an appropriate nominalized stative verb with a locative
complement, something like ‘that girl who is standing over there.’ The possessor
of alienably possessed nouns is also indicated with an RC. With plural possessors,
this is completely transparent and regular, as illustrated by hoyaat in (31) below.
Singular forms have become slightly irregular. All these facts increase the number
of RCs and especially subject RCs that one might expect to find in natural dis-
course. The fact that they are nonetheless relatively infrequent is therefore all the
more striking.
2. Morphology
RCs in Seri all utilize deverbal nouns, which come in three types: deverbal noun
oriented toward the subject, regardless of its semantic role (DN.S), deverbal noun
oriented towards the direct object (DN.O), and deverbal noun oriented toward
other nominals, including indirect objects and various kinds of obliques (DN.I).
The choice between different deverbal nouns depends on the grammatical relation
that the head noun has in the RC.5 Thus DN.S is used when the head noun cor-
responds to the subject of the RC, DN.O when it corresponds to the direct object
(whether primary object or secondary object, to use the terminology of Dryer
1986), and DN.I when it is none of the above. So-called possessor relatives use the
form that is appropriate for the possessed noun in the RC. These deverbal nouns
(especially the DN.I forms) also have other uses in the language than in RCs;6
therefore these deverbal nouns are not referred to as relativized forms.
5. By way of comparison, Diegueño, a Yuman language spoken in Baja California Norte
and southern California, has the simple nominalizer /kw/ for subject relatives based on non-
future clauses (Langdon 1970: 171, 176–177; see also Dryer 2008). Seri has been claimed in
the past to have a genetic relationship with Yuman languages (see Campbell 1997: 290–296
for an overview of the proposals). Some Uto-Aztecan languages, some of which have been in
contact with Seri, such as Yaqui, also have nominalized RCs formed by suffixation (Langacker
1977: 179–181).
6. One very common use is as the main predicate when used with a modal. Example (16) in
this paper illustrates this use, which is not limited to stative predicates by any means.
Relative clauses in Seri
The morphology of the forms is interesting in its own right and is important
since it provides the clues for the interpretation of the examples and for prevent-
ing ambiguity as to who is doing what in most cases. These facts are only briefly
reviewed in this section, however; see Marlett (1981, in preparation) for more
extensive treatment and Table 1 for a summary.
Realis forms are different morphologically from irrealis forms, not surpris-
ingly; and realis forms are much more complicated. Realis forms essentially have
non-future interpretations and irrealis forms have future interpretations. All
irrealis forms utilize the same prefix that appears on independent (main clause)
finite irrealis verbs, but unlike the realis deverbal nouns, the irrealis forms occur
with what has been called an auxiliary verb, and they have other complications
in their morphology that are briefly discussed below. The auxiliary forms that
occur with irrealis verb forms in this construction are ca (glossed AUX.S, since
it goes with subject-oriented irrealis clauses and begins with a /k/ as do the basic
subject-oriented realis deverbal nouns) and ha (glossed AUX.O, since it goes with
object-oriented and other irrealis clauses).7
7. This auxiliary is consistently written as ha in this article even though the glottal stop is
very lenis and sometimes omitted. The AUX that appears with independent irrealis verbs in
main clauses is undoubtedly related etymologically.
Stephen A. Marlett
The corresponding irrealis forms are given in (9), all of which occur with the
auxiliary verb ca.
(9) ‘drink’ DN.S (irrealis)
a. smasi ca /smasi ka/ {si-m-asi ka}
‘who will not drink it’ ir.id-neg-drink aux.s
b. sompasi ca /sompasi ka/ {si-m-p-asi ka}
‘that/what will not be drunk’ ir.id-neg-pv-drink aux.s
c. spasi ca /spasi ka/ {si-p-asi ka}
‘that/what will be drunk’ ir.id-pv-drink aux.s
d. siisi ca /siisi ka/ {si-asi ka}
‘who will drink it’ ir.id-drink aux.s
See example (2) for a simple illustration of a DN.S form modifying a simple noun.
The DN.S form is used in example (10), as expected, because the possessed
noun iiquet is the subject of the RC. (This possessed noun is itself a deverbal noun
in a type of expression discussed in Section 2.2 meaning ‘whom she was pregnant
with’, a common expression for the child of a woman.) The deverbal noun cmiih
(part of the euphemistic idiom meaning ‘die’) is underscored and the head noun
iiquet is in boldface in this example. (I use this convention generally from this
point forward.)
The corresponding irrealis forms are given in (13)–(14). Notice that these all occur
with the auxiliary verb ha that was mentioned above. The irrealis verb forms in
this context are inflected for the person of the subject in the same way that finite
verbs are inflected; they do not use the possessor prefixes. Therefore these forms
alone among all of those used in RCs might in fact be considered finite.
8. Plural stem formation in this case (like in some others) involves some change to the root
itself. The change is often the truncation of the root (deleting the vowel of the final syllable).
In the case of the root {asi}, the change is to change /i/ to /j/. The common suffix {-tox} that
appears on verbs with plural subjects loses its initial consonant in its use with this verb. Not
many of these alternations are very regular.
Stephen A. Marlett
The DN.O form is used in example (15), as expected, because the noun phrase
iiquet ‘her child’ is the direct object of the RC.
In example (16) the head noun canoaa is not the primary (or, direct) object but
rather the secondary object (analyzed as a chômeur in Marlett 1981 & Marlett 1984
within the framework of Relational Grammar (Perlmutter 1983)). Therefore the
verb in the RC is in the DN.O form.
(17) DN.I (realis) (all with various translations relating to circumstances and the
proposition itself, including ‘with which …’, ‘where …’, ‘how …’, and ‘the fact
that …’; plural forms could also be listed)
9. The DN.I form has another major use in the language. It is the common form used in
complement clauses to express the proposition itself.
Relative clauses in Seri
The corresponding irrealis forms are given in (18), all of which occur with the
auxiliary verb ha. The irrealis verb forms in this context are inflected for possessor
to indicate the person of the subject.
(18) DN.I (irrealis) (all with various translations relating to circumstances and
the future proposition itself, including ‘with which …’, ‘where …’, ‘how …’,
and ‘the fact that …’; plural forms could also be listed)
a. hisiisi ha /ʔisiisi ʔa/ {ʔi-si-asi ʔa} 1.pos-ir.id-drink aux.o
b. hiscmasi ha /ʔiskmasi ʔa/ {ʔi-si-m-asi ʔa} 1.pos-ir.id-neg-drink
aux.o
c. ispasi ha /ispasi ʔa/ {i-si-p-asi ʔa} 1.pos-ir.id-pv-drink
aux.o
d. iscompasi ha /iskompasi ʔa/ {i-si-m-p-asi ʔa} 1.pos-ir.id-neg-pv-
drink aux.o
Two simple examples of realis DN.I forms from a text are given in (19) and (20).
Example (20) has a null head.
3. Syntax
In this section several syntactic facts about the RC constructions are presented.
The head may also be more complex. I presume that example (22) illustrates recur-
sion, or ‘stacking’, as it has been called,10 with haaco ‘house’ as the direct head only
of the RC caacoj ‘that is big’. The RC caziim ‘that is beautiful’ modifies the nominal
haaco caacoj ‘big house’, and the RC cmaa quiih ‘that is new’ modifies the nominal
haaco caacoj caziim ‘beautiful big house’.
(22) [[[[haaco c-aacoj] c-aziim] cmaa qu-iih] zo]
house sn-big sn-pleasant now sn-be.located.fl a
‘a new beautiful big house’
The noun haaco ‘house’ is, however, the ultimate head of all three RCs. The recur-
sion has nested a subject RC inside of a subject RC inside of another subject RC.
In (23) recursion embeds a subject RC (siimet caaitic ‘bread that was soft’)
inside an object RC (zixcám com... oohit ‘... that the fish ate’). The noun siimet
‘bread’ is ultimately the head of both RCs.
(23) [[Zixcám com [siimet c-aaitic] oo-hit] quih] m-ooxp.
fish the.hz bread sn-soft 3.pos+on-eat the.fl px-white
‘The bread that was soft that the fish ate was white.’11
10. “Relative clauses are said to be stacked if a structure exists such that the first clause modi-
fies the head noun, the second modifies the head noun as already modified by the first clause,
the third modifies the head noun as already modified by the first clause as in turn modified by
the second clause, and so on” (Stockwell, Schachter & Partee 1973: 442).
11. The acceptability of this example might be affected by the fact that siimet caaitic is a
conventionalized term for white store-bought packaged type.
Relative clauses in Seri
Example (28) contains another RC with a null head. It is somewhat more compli-
cated since it includes the DP that means ‘rain seer’ (a person with supernatural
power related to rain). That phrase also has an RC, but it is not relevant at this
point.
(28) [[Ø Ziix cmiique caii cmaam ipca cö-c-oos quij
thing person mature woman rain 3.io-sn-sing the.cm
cö-c-azcam] tamocat] ox t-ooza, yo-qu-e.
3.io-sn-arrive/pl md.aw.pl thus rl-say/pl dt-us-say
‘That’s what those who went to the “rain seer” woman said.’
(Cmaam_Ipca_Quiho 06)
Some examples with null heads may not be RCs by the definitions and character-
izations presented in Section 1; see the discussion in Section 4.1.
12. Keenan’s (1985: 161–163) brief introduction to “internal RCs”, as he called them, included
a few observations about the “relatively few examples” that were cited. Like the languages in
his sampling, Seri has a basic word order SOV. Unlike the languages he cites, however, there
are no prenominal RCs in Seri. Whereas it is also true in Seri that the head is not “distinctively
marked,” the high degree of ambiguity that some other languages seem to exhibit does not
occur because of the specific morphology that each type of RC requires in Seri.
Stephen A. Marlett
of the order Head followed by Deverbal Noun. The claim here, however, is that
they have the structure shown informally in (29a) rather than that shown in (29b):
(29) a. [[Nominal Deverbal]Clause Determiner]
b. [Nominal [Deverbal]Clause Determiner]
More formally, the syntactic analysis is claimed to be something like that proposed
in Williamson (1987) for Lakhota (Sioux) and in Basilico (1996), using the DP
Hypothesis (Abney 1987).13 This analysis is sketched in Figure 1, which shows that
the head of the RC does not appear externally to the RC and in fact may occur with
its determiner. Details are discussed below.
Example (30) is an example with stacked relatives and overt determiners
inside the RC. The constituents in this example have been explicitly labeled.
(30) …[[xepe quih ]DP c-xatlc ]S quih]DP c-meque]S com]DP …
sea the.fl sn-thin the.fl sn-warm the.hz
‘… the shallow and warm sea…’ (Alo_Quicös 3)
Example (31) has a transitive verb quexl with an overt direct object (canoaa hoyaat
quih ‘our boat’ – itself an RC, but this is irrelevant here); this direct object DP
occurs between the head (subject) and the deverbal noun. This order is what is
expected for a language with basic SOV word order.
(31) [[Ctam [canoaa h-o-yaa-t quih] qu-exl] quih]
Man boat 1.pos-on-own-pl the.fl sn+tr-buy the.fl
háqui t-iih?
where? rl-be.located.fl
‘Where is the man who bought our boat?’
Example (32) has a deverbal noun based on an intransitive verb with an overt DP
as complement of the relational preverb iti, which also occurs before the deverbal
noun, as expected.
(32) [[ziix.quiisax14 xepe quih i-teel com i-ti
person sea the.fl 3.pos-edge the.hz 3.pos-on
c-aap] cop
sn-stand the.vt
‘…the person standing on the seashore …’ (Siete_Filos 35)
13. More elaborate analyses are proposed in different theoretical frameworks. I do not
review all of the proposals in the literature here. See Cinque (2009) for one recent interesting
alternative analysis.
14. This is a lexicalized expression of the type described in (65) in section 4.1.
Relative clauses in Seri
DPi
(Relative Clause) S D
DPi
NP D
(Head) N
The head noun is not always exactly where the basic linearization principles would
place it, however. RCs show some variability in word order, just as simple clauses
do. In example (33a) the head noun is in the direct object position, as expected
by the basic linearization principles of Seri, but in (33b) it is in clause-initial
position.15
(33) a. [[María quih cafee oo-si] cop] c-matj iha.
m. the.fl coffee 3.pos+on-drink the.vt sn-hot dcl
b. [[Cafee María quih oosi] cop] cmatj iha.
‘The coffee that María is drinking/drank is/was hot.’
Examples (34a-b) show similar variability in the order of a primary object (Juan)
and secondary object (canoaa ‘boat’) (or direct object and chômeur, respectively,
as analyzed in Marlett 1981), when the relativized noun is the secondary object.
This variability is not possible, apparently, for unknown reasons, when the relativ-
ized object is the primary object, as shown in (35a-b).
(34) a. [[Canoaa Juan quih ma iiy-e] com] qu-iha ha.
boat J. the.fl 2s.do-3.pos+on-give the.hz sn-fast dcl
b. [[Juan quih canoaa ma iiye] com] quiha ha.
‘The boat that John gave you is fast.’
15. Examples such as (33b) are the strongest contenders for externally headed RCs, but there
is no evidence that they are in fact such.
Stephen A. Marlett
honetically only [k].16 It occurs after the heads of intransitive RCs when one
p
wants to emphasize the characteristics or descriptions provided. A very simple
example is given in (39); others can be seen above in (30) and (38). The article
under discussion is the one that follows canoaa in (39) and the one following
caaytaj in (38).
(39) [[[ [Canoaa]NP quih]DP [qu-isil]VP]S com,]DP tiix
boat the.fl sn-small the.hz dem.dt
ih-s-exl a-ha.
1s.subj.tr-ir.id-buy aux-dcl
‘I will buy the small boat.’
The analysis of the complete nominal in example (39), shown in Figure 2, falls out
quite directly from the structure of internally headed RCs that has been presented.
DP
S D
DP VP
NP D VNom
Examples (40)–(42) also have stacked relatives and multiple instances of the post-
head definite article, with (40) also including a predicate noun RC of the type
discussed in Section 3.6.
(40) [[[[[ Xazlc]NP quih]DP [ctam-cö]NP quih]DP [c-aacöl]VP ]S
puma+pl the.fl male-pl the.fl sn-big+pl
quih]DP h-t-aco-tim, ...
the.fl 1s.subj.tr-rl-kill-impf
‘I have killed big, male pumas, …’ (Conejo_Puma 09.1)
16. This phonetic reduction perplexes novice Seri writers since they hear the [k] that they
are pronouncing but are uncertain what to do about it.
Stephen A. Marlett
The examples given in (43), which are ungrammatical variations on (42), illustrate
an important fact for all such cases. A robust determiner, whether a demonstrative
or an article other than quih, is not permitted internally to the DP in the positions
occupied by quih, although quih may be the final determiner under appropriate
circumstances. (The choice of final determiner is based on the shape or orien-
tation or noun class of the head noun; see Marlett & Moser 1994 & Marlett, in
preparation.)
(43) a. *Hap coi caacöl coi quiipe tacoi …
b. *Hap tacoi caacöl quih quiipe quih …
c. *Hap quih caacöl coi quiipe tacoi …
d. *Hap quih caacöl tacoi quiipe quih …
The examples above in this section have intransitive verbs in the RCs. The situa-
tion with transitive clauses is more complicated and more interesting. When there
are two nominals present (even if one is null), a definite head noun obligatorily
does not have an article with it. This fact unambiguously signals it as the head
of the RC despite variability in word order, which is completely acceptable. See
examples (44)–(47).
(44) a. haxz ctam quih oco-ho cop
dog man the.fl 3.pos+on-see the.vt
b. ctam quih haxz ocoho cop
c. *haxz quih ctam quih ocoho cop
d. *ctam quih haxz quih ocoho cop
‘the dog that the man saw’
(45) a. Ø haxz ocoho cop
b. *Ø haxz quih ocoho cop
‘the dog that s/he/it saw’
Relative clauses in Seri
17. /kwiiʔka/ {ko-i-Ø-aʔka} 3.io-3.pos-pon-be.located. The syntax of the idiom is not clear
to me.
Stephen A. Marlett
DP
S D
DP NP
NP N
This analysis accounts for the semantics and the syntax of these examples in a
straightforward way. Example (55) illustrates that the order superficially looks like
the head noun is preceding its “modifier”, which is also a noun. For a language
that is otherwise strictly head-final, this would be anomalous. The RC analysis
accounts for the word order directly by claiming that the noun ctam is actually a
predicate noun and thus appropriately in final position for this SOV language that
regularly does not use copular verbs in such constructions. As a predicate noun,
its semantic relationship to the head is therefore clear as well. This is especially
helpful in an example such as (56) where we have a construction consisting of a
pronoun and noun, which is given semantic interpretation without any additional
machinery. Pronouns are not usually thought of as having “modifiers”, but they
regularly occur as subjects of sentences with predicate nouns.
In example (60) the subject of the embedded clause is different from the head
noun of the RC; it is mentioned explicitly in dislocated, sentence-final position.
Relative clauses in Seri
The topic of the paragraph is the ghost shrimp (hant quixoaa). The verb pootax
(‘it goes’) is in the irrealis mood for obvious reasons, and the presence of different
subject marking (ta) is important since the one walking and the one seeing are two
different entities.
(60) Ziix.quiisax [poo-tax ta qu-i-ho]S zo haquix
person ir.dp-go ds sn-tr-see a somewhere
i-m-iih iha, hant qu-ixoaa quij.
sn-neg-be.located.fl dcl land sn-plan.to.figh the.cm
‘No one has ever seen a ghost shrimp walking.’ (More literally, ‘The person
who has seen the ghost shrimp walking doesn’t exist.’) (Hant_Quixoaa 03)
Such examples illustrate that the only part of the RC that has special properties
is the verb since it is in a deverbal noun form. Such clauses may have embedded
subordinate clauses like any finite clause.
In this section I discuss the frequency of RCs in Seri discourse since this relates to
a general picture of information flow in the language. In order to prepare for the
small statistical study that is presented in this section, however, it was necessary to
first identify the RCs in the texts that were used. The difficulty with doing this in a
simple, mechanical way is that various other constructions occur in the language
use the same denominal verb forms described in Section 2. If these RC-looking
constructions were not distinguished from the RCs that meet the definitions and
characterizations presented in Section 1, the statistics would look very different,
and a distorted picture of the facts would emerge. These other RC-looking con-
structions do not have the same conditions on their distribution as RCs. They
should be counted quite differently in any careful presentation of information
flow. For one simple example, consider the words in (61).
(61) zixcám c-aacoj
fish sn-big
‘giant sea bass’ (Stereolepsis gigas) (Moser & Marlett 2005: 634)
This pair of words looks just like the internally headed RC meaning fish that is
big, which is how it could be translated in some particular context. However, this
phrase is also lexicalized to mean specifically the giant sea bass. The conditions on
its use in discourse are no different than those of a simple noun such as caanj ‘Gulf
grouper’ (Mycteroperca jordani). Those conditions are quite different from the dis-
course conditions governing the presentation of a noun phrase meaning fish that
Stephen A. Marlett
is big. In Section 4.1, I discuss four sets of data that need to be excluded from a sta-
tistical study of RCs as defined in Section 1. This list is not exhaustive. For exam-
ple, as mentioned at the very end of Section 1, possessors of alienably possessed
nouns are expressed by RCs. While these are true RCs, their occurrences might in
some texts inflate the number of occurrences of RCs in uninteresting ways.
A great many RCs have become lexicalized with their head nouns, resulting in
phrasal compounds that refer to very specific items or to classes of items; see
examples (64)–(65) as well as (61) above.
(64) hehe qu-iinla
plant sn-ring
‘desert senna’ (Senna covesii) (Moser & Marlett 2005: 376)
Another common type of RC-looking structure has no overt head noun; I present
them as having null heads below. The examples are pragmatically quite different
from RCs because the nominalized clauses are giving all of the pertinent informa-
tion about the identification of the referent, not just some additional information,
since the head is null. Huddleston, Pullum, and Peterson (2002: 1035–1036) refer
to this type as the fused relative construction, which they treat separately from the
19. The morphologically explicit plural form of the noun hapxa, which is hapxalc (Moser &
Marlett 2005: 342), is not commonly used.
Relative clauses in Seri
integrated relative clause in English.20 In many cases, if one were not aware of the
morphological makeup of the words and the related finite verb forms, one would
think they were simple nouns. See the examples in (66)–(69).
(66) [Ø icozim] < {i-Ø-koʃim}
‘summer’, ‘(the time) when it is hot’ 3p-pon-be.hot.weather
Many fixed expressions now used as names (Marlett 2008a) obviously originated
as nominals with RCs, but they are also lexicalized and therefore pragmatically
distinct from the types of structure discussed as RCs above. Such names are illus-
trated in (70)–(71).
(70) Cmaam C-oos-tim
woman sn-sing-impf
‘Singing Woman’ (Topete 042)
20. They point out that in the fused relatives in English, “it is not possible to separately
identify antecedent and relative clause” (p. 1036). They thus disagree with earlier work on
such constructions, sometimes referred to in the literature as ‘free relatives’ (see McCawley
(1988: 431–432) and Radford (2007: 233)); see also the discussion of headless relatives in
Givón (2001: 205). Two of the many examples in English discussed in Huddleston, Pullum
and Peterson are what he did and whoever devised this plan. Since this relative construction in
English “is so different from the integrated, supplementary, and cleft relative clause construc-
tions” (loc. cit.), Huddleston, Pullum and Peterson treat it separately from those constructions.
21. A singular form is used in this place name instead of the plural form xaaslca, probably
because the cardon cactuses are in a group.
Stephen A. Marlett
Table 3. Ratio of RCs to clauses and RC-modified DPs to other DPs
Short title # of RCs Ratio of RCs to Ratio of RC-modified
other clauses DPs to other DPs
None of the RCs in these texts used an object relative form. The great major-
ity were subject relatives (DN.S), but DN.I forms are also found in one text. Fur-
thermore, RCs based on intransitive clauses greatly outnumbered those based on
transitive clauses, as shown in Table 4. (Examples with passives of transitive verbs
such as those translated as ‘who was named’ were counted as intransitive clauses.)
At this point we do not know how these statistics might compare with European
22. The count of DPs/NPs requires decisions that another person might make differently.
While most of these nominals are DPs, a few are NPs (without a determiner).
Relative clauses in Seri
languages (see Roland, Dick & Elman 2007 for some relevant facts, however, for a
much larger corpus in English), nor with other languages spoken in and around
the Seri area.
It remains to be studied how RCs are actually distributed in Seri discourse.
5. Conclusion
In this article a general picture of RCs in Seri has been presented. These RCs are
similar to those of many languages of the world in that they are nominalized. The
morphology is not simple, but rather very explicit regarding the grammatical
function of the head noun in the RC, except for obliques, where some ambiguity is
always potentially present. It has been argued that RCs are always head-internal in
Seri, which seems to be a less common situation cross-linguistically. The head of
an RC is not always formally indefinite, pace earlier cross-linguistic claims about
internally headed RCs. Nevertheless, the head of an RC typically occurs without
a determiner, obligatorily so when the RC is transitive and variably so when it is
intransitive. A certain definite article occurs after the head of an intransitive RC
when particular emphasis is given to the content of the RC. Some recursion is
possible, especially when the RCs are ultimately modifying the same head. How-
ever, some kinds of center-embedding are not possible. Future work is needed to
understand the distribution of RCs in texts since it was shown that they are used
relatively infrequently.
References
Abney, S.P. 1987. The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, MIT.
Andrews, A. 2007. Relative clauses. In Language Typology and Syntactic Description, Vol. 2,
T. Shopen (ed.), 206–236. Cambridge: CUP.
Basilico, D. 1996. Head position and internally headed relative clauses. Language 72(3): 498–532.
Campbell, L. 1997. American Indian Languages: The Historical Linguistics of Native America.
Oxford: OUP.
Stephen A. Marlett
Cinque, G. 2009. The pronominal origin of relative clauses. Presentation at Ealing VII, Paris.
〈http://ealing.cognition.ens.fr/ealing2009/handouts/Cinque.RCs.handout.pdf〉 (14
September 2010).
Comrie, B. 1989. Language Universals and Linguistic Typology: Syntax and Morphology, 2nd edn.
Chicago IL: University of Chicago Press.
Comrie, B & Kuteva, T. 2008. Relativization on subjects. In The World Atlas of Language Struc-
tures, Ch. 122, M. Haspelmath, M.S. Dryer, D. Gil & B. Comrie (eds). Munich: Max Planck
Digital Library. 〈http://wals.info/feature/122〉 (14 September 2010).
de Roeck, A., Johnson, R., King, M., Rosner, M., Sampson, G. & Varile, N. 1982. A myth about
center-embedding. Lingua 58: 327–340.
Dryer, M.S. 1986. Primary objects, secondary objects, and antidative. Language 62(4): 808–845.
Dryer, M.S. 2008. Order of relative clause and noun. In The World Atlas of Language Structures,
Ch. 90, M. Haspelmath, M.S. Dryer, D. Gil & B. Comrie (eds). Munich: Max Planck Digital
Library 90. 〈http://wals.info/feature/90〉 (14 September 2010).
Enrico, J. 2003. Haida Syntax. Lincoln NB: University of Nebraska.
Givón, T. 2001. Syntax: An Introduction, Vol. 2. Amsterdam: John Benjamins.
Hale, K.L. & Craig, C. 1988. Relational preverbs in some languages of the Americas. Language
64 (2): 312–344.
Handbook of the International Phonetic Association: A Guide to the Use of the International
Phonetic Alphabet. 1999. Cambridge: CUP.
Huddleston, R., Pullum, G.K. & Peterson, P. 2002. Relative constructions and unbounded
dependencies. In The Cambridge grammar of the English language, R. Huddleston &
G.K. Pullum (eds), 1031–1096. Cambridge: CUP.
Hudson, R. 1996. The difficulty of (so-called) self-embedding structures. UCL Working Papers
in Linguistics 8: 1–33.
Jacobsen Jr., W.H. 1998. Headless relative clauses in Washo. In Studies in American Indian
Languages: Description and Theory, L. Hinton & P. Munro (eds), 102–116. Berkeley CA:
University of California Press.
Kaufman, T. 1989. A research program for reconstructing proto-Hokan: First gropings. In
Papers from the 1988 Hokan-Penutian Languages Workshop [University of Oregon Papers
in Linguistics, Publications of the Center for Amerindian Linguistics and Ethnography 1],
S. Delancey (ed.), 50–168. Eugene OR: University of Oregon.
Keenan, E. L. 1985. Relative clauses. In Language Typology and Syntactic Description: Complex
Constructions, Vol. 2, T. Shopen (ed.), 141–170. Cambridge: CUP.
Kuno, S. 1974. The position of relative clauses and conjunctions. Linguistic Inquiry 5: 117–136.
Langacker, R.W. 1977. Studies in Uto-Aztecan Grammar. Dallas TX: Summer Institute of
Linguistics and University of Texas at Arlington.
Langdon, M. 1970. A Grammar of Diegueño: The Mesa Grande Dialect. Berkeley CA: University
of California Publications in Linguistics.
Langdon, M. 1977. Syntactic change and SOV structure: The Yuman case. In Mechanisms of
Syntactic Change, C.N. Li (ed.), 255–90. Austin TX: University of Texas Press.
Malchukov, A.L. 2004. Nominalization/ verbalization: Constraining a typology of transcatego-
rial operations. Munich: Lincom.
Marlett, S.A. 1981. The Structure of Seri. Ph.D. dissertation, University of California at San
Diego.
Marlett, S.A. 1984. Personal and impersonal passives in Seri. In Studies in Relational Grammar
2, D.M. Perlmutter & C. Rosen (eds), 217–239. Chicago IL: University of Chicago Press.
Relative clauses in Seri
Marlett, S.A. 2005. A typological overview of the Seri language. Linguistic Discovery 3: 54–73.
Marlett, S.A. 2007. La relación entre lenguas “hokanas” en México: ¿Cuál es la evidencia? In
Memorias del III Coloquio Internacional de Lingüística, M. Swadesh, C. Buenrostro et al.
(eds), 165–192. Mexico City: Universidad Nacional Autónoma de México and Instituto
Nacional de Lenguas Indígenas.
Marlett, S. A. 2008a. The form and use of names in Seri. International Journal of American Lin-
guistics 74 (1): 47–82.
Marlett, S.A. 2008b. The Seri-Salinan connection revisited. International Journal of American
Linguistics 74 (3): 393–399.
Marlett, S.A. In preparation. Seri grammar.
Marlett, S.A. & Moser, Mary B. 1994. El desarrollo de clases nominales en seri. In Estudios de
Lingüística y Sociolingüística, G. López Cruz & J.L. Moctezuma Zamarrón (eds), 97–103.
Hermosillo: Universidad de Sonora and Instituto Nacional de Antropología e Historia.
McCawley, J.D. 1988. The Syntactic Phenomena of English, Vol. 2. Chicago IL: University of
Chicago Press.
Miller, G. & Chomsky, N. 1963. Finitary models of language users. In Handbook of Mathemati-
cal Psychology, Vol. 2., R.D. Luce, R.R. Bush & E. Galanter (eds), 419–493. New York NY:
Wiley.
Montaño Herrera, R. In press. Leatherback sea turtle – Xiica Cmotómanoj. In Inside Dazzling
Mountains: Contemporary Translations of Southwest Native Verbal Arts, D. Kozak (ed.).
Lincoln NB: University of Nebraska Press.
Montaño Herrera, R., Moreno Herrera, F.X. & Marlett, S.A. (eds). 2007. Comcaac quih ziix quih
ocoaaj hac. 〈http://lengamer. org/admin/language_folders/seri/user_uploaded_files/links/
File/Enciclopedia/EncicloSeri.htm〉 (14 September 2010).
Moser, M.B. & Marlett, S.A. (compilers). 2005. Comcáac quih Yaza quih Hant Ihíip hac: Dic-
cionario Seri-Español-Inglés. Hermosillo & Mexico City: Universidad de Sonora & Plaza y
Valdés Editores.
Moser, M.B. & Marlett, S.A. (compilers). 2006. Seri texts; Textos seris. 〈http://lengamer.org/
admin/language_folders/seri/user_uploaded_files/links/File/Textos/SeriTexts.htm〉 (14
September 2010).
Perlmutter, D.M. (ed.). 1983. Studies in Relational Grammar 1. Chicago IL: University of Chicago
Press.
Radford, A. 2004. Minimalist Syntax: Exploring the Structure of English. Cambridge: CUP.
Roland, D., Dick, F. & Elman, J.L. 2007. Frequency of basic English grammatical structures:
A corpus analysis. Journal of Memory and Language 57: 348–379.
Stockwell, R.P., Schachter, P. & Partee, B.H. 1973. The Major Syntactic Structures of English. New
York NY: Holt, Rinehart and Winston.
Williamson, J. 1987. An indefiniteness restriction for relative clauses in Lakhota. In The Repre-
sentation of (In)definiteness, E.J. Reuland & A.G.B. ter Meulen (eds), 168–190. Cambridge
MA: The MIT Press.
Relative clauses in Gavião of Rondônia
Denny Moore
Museu Paraense Emílio Goeldi-MCTI
1. Introduction
The Mondé branch of the Tupi language family contains three languages: Surui
of Rondônia, Salamãy (which has only two known semispeakers), and a language
which is composed of the four mutually intelligible dialects spoken by the Gavião
of Rondônia, the Zoró, the Cinta Larga, and the Aruá. The present population of
the Gavião is approximately 450 persons, all of whom speak the language. The
language of the Gavião contains syntactic nominalization which produces con-
structions which translate as relative clauses or as sentential complements. Comrie
and Horie (1995) observe that not all languages have a distinction between relative
clauses and complement clauses, citing evidence from Japanese and from Khmer.
Gavião syntactic nominalizations exemplify this lack of distinction.
In Gavião, two particles derive nominals: mát ‘concrete nominalization’ and
méne ‘abstract nominalization’. In the dialect of the Aruá, the form for abstract
nominalization is máne, which suggests as an etymology mát+ve, in which the
suffix -ve (which also occurs with verbs and adjectives) has the meaning ‘abstract
nominalization’. It appears that syntactic nominalization using these two par-
ticles is an innovation in the language of the Gavião, Zoró, Cinta Larga, and
Denny Moore
Aruá, not shared by the other two languages of the branch. The nominalization
particles probably were derived diachronically from the homophonous discourse
pronouns.
The two discourse pronouns, mát (plural: máày) ‘concrete discourse pronoun’
and méne ‘abstract discourse pronoun’, always occur sentence initially. Like other
pronouns in Gavião, they can be the nucleus of a NP but cannot modify follow-
ing nominals. The discourse pronouns refer to immediately preceding sentences,
which are independent prosodic and syntactic units. Examples:
(1) mát sô̦òt va bó pa̦-ága pazé-èy tá eé-na
that fermented ingest focus 1pi-aux other-pl with that-manner
‘Fermented like that we drink it with others’. (Previous sentence: ‘There it
ferments.’)
Note that in example (1) mát refers to the subject of the preceding sentence, while
in example (2) it refers to something understood, but not explicitly present in the
preceding sentence. In example (3) méne refers to the birds, which are the object
Relative clauses in Gavião of Rondônia
of the preceding sentence, in their rather abstract capacity as signs. In example (4)
méne refers to a manner of hunting, which is understood but not mentioned in the
preceding sentence.
As can be seen, these pronouns do not necessarily have a coreferential NP in
the preceding sentence. The hearer must infer the referent from context. This same
ambiguity (or flexibility) characterizes the nominalizations formed by mát and
méne in their role as nominalization particles.
As a diachronic hypothesis, the discourse pronouns extended their distribu-
tion and functions, becoming nominalization particles. According to Noonan
(1985: 47), complementizers frequently are derived from pronouns, for example
as in the case of the complementizer that in English, which is derived from the
pronoun that. The Gavião case is the mirror image of the complements introduced
by that in English, in which the complementizer precedes the clause: both mát and
méne follow the material to which they refer as pronouns and follow the material
(verb phrases and clauses) which they nominalize. The sequence below represents
this extension of the functions of mát and méne.
immediately preceding discourse – mát/méne > VP or CLAUSE – mát/méne
The particle mát appears also in conjunctions, as described by Moore (1984, 1989).
In their role as nominalizers mát and méne are considered particles, since (a)
there is a general pattern in Gavião of syntactic derivation by means of par-
ticles and (b) no noun, pronoun, or demonstrative in this language can form
a construction with a VP which precedes it. The scope of the nominalization
particles is the verb phrase, not the verb, since particles with phrase scope,
for example ó̦òp ‘negative’ or terè ‘true’, can occur between the verb and the
nominalizing particle. The concrete nominalizations are substantives, places,
events, etc. The abstract nominalizations are facts, reasons, or manners. Words
which translate as postpositions in Gavião are formally transitive verbs since
they have the same distribution and can be negated, intensified, or nominalized
in the same manner as verbs. Some examples of nominalized VPs are presented
below (underlined):
(5) “me-tá mát ká téét méèy-ka
2p-live nmnlz.concrete in exact 2p-(aux.imperat.def)-go
paágáá kára-ále-á” máà tá-kay-á
(3s)-open yet-future-end (3s)-aux.past 3p-involve-end
‘“Go open (it) where you live”, he told them.’
Denny Moore
(11) gô sep
1s+mouth leaf.obj
‘my book’
Constructions of the type vit ma’á̦p in (10) are frequently considered genitives
(Noonan 1985: 60). However, in Gavião these constructions are syntactic nouns
and often modify following noun stems, which a genitive construction cannot do
in Gavião. The examples (9)–(11) are compounds and not genitives; the noun stem
sep ‘object in the form of a leaf ’ is one of the noun stems which can be modified
but never possessed in Gavião (*gaáy sep ‘my mother’s leaf-like object’). The nomi-
nalizations can also constitute the first NP of an apposition, in which the second
member is a full NP, not just a noun stem:
(12) a̦á kávo ká mát e̦-gá
this year in nmnlz.concrete 2s-field
‘your field of this year’
One type of subordinate clause, whose auxilary is marked with the suffix //-néè//,
is always nominalized by mát or méne. According to Givón (1990), “In many lan-
guages, REL-clauses as well as verb complements and adverbial clauses are all
nominalized, so that only main clauses have fully finite syntax”. This is not the
case of the nominalized clauses in Gavião, which have the same composition pos-
sibilities as other clauses, except that certain particles only occur in the matrix
sentence. In the syntactic nominalizations there are three contrasts of time/aspect
marked on the auxiliary: unmarked, past indefinite, and past definite. Nominal-
ized clauses, like any other clauses in Gavião, contain a subject and an auxiliary and
can also contain various VPs and embedded clauses, whose order can be switched.
A nominalized clause frequently has as its internal head the subject of the clause.
However, the internal head can also be an object of a verb or, apparently, a genitive.
It is also possible that the clause contains no internal head. Examples of clauses
without external heads:
(13) “eé bó pazé-èy máà sóp abí palí sábéè
then focus other-pl aux.past clay.pit beside paxiúba board
ánéè a-vé-pea
(aux.past.def)-nominal 3c-intrans-beat.pl.obj
mát picaá mága-á” kí-ip
nmnlz.concrete upright put-end evidence-past
‘Then others put beaten paxiuba boards upright beside the claypit.’
Denny Moore
In the above example, the internal head is ‘beaten paxiuba boards’, the subject of
the embedded clause, and the nominalization is modified by the adjective ‘upright’.
In (14), the internal head is the third person subject and the nominalized clause is
the subject of the matrix sentence.
In (15) there is no internal head in the nominalized clause, which is the object of
the verb ‘use’
In (16), the nominalization has no internal head and is the object of ‘in’. In e xample
(17) below the concrete nominalization has no internal head and possesses the
noun stem ‘back’.
(17) “‘jè tá-kaypaà meé-néè mát
there 3p-call 2p-(aux.past.def)-nominal nmnlz.concrete
ábíìt abí teé mâà ma’á̦-á’
back.dimin on.side.of continuing 1s.aux.past (3s)-get-end
Relative clauses in Gavião of Rondônia
Note that the nominalized clause in (17) is interpreted as a place, with no explicit
marking to indicate this interpretation. Nominalized clauses in Gavião do not
have relativizers of location (where), manner (how) or time (when).
In the two following examples the nominalized clauses have an internal head
(in these examples it is an object) and an external head identical to the internal
head. These cases are more similar to the classic idea of a relative clause:
In (18), the internal and external head is ‘leaflike object’, which designates photos.
In (19), the internal and external head of the nominalization construction is ‘cloth-
ing’ and the nominalization is the subject of an existential sentence, with no verb.
The use of external heads is a possibility which permits explicit indication of
the head of a relative, eliminating ambiguity. Abstract nominalizations, as well as
concrete nominalizations, can have internal and external heads also, though the
abstract cases are rare for semantic reasons. One example permitted by our con-
sultant is the following:
In this example, the internal head is ‘1pi-photo’ and the external head is ‘leaflike
object’.
According to the “Accessibility Hierarchy” of Keenan and Comrie (1977), the
expectation would be that subjects and objects would be more easily relativized
than other arguments of the embedded clauses – which seems to be the case. The
occurrence of coreferential external heads facilitates the relativization of internal
arguments. Aside from subjects and objects, the Gavião accept at least examples of
constructions in which a genitive is the internal and external head:
(21) ávɨlɨ pí ánéè pa-záp sígɨ
dog foot (aux.past.def)-nominal 1pl-house near.to
mát ávɨlɨ máà
paderè va-á
nmnlz.concrete dog aux.past person bite-end
‘The dog whose track was near our house bit someone.’
5. Discussion
As can be seen from the above, the Gavião language does not have a construc-
tion which specifically forms relative clauses. The syntactic nominalizations are
Relative clauses in Gavião of Rondônia
very general constructions which translate as relative clauses with a head, headless
relative clauses, or complement clauses. Their composition is simple and elegant.
Where N’ is a syntactic (bar-one) noun, S’ is an embedded clause, and [+NOM]
indicates that the functional type of the clause is nominal (marked by //-néè//on
the auxilary) the composition rule is:
N′ → VP – mát
S′ méne
[+NOM]
In common language, the rule states that one possible composition of a syn-
tactic noun consists of a verb phrase or a clause of the nominal type followed by
one of two particles, mát or méne. This syntactic noun, like any other noun, can
modify a noun stem which follows it, as in the examples (9) and (10) above, form-
ing a larger syntactic noun. The syntactic noun can also be the head of a noun
phrase by itself. One consequence of this simplicity is the ambiguity demonstrated
in (22a–c) above. This ambiguity can be reduced by the use of external heads as
in (18) above.
One definition of a restrictive relative clause is given by Givón (1990: 646): “A
relative clause codes a proposition one of whose participants is coreferential with
the head noun that is modified by the clause”. This definition would seem to elimi-
nate all the constructions in the examples above, except those which have coref-
erential external and internal heads, eliminating even those which have obvious
internal heads, since it would be strange to say that an internal head is modified
by the clause of which it is an argument and from which it was not extracted. The
effect of the syntactic nominalization is, more precisely, to give a nominal distribu-
tion to the syntactic material represented by the verb phrases and clauses which
are nominalized. The question of an internal head is a question of interpretation
in context and not of the structure of the construction.
According to a definition of a relative clause that is less Eurocentric, in the
terminology of Keenan (1985: 161–63), the syntactic nominalizations which have
an internal head but not an external head would be what he calls internal rela-
tive clauses (“internal RCS”). Typologically, the Gavião constructions share two
properties with the internal relative clauses of Keenan: they occur in SOV lan-
guages and lack explicit marking of the internal head. Keenan does not mention
clauses with neither internal nor external heads, such as (15) or (17). Typically
these translate as complement clauses when they are abstract and have an external
head which treats a preceding NP as an argument, as in (23), or are objects of verbs
which accept abstract objects, for example ‘want’ or ‘await’.
Denny Moore
Notes on transcription
The symbols c and j denote palatal affricates, y the palatal glide, and s and z dental
affricates. The voiced bilabial fricative is indicated by v and the glottal stop by an
apostrophe. Long vowels are represented by sequences of two vowels. Low tone is
unmarked; high tone is marked with an acute accent, rising tone by a circumflex,
and an alternating tone is marked by a grave accent. Quotation marks indicate
direct quotes or thoughts in Gavião, which occur with no explicit word for ‘say’ or
‘think’. Vowel nasalization is marked by the ogonek beneath the vowel.
Grammatical glosses which are not obvious are:
1pe – first person exclusive, 1pi – first person inclusive, 3c – third person coreferen-
tial/ crossreferencing, aux – auxiliary, def – definite, indef – indefinite, intrans –
intransitivization, nmnlz – nominalization, pl – plural, trans – transitivization
All examples are from texts, except the last four, which were checked with at
least two consultants.
References
Comrie, B. & Horie, K. 1995. Complement clauses versus relative clauses: Some Khmer evidence.
In Discourse Grammar and Typology: Papers in Honor of John W.M. Verhaar, W. Abraham,
T. Givón & S. Thompson (eds), 65–75. Amsterdam: John Benjamins.
Givón, T. 1990. Syntax: A Functional-Typological Introduction, Vol. 2. Amsterdam: John
Benjamins.
Keenan, E.L. 1985. Relative clauses. In Language Typology and Syntactic Description: Complex
Constructions, Vol. 2, T. Shopen (ed.), 141–170. Cambridge: CUP.
Keenan, E.L. & Comrie, B. 1977. NP accessibility and universal gramar. Linguistic Inquiry
8: 63–100.
Moore, D. 1984. Syntax of the Language of the Gavião Indians of Rondônia (Brazil). Ph.D. dis-
sertation, City University of New York.
Moore, D. 1989. Gavião nominalizations as relative clause and sentential complement equiva-
lents. International Journal of American Linguistics 55 (3): 309–325.
Noonan, M. 1985. Complementation. In Language Typology and Syntactic Description: Complex
Constructions, Vol. 2, T. Shopen (ed.), 42–140. Cambridge: CUP.
Relative clauses in Yucatec Maya
Light heads vs. Null domain*
Rodrigo Gutiérrez-Bravo
El Colegio de México
1. Introduction
* I would like to thank Zarina Estrada, the audience at the 2009 Seminario de Complejidad
Sintáctica (Hermosillo, Universidad de Sonora), and two anonymous reviewers for helpful
feedback that greatly contributed to improving this paper. All errors that remain are my own.
Rodrigo Gutiérrez-Bravo
(1) polish
Jan czyta to, [co Maria czyta].
Jan reads this what Maria reads
‘Jan reads what Maria reads.’ (Citko 2004: 96)
Citko provides evidence that relatives like (1) are different from both headed and
headless relatives. She further notes that these constructions do not seem to be
specific to Polish. A preliminary overview of this phenomenon in European lan-
guages indicates that a number of them could also have light headed relatives.
Spanish is amongst these languages, as illustrated in (2), where the only element
that introduces the relative clause is the article la.
(2) spanish
He visto a la [que me presentaste].
have-1sg seen acc the that to.me introduced-2sg
‘I have seen the one that you have introduced to me.’ (Citko 2004: 97)
Yucatec Maya, the Mayan language from the Yucatán Peninsula, México, shows
relative clauses that are similar to the Polish and Spanish examples in (1) and (2).
This is shown in the constructions in (3) and (4), where the only element introduc-
ing the relatives is the demonstrative determiner le.1
Example (3) is like the Polish example in (1) in that it corresponds to a pro-
nominal relative, whereas (4) is like the Spanish example in (2) in that it corre-
sponds to a gap relative.
(3) Ma’ táan u man-ø-o’ob le ba’ax
neg dur erg.3 buy-abs.3sg-3pl dm what
u
k’áat-ø-o’ob-e’.
erg.3 want-abs.3sg-3pl-cl
‘They do not buy what they want.’ (MDG-B: 113)
1. Yucatec does not have a copula, so predicative constructions like the cleft in (4) are
constructed by simple concatenation of the subject and the constituent that functions as
its predicate. All examples are presented according to the orthographic conventions of the
Academia de la Lengua Maya de Yucatán and so they do not necessarily reflect their phonetic
form accurately. The name after each example corresponds to the text in my corpus that the
example is taken from; all texts correspond to oral narratives. The abbreviations used in
the examples are the following: abs – absolutive, asv – assurative, caus – causative, cit –
reportative, cl – clitic, cp – completive, dm – demonstrative, dur – durative, ep – epenthesis,
erg – ergative, ex – existential, fem – feminine (biological), foc – focus, hab – habitual,
ind – indicative, irr – irrealis, loc – locative, nex – negative existential, numc – classifier,
pass – passive, pl – plural, prep – preposition, prf – perfect, sg – singular, term – termina-
tive, top – topic, trns – transitive.
Relative clauses in Yucatec Maya
As such, one could consider the possibility that Yucatec is a language that has
light-headed relatives. However, in what follows I argue that constructions like
(3) and (4) are not instances of light-headed relatives. Rather, they correspond to
cases where the domain of the relative clause (in the sense of Andrews (2007)) is
reduced, a possibility in fact considered in Citko (2004). The conclusion I arrive at
is that light-headed relatives and relatives with a reduced or null nominal domain
correspond to different phenomena. I argue that Yucatec in fact shows the latter,
in spite of the superficial resemblance of (3) and (4) to the light-headed relatives
of Citko (2004). Before addressing this issue, in the following section I provide a
basic description of relative clauses in this language.
2. The precise characterization of the basic word order of Yucatec is still subject to much
debate, with numerous works assuming that the language’s unmarked word order is instead
VOS. This issue is tangential to our discussion of relative clauses, and so it will not be addressed
any further here. In contrast, there is agreement in the literature that the unmarked order
of mono-valent constructions in this language is VS. See Skopeteas and Verhoeven (2005),
Gutiérrez-Bravo and Monforte (2010) and Skopeteas and Verhoeven (2009).
Rodrigo Gutiérrez-Bravo
precedes it. Full argument and adjunct phrases in turn appear to either the right or
the left of this basic structure of the clause, as illustrated in (5).
(5) U y-íichami [yaan ui taa-s-øj] jun p’íit centaboj.
erg.3 ep-husband aux erg.3 come-caus-abs.3sg one bit money
‘Her husband must bring a little bit of money.’ (MDG-B: 189)
Example (6) also shows another important property of relative clauses in Yucatec,
namely, that they are clearly embedded inside the noun phrase. The clitic -o’
belongs to a set of distal clitics (-o’ can be translated approximately as ‘that’). These
phrasal clitics always attach to the right edge of the noun phrase and their presence
. In traditional Mayan linguistics, agent focus refers to a special form of the verb observed
when the transitive agent is focused, questioned or relativized. When relatives in Yucatec show
the agent focus form, they lack the ergative proclitic and the aspect auxiliary that precedes
it, as in (4) and (9). See Stiebels (2006) for a recent summary of agent focus across Mayan
languages.
Relative clauses in Yucatec Maya
is obligatory when the noun phrase is introduced by the determiner le, as in (7).
Hence the presence of this clitic to the right of the relative in (6) indicates that the
relative is a constituent inside the noun phrase.4
(7) Le kajtalil way-a’.
dm hamlet here-cl
‘This hamlet here.’ (MDG-B: 23)
4. Space considerations do not allow me to go into detail of other properties of relative
clauses in Yucatec, such as the external nature of the head of the relative and the specific
position of relative pronouns inside restrictive relative clauses. These issues are dealt in detail
in Gutiérrez-Bravo and Monforte (2009), to which I refer the reader for details.
5. Both interrogative and relative pronouns in Yucatec are in turn identical to indefinite
quantifiers (see Tonhauser 2003). Indefinite quantifiers, however, do not undergo obligatory
fronting to the left edge of the clause, whereas homophonous relative (and interrogative)
pronouns do. This difference in the behavior of these two kinds of elements is evidence
against an alternative analysis where the constructions considered in this paper are instead
analyzed as adverbial clauses or as purely nominal constructions headed by an indefinite
quantifier.
Rodrigo Gutiérrez-Bravo
(11) Jach raro persona [RC [NP máax] ti’ k-u si’ib-il]
very rare person who prep hab-erg.3 grant+pass-ind
‘He’s unusual, a person that it (this power) is granted to.’ (MDG-B: 62)
6. Pronominal relatives where a prepositional phrase is relativized, such as (11), display the
phenomenon known as pied-piping with inversion (see Smith-Stark (1988) for a survey and
Aissen (1996) for an analysis). As a head-initial language, Yucatec has prepositions and so
PPs canonically display the order P+NP. In pied-piping with inversion, however, a relative or
interrogative pronoun inverts its position with respect to the preposition, as in (11), where the
relative pronoun now appears to the left of the preposition. See Gutiérrez-Bravo (2010) for
detailed discussion of this phenomenon in relatives in Yucatec.
7. The particle ken in (12) and (19) is analyzed in Bohnemeyer (2002) as a subordinator, on
the basis that it appears exclusively in embedded contexts. It seems to me, however, that this
particle is instead an auxiliary that indicates optative aspect. Evidence for this can be found
in the fact that in some dialects of Yucatec this particle is inflected with the set of absolutive
suffixes to cross-reference the subject (as in (19)). As such, Yucatec resembles Mam, a Mayan
language from Guatemala, which has aspect auxiliaries that are only observed in subordinate
clauses (England 1983).
Relative clauses in Yucatec Maya
As such, except for the fact that they show an obligatory gap corresponding to the
relativized constituent, these clauses are identical in their structure to matrix clauses
like (5). Besides having gap relatives for subjects and objects, Yucatec also has gap
relatives for indirect objects and PP complements, temporal expressions and pos-
sessors, illustrated in the examples below. Gap relatives where the complement of
a preposition is relativized display “preposition stranding”, as in (18). Observe that
Rodrigo Gutiérrez-Bravo
these again are contact relatives in that there is no complementizer or any other
morphosyntactic element separating the head from the relative.
(18) Yaan-ø kex óox túul ko’olel [RC k-u
ex-abs.3sg even.though three numc woman hab-erg.
ts’a-ik-ø ti’ ____]-e’.
3 give-ind-abs.3sg prep -cl
‘There were even three women he gave it (his money) to.’ (MDG-B: 32)
(19) Le día [RC ken-o’on k wa’alkun-t-ø ____]-o’…
dm day aux-abs.1pl erg.1pl erect-trns-abs.3sg-cl
‘The day on which we erect them (the cross bars).’ (Bohnemeyer 2002: 258)
(20) Ti’ a nal [RC tun jóok’-ol
prep erg.2 corn dur+erg.3 come.out-ind
u yi’ij-o’ob ____]-o’.
erg.3 tip-plur -cl
‘To the corn (of yours) whose tips are just sprouting.’ (MDG-B: 13)
Finally, it is worth noting that gap relatives are not observed when a location is
relativized; relativization of a location in Yucatec necessarily requires the relative
pronoun tu’ux and so only pronominal relatives are observed in these cases. With
this I conclude the preliminary description of relatives in Yucatec. In the following
section I introduce the cases where the nominal head of the relative is absent. Here
I argue that these are not cases of light-headed relatives, but rather relative clauses
that have a domain that is partially or totally null.
In this section I argue that Yucatec relatives like (3) and (4), repeated here as (21)
and (22), are not instances of light-headed relatives in the sense of Citko (2004).
Instead I argue that these relatives result from the possibility of having some or all
of the elements in the domain of the relative being null. The gist of the argumen-
tation will be that this possibility is independent of relative clause formation and
instead depends on the general properties that are regularly observed in the noun
phrase in Yucatec.
(21) Ma’ táan u man-ø-o’ob le [ba’ax
neg dur erg.3 buy-abs.3sg-3pl dm what
u
k’áat-ø-o’ob]-e’.
erg.3 want-abs.3sg-3pl-cl
‘They do not buy what they want.’ (MDG-B: 113)
Relative clauses in Yucatec Maya
In the remainder of this section I present the evidence that favors the analysis in
(23), i.e. a null head analysis, over a light head analysis like the one argued by Citko
for Polish.
The fact that the range of elements that introduce the relatives under consideration
is reduced when compared to that of true light-headed relatives in Polish, and
the fact that they correspond precisely to the set of pre-nominal modifiers other-
wise observed in the language, is what is expected if a null nominal is the head of
these structures.
The same situation is observed for the relative pronoun máax ‘who’, in the relative
in (9) when compared to the headed relative in (11). In other words, an important
property that distinguishes headed from light-headed relatives in Polish is absent
in Yucatec. I take this as further evidence that the Yucatec relatives considered here
are not a different kind of relative, but instead simply appear to be so because of the
independent property of Yucatec that allows the head of a noun phrase to be null.
which typically can be found after the existential yaan, ‘existing’. Examples are
presented below.
(30) Ka’ t-u k’áat-aj-ø -e’ wa yaan-ø meyaj.
and cp-erg.3 ask-mod-abs.3sg-cl if ex-abs.3sg work
‘And he asked if there was work (available).’ (Gigante)
(31) Wa yaan-ø k’oja’anil-e’…
if ex-abs.3sg disease-cl
‘If there are diseases…’ (MDG-B: 205)
Observe that the clausal constituent that follows the existential in these examples is
no different from the canonical matrix clause in Yucatec first illustrated in (5). Yet
these clauses are not interpreted as matrix or complement clauses, but as relative
clauses. Furthermore, the relation between these clausal constituents and the exis-
tential is in itself puzzling. As illustrated in (30) and (31), existential yaan selects
a noun phrase as its argument, not a clause. The null head analysis I propose here
provides a straightforward solution for both problems. As illustrated in (34), in
this analysis the existential constructions considered here are no different in their
basic structure and properties from existential constructions with overt nominal
8. Yaan can also select for a dative (oblique) phrase besides this noun phrase. This results in
a dative-possessor construction of the kind ‘X exists to Y’, which is the standard way in which
predicative possession is expressed in Yucatec. This kind of construction, however, is not rel-
evant for our discussion of relative clauses.
Relative clauses in Yucatec Maya
arguments like (30) and (31), except for the fact that the null nominal head is
modified by a gap subject relative.
(34) Yaan-ø [NP Ø [RC k-u wéej taal ____ bejla]]-e’.
ex-abs.3sg hab-erg.3 still come today-cl
‘There are those that still come today.’
Descriptively, these are cases where the entire domain of the relative happens to
be null because; (a) the noun that heads the noun phrase is null and, (b) none of
the ordinary modifiers that precede the noun in Yucatec are present. Hence an
appropriate descriptive label for this kind of relative clause would be null-domain
relative. Observe that this is a natural extension of the analysis presented so far.
Specifically, in my proposal the relatives in (32) and (33) do not need to be ana-
lyzed as a kind of relative different from the gap relative in (23). In contrast, it is
unclear if an alternative light-head analysis can unify all these data as part of the
same phenomenon, since in (32) and (33) there is simply no element that could
function as the light head to begin with.
Lastly, the behavior of these constructions in discourse provides further sup-
port for the proposal that these relatives are indeed headed by a null nominal.
It has been widely observed that once a nominal is introduced in the discourse,
further reference to it is made by means of reduced or null pronominal forms
(see for instance Lambrecht (1994)). The heads of relative clauses appear to also
be subject to this condition, as first observed by Rojas (2006) for a number of
Zapotec relative constructions similar to the ones observed in Yucatec. This is
illustrated in (35), from which (32) is originally taken. In this sample of the nar-
rative, the referent of the null head of the relative, máak ‘people’, is introduced
two clauses before.
(35) Cuarenta y cinco máaki k-u meyaj, u personal le
forty and five people hab-erg.3 work erg.3 personnel dm
maquina-o’. Tumen k-u jo’och-kij-o’ob. Yaan-ø
machine-cl because hab-erg.3 grate-henequen-plur ex-abs.3sg
[NP Øi [RC k-u púut-ik-ø-o’ob le fibra]]-o’. . .
hab-erg.3 carry-ind-abs.3sg-plur dm fiber-cl
‘Forty five people worked (there), they were the machine’s personnel.
Because they used to harvest henequen. There were those that carried the
fiber…’ (MDG-B: 101)
The text goes on to list and describe the different groups of workers and their
specific roles in the production of henequen fiber. For our purposes what is
important is that (35) illustrates that the otherwise peculiar constructions
Rodrigo Gutiérrez-Bravo
like (32) and (33) appear precisely in those contexts where nominals are
expected to be phonologically reduced or null. Specifically, they are observed
when their antecedents have already been introduced in the previous discourse
and so the nominal can be taken to be discourse-old. I take this as further evi-
dence in favor of the analysis where these clauses are headed by a null nominal
in their domain.
In summary, in this section I have presented evidence in favor of a null-head
analysis of the relative constructions discussed so far. I have argued that three
properties of these constructions point to the conclusion that they are not
light-headed relative clauses. These three properties are; (a) the absence of a dis-
tinct set of elements that function as light heads; (b) the absence of a distinct set
of relative pronouns for the apparent light-headed relatives, and; (c) the behavior
of these relative constructions in existential contexts. I have further argued that
the analysis I propose extends naturally to a number of relative constructions in
Yucatec where there is no overt element whatsoever in the domain of the relative,
which I have labeled as null domain relative clauses.9
4. Conclusions
9. An anonymous reviewer asks how compatible this analysis is with an alternative anal-
ysis where the constructions I have analyzed here are taken to be nominalizations instead
of relative clauses, as in Shibatani (2010). The reviewer points out that, in such an alterna-
tive analysis, the determiner could simply be modifying the nominalized constructions
as it would modify any noun. It seems to me that there are two pieces of evidence that
make the nominalization account problematic. First, recall that pronominal relatives in
Yucatec show pronouns that are just like interrogative pronouns in both their form and the
requirement that they appear at the left edge of the clause (see also fn. 5). The nominaliza-
tion analysis would fail to capture this parallelism between the relative constructions and
pronominal interrogative constructions, which are clearly clausal in nature. Secondly, the
nominalization analysis would need to explain why, in the constructions I have analyzed
here, there is always one argument that is either pronominalized with a wh-pronoun (in
pronominal relatives) or missing altogether (in gap relatives). These properties have been
well attested for headed relative clauses crosslinguistically but not, to the best of my knowl-
edge, for nominalizations.
Relative clauses in Yucatec Maya
olish. I have argued that these relative constructions in Yucatec show most of the
P
characteristic properties of fully headed relatives and not those of the light-headed
relatives of Polish. As such, I have proposed that the Yucatec relative clauses under
consideration are best understood as headed relative clauses where the domain of
the relative is partially or entirely null.
References
Aissen, J. 1996. Pied-piping, abstract agreement and functional projections in Tzotztil. Natural
Language and Linguistic Theory 14: 447–491.
Andrews, A. 2007. Relative clauses. In Language Typology and Syntactic Description, 2nd. edn.,
T. Shopen (ed.), 206–236. Cambridge: CUP.
Bohnemeyer, J. 2002. The Grammar of Time Reference in Yukatek Maya. Munich: Lincom.
Bresnan, J. & Grimshaw, J. 1978. The syntax of free relatives in English. Linguistic Inquiry
9: 315–391.
Briceño Chel, F. 2002. Topicalización, enfoque, énfasis y adelantamiento en el maya yukateco. In
La organización social entre los mayas prehispánicos, coloniales y modernos, V. Tiesler Blos,
R. Cobos & M. Greene Robertson (eds), 374–387. Mexico City/Mérida: INAH/UADY.
Bricker, V. 1978. Wh-questions, relativization and clefting in Yucatec Maya. In Papers in Mayan
Linguistics, Laura Martin (ed.), 109–139. Columbia MI: Lucas Brothers.
Citko, B. 2004. On headed, headless, and light-headed relatives. Natural Language and Linguistic
Theory 22: 95–126.
Comrie, B. 1989. Language Universals and Linguistic Typology. Chicago IL: The University of
Chicago Press.
England, N. 1983. A Grammar of Mam, a Mayan Language. Austin TX: University of Texas
Press.
Gutiérrez-Bravo, R. 2002. Formas verbales incorporadas transitivas en maya yucateco. In Del
cora al maya yucateco: Estudios lingüísticos sobre algunas lenguas indígenas mexicanas,
P. Levy (ed.), 131–178. Mexico City: UNAM.
Gutiérrez-Bravo, R. 2010. Free relative clauses in Yucatec Maya. Ms, El Colegio de México.
Gutiérrez-Bravo, R, & Monforte, J. 2008. La alternancia sujeto-inicial/verbo-inicial y la Teoría de
Optimidad. In Teoría de Optimidad: Estudios de sintaxis y fonología, R. Gutiérrez-Bravo &
Herrera Zendejas, E. (eds), 61–99. Mexico DF: El Colegio de México.
Gutiérrez-Bravo, R. & Monforte, J. 2009. Focus, agent focus and relative clauses in Yucatec
Maya. In New Perspectives in Mayan Linguistics, H. Avelino, J. Coon & E. Norcliffe (eds),
83–96. Cambridge MA: MIT Working Papers in Linguistics.
Gutiérrez-Bravo, R. & Monforte, J. 2010. On the nature of word order in Yucatec Maya. In
Information Structure in Languages of the Americas, J. Camacho, R. Gutiérrez-Bravo &
L. Sánchez, 139–170. Berlin: Mouton de Gruyter.
Kroeger, P. 2005. Analyzing Grammar. Cambridge: CUP.
Lambrecht, K. 1994. Information Structure and Sentence Form. Cambridge: CUP.
Lehmann, C. 1984. Progress in general comparative linguistics. Studies in Language 8: 259–286.
Monforte, J., Dzul, L. & Gutiérrez-Bravo, R. 2010. Narraciones Mayas. Mexico City: Instituto
Nacional de Lenguas Indígenas.
Rodrigo Gutiérrez-Bravo
van Riemsdijk, H. 2006. Free relatives. In The Blackwell Companion to Syntax, Vol. II, M. Everaert
& H. van Riemsdijk (eds), 338–382. Oxford: Blackwell.
Rojas, R. M. 2006. Construcciones relativas en el zapoteco de Santa Ana del Valle, Oaxaca. Paper
presented at Primer Coloquio de Exalumnos de la Maestría en Lingüística Indoamericana
del CIESAS, Mexico City.
Shibatani, M. 2010. Nominalizations sí, headless relatives no! Paper presented at the XI Encuen-
tro de Lingüística del Noroeste, Universidad de Sonora, Hermosillo.
Skopeteas, S. & Verhoeven, E. 2005. Postverbal argument order in Yucatec Maya. Sprachtypolo-
gie und Universalienforschung (STUF) 58: 347–373.
Skopeteas, S. & Verhoeven, E. 2009. Distinctness effects on VOS order: Evidence from Yucatec
Maya. In New Perspectives on Mayan Linguistics, H. Avelino, J. Coon & E. Norcliffe (eds),
157–174. Cambridge MA: MIT Working Papers in Linguistics.
Smith-Stark, T. 1988. ‘Pied piping’ con inversión en preguntas parciales. Ms, El Colegio de
México.
Stiebels, B. 2006. Agent focus in Mayan languages. Natural Language and Linguistic Theory
24: 501–570.
Tonhauser, J. 2003. F-constructions in Yucatec Maya. In The Proceedings of SULA 2, J. Anderssen,
P. Menéndez-Benito & A. Werle (eds), 203–223. Amherst MA: GLSA.
Questionable relatives
Marianne Mithun
University of California, Santa Barbara
1. Introduction
In a number of languages, interrogative and relative pronouns show the same forms:
Who came?; I saw the man [who came]. The similarity raises intriguing q uestions
Marianne Mithun
about the relationship between the two: perhaps some shared fundamental seman-
tic feature, or some recurring diachronic pathway by which one develops into the
other. Haspelmath (2001), Heine and Kuteva (2006), and others point out, how-
ever, that the pattern is not distributed evenly around the globe: it is highly con-
centrated in Europe.
The pattern is not entirely nonexistent elsewhere. It occurs in some indig-
enous languages of the Americas. It has been observed, for example, in South
America in Tariana, an Arawakan language of Brazil (Aikhenvald 2002). It
has been noted in Mesoamerica in Nahuatl and Pipil, Uto-Aztecan languages
(Karttunen 1976; Hill & Hill 1986; Campbell 1987). It also occurs in North
America in Iroquoian languages. The examples below are from Tuscarora, a
Northern Iroquoian language spoken in the 16th century in what is now North
Carolina.
(1) Tuscarora káhne’ ‘who’ question: Elton Greene, speaker
Kahné’ weθatkáhri’θ?
who one told you
‘Who told you?’
Matches also appear in other Iroquoian languages. The examples in (3) and (4)
are from Mohawk, spoken in the 16th century in what is now eastern New York
State.
(3) Mohawk nahò:ten’ ‘what’ question: Sha’tekenhátie’ Marian Phillips, speaker
Nahò:ten’ sá:ton?
what you are saying
‘What are you saying?’
The Iroquoian languages differ strikingly in their morphological, syntactic, and dis-
course structures from those of the European languages famous for interrogative/
relative pronoun matches, but they actually have much to contribute to our under-
standing of the pattern. Here we look more closely at what we can learn from
them, with a focus on Tuscarora.
Questionable relatives
The genetic relationships among the modern Iroquoian languages for which we
have documentation of connected speech are sketched in Figure 1.
Iroquoian
We do not know when the various branches split, but estimates have placed
the separation of Southern and Northern Iroquoian at three and a half to four
millennia, and that of Tuscarora from the other Northern Iroquoian languages at
about two to two and a half (Lounsbury 1961).
The only known representative of the Southern branch is Cherokee. In the
16th century the Cherokee inhabited a wide area of the Southeast, covering parts
of what are now Tennessee, North Carolina, Virginia, South Carolina, Georgia,
and Alabama. In 1838 most Cherokee were forced to march to Oklahoma, where
their descendants live today, though some managed to remain in western North
Carolina. The Cherokee language does not show the interrogative/relative match.
Relative constructions are formed instead with a subordinating prefix ji- and/or
a special tonal pattern on the verb marking subordination. There are no relative
pronouns.
(5) Cherokee káako ‘who’ question: Montgomery-Anderson 2008: 478
Káako uùtlv́vka?
who 3 is sick
‘Who is sick?’
(6) Cherokee ji- relative: Montgomery-Anderson 2008: 132
Haatlv naʔ achúúja ji-tvvsúúhwisi aàsoóy v̋.
Where that boy rel-he.will.paint.subor fence
‘Where’s that boy who will paint the fence?’
Marianne Mithun
It is becoming ever clearer that speakers can transfer grammatical patterns from
one language to another without actual morphological or lexical substance.
Bilinguals can perceive a pattern in one of their languages and replicate it in the
other, using only native material. Such a process apparently occurred in Tariana
(Aikhenvald 2002: 183, Heine & Kuteva 2005: 3, 2006: 213). Relative clauses were
originally formed in Tariana by means of a prefix ka-.
(9) Tariana traditional relative clause: Aikhenvald 2002: 183
ka-yeka-kanihi̜ kayu-na na-sape.
rel-know-dem.anim thus-rem.p.vis 3pl-speak
‘Those who knew used to talk like this.’
Younger speakers are now producing the Tariana pattern in (11), simply adding
Tariana interrogative pronouns like kwana ‘who?’ to the existing Tariana relative
clause structure.
(11) Tariana innovated relative clause: Aikhenvald 2002: 183
kwana ka-yeka-kanihi̜ kayu-na na-sape.
who rel-know-dem.anim thus-rem.p.vis 3pl-speak
‘Those who knew used to talk like this.’
But the transfer of grammatical patterns can be more complex. Through care-
ful examination of the literature on the development of relative pronouns, Heine
and Kuteva (2006) hypothesize that an interrogative/relative pronoun match can
develop gradually through a sequence of steps, paraphrased in (12).
(12) Heine and Kuteva Stages of Grammaticalization: 2006: 209
Stage 1 The marker begins in lexical gap questions.
Who came?
Stage 2 The marker is extended to introducing indefinite
subordinate clauses
I don’t know who came.
Stage 3 The marker is extended further to definite subordinate clauses.
You also know who came.
These structures may be interpreted as headless relative clauses.
You know the one who came.
Stage 4 The marker is extended still further to headed relative clauses.
Do you know the woman who came?
There is, however, some very pertinent Tuscarora material. At the beginning of the
18th century, most Tuscarora people began leaving North Carolina for the north,
eventually settling in western New York State near Niagara Falls, and in southern
Ontario on the Grand River Reserve. In 1858 a man named Jonathan Napoleon
Brinton Hewitt was born on the Tuscarora reservation in New York to a Tuscarora
mother. He learned English as his first language, but acquired Tuscarora from
school friends at around age 11. In 1880 he was hired by the Bureau of Ethnology to
assist in the documentation of Tuscarora and other Iroquoian languages. Between
1888 and 1897, with the assistance of Lucinda Thompson, a fi rst-language Tusca-
rora speaker, he collected 36 texts in Tuscarora. The texts have been edited and
published in a volume by Blair Rudes and Dorothy Crouse (1987). It is a signifi-
cant collection, running 621 pages exclusive of appendices. There is also a second
body of texts, collected from the mid-20th century to the present, primarily from
speakers born near the end of the 19th century. It includes material from speaker
David Hewitt collected by A.F.C. Wallace and W. Reyburn in 1948 and 1950; from
Nellie Gansworth by Wallace in 1948 and 1949 and by F. G. Lounsbury in 1952
and 1954; from Edith Jonathan in 1950 by Lounsbury; and from Elton Greene
through the early 1970’s by M. Mithun. More recent material has been provided
by speaker Howard Hill to F. Patterson, B. Bissell, and M. Mithun. All of these
speakers learned Tuscarora as a first language, then later became fluent in English.
Not all of the speakers recorded by Hewitt are identified by name, but for those
that are, it has been possible to ascertain their dates of birth through cemetery
records and with the help of Wendy Rae Bissell, Tuscarora genealogist. The known
speakers represented in the two sets of texts, along with their dates of birth, are
listed in (13).
The relatively short period of documentation available for Tuscarora, little more
than a century, provides compelling evidence for exactly the scenario proposed
by Heine and Kuteva. Significantly, this period coincides with the spread of
Questionable relatives
ilingualism in English. After spending the summers of 1948 and 1949 at the Tus-
b
carora reservation in New York State, Anthony F.C. Wallace described the transi-
tion from Tuscarora to English.
The next 80 years – roughly from 1865 to 1948 – saw the consolidation
of the economic adjustment so successfully made. Further social integration with
the surrounding Whites has been necessary. This has accelerated the decline of
the native language and its progressive replacement by English; … the Tuscarora
language is now spoken by preference only by the older people, and scarcely at all
by the youngsters. (Wallace 1952: 16).
The language shift described by Wallace began after the first group of speakers had
reached adulthood, but before those in the second were born.
Heine and Kuteva point out that all of the indefinite pronominal m arkers
in a language need not evolve in lockstep; ‘each can exhibit a different
grammaticalization behavior’ (2006: 210). Indeed, Tuscarora shows the indepen-
dent development of individual markers. The next sections examine the evolution
of each of the Tuscarora interrogative pronouns: Inanimate té’ or tawé̜:te ‘what?’,
Human káhne’ ‘who?’, Space hè̜:we ‘where?’, and Time kahné̜’kye ‘when?’. (Manner,
quantity, and attributive questions are based on the Inanimate té’ ‘what?’).
The particle té’ ‘what?’ sometimes appears in the 19th century texts in combination
with the word awé̜:te ‘thing(s)’, based on the verb root -ęte ‘be a certain one’.
Sometimes Hewitt wrote the combination as two words te’ awé̜:te, sometimes as
té’awé̜:te and sometimes as ta’awé̜:te.
Marianne Mithun
(16) 19th century té’ awé̜:te ‘what thing?’ question: 1888 ms 422, RC 1987: 429
Te’ awé̜:te káha’w?
what thing it takes
‘What did she take with her?’
(17) 19th century ‘what?’ question: 1888 ms 432, RC 1987: 12–13
Te’awé̜:te θačhú:ri?
what you have eaten
‘What have you eaten?’
In the modern language, the form tawé̜:te is pervasive, a longer alternate of té’ for
‘what?’.
(18) 20th century tawé̜:te ‘what?’ question: Elton Greene, speaker p.c. 1971
Tawé̜:te weθatkáhri’θ?
what he told you all
‘What did he tell you?’
The use of te’/tawé̜:te in questions corresponds to Heine and Kuteva’s Stage 1, the
point of departure.
The 19th century texts also show the use of both té’ and tawé̜:te ‘what’ intro-
ducing complements of verbs of speech, cognition, and perception, where the
speakers, thinkers, and perceivers do not know the identity of the referents of the
complement clauses. These constructions are termed indefinite complements by
Heine and Kuteva.
(19) 19th century indefinite complement of speech: Thompson 1888 ms 432:
RC 1987: 579
Ha’ kayę’na’né̜:’nyu:t kwęhs akayeyę’nè:rik te’ her é̜’ru’ uhtá’kę’.
the they invited them not could the know what also until behind
‘The invitees were unaware of what had transpired previously.’
These uses have persisted through the 20th century into the modern language.
Questionable relatives
The marker té’ does not appear in definite complements in the 19th century
texts, where the speakers, thinkers, or perceivers can identify the referent of the
complement. This use is frequent in the 20th century, however.
(25) 20th century definite complement of speech: Nellie Gansworth 1948,
speaker, to Wallace
Wahratkáhrye’ te’ thwahrá:’nye’r.
he told what he did
‘He told them what he had done.’
(27) 20th century complement of cognition: Elton Greene, speaker p.c. 1972: 421
Kyę’né:ri: te’ tì:wa’θ ruhwístayę’.
I know what so it amounts he money has
‘I know how much money he has.’
As referring expressions, the headless relatives can be preceded by the article ha’.
(29) 20th century headless relative with article: Elton Greene, speaker
p.c. 1972: 326
É̜ kyekwarihé̜:tyę’ kyení:kę: kayetá:kre’,
I will teach you this they dwell
‘I will teach you, this tribe,
ha’tawé̜:te neyawętahwęčúhę kyení:kę: ęθwayę’né:ri:k.
the what it is necessary this you all will know
that which it is necessary for you to know.’
Questionable relatives
(30) 20th century headless relative with article: Elton Greene, speaker
p.c. 1972: 186
Nyękwa’tikęhriyúhθe ha’ tawé̜:te, kakurihwíhs’ę.
it is pleasing to us the what they have promised
‘We are pleased with what they promised.’
(31) 20th century headless relative with article: Elton Greene, speaker
p.c. 1971: 131
Čhé̜’ kwà:nę wakyehserhá:r’ę ha’ tawé̜:te tika’nyé:rhę’.
just much it keeps me busy the what I do here and there
‘The things I do keep me very busy.’
Further extension of the pronoun té’ ‘what’ into headed relative clauses, Heine and
Kuteva’s Stage 4, has not taken place in Tuscarora. The somewhat rare example
below might at first glance be taken as a headed relative clause.
(32) Tuscarora headed relative?: Elton Greene, speaker p.c. 1972: 186
Ha’ ękayętęhnin é̜ hek hé’thu ha’.. ę:
the they will be selling there the uh
‘They’ll be selling there the
tawé̜:te kayakyetì:yahs u’tíkste tawé̜:te hé’thu.
what they make beadwork what there
beadwork they make.’
The prosody of this sentence, visible in Figure 3, indicates that the second line is
actually composed of a series of appositives: ‘what they make, beadwork, whatever’.
There is a pause after ‘they make’ and a pitch reset on the following noun ‘beadwork’.
The interrogative use of the pronoun káhne’ ‘who?’ has remained unchanged over
the past century.
Marianne Mithun
(34) 20th century ‘who?’ question: Elton Greene, speaker, p.c. 197
Kahné’ weθatkáhri’θ?
who one told you
‘Who told you?’
(35) 20th century ‘who?’ question: Elton Greene, speaker, p.c., 1971
Kahne’ wa’na’natkáhri’θ?
who one told one
‘Who did he tell?’
The same pronoun appears in indefinite complements in the 19th century texts,
where the speaker, knower, or perceiver does not have a specific referent in mind,
but it is rare.
(36) 19th century indefinite ‘who’ complement: Hewitt, RC 1987: 50
Í: ’ętkę’tiké̜ hnę’ ha’ káhne’ ęθtí:tya:k.
I I will decide for you the who you two will marry
‘I will decide for you who you shall marry.’
(41) 0th century headless ‘who’ relative: Elton Greene, speaker p.c. 1972: 183
’kayę’na’nit’úhthahs ha’ káhne’ kayę’nęné̜ hyar.
caused them to sleep the who they are guarding him
‘He put to sleep those who were guarding him.’
The use of the interrogative pronoun hè̜:weh ‘where?’ in questions shows little
change over the past century. (The initial h had disappeared from the speech of
Elton Greene, but it remains in that of Howard Hill.).
(43) 20th century è̜:we ‘where?’ question: Elton Greene, speaker, p.c. 1971
È̜:we nyé̜:kye:t?
where there I shall go
‘Where shall I go?’
The 19th century texts show no other uses of this marker. It is used in the modern
language, however, to introduce complements.
(55) 20th century ‘when’ complement: Elton Greene, speaker p.c. 1972: 413
Kyę’né:ri: kahné̜’kye twahrayé̜:thu’.
I know when so he planted
‘I know when he planted.’
(56) 20th century ‘when’ adverbial clause: Elton Greene, speaker p.c. 1972: 434
Ha’ ù:nę kahné̜ ’kye, … ha’ ęki’rwé̜ hθę’,
the now when the I will tail drop
θehyáhra:k ęθwa’né’ku’.
you all remember you all will run away
‘Now when I drop my tail, remember to run away.’
(57) 20th century ‘when’ adverbial clause: Howard Hill, speaker to Francene
Patterson 2000
Kahné̜’kye ahsku’čhè̜:ni’, thwé:’n ęhsne’rawíhsi’.
when you would find it all you will root un give
‘When you find it, pull out the whole root.’
The Tuscarora patterns raise some interesting questions. In the 19th century,
the pronouns ‘what’, ‘who’, ‘where’, and ‘when’ differed in their extensions to new
contexts. The indefinite ‘what’ was already used not only in questions, but also in
indefinite and definite complements (i, ii, ii). Human ‘who’ and locative ‘where’ were
used only in questions and indefinite complements (i, ii). Temporal ‘when’ was used
only in questions (i). By the late 20th century, all had expanded through all steps
to headless relatives (i, ii, iii, iv), but none has yet moved into headed r elatives (v).
One issue taken up by Heine and Kuteva (2006: 226–229) is directional-
ity. Citing Lehmann 1984; Thomason and Kaufman 1988: 320, Schafroth 1993;
Matras 1996: 64, Kortmann 1998: 554, Le Goffic & Wang 2002, and Heine and
Questionable relatives
Kuteva 2002, they write, ‘That the polysemy between question and subordina-
tion markers that we are concerned with here is the result of a unidirectional pro-
cess from the former to the latter has been claimed independently by a number
of authors’ (2006: 226). The Tuscarora situation summarized in Figure 4 provides
especially strong evidence for this directionality. All have Stage i as a point of depar-
ture (questions), and at each point in time, each pronoun shows only contiguous
stages of development.
The fact that the individual Tuscarora pronouns had reached different stages of
development by the 19th century raises the question of how universal the order of
development among the various indefinite pronouns might be cross-linguistically.
Comparisons with sequences discussed by Heine and Kuteva indicate that the
order varies.
High German was ‘what’, wer ‘who’, and wann ‘when’ have progressed into
headless relatives, while wo ‘where’ has expanded one step further into headed
relatives but only for some speakers. Thus the German inanimate, human, and
time markers are still solidly at Stage 3, and the space marker is somewhere
between their 3 and 4 (2006: 210–211). English, who, where, when and which have
completed the path, all now appearing in headed relative clauses, but what still has
not taken that final step. ‘What’ expanded first in Tuscarora but last in English.
The differences in rates of development of individual markers within lan-
guages, and in the order of development of categories across languages, raise ques-
tions about what factors might retard or hasten expansion along this pathway. A
well-known frustration is that though we can sometimes point to motivations
behind a change once it has happened, we cannot predict whether or not a change
will take place when the motivations are present. We can, however, begin to
assemble hypotheses about factors that may affect rates of change. The h ypotheses
Marianne Mithun
Expansion of the more advanced Tuscarora markers ‘what’ and ‘who’ created more
specific constructions than those that had existed before. Both form complement
clauses. The less advanced markers ‘where’ and ‘when’ replaced demonstratives in
established constructions. These usually serve as adverbial clauses.
In the 19th century, speakers communicated such ideas with two other
constructions.
The most common was simply a juxtaposed finite clause. (Clauses can consist
of just a verb, or a verb plus additional material.) The clause is typically preceded
by the article ha’, which signals both identifiability and the dependent status of the
following clause.
(58) 19th century clausal nominal: Joseph Williams, speaker 1888, Hewitt
in RC 382
Há:ne:’ nęká:ye:’r ha’ tyurihú’nę:.
that it will happen the it is customary
‘What is customary must take place.’
Questionable relatives
The 19th century juxtaposed clause, even with an article and prosody that
probably linked it to an adjacent clause, was more general in its grammati-
cal function, indicating dependency but not specifying a particular semantic or
grammatical role.
In the 19th century, such clauses sometimes occurred in apposition to the
word awé̜:te ‘thing(s)’.
(61) 19th century ‘the things’: 1897 ms 411, RC 1987: 359
Wa’thrathnyaré:tya’t ha’awé̜:te wahrá:kę’.
he news spread the thing he saw
‘He reported what he had seen.’
(62) 19th century ‘the things’: 1888 ms 432, RC 1987: 603
Ù:nę wa’na’natkáhri’θ ha’awé̜:te katíhu’θ yétkwakęw…
now she told her the things it exists ..her stomach interior
‘She told her what was living in her stomach, [many hundreds of snakes].’
By the 20th century, the word awé̜:te ‘thing(s)’ had generally been replaced by the
indefinite pronoun tawé̜:te ‘what’ in the headless relative constructions seen earlier.
(30) 20th century headless ‘what’ relative: Elton Greene, speaker p.c. 1972: 186
Nyękwa’tikęhriyúhθe ha’ tawé̜:te, kakurihwíhs’ę.
it is pleasing to us the what they have promised
‘We are pleased with what they promised.’
In the modern language, headless relatives based on tawé̜:te ‘what’ are now the
majority pattern.
The innovated ‘what’ headless relative construction ‘what they promised’
in (30) could be viewed as the result of adding tawé̜:te ‘what’ to existing clausal
constructions like that in (58) ‘what is customary’. It could also be viewed as the
replacement of the nominal awé̜:te ‘thing(s)’ with the new indefinite pronoun
tawé̜:te ‘what’ and expansion to a majority pattern.
Marianne Mithun
When we examine further data we see that this is actually a different kind of
construction. Demonstrative pronouns kyení:kę ‘this one, these’ and hení:kę:‘that
one, those’ are frequent in Tuscarora speech. They occur especially often at the
ends of simple sentences, as below. We can see from the pitch trace and waveform
in Figure 7 that both demonstratives hení:kę:‘that one, those’ were grouped pro-
sodically with the preceding clause. Both showed a final fall in pitch, followed by a
pause, then a pitch reset on the following clause.
(64) Demonstrative construction: Elton Greene, speaker p.c. 1972: 257
Wahękhé̜hsyę’ hení:kę:.
I heard that.
‘I heard that
Nahrakwa’nętí:ye:t hé:snę:, ękáhnę’t hení:kę:.
he sent me here hence I will destroy those
He sent me to destroy them, those things.’
Figure 7. ‘I heard it, that thing. He sent me to destroy them, those things’
Questionable relatives
Sound recordings are not available for the sentence in (63), but we do have
recordings of similar constructions. The prosody indicates that these consist of
a sequence of referring expressions in apposition, rather than a single relative
clause.
(65) Demonstrative construction with prosody: Elton Greene, speaker
p.c. 1972: 258
Wa’tkaháhihθ hení:kę:, ruya’kwáher.
it met it that one he body carries
‘It met it, that one, a dinosaur.’
(74) 20th century demonstrative structure: Elton Greene, speaker p.c. 1972: 183
Yahwahrárku’ kyení:kę: Čá:ks rayá:θę.
he went there this one Obadiah he is named
The actual structure is clearer when intonation is brought into the picture. Each
line of transcription below represents a separate prosodic phrase.
The demonstratives were grouped prosodically with the preceding clause in each
case. (The sounds between phrases 2 and 3, and between 3 and 4, are breaths.).
0 8.499
Time (s)
Kę’ also appears pervasively in the 19th century in headless relative constructions,
meaning ‘the place where’.
(78) 19th century kę’ headless relative: 1888 ms 432, RC 1987: 192–3
Hé’thu yawáhe:t kę’ ru’nihsúhe’.
there it went there where he has hidden
‘It went over to where the hunter was hiding.’
The sentence below is from a traditional ceremony which takes place when a chief
dies and his wampum is passed on to his successor.
(80) 20th century kę’ headless relative: Elton Greene, speaker p.c. 1972: 444
Kyení:kę: čuhtíčhe:θ yé̜:θwe:t ha’ ké̜’ sęr’ é̜ hsayę’.
this wampum it will go back there the where you clan have
‘This wampum will go right back to your own clan.’
Marianne Mithun
(83) 19th century adverbial clause: Joseph Williams, speaker 1888 ms 438, RC
1987: 376
Kę:θ ha’ wa’ęktányę’θ wa’nyakyá’thnę’.
customarily the one village enters for me we two play ball
‘When anyone visits me, he and I play a game of ball.’
The temporal kahné̜’kye ‘when’ has just begun to reach Heine and Kuteva’s
Stage 3, but it is still a minority pattern with a specialized sense. It is replacing a
temporal demonstrative in an established adverbial clause construction. Discuss-
ing the replacement of demonstratives by interrogative pronouns ancestral to who,
which, etc. in Middle English, Romaine notes that ‘The transition from interroga-
tive to relative pronoun began in types of indirect questions where the interroga-
tive character of the pronoun became weakened, and the pronouns so used were
generalizing relatives’ [‘whoever’, ‘whatever’, ‘whichever’ etc.] (Romaine 1984: 449,
cited in Heine and Kuteva 2006: 220). The contexts in which the temporal kahné̜
’kye ‘when’ occurs in modern Tuscarora suggest that at least this marker went
through such a stage.
still are today, ‘what’ and ‘who’, created new structures. They did not simply replace
other markers in existing constructions. The two interrogative pronouns that were
less advanced in the 19th century and are still competing with other construc-
tions, ‘where’ and ‘when’, are replacing demonstrative adverbs in well-established
adverbial clause constructions. How this fact might have affected developments
can only be a matter of conjecture. Bilingualism could have brought an awareness
of distinctions the language had left unspecified up to that point, those accom-
plished by ‘what’ and ‘who’ subordinate clauses in English. The awareness might
have prompted bilingual speakers to fill the newly-perceived lack with construc-
tions based on native ‘what’ and ‘who’ pronouns. Place and time constructions
were already established, so there was less motivation for expanding the functions
of the indefinite ‘where’ and ‘when’ pronouns.
Such a hypothesis of course pushes the question back one step. Why did place
and time constructions develop first? In both the 19th and 20th century material
(as well as in all related languages), constructions persist that would provide likely
sources for their development. A very common pattern of expression in Northern
Iroquoian languages involves a sequence of clauses or sentences, the second of
which begins with ‘there’, sometimes translated ‘that’s where’.
(88) 19th century ‘there’: Joseph Williams, speaker 1888, ms 438, RC: 401
Wahrá:kę’ neyu’niyháknę yuyené̜’ę karatkwár’u’y.
he saw it is creek between it is dead tree fallen large elm
‘He saw a large elm tree lying across a small stream of water.
Hé’thu kwè:ni’ wahrá:kę’ yętkyéhnač uyękwì:re
there simply he saw one burden carries wood
Near by this fallen tree he beheld a diminutive old woman
tiwathwaritá’θ’a kahskwarí’a.
it is backpack small it is feeble small
loading up her forehead strap with fagots and pieces of wood.’
(89) 20th century ‘there’ construction: Elton Greene, speaker p.c. 1972: 181
Yahwahrárku’ kyení:kę kayeyękí:rya’ks
he went away this one they are chopping wood
‘He went away, this one, to a chopping bee
hé’thu yahwáhre:t.
there he went there
that’s where he went.’
Similarly, a demonstrative adverb ù:nę or áθe’ ‘at that time, then’ often occurs at
the beginning of a sentence, situating it temporally with respect to the preceding
sentence.
Marianne Mithun
(90) 20th century ù:nę ‘then’ sentence: Elton Greene, speaker p.c. 1972: 220
‘Now then, the one that was slow, it caught him.
It carried it back and threw his body somewhere.’
Ù:nę nektí:ha’nę:’t, wahrú:tkaht hení:kę:.
now second one it chased him that one
‘Then it chased the second one.’
100
Pitch (Hz)
70
Figure 14. ‘The mist covered her when she went down’
10. Conclusion
The proposal by Heine and Kuteva (2006), that interrogative pronouns can
expand their range of uses along a pathway from simple questions to indefinite
complements to definite complements to headless relatives to headed relatives,
helps us make sense of the patterns we find in a number of languages, among them
Tuscarora. The recurring matches we find between interrogative and relative pro-
nouns are no accident: they can result from recurring pathways of development.
Such developments are made possible by a fundamental semantic feature of the
Questionable relatives
arkers themselves: all are indefinite pronouns. The proposal by Heine and Kuteva
m
that progress along the pathway can be stimulated by language contact allows us
to account for the apparently accelerated development of the Tuscarora pronouns
within less than a century. In turn, Tuscarora provides especially robust support
for the Heine and Kuteva proposals. Within this language alone, it is possible to
document the development of all of the major interrogative pronouns words along
the same trajectory, step by step. The perfect coincidence of these Tuscarora devel-
opments with bilingualism in English adds evidence of the potential effect of con-
tact in stimulating such evolution.The recognition of this recurring pathway of
development provides some explanations, but it also raises some intriguing new
questions. We now know, for example, that individual pronouns do not all evolve
at the same rate within individual languages, and that they do not evolve in the
same order cross-linguistically. One future challenge could be to discover what
kinds of factors are necessary for such developments to take place, and, once they
are present, what additional factors might accelerate or retard them.
References