Relative Clauses in Languages of the Americas

Typological Studies in Language (TSL)

A companion series to the journal Studies in Language. Volumes in this
series are functionally and typologically oriented, covering specific topics in
language by collecting together data from a wide variety of languages and
language typologies.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/tsl

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

John Benjamins Publishing Company

Amsterdam / Philadelphia
 TM
The paper used in this publication meets the minimum requirements of
8

the American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Relative clauses in languages of the Americas : a typological overview / edited by

Bernard Comrie, Zarina Estrada-Fernández.
p. cm. (Typological Studies in Language, issn 0167-7373 ; v. 102)
Selected papers from Seminar on Linguistic Complexity held at the University of Sonora,
Hermosillo, Mexico.
Includes bibliographical references and index.
1. Grammar, Comparative and general--Relative clauses. 2. Complexity (Linguistics) 3.
America--Languages--Research--History. I. Comrie, Bernard, 1947- II. Estrada
Fernández, Zarina.
P297.R45   2012
497--dc23 2012022781
isbn 978 90 272 0683 1 (Hb ; alk. paper)
isbn 978 90 272 7339 0 (Eb)

© 2012 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
 Table of Contents

Map of Languages Mentioned vii

Introduction ix

part i. Diachrony, typology, and theory

Toward a diachronic typology of relative clause 3
T. Givón
The evolution of language and elaborateness of grammar: The case
of relative clauses in creole languages 27
Tania Kuteva & Bernard Comrie
Some issues in the linking between syntax and semantics in relative clauses 47
Robert D. Van Valin, Jr.

part ii. Uto-Aztecan

Relative clauses and nominalizations in Yaqui 67
Albert Álvarez González
On relative clauses and related constructions in Yaqui 97
Lilián Guerrero
From demonstrative to relative marker to clause linker: Relative clause
formation in Pima Bajo 127
Zarina Estrada-Fernández
Functional underpinnings of diachrony in relative clause formation:
The nominalization-relativization connection in Northern Paiute 147
Tim Thornes

part iii. Elsewhere in the Americas

Clauses as noun modifiers in Toba (Guaycuruan) 173
María Belén Carpio & Marisa Censabella
Between headed and headless relative clauses 191
Patience Epps
 Relative Clauses in Languages of the Americas

Relative clauses in Seri 213

Stephen A. Marlett
Relative clauses in Gavião of Rondônia 243
Denny Moore
Relative clauses in Yucatec Maya: Light heads vs. Null domain 253
Rodrigo Gutiérrez-Bravo
Questionable relatives 269
Marianne Mithun
Language and language family index 301
Name index 303
Subject index 305
 Introduction

Bernard Comrie & Zarina Estrada-Fernández

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 

Guerrero’s contribution on the same language emphasizes the differences

among relative clauses depending on the syntactic-semantic role being relativized,
including discussion of different degrees of nominalization in relative clauses. It
also places significant emphasis on the similarities and differences between rela-
tive clauses and other formally similar constructions, for instance nominalizations
used as complement clauses.
Pima Bajo, discussed by Estrada-Fernández, is interesting in that its rela-
tive clause structures involve morphological material that is innovative in mark-
ing relative clauses in comparison with other, even closely related Uto-Aztecan
languages. Relativizing subjects and direct objects involves nominalization, while
relativizing other syntactic-roles requires a paratactic structure. It would be inter-
esting to trace the development of these possibilities historically, perhaps involving
the loss of inherited possibilities and the development of new ones, in particular
new paratactic possibilities.
Finally within the Uto-Aztecan chapters, Thornes examines the structure of
relative clauses and their links to nominalization in Northern Paiute, including
consideration of the functional role of relative clauses within Northern Paiute dis-
course – he notes that they occur rarely, frequently without heads. (It would be
interesting to examine more broadly whether there is a correlation between fre-
quent headlessness of relative clauses and an origin in nominalizations.) He also
expands the discussion by considering in detail the interaction of relative clause
formation and valency-changing processes such as the passive voice.
The use of nominalizations to express relative clauses recurs in four of the
other contributions. Carpio and Censabella discuss three different devices for
coding noun modifying clauses, i.e. relative clauses and noun complements, on
the basis of a corpus of narrative texts in Toba, a Guaycuruan language spoken
in the Chaco area of Argentina. While one of these devices is specific to relative
clauses, the other two are also used with noun complements. It is shown that
these coding devices are selected according to the type of information encoded
in the dependent clause – pragmatic factors – rather than, in relative clauses, the
­syntactic function of the head noun within the dependent clause, which can only
be inferred from the absence of that argument from the clause. In addition, the
synchronic polyfunctionality of the subordinators is described.
Three further contributions discuss, in addition to nominalization, the ques-
tion of the headedness of relative clauses. Patience Epps is interested in showing
that the traditional division of relative clauses into two discrete categories – headed
vs. headless – does not work for Hup, a Nadahup (Makú) language of northwestern
Amazonia. She deals with a complex range of relative clauses constructions, all of
them nominalized and in apposition to a domain nominal. The various types not
only resist a division into two categories, but also call into q ­ uestion the ­positing
 Bernard Comrie & Zarina Estrada-Fernández

of a discrete intermediate category (of light-headed relatives). Rather, they form a

continuum from fully headed to fully headless. The discussion takes into consider-
ation not only the synchronic facts but also the diachronic evidence that supports
her analysis.
Marlett shows that Seri relative clauses have a number of unusual properties,
in addition to some that are more widespread in the Americas, such as nomi-
nalized internally-headed relative clauses. The language distinguishes three kinds
of such relative clauses depending on whether the element relativized is s­ ubject,
direct object, or neither of these, with subject-oriented relative clauses being
more frequent since they are the translation equivalents of a range of modifying
­constructions in other languages, including attributive adjectives. Potential ambi-
guity resulting from the head-internal nature of the construction, e.g. between ‘the
dog that the man saw’ and ‘the man that the dog saw’, is avoided by the require-
ment that the head lack an article. Seri relative clause constructions can involve
recursion, including center-embedding. Nonetheless, relative clauses are of low
frequency in the language, posing interesting questions for future research on how
speakers process them and how children acquire them.
Moore presents an elegant study of relative clauses in Gavião of Rondônia, a
Tupian language. In this language all relative clauses are nominalized construc-
tions, marked by one of two particles, both of which developed through the
grammaticalization of discourse pronouns. This construction can serve both as a
relative clause and as a complement clause, the latter also when there is an external
head (e.g. ‘the sign that I went fishing’). Relative clauses may have internal and
external heads, though either or both may be absent; since the internal head is not
marked explicitly as a head, interesting ambiguities arise, as between ‘the house of
the person that the snake bit’, ‘the house of the snake that bit the person’, and ‘the
house where the snake bit the person’.
Finally with respect to headedness of relative clauses, Gutiérrez-Bravo’s study
of Yucatec Mayan argues that apparently headless relative clauses, translation
equivalents of English sentences like they do not buy what they want, must in fact
be analyzed with a head, albeit phonologically null, since this head can host nomi-
nal modifiers such as determiners, quantifiers – and further relative clauses! The
whole issue of heads of relative clauses thus emerges as a fascinating topic for fur-
ther cross-linguistic investigation, perhaps particularly in languages where relative
clauses have their origin in nominalizations.
The article by Mithun on Tuscarora brings us back to the question of the ori-
gin of relative clauses other than through nominalization, though this time in the
case of a language in a heavy contact situation, namely with English. The kind of
relative clause frequent in English and many other European languages whereby
relative clauses are introduced by a pronoun identical in form to ­interrogative
 Introduction 

­ ronouns of the language is known to be rare ­cross-linguistically, virtually

p
restricted to European languages other than as the result of recent contact with
such European languages. The comparison of contemporary and nineteenth-­
century Tuscarora enables tracing of several of the details of the spread of the
­English construction into Tuscarora, but perhaps more importantly for the
spread of the construction within Tuscarora, including the observation that
while this use of the interrogative pronoun is found in contemporary (but not
nineteenth-century) Tuscarora headless relatives, it has not yet spread to headed
relative clauses. Even though the use of interrogative pronouns to introduce
relative clauses is a borrowing from English, its spread in Tuscarora has followed
established paths of grammaticalization rather than immediately filling the full
range where interrogative/relative syncretism is found in the donor language.
In this brief Introduction we have necessarily had to restrict ourselves to
­certain  major, and especially recurrent themes. The articles that you are about
to read also include a wealth of other descriptive, analytical, and methodological
riches. Before taking our leave of you, we would only draw your attention to the
increasing use of naturalistic data, perhaps particularly important in a domain
like relative clauses where speakers are often tempted to give rather literal transla-
tions in the absence of an obvious idiomatic equivalent; and, in Mithun’s contribu-
tion, of instrumental phonetic analysis to back up impressionistic accounts of the
­prosodic properties of boundaries between less and more tightly linked clauses.
part i

Diachrony, typology, and theory

 Toward a diachronic typology
of relative clause*

T. Givón
Linguistics Department University of Oregon
and White Cloud Ranch Ignacio, Colorado

A diachronic typology of relative clauses points toward two major pathways

towards embedded (restrictive) REL-clauses. The first starts from chained
(conjoined) clauses, the second from parenthetical non-restrictive REL‑clauses.
In both pathways, embedded (restrictive) REL-clauses start their life as paratactic
clauses under their own separate intonation contour. In both pathways, the first
step in creating a complex embedded construction is merger of the erstwhile
separate intonation contours of main and subordinate clause. Other dimensions
of REL-clause typology, such as nominalization, demonstrative pronouns, zero
anaphora, etc. intersect with these two diachronic pathways. The synchronic
syntactic properties of REL-clauses are to a large extent predictable from
their diachronic source, so that universals of grammar are fundamentally
developmental, expressing themselves through emergence.

Keywords:  REL-clauses; diachronic typology; universals of grammar

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

Syntaxis: The two clauses condensed under a single intonation

b. 
­contours
c. Lexis: The two verbs co-lexicalize into a single word.1

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

In my earlier work on the typology of REL-clauses (Givón 1990, 2001), I did

not distinguish sufficiently between the two major diachronic pathways of
­complex-clause genesis (2). While outlining an essentially diachronic typology
(7–8 main types), I focused on the source of the mechanisms used to recover the
case-role of the missing co-referent argument inside the REL-clause. This yielded
a fairly coherent synchronic typology, but in retrospect it seems that I could have
divided the 7–8 types into the two mega-types as in (2). In this paper I would like
to investigate the feasibility of this more comprehensive approach to the ­diachronic
typology of REL-clauses.

1.2  Reconstruction methods

A note is perhaps in order concerning the methodology most commonly used
in diachronic reconstruction of syntax. There are three useful reconstructive
­methods in historical syntax:

i. The study of historical records of contiguous developmental stages;

ii. The study of synchronic variation of co-existing related constructions;
iii. Doing internal reconstruction by studying surviving ‘relic’ clues.

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).

2.  From clause-chaining (‘conjunction’) to embedding

In earlier discussion, I labeled this pathway, found in serial-verb languages all

over the world, the “non-embedding strategy”. This was clearly an imprecise char-
acterization. More accurately, the early paratactic stage of this pathway involves
two (or more) clause in a chain, each under its own intonation contour. But in
almost all the serial-verb languages that employ this strategy, one already finds
the c­ o-existing syntactic (‘condensed’, ‘embedded’) variant, where the REL-clause
falls under the same intonation contour as the main clause. And in many cases no
­re-ordering or restructuring is done − beyond the change of intonation.
I will illustrate this diachronic route to embedded REL-clauses first with
­examples from Bambara (Mandeic; Niger-Congo). The data is originally from
Charles Bird (1968) and Ibrahima Coulibaly (i.p.c.). Consider first the ­paratactic,
un-embedded variants, where the demonstrative min ‘that’ modifies the ­co-referent
noun inside the would-be REL-clause (3a–e). One could consider min now the
REL-clause maker, but it is still used in the language as a demonstrative modifier
or demonstrative pronoun, and its position in the clause is still compatible with
the original use (tone markings not shown):
(3) a. Unembedded, pre-posed (subj-rel, obj-main):
cε min ye muru san, n ye. o ye
man rel past knife buy I past him see
‘The man who bought the knife, I saw him’.
b. Unembedded, post-posed (subj-rel, obj-main):
n ye o ye, cε min ye muru san.
I past him see man rel past knife buy
‘I saw him, the man who bought the knife’.
(Hist.: ‘I saw him, that man bought the knife’.)
c. Unembedded, pre-posed (obj-rel, obj-main):
n ye so min ye, cε be o dyɔ.
I past house rel see man prog it build
‘The house that I saw, the man is building it’.
(Hist.: ‘I saw that house, the man is building it’.)
 T. Givón

d. Unembedded, post-posed (obj-rel, obj-main):

cε be o dyɔ, n ye so min ye.
man prog it build I past house rel see
‘The man is building it, the house that I saw’.
(Hist.: ‘The man is building it, I saw that house’.)
e. Unembedded, extraposed:
cε ye muru san, n ye min ye.
man past knife buy I past rel see
‘The man bought the knife, that one I saw’.
(‘Hist.: ‘The man bought the knife, I saw that one (the knife)’.)
No re-ordering of elements occurs in such un-embedded ‘REL-clauses. Both the
anaphoric pronoun o (‘s/he’, ‘it’) and the demonstrative min (‘that’) are used the way
they are used in normal clause-chaining in discourse. But Bambara can also place
both clauses under a joint intonation contour, in a configuration that is clearly an
early form of embedding. This relativization strategy is much less common (Bird
1968). It involves placing the entire ‘relative’ clause at the location inside the main
clause where the head-noun should have been (Bird 1968):
(4) a. Simple (main) clause:
n ye cε ye.
I past man see
‘I saw the man’.
b. With rel-clause:
n ye cε min [Ø] ye muru san ye.
I past man rel past knife buy see
‘I saw the man who bought the knife’.
(Hist.: ‘I [,] that man bought the knife [,] saw’.)
Finally, in some configurations, and with the anaphoric pronoun omitted under
the merged intonation contour, the old chained structure looks more and more
like a truly embedded one. The transition from (5b) to (5c) below involves no
re-ordering, just merging of the intonation contours and dropping the anaphoric
pronoun (Bird 1968):
(5) a. Simple (main) clause:
cε ye muru san.
man past knife buy
‘The man bought the knife’.
b. Chained (paratactic) configuration:
n ye cε min ye, o ye muru san.
I past man rel see he past knife buy
‘The man that I saw, he bought the knife’.
(Hist.: ‘I saw that man, he bought the knife’.)
 Toward a diachronic typology of relative clause 

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’.)

A similar pattern of relativization is found in Supyire (Senufu; Niger-Congo),

another clause-chaining, serial-verb language (Carlson 1994). An erstwhile
demonstrative pronoun has become the generalized REL-clause suffix, while a
full-size demonstrative pronoun is often use inside the REL-clause, as in Bambara.
Thus (with tone-marking not shown):

(6) a. Nami ɳ ge-mu u a pa-ge, mii a mye

man dem-rel he perf come-rel I perf see
‘The man who came, I saw (him)’
b. Nami ɳ ge-mu mii a ɳ ye-ge, u a pa
man dem-rel I perf see-rel he perf come
‘The man I saw, he came’

One REL-clause type, which Carlson (1994; pp. 513–514) calls “semi-embedding”,
represents the beginning of a syntactic, embedded, pattern:

(7) a. Ka pi i bage e u a, titige-ke

and they narr house.def in he perf descend-rel
‘Then they the house in which he had descended,
b. ka pi i kuru cyee mii na
and they narr that show me to
then they showed that-one to me’.

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

(8) a. Main clause:

otoka-ga onna-ni tegami-o kaita
man-subj woman-dat letter-acc sent
‘The man sent a letter to the woman’.
b. Syntactic: subject rel-clause:
onna-ni tegami-o kaita otoka-wa Kobe-ni ikimashita
woman-dat letter-acc sent man-top Kobe-to went
‘The man who sent a letter to the woman went to Kobe’.
c. Paratactic version:
onna-ni tegami-o kaita, otoka-wa Kobe-ni ikimashita
woman-dat letter-acc sent man-top Kobe-to went
‘He sent a letter to the woman, the man (who) went to Kobe’.
d. Syntactic: object rel-clause:
otoka-ga onna-ni kaita tegami-wa Kobe-ni ikimashita
man-subj woman-dat sent letter-top Kobe-to went
‘The letter that the man sent to the woman went to Kobe’.
e. Paratactic version:
otoka-ga onna-ni kaita, tegami-wa Kobe-ni ikimashita
man-subj woman-dat sent letter-top Kobe-to went
‘The man sent it to the woman, the letter (that) went to Kobe’.
f. Syntactic: dative rel-clause:
otoka-ga tegami-o kaita onna-wa Kobe-ni ikimashita
man-subj letter-acc sent woman-top Kobe-to went
‘The woman to whom the man sent a letter went to Kobe’.
g. Paratactic version:
otoka-ga tegami-o kaita, onna-wa Kobe-ni ikimashita
man-subj letter-acc sent woman-top Kobe-to went
‘The man sent her a letter, the woman (who) went to Kobe’.

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 

indeed, such supra-segmental changes are probably universal, transcending

typological boundaries.
The last language I’ll cite as a representative of this clause-chaining pathway,
Hittite, has been dead for millennia. But essentially the same Bambara-type rela-
tivization strategy has been shown to exist here (Justus 1973, 1976). The examples
below are taken from Robert (2006), where both paratactic clauses are marked
with conjunctions:
(9) a. nu ku-it LUGAL-uš teez-zi, nu apaa-at luzzi karap-zi
conj rel-acc king-nom say-3sg conj that-acc luzzi do-2sg
‘Whatever the king says, that the luzzi shall perform’.
b. ku-u-ša-ta-ma ku-it píddaa-i, na-aš-kan šameen-zi
bride-price-ptc-conj rel-acc give-3sg conj-he-ptc forfeit-3sg
‘What (ever) bride-price he gave, he forfeits (it)’.

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’.

And the second conjunctions may also be dropped:

(11) pa-apreez-zi ku-iš, 3 gín ku.babbar paa-i
be-impure-3sg rel-nom three shekel silver pay-3sg
‘The one who is impure, (he) pays three silver shekels’.
The case-marked ku- inside the REL clause is quite analogous to the Bambara min
and was probably a demonstrative determiner, used naturally as a demonstrative
pronoun (see discussion in Justus 1976, as well as the German data further below).
One pre-posed paratactic REL-clause may be followed by more than one main
(‘resumptive’) clause, in a typical clause-chaining pattern (Robert 2006):

‘­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

(12) a. lu-meš Ubaru, lu-aš ku-iš lugal-wa-aš

men-nom Ubaru man-nom rel-nom king-gen
pé-ra-an eeš-zi,
in-front be-3sg
‘Men of Ubaru, whatever man that is in front of the kind,
b. ne šaraa tie-enzi,
conj upward step-3pl
they step forward,
c. nu aappa tie-nzi,
conj backward step-3pl
then (they) step backward,
d. ne araanda.
conj stand.3pl
and then (they) stand’.

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):

(13) a. gan moi-n-g-u-a,

child be-he-nom-him-perf
‘The child (who) is here
b. Gomia tei awi d-i.
Gomia there send leave-impf
send (him) to Gomia’.

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:

...The distinction between sentences with both [conjunctions] and sentences

with neither points to a structural distinction between adjoined [paratactic] and
embedded [syntactic] relative clauses. After Old Hittite, it is no longer necessary
for the resumptive [main] clause to include either both resumption [explicit
anaphoric pronouns] and conjunction... (2006, p. 17).

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):

(14) MOT: They both are wearing earrings

And what else is this dolly wearing?
NIN: A blouse like that one.
Louise gave me that one. (p. 42, Nina-III transcripts)

The use of the demonstrative pronoun (‘that one’) by Nina is reminiscent of

­Bambara and Hittite. The communicative goal, given clearly in the context and
negotiated over successive-adult-child turns, is that of identifying a referent by
citing an event in which it was a participant, a standard communicative motiva-
tion for using restrictive REL-clauses. But the construction is spread paratactically
over two adjacent turns and three intonational clauses. In tightly-edited written
English, a single person would have restored the ellipsis and merged the mother’s
and daughter’s contributions into:
(15) She is wearing a blouse like the one Louise gave me.

3.  From parenthetical to non-restrictive to embedded REL-clauses

The second parataxis-to-syntaxis pathway to embedded REL-clause is clearly illus-

trated in the extant synchronic variation in present-day German. In this language
an older generation of cumbersome nominalized REL-clauses has been replaced
by a new one. The genesis of the new embedded structure involves the following
steps:
 T. Givón

i. Recruit the still-extant Y-movement construction with case-marked stressed

demonstrative pronoun.
ii. Insert it post nominally after a head noun, with an intonation break, thus
­creating a non-restrictive REL-clause.
iii. Merge the intonation contours and de-stress the demonstrative, thus arriving
at a restrictive REL-clause.

As illustration, consider (Theo Vennemann, Charlotte Zahn & Christa Toedter,

i.p.c.; see also Borgert & Nyhan 1976):
(16) a. Simple clause:
Martin hat dem Mann das Buch gegeben
M. has the/dat man the/acc book given
‘Martin gave the book to the man’.
b. Y-movement clause, nom:
DER hat das Buch dem Mann gegeben
THAT/nom has the/acc book the/dat man given
‘THAT one gave the book to the man’.
c. Y-movement clause, acc:
DAS hat Martin dem Mann gegeben
THAT/acc has Martin the/dat man given
‘THAT one Martin gave to the Man’.
d. Y-movement-dat:
dem hat Martin das Buch gegeben
THAT/dat has Martin the/acc book given
‘To THAT one Martin gave the book’.
(17) Non-restrictive (parenthetical) rel-clauses:
a. Nominative:
Ich kenne die Frau, DIE hat dem.
I know the woman, THAT/nom has the/dat
Mann das Buch gegeben
man the/acc book given
‘I know the woman, the one gave the book to the man’
(Hist.: ‘I know the woman. THAT one gave the book to the man’).
b. Accusative:
Ich kenne das Buch, DAS hat Martin dem
I know the book, THAT/acc has Martin the/dat
Mann gegeben
man given
I know the book, the one Martin gave to the man’.
(Hist.: ‘I know the book. THAT one Martin gave to the man’).
 Toward a diachronic typology of relative clause 

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):

(18) Restrictive rel-clauses

a. Nominative:
Ich kenne die Frau die dem.
I know the woman that/nom the/dat
Mann das Buch gegeb]en hat
man the/acc book given has
‘I know the woman who gave the book to the man’.
b. Accusative:
Ich kenne das Buch das Martin dem.
I know the book that/acc Martin the/dat
Mann gegeben hat
man given has
‘I know the book that Martin gave to the man’.
c. Dative:
Ich kenne den Mann dem Martin
I know the/acc man that/dat Martin
das Buch gegeben hat
the/acc book given has
‘I know the man to whom Martin gave the book’.

Essentially the same pathway is described in other Germanic languages, such

as Old Norse and Old English (Heine & Kuteva 2007). In other languages, this
­pattern in whole or part may be used to augment an existing REL-clause structure.
Thus for example, in spoken informal Israeli Hebrew, one finds the following use
of demonstrative-marked headless REL-clauses invading the regular REL-clause
paradigm:
 T. Givón

(19) a. Standard rel-clause structure:

Ha-’ish she-pagash-ti ’oto ’etmol...
the-man rel-met-1sg him yesterday
‘the man I met yesterday...’
b. Standard headless rel-clause:
zé she-pagash-ti ’oto ’etmol
dem rel-met-1sg him yesterday
‘the one I met yesterday...’
c. Standard non-restrictive rel-clause:
ha-’ish, zé she-pagash-ti ’oti ’etmol,....
the-man dem rel-met-1sg him yesterday
‘the man, the one I met yesterday...’
d. Non-standard condensation to restrictive rel-clause:
ha’ish ze-she-pagash-ti ’oto ’etmol....
the-man dem-rel-met-1sg him yesterday
‘the man I met yesterday...’

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:

i. The co-referent element has to be anaphoric.

ii. The co-referent element has to be contrastive/emphatic.

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 

4.  Making sense of nominalized REL-clauses

In many language families – Turkic, Caribbean, Bodic (Tibeto-Burman), No.

­Uto-Aztecan, Sumerian, to cite but a few – all subordinate clauses are n
­ ominalized,
at least historically. Such structures may re-acquire finite properties over time
(Givón 1994; Watters 1998), but the morphology retains, for a long time, the tell-
tale marks – clear fossil evidence – of earlier nominalized structure. The question
is: can the process that created such structure to begin with be shown to conform
to our parataxis-to-syntaxis prediction?
I will illustrate how such a pathway can be reconstructed by citing the
nominalized REL-clauses of Ute (Numic, Uto-Aztecan). Ute marks consis-
tently the case-distinction subject (nominative) vs. non-subject (object, geni-
tive, oblique). The verb in subject REL-clauses is marked with the subject
nominalizing ­suffix  -tu. The verb in object REL-clauses is marked with the
non-subject nominalizing suffix -na, and the subject then takes the genitive
case. In i­ndirect-object REL-clauses, the subordinator pu- carries the relevant
­post-position. Thus (Givón 2011):

(20) a. Main clause:

Mamachi tupuych-i tuka’na-pu-vwan wacu-ka
woman/s rock-o table/o-on put-perf
‘The woman put the rock on the table’.
b. Restrictive rel-clause–subj
mamachi ’u tupuych-i tuka’napu-vwan
woman/s the/s rock-o table/o-on
wacu-ka-t(u)…
put-perf-nom
‘the woman who put the rock on the table’
(Hist.: ‘the woman putter of rock on the table...’)
c. Non-restrictive rel-clause: subj:
mamachi ’u, (’ú) tupuych-i tuka’na-pu-vwan wacu-ka-t(u)...
woman/s the/s (that/s) rock-o table/o-on put-perf-nom
‘the woman, (that one) who put the rock on the table...’
(Hist.: ‘the woman, (that) putter of rock on the table...’)
d. Restrictive rel-clause:obj:
tupuy-chi uru mamachi tuka’napu-vwan wacu-ka-n(a)...
rock/s the/s woman-gen table/o-on put-perf-nom
‘the rock that the woman put on the table...’
(Hist.: ‘the rock of the woman’s putting on the table...’)
 T. Givón

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’).

Just as in German or Hebrew, all it takes in Ute to move from non-restrictive

(paratactic) to restrictive (syntactic/embedded) REL-clause is the merger of into-
nation contours. We have just subsumed the nominalized REL-clause pattern, at
least in principle, under our second parataxis-to-syntaxis diachronic pathway –
from non-restrictive to restrictive REL-clause.
In a nominalizing language, the etymological source of the nominalizers may
shed some light on the evolution of nominalized REL-clauses. In general, nominal-
izers are most commonly derived from head nouns in genitive noun compounds.
This is, for example, the etymology of the nominalizing suffix in English ‘free-dom’
(Old High German tuom ‘judgement’) or ‘child-hood’ (Gothic haidus ‘quality’).
In Lhasa Tibetan, four nominalizing suffixes are used in relativization: -mkhan
is used for actor; -sa for the locative (and, by extension, dative-benefactive); -yag
(the ‘default’ suffix) for patient and instrumental (in the non-perfective), and -pa
for non-actor or patient (in the perfective).
In lexical nominalizations in Tibetan, -mkhan (historically ‘teacher’ or
‘expert’) is used in many actor derivations, and -sa (historically ‘earth’, ‘ground’,
and by extension ‘place’) in locative derivations. Thus (DeLancey 1988):

(23) a. Actor/subject derivation:

s’i’n-mkhan ‘carpenter’
wood-expert
 T. Givón

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):

(24) a. Actor rel-clause:

stag gsod-mkhan mi...
tiger kill-nom man
‘the man who killed the tiger...’
(Hist.: ‘the tiger kill-expert man...’)
b. Locative rel-clause:
kho sdod-sa-’i khan’.pa...
he/abs live-nom-gen house
‘the house where he lives...’
(Hist.: ‘his living-place house...’)
c. Instrument rel-clause:
kho-s stag gsod-yag-gi mem.da...
he-erg tiger kill-nom-gen gun
‘the gun with which he killed the tiger...’
(Hist.: ‘his tiger killing-tool gun...’/)
d. Patient rel-clause:
kho-s bsad-pa-’i stag...
he-erg kill-nom-gen tiger
‘the tiger that he killed...’
(hist.: ‘his killing-victim(?) tiger...’)

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 

5.  A note on word-order

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):

(25) a. Subject nominalization:

mài qìche de dàbàn dou shì hâo rén
sell car nom majority all be good person
‘Car sellers are mostly good people’.
b. Object nominalization:
mài gêi Lîsì de shì zuì guì de
sell to L. nom be most expensive nom
‘What is sold to Lisi is the most expensive’.
c. Subject rel-clause:
jintian yíng de rén yùnqì hâo
today win nom person luck good
‘The people who won today had good luck’
d. Object rel-clause:
jintian yíng de qián fù fáng-zu
today win nom money pay house-rent
‘The money (we) won today goes to pay the house rent’.

Mandarin displays the same zero-anaphora case-recoverability strategy as ­Japanese,

whose REL-clauses were also historically nominalized (M. Shibatani, i.p.c.). One
may thus suggest that Mandarin restrictive REL-clauses are ­pre-nominal because
they arise from the same clause-chaining paratactic source suggested above for
Japanese and Tibetan.
There is another type of Rel-clause in Mandarin, Li and Thompson’s (1981)
“descriptive” clause, that is post-nominal, and is used in presentative construc-
tions with REF-indefinite head noun. Its origin from clause-chaining is transpar-
ent, involving the merger of the two intonation contours:

(26) a. Paratactic clause-chain source:

wo you yi-ge meimei, xihuan kan dianyin
I have one-cl sister like see movie
‘I have a sister, [she] likes to see to see movies’.
 T. Givón

b. Syntactic presentative with post-nominal rel-clause:

wo you yi-ge meimei xihuan kan dianyin
I have one-cl sister like see movie
‘I have a sister who likes to see to see movies’.

Finally, non-restrictive REL-clauses in Mandarin are indistinguishable from

chained clauses. Given that both are equally asserted (rather than presupposed),
and given the zero-anaphora of Mandarin, this is only to be expected.
One may now suggest that nominalized REL-clauses in Ute, a fairly classical
SOV language till recently, are post-nominal not in violation of any Greenbergian
universals, but because they arose through the non-restrictive paratactic chan-
nel. While nominalized REL-clauses in Tibetan and Mandarin are pre-nominal
because they arose through the clause-chaining channel. Nominalization as a
case-recoverability strategy does not correlate, but rather intersects freely, with
either major paratactic source of embedded REL-clauses.
A by-product of this discussion, I think, is that the oft-cited Greenbergian
correlation between SOV word-order and pre-nominal REL-clauses is probably
a typological accident, due to the high correlation between the SOV order and
clause-chaining. And this correlation is, in turn, due to the high synchronic – and
most likely also diachronic (Givón 1979b) – prevalence of the SOV word-order; as
well as to the high prevalence – often unremarked – of clause-chaining.

6.  Cleft and WH-questions: From parataxis to syntaxis

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):

(27) a. Main clause:

Kipes ka-swiimin kit
K. 3sg-buy-past chair
‘Kipes bought a chair’.
b. Restrictive rel-clause:
kit ki a-swiim-in Kipes...
chair dem 3sg/rel-buy-past K.
‘the chair that Kipes bought...’
 Toward a diachronic typology of relative clause 

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?’

Presumably, Kihungan already had a restrictive REL-clause construction before

recruiting it to fashion cleft and WH-question constructions. In a way, however,
the parataxis-to-syntaxis trajectory of the latter two recapitulates the presumed
diachronic trajectory of REL-clauses.

7.  Some tentative conclusions

Of the 7–8 major relativization strategies considered earlier (Givón 2001), I have
suggested here that at least four:

i. The non-embedding strategy (Bambara, Supyire, Hittite)

ii. The zero-anaphora strategy (Japanese, Mandarin)
iii. The case-marked demonstrative-pronoun strategy (German)
iv. The nominalization strategy (Ute, Tibetan)

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:

i. From paratactic to syntactic complexity.

ii. From syntactic to lexical complexity.

In the heydays of Generative Semantics, and before Shibatani’s (1972) paper on

the semantics of causative constructions, both packaging steps were considered
trivial, a matter of mere surface structure. Causative clause-union was the prime
example cited by proponents of GS:
(28) a. Paratactic: She let him, and he went.
b. Syntactic: She let him go.
c. Co-lexicalized: She let-go of him.

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 

condensation or merely an unintended consequence remains to be seen.4 Still,

the rise of hierarchic structure is, in general, part of the ­mechanism of increased
automaticity.

Abbreviations

acc accusative past past

ant anterior pass passive
cl classifier perf perfective
conj conjunction pl plural
dat dative pred predicative
def definite pres present
dem demonstrative prog progressive
erg ergative ptc participle
gen genitive prt preterite
hab habitual rel relativizer
impf imperfective s subject
narr narrative mode sg singular
nom nominative subj subject
o onbject top topic.
obj object

References

Bird, C. 1968. Relative clauses in Bambara. J. of West African Languages 5.

Borgert, U. H. G. & C. A. Nyhan. 1976. A German Reference Grammar. Sydney: Sydney
­University Press.
Carlson, R. 1994. A Grammar of Supyire. Berlin: Mouton de Gruyter.
Delancey, S. 1988. Relativization and nominalization in Tibetan and Newari. Ms, University of
Oregon at Eugene.
Diessel, H. 2005. The Acquisition of Complex Sentences. Cambridge: CUP.
Everett, D. 2005. Cultural constraints on grammar and cognition in Piraha. Current ­Anthropology
46(4): 621–634.
Gildea, S. (ed.). 2000. Reconstruction Grammar: Comparative Linguistics and G
­ rammaticalization
[Typological Studies in Language 43]. Amsterdam: John Benjamins.

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*

Tania Kuteva & Bernard Comrie

University of Düsseldorf and University of London / Max Planck Institute
for Evolutionary Anthropology and University of California Santa Barbara

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.

Keywords:  Relative clause; creole languages; language contact; language

evolution

1.  Introduction

A possible assumption about language evolution is that language structures were

simple before they became complex. Thus Comrie (1992: 205) argues that certain

*  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 

to language evolution makes regarding the simple-before-complex grammar of

­language against data from creole languages. For this purpose, we investigate one
particular construction in creoles which McWhorter (2001) leaves out of consid-
eration, the relative clause construction.
In this paper, we restrict ourselves to relativization upon the subject: the data
on subject relativization is most readily accessible from existing descriptions,
therefore it makes a good starting point for our investigation.
On the basis of a sample of all creole languages for which we were able to
find conclusive data on subject relativization, we show that the above prediction
with regard to elaborateness in marking of the relative clause construction is
met: creole languages regularly have simply marked relative clause construc-
tions, encoded by no more than one relativization marker. This typological
result stands out as particularly significant if we view it against the background
of n
­ on-creole languages, for which we have been able to identify cases with two
or three or four or even five relativization markers (Comrie & Kuteva 2005;
Kuteva & Comrie 2005).
The typological fact about creoles consistently marking the relative clause
construction by one marker only is rather puzzling in the light of recent fi ­ ndings
in contact linguistics. Thus recent research indicates that contrary to traditional
assumptions (Trudgill 1983; Gumperz & Wilson 1971), language contact often
brings about diversification and complexification of grammar (Heath 1978;
Aikhenvald 2002; Heine & Kuteva 2005; Kuteva 2008). Our own findings based on
a sample of high-contact, non-creole language varieties – which we examine for
elaborateness of marking for the relative clause construction – supports the thesis
that one result of language contact involves increase in elaborateness and redun-
dancy of marking: 50% of our sample varieties exhibit more than one relativization
marker. One would expect then that such extreme cases of language contact as
creoles would also involve elaborateness of marking.
Assuming that:
a. a high degree of language contact often leads to increase in elaborateness of
marking of grammatical categories, and;
b. creoles result from extreme language contact
in this study we propose a functional explanation of the present finding about the
lack of elaborateness of marking for the relative construction in creoles.
The paper is organized as follows. In Section 2, a brief overview is offered of
how non-creole languages mark the relative clause construction. In Section 3, we
present the main results of our study, the degree of elaborateness of ­relativization
marking in creoles. Section 4 will then discuss some recent findings of ­contact
linguistics relevant to marking of grammatical categories as well as the result
 Tania Kuteva & Bernard Comrie

of our own study of marking of the relative clause construction in high-contact

­(non-creole) language varieties. Finally, in Section 5, we offer a functional expla-
nation of the present findings.

2.  Encoding the relative clause construction in the languages of the world

On the basis of a genetically and geographically balanced sample representative

of the languages of the world, we have shown elsewhere (Comrie & Kuteva 2005)
that languages may be ordered along a continuum depending on the number of
­elements they employ for marking their relative construction. In Comrie and
Kuteva (2005), this continuum was termed the elaborateness of expression scale.
At one end of the elaborateness scale, there are languages which have no overt
marker of relativization at all. Maale (Omotic, Afroasiatic, spoken in Ethiopia)
illustrates this situation:
(1) Maale (Azeb Amha 2001: 160)
ʔííní [[ziginó mukk-é] ʔatsi] za-é-ne.
3ms:nom   yesterday come-pf person:m:abs see-pf-a:dcl
‘He saw the man who came yesterday.’

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:

i. a relative conjunction/determiner (varying for number and nominal class)

ii. a complementizer marker -bah
iii. pronoun-retention
iv. a verbal suffix-ne (a multipurpose marker for topicalization, nominalization
and relativization)
v. a sentential definitizer -la (related to the determiner system), see Chumbow
(1977: 296–297, 302):
(2) Ngemba (Chumbow 1977: 290)
nyung wá bah a-keshung-ne mung wa la a kung atsang.
man rel bah he-tns.beat-ne child det la he enter into prison
‘The man who beat the child went to prison.’

The complementizer -bah is optional, the other four relativization markers are
obligatory, however.
 The evolution of language and elaborateness of grammar 

Between the two end-points of the elaborateness of expression scale there

are languages which exhibit either one, or two, or three, or four markers of
­relativization (cf. Comrie & Kuteva 2005; Kuteva & Comrie 2005 for details).

3.  How do creoles mark relativization?

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.’

At a closer look, however, it becomes clear that the subject pronoun em is

­optionally used after the subject in any main clause, too, depending on the degree
of h
­ ighlighting intended by the speaker. This fact then “neutralizes” the use of
the subject pronoun after the relativization marker in the relative clause because
 The evolution of language and elaborateness of grammar 

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):

(10) Tayo (Corne 1994: 289)

fiy- la sa la okipe de nu.
girl- art rel 3sg care.for prep 1pl
‘The girl who looked after us.’
 Tania Kuteva & Bernard Comrie

(11) Tayo (Corne 1994: 287)

Chya- la la rule par ter.
dog- art 3sg roll prep ground
‘The dog is rolling on the ground.’
One of our sample creoles, Negerhollands – a Dutch-based creole, which was
­spoken until 1987 in what is now the US Virgin Islands (Hinskens & van Rossem
1995: 63) – presents a very interesting case. Based on a broad empirical survey of
three eighteenth-century Negerhollands translations of the same text, a so-called
“Gospel Harmony” (the oldest version written before 1780, the other two around
1795), Hinskens and van Rossem identify three possible ways to mark relative
clauses in Negerhollands. The first of these does not differ from what we regularly
find in other creoles as well: a single marker, die – which Hinskens and van ­Rossem
identify as “the relative pronoun” – encodes the relative clause construction:
(12) Negerhollands (Hinskens & van Rossem 1995: 74)
Voordoor een Skrik a vall op em, en op allemaal,
for a fright pa fall onto him, and onto all,
die a wees met em, over deese VischVang.
who pa be with him, about this catch
‘For he, and those who were with him, were terrified by this catch.’
(i.e. the miraculous draught of fishes).
Second, we have the relative pronoun die followed by a multifunctional word,
sender, which derives from the third person plural subject pronoun and which
is convincingly shown to serve not only as an object/oblique pronoun, a reflexive
pronoun, a possessive pronoun, but also as a plural marker on both nouns and
relative pronouns:
(13) Negerhollands (Hinskens & van Rossem 1995: 68)
En as Herodes a kik, dat em a wees bedrogen van die
And as Herodes pa see that he pa be cheated by the
Wies Mann sender da em a kom moeschi quaat, en a
wise man plur then he pa become very angry and pa
stier ut, en a lat mattaan allemaal die a wees Bargad
send out and pa let murder all who pa be border
met sender, die sender a wees tweejaar
with them(obj/obl) them who(rel+pl) pa be two year
oud en minder.
old and less
‘And when Herodes saw that he had been cheated by the wise men, he got
very angry and [he] ordered to kill all [children] younger than two years old
from Bethlehem and the surrounding areas.
 The evolution of language and elaborateness of grammar 

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…’):

(14) Negerhollands (Hinskens & van Rossem 1995: 69)

Die Samaritanse Vrow a see na em: hoesoo
The Samaritan woman pa say to him: why[is it that]
joe, die joe ben een Hoodio, bed Drink van mi,
you(sg) who you(sg) are a Jew, ask drink from me,
die mi ben een Samaritans Vrow?
who I am a Samaritan woman
‘The Samaritan woman said to him: “Why do you, as a Jew, ask a drink from
me, a Samaritan woman?”’

Examples like these testify to double marking of the relative clause construction.
Note, however, that:

i. Hinskens and van Rossem (1995: 69–71) trace this kind of construction to

German,2 which was the mother tongue of most Moravian Brethren, the
­missionaries who translated the texts investigated by Hinskens and van
­Rossem, and that this construction is especially typical of “more elevated
styles” as the following example from present-day German shows:

(15) German (Hinskens & van Rossem 1995: 71)

Und ich, der ich  
selber Jude bin, bin verpflichtet
And I, who I [my]self Jew am, am obliged
die Wahrheit zu sagen.
the truth to say
‘And I, who am myself a Jew, am obliged to tell the truth.’

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 regularity with which creoles exhibit lack of elaborateness of r­ elativization

marking stands out particularly clearly if we recall the fact that non-creole
­languages exhibit cases of two, three, four and even five relativization markers (see
Section 2).

4.  Simplification/complexification and language contact

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

For the purposes of the present study, we examined 12 high-contact

­language varieties for the marking of the relative clause construction. This –
again  – is a convenience sample since it includes only varieties for which we
were able to find detailed grammatical description, see Appendix 2. In 6 of
our sample v­ arieties, i.e. in 50% of our sample languages, it is possible to iden-
tify more than one marker of the relative clause construction. One of these
­varieties encodes subject r­ elativization by means of zero, or one, or three mark-
ers; five of them by means of two markers. Let us illustrate the latter variet-
ies (with two ­markers) with an example from the Saraswat Brahmins Konkani
variety (­Nadkarni 1975: 680; Heine  & Kuteva 2005: 128–129). Konkani is an
­Indo-Aryan language. The ­Saraswat ­Brahmins Konkani variety is in close
­contact with Kannada, a ­Dravidian language. In this contact situation, the
model l­anguage ­Kannada (see (19b) below) has two relative clause markers
(the q ­ uestion word yāva ‘which?’, and the polar question marker ō in the rela-
tive clause of a ­correlative ­construction). ­Konkani, on the other hand, has an
already existing relative c­onstruction which is only marked by one r­ elativization
marker, the relative pronoun jo ­(Nadkarni 1975: 678). In the h ­ igh-contact vari-
ety, ­Saraswat ­Brahmins Konkani (see (19a) below), ­however, the new relative
­construction involves more markers than the old one; in fact, the Saraswat
­Brahmins ­Konkani speakers are using an isomorphic relative ­construction
whereby instead of the Kannada question word, yāva, they are using their own
question word, khanco, and instead of the Kannada polar question marker ō,
they are using the ­Konkani polar question marker ki, see also Heine and Kuteva
(2005: 128–129). In other words, the speakers of Saraswat Brahmins K ­ onkani
have increased the marking of their relative construction due to the c­ontact
with Kannada:

(19) Saraswat Brahmins Konkani and Kannada

 (Nadkarni 1975: 674–675; Heine & Kuteva 2005: 128–129);
a. [khanco mhāntāro pepar vāccat āssa- ki] to Dāktaru āssa.
b. [yāva mudukanu pēpar ōdutta iddān- ō] avanu Dāktaranu iddāne.
 which old.man paper reading is q that doctor is
‘The old man who is reading a newspaper is a doctor.’

Another example is Tariana of northwest Brazil, which employs the rela-

tivization markers of the two languages involved in this particular language
contact situation, the North Arawak language Tariana, on the one hand, and
the ­Indo-European ­language Portuguese, on the other. Portuguese, being the
official language of B
­ razil, has influenced Tariana of northwest Brazil in vari-
ous ways (Aikhenvald 2002). One of the results of this influence is that young
and innovative T ­ ariana speakers have started to also use their interrogative
 The evolution of language and elaborateness of grammar 

­ ronoun as a marker of the relative clause construction on the pattern of the

p
model ­language Portuguese:
(20) Tariana (Aikhenvald 2002: 183)
ka- yeka- kani hĩ kayu- na na- sape.
rel- know- past.rel.pl dem.anim thus- remp.vis 3pl- speak
‘Those who knew used to talk like this.’
The new relativization marker has not replaced, however, the existing Tariana
relative construction: What younger Tariana speakers do is simply add their rela-
tive pronoun (e.g. kwana ‘who?’) to the already existing Tariana relative clause
­construction, the result being a more elaborate marking of relativization:
(21) Younger Tariana speakers (Aikhenvald 2002: 183)
kwana ka- yeka- kani hĩ kayu- na
who rel- know- past.rel.pl dem.anim thus- remp.vis
na- sape.
3pl- speak
‘Those who knew used to talk like this.’
In other words, the contact between Tariana and Portuguese has led to increase in
the phonological material used to encode the relative clause construction in Tariana.
The fact that high-contact varieties may exhibit more elaborate m ­ arking of sub-
ject relativization is not particularly interesting if viewed against the b
­ ackground of
the sample representative of the languages of the world, since as m ­ entioned already,
some languages exhibit not only two but three, or four or even five markers of the
relative clause construction. However, if we compare the high-contact varieties in
our sample to the respective non-contact varieties involved in each particular lan-
guage contact situation, the result is extremely interesting: each high-contact vari-
ety exhibits elaborateness of marking either equal to (that is, the same number of
relativization markers as) or greater than (that is, a greater number of relativization
markers than) in one or both of the respective non-contact varieties.
In other words, our own results about the elaborateness of marking of the rela-
tive clause construction in high-contact language varieties supports the thesis –
articulated in recent research – that language contact may well involve increase in
elaborateness of marking.

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

of language contact. In a recent survey of current theories of creole formation,

Migge (2003: 23, 130) observes that creole formation has been ­compared not only
to borrowing (Myers-Scotton 1997) and interference through shift (Thomason &
Kaufmann 1988; Mufwene 1990) but also to convergence in situations involving
language maintenance (Gumperz & Wilson 1971). One might then expect that
creoles would also exhibit elaborateness of marking. As we showed in Section 3,
however, with respect to one particular structure at least, the relative clause con-
struction, the phonological material for the marking of this grammatical structure
in creole languages is manifested in no more – and most of the time, no less – than
one marker. How is this fact to be explained?
In the present study, we propose a functional explanation for this. We assume
that what is functional involves a simple, iconic 1-to-1 form:meaning pairing,
and starting from this assumption, we propose that creoles show the simplest
and most functional structure of the relative clause because their grammars are
recent creations, which have not had enough time to produce “junk”/redundancy/
elaborateness.
We are aware that one possible objection to our analysis may be that creoles
usually start out with no relativization marker at all, and it is only later that they
develop one. One may even go further and claim that the zero marker – in the
incipient stages in some Atlantic creoles, at least – is due to substrate influence
from African languages. Gullah, for instance, exhibits sentences like the one in
(22) (see also (3) above), which is reminiscent of serial verb constructions like the
ones in the substrate Kwa languages:

(22) Gullah (Mufwene 1986: 15)

B. and H. convince L. to get in a trunk bin out in d’yard.
‘B. and H. convinced L. to get in a trunk [that] was in the yard.’

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

a affirmative dem demonstrative

abs absolutive det determiner
anim animate m masculine
ant anterior ms masculine singular
art article neg negative
cop copula nom nominative
dcl declarative obj object
 Tania Kuteva & Bernard Comrie

obl oblique q question

pa past rel relative
pf perfect remp remote past
pl plural sg singular
plur plural sm subject marker
pred predicate tns tense
prep preposition vis visual.

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 

Palenquero 1/2 (Schwegler & Green 2007)

Panamanian Creole English 1 (Holm 1986: 484)
Philippine Creole Spanish 1 (Holm 1989: 321)
Reunion Creole4 1/0 (Corne 1995: 57; 62)
San Andrés Islander
Creole English 1 (Angela Bartens p.c.)
Santome 1 (Tjerk Hagemeijer p.c.)
Sango 1 (Hutchinson 1986: 82)
Saramaccan 1 (Byrne 1988: 348)
Seselwa 1 (Adone 2001: 64; Holm 1989: 402)
Solomons Pijin 1 (Crowley 1990: 13)
Sranan 1/0 (Bruyn 1995: 65)
Sri Lancan Creole Portuguese 1 (Holm 1989: 290)
St. Kitts Creole 1 (Philip Baker p.c.)
St. Lucian Creole 1 (Carrington 1984: 145)
Tayo 1 (Corne 1995: 57; Corne 1994: 287, 289–290)
Ternate Chabacano 1/0 (Eeva Sippola p.c.)
Tok Pisin 0/1 (Romaine 1992: 157–160)
Trinidadian English Creole 1 (Susanne Mühleisen p.c.)
Tugu 1 (Philippe Maurer p.c.)
Vincentian Creole 1/0 (Prescod 2004: 228; Paula Prescod p.c.)

Appendix 2

Language varieties (high contact)

Asian English 1 (Hansen et al. 1996: 226)
Basque spoken in Spain 2 (Heine & Kuteva 2003)
Colloquial Singapore English 2 (Alsagoff & Lick 1998: 129)
Chicano English 1 (Fought 2003: 103)
Frisian spoken in Schleswig-Holstein 2 (Hoekstra 2002: 71)
Hiberno-English 0 (Dolan 1999: xxvi)
Maori and Pakeha English 1 (Bell 2000: 235)
Nepali English 1 (Hartford 1996: 92, 99, 100)
Pennsylvania German 1 (van Ness 1990: 106)
Saraswat Brahmins Konkani 2 (Nadkarni 1975: 674–675)
South African Indian English 0/1/3 (Mesthrie 1991: 466)
Tariana of northwest Brazil 2 (Aikhenvald 2002: 183)

References

Adone, D. 2001. A cognitive theory of creole genesis, Habilitationsschrift. Universität Düsseldorf.

Aikhenvald, A. 2002. Language contact in Amazonia. New York: Oxford University Press.

4.  Reunion Creole is given as example of partial creolization in Holm (1989: 391).

 Tania Kuteva & Bernard Comrie

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

Robert D. Van Valin, Jr.

Heinrich-Heine University Düsseldorf & University at Buffalo,
the State University of New York

Relative clauses present an interesting challenge for theories of the

syntax‑semantics interface, because one element functions simultaneously
in the matrix and relative clauses. The exact nature of the challenge depends on
whether the relative clause is externally-headed or internally‑headed.
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 its typological rarity,
and such accounts typically involve movement rules, both overt and covert, and
phonologically null elements. The analysis of internally-headed relative clauses
often involves the positing of an abstract structure including a null external
head, with covert movement of the internal head to that position. The purpose
of this paper is to show that the essential features of both types of relative
clause can be captured in a syntactic theory that eschews movement rules and
phonologically null elements, Role and Reference Grammar. It will be argued
that a single set of linking principles can handle the syntax-to-semantics linking
for both types.

Keywords:  Externally-headed relative clauses; internally-headed relative

clauses; Role and Reference Grammar; linking syntax and semantics

1.  Introduction1

Relative clauses present an interesting challenge for theories of the syntax-­

semantics interface, because one element functions simultaneously in the matrix
and relative clauses. The exact nature of the challenge depends on whether the
relative clause is externally-headed, as in (1a), or internally-headed, as in (1b).

1.  I would like to thank Ranko Matasović and Dejan Matić for comments on an earlier draft.
 Robert D. Van Valin, Jr.

(1) a. [nuna-Ø ranti-shaq-n] bestya-Ø alli

 man-nom buy-perf-3 horse-nom good
bestya-m ka-rqo-n Ancash Quechua
horse-evid be-past-32 (Cole 1987)
b. [nuna-Ø bestya-ta ranti-shaq-n] alli bestya-m ka-rqo-n
 man-nom horse-acc buy-perf-3 good horse-evid be-past-3
‘The horse that the man bought was a good horse.’

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 

(Jackendoff 2002), featuring a monostratal syntax, with the single morphosyntac-

tic representation given to a sentence in a language being concrete, not abstract, in
the sense that it should represent the actual form of the sentence, including the lin-
ear sequence of its constituent elements and their morphological properties; there
are no phonologically null elements in the syntax. Having a monostratal syntax
excludes movement rules, both overt and covert, and the prohibition against pho-
nologically null elements rules out traces, null heads, and empty WH-operators.
The organization of RRG is given in Figure 1.

syntactic representation

Discourse-pragmatics
Linking
algorithm

semantic representation

Figure 1.  General structure of Role and Reference Grammar

There is a direct mapping between the semantic representation and the

­syntactic representation, unmediated by abstract syntactic representations, and
this mapping is codified in the RRG linking algorithm. The system maps between
syntax and semantics in both directions, i.e. from the semantic representation to
the syntactic representation, and from the syntactic representation to the s­ emantic
representation. This mirrors what speakers and hearers do in speech produc-
tion and comprehension.3 The question of determining the function of the head
noun in the clause in which it does not occur is a problem for the hearer, not
for the speaker, and therefore it is a problem for the syntax-to-semantics link-
ing ­system. Accordingly, we will limit this discussion to the syntax-to-semantics
linking in both types of relative clauses. A central question is whether the two
types of ­relative clause require different linking rules. One of the motivations for
positing a null external head in generative analyses is to assimilate the analysis of
­internally-headed relative clauses to that of externally-headed ones. Can a mono-
stratal ­syntactic theory which rejects movement rules and phonologically null
­elements give a unified treatment of the two types of relative clause?
The discussion will proceed as follows. The basics of the RRG representation
of relative clauses and of the linking algorithm from syntax to semantics will be

3.  See Van Valin (2006) for discussion of how the RRG linking system fits into models of
sentence processing.
 Robert D. Van Valin, Jr.

summarized. Section 3 will investigate the linking in externally-headed relative

clauses, and in Section 4 the focus will be on internally-headed relatives. Summary
and conclusions will be given in Section 5.

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

PrCS CORE PERIPHERY

RP NUC PP

PRED

RP V PP ADV

What did Robin show to Pat in the library yesterday?

Figure 2.  The layered structure of an English WH-question

 Some issues in the linking between syntax and semantics in relative clauses 

RPs also have a layered structure, analogous to that of the clause.

The semantic representation of sentences is grounded in an Aktionsart-based
system of lexical decomposition, the main details of which are beyond the scope
of this discussion. The representation of a predicating element is called its ‘logical
structure’[LS]. Examples of simplified semantic representations for four English
sentences are given in (2).
(2) a. Kim is tall. be¢ (Kim, [tall¢])
b. Kim is singing. do¢ (Kim, [sing¢ (Kim)])
c. I saw the window. see¢ (1sg, window)
d. Kim smashed the window. [do¢ (Kim, Ø)] CAUSE [INGR
smashed¢ (window)]

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.

be¢ (1SG, [tall¢ ])

I am tall Ma-há˛ske Lakhota

1SG-tall

Figure 3.  Attributive predications in English and Lakhota

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.

As mentioned above, the linking algorithm maps from semantics to syntax

and from syntax to semantics. The linking from semantics to syntax for a sentence
like (2d) would be roughly as in the following oversimplified description. The
semantic representation is composed in the lexicon, and then actor and under-
goer are assigned. Then the PSA is chosen, which is the actor Kim in this case. The
appropriate syntactic structure is selected, and the PSA is linked to the core-initial
argument position, and the undergoer is linked to the immediately post-nuclear
position.5 This is represented in the tree structure in Figure 4. The main concern in
this paper, however, is the syntax-to-semantics linking, and this linking in simple
sentences is summarized in (3)
(3) Linking from syntax to semantics (summary)
a. The parser outputs a labeled tree structure.
b. The first step is to derive as much information from the overt
­morphosyntactic features of the clause: case marking/word order, the
voice of the verb, adpositions.
c. The second step is to retrieve the LS of the verb from the lexicon and
assign macroroles where possible.
d. The information from these steps should link everything in the core to
the argument positions in the LS; if there is an element in the special
clause-initial position (the PrCS), it will be linked last to the remaining
unlinked argument position in the LS.

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.

3.  Externally-headed relative clauses

As noted in the introduction, the issue with respect to externally-headed rela-

tive clauses [EHRC] is the determination of the function of the head noun within
the relative clause. The problem is exemplified by English relative clauses, which
are in square brackets.
(4) a. The man [who/that won the lottery] ended up broke.
b. The man [who/(that) the police interviewed] had no helpful information.
c. The man [who/(that) the police showed the photo to] could not identify
anyone in it.

An English finite relative clause can be introduced by either a relative pronoun

or a complementizer (that). Given the demise of whom, the form of the relative
 Robert D. Van Valin, Jr.

­ 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 

b. Na-hita ny zaza [(izay) nan-asa

prfv.atv-see det child   clm prfv.atv-wash
ny vehivavy] Rakoto.
det woman Rakoto
‘Rakoto saw the child that washed the woman.’
*‘Rakoto saw the child that the woman washed.’
c. Na-hita ny zaza [(izay) sas-an’ny vehivavy] Rakoto.
prfv.atv-see det child  (clm) wash-pass-det woman Rakoto
‘Rakoto saw the child that was washed by the woman.’

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

CORE PrCS CORE

N N
RP NUC RP NUC
RP
PRED PRED

V V
I saw the window (that) Kim smashed I saw the window which Kim smashed

Figure 5.  Structure of English EHRCs

(7) a. I saw the window (that) Kim smashed.

a′.  see¢ (1sg, [be¢(windowi, [[do¢ (Kim, Ø)] CAUSE [INGR smashed¢
(yi)]])])
b. I saw the window which Kim smashed.
b′.  see¢ (1sg, [be¢(windowi, [[do¢ (Kim, Ø)] CAUSE [INGR smashed¢
(whichi)]])])

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 

(8) Rules governing linking from syntax to semantics in EHRCs

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 either the
­unlinked argument position in the relative clause LS, or, if there is
a relative pronoun, to the argument position linked to the relative
­pronoun.
c. Insert the attributive LS into the argument position in the matrix
LS occupied by the head noun, replacing the variable in the first
­argument position in the attributive LS with the head noun.

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

Actor Undergoer Actor 3d

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)]]

Figure 6.  Syntax-to-semantics linking in English without relative pronoun

 Robert D. Van Valin, Jr.

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

Actor Undergoer Actor 3d

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)]]

Figure 7.  Syntax-to-semantics linking in English EHRC with relative pronoun

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.

4.  Internally-headed relative clauses

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

Ne ye so min ye tye ye san

1SG PAST horse REL see man PAST buy
‘The man bought the horse that I saw.’

Figure 8.  The structure of IHRC in Bambara in (9a)

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

Ne ye so min ye tye ye san 3b

3b

Actor Undergoer Actor

3d 3d

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)])]

Coindexing (10b) RC LS substitution (10c)

[do¢ (tye, Ø)] CAUSE [BECOME have¢ (tye, [be¢ (xi, [see¢ (ne, soi)])])]

Figure 9.  Syntax-to-semantics linking in Bambara IHRC in (9)

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

Albert Álvarez González

Universidad de Sonora

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.

Keywords:  Relativization; nominalization; finiteness; referentialization

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.  Relative clauses

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.

2.2  Relativization types

Typological variation in RCls includes, for instance, the position of the RCl with
respect to the head (Payne 1997: 326). Using this criterion, RCls can be prenominal
(the clause occurs before the head), postnominal (the clause occurs after the head),
internally headed (the head occurs within the RCl) or headless.
Another typological parameter that shows variation in RCls is the mode of
expression of the relativized noun phrase. The so-called ‘case recoverability’ prob-
lem (e.g. Keenan 1985) is associated with the way in which the role of the refer-
ent of the head noun is identified within the RCl. Although the head noun has

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):

a. The gap strategy

b. The non-reduction strategy
i. Correlative clauses
ii. Internally headed relative clauses
iii. Paratactic relative clauses
c. The pronoun-retention strategy
d. The relative pronoun strategy

Finally, another important parameter in which languages vary in their formation

of RCls has to do with which grammatical relations can be relativized. The well-
known “noun phrase accessibility hierarchy” proposed by Keenan and Comrie
(1977) (see also Comrie 1981 or Lehmann 1986) shows, however, that variation
regarding which syntactic functions in a RCl are relativizable is not random but
rather follows a clear pattern. Certain syntactic functions are inherently easier to
relativize than others. These differences in accessibility to relativization are pre-
sented in the following continuous hierarchical segment:
subject > direct object > indirect object > oblique > possessor

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.  Relative clauses in Yaqui

Yaqui is a Uto-Aztecan language within the Taracahitan branch. It is s­poken

in the northwest of Mexico in the state of Sonora by almost 16,000 speak-
­ ordering state of Arizona by approximately 5,000.3 Yaqui is a
ers and in the b

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 

­ ominative-accusative language as can be seen from the pronoun system illus-

n
trated in the following table:

Table 1.  Pronominal system in Yaqui

Nominative Accusative Genitive

1Sg inepo, =ne nee in, nim

2Sg empo, =’e enchi em
3Sg aapo aapo’ik, a aapo’ik, a
1Pl itepo, =te itom itom
2Pl eme’e, =’em enchim em, enchim
3Pl bempo aapo’im, am bem, bempo’im

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.

(1) U yoeme-Ø uka kari-ta jinu-k

det man-nom det.acc house-acc buy-perf4
‘The man bought the house.’

(2) U yoeme-Ø u-me kari-m jinu-k

det man-nom det-pl house-pl buy-perf
‘The man bought the houses.’

In the following subsections, I will present the different possibilities in Yaqui

to relativize subjects, direct objects, indirect objects, obliques and locative
­complements. Special attention will be paid at this point to the positional type
of RCls, the relativization markers, the relativization strategies, and the pres-
ence or absence of agreement between the head noun and the RCl.

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

3.1  Subject relativization

Subject relativization is done by means of the marker -me, which is suffixed to the
verb within the RCl. The head noun is external to and precedes the RCl. There is
no overt case-marked reference to the head noun within the RCl (the ‘gap’ or ‘zero’
strategy). In (3), the head noun yoeme is the subject of both the main clause and
the RCl. The object of the RCl is marked with the accusative suffix -ta and precedes
the relativized verb. This is the word order expected for a SOV language.
(3) U yoeme-Ø [kari-ta jinu-ka-me] ousi tom-ek
det man-nom  house-acc buy-perf-rel a_lot_of money-poss
‘The man who bought the house has a lot of money.’

The construction in (4) shows the absence of number agreement between

the ­subject head noun yoemem and the postnominal RCl karim jinukame. Once
again, we can see that the plural object within the RCl doesn’t take the accusative
marker.
(4) U-me yoeme-m [kari-m jinu-ka-me] ousi tom-ek
det-pl man-pl  house-pl buy-perf-rel a_lot_of money-poss
‘The men who bought the houses have a lot of money.’

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 

(7) U jamut yoem-ta-u [Bikam-me-u jo’aka-m]-ta-u

det woman man-acc-dir  Vicam-pl-dir live-rel-acc-dir
rebo’osa-m jinu-k
mantilla-pl buy-perf
‘The woman bought the mantilla from the man who lives in Vicam.’

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.’

3.2  Direct object relativization

The example in (9) shows the direct object relativization of a head noun that is,
in turn, the subject of the main clause. This head noun is the direct object of the
RCl, and it is relativized by no morphosyntactic element, i.e. by means of a gap.
As in the subject relativization cases just seen, the RCl follows the head, but unlike
­subject relativization cases, its verb is now marked by the suffix -’u. The pronomi-
nal subject of the RCl appears in the genitive form.
(9) U bisikleeta [in jinu-ka-’u] sikili
det bicycle  1sg.gen buy-perf-rel red
‘The bicycle that I bought is red.’

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.’

3.3  Indirect object relativization

Indirect object relativization is realized by means of the same suffix-’u, which
was involved in direct object relativization but, contrary to the latter, there is an
overt case-marked reference to the head noun within the RCl. In the RCl in (15),
we can observe a third person singular dative pronoun (au), which co-refers
with the subject head noun (u yoeme ‘the man’). This resumptive personal pro-
noun illustrates another relativization type: the pronoun retention or the ana-
phoric pronoun strategy. Additionally, we can mention, again, the postnominal
position of the RCl with regard to the head noun and the genitive form of the
subject in the RCl.
(15) U yoeme [em rebo’osam au jinu-ka-’u] naamuk-ia
det man  2sg.gen mantilla 3sg.dat buy-perf-rel get_drunk-res
‘The man from (lit. to) whom you bought the mantilla is drunk.’

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.’

(18) U chu’u yoeme-me [em rebo’osam ameu

det dog man-pl   2sg.gen mantilla 3pl.dat
jinu-ka-’u]-m ke’e-ka
buy-perf-rel-pl bite-perf
‘The dog bit the men 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.’

(20) Juan yoeme-me-u [em rebo’osam ameu

John man-pl-dir   2sg.gen mantilla 3pl.dat
jinu-ka-’u]-m karo-ta nenka-k
buy-perf-rel-pl car-acc sell-perf
‘John sold the car to the men from whom you bought the mantilla.’

3.4  Oblique relativization

It is also possible to relativize an oblique complement in Yaqui. The example in (21)
shows the relativization of an instrumental complement whose head noun is the
subject of the main clause. As in the cases involving indirect object ­relativization
 Albert Álvarez González

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 example in (22) shows the relativization of a comitative complement. The

same characteristics explained earlier can be observed in this example.
(22) Wa’a yoeme [nim a-mak yepsa-ka-’u] naamuk-ia
dem man  1sg.gen 3sg.acc-com come-perf-rel get_drunk-res
‘That man I came with is drunk.’

Contrary to what happens in indirect object relativization, there is no number

agreement between the head noun and the RCl verb. This is evident in the example
in (23), where a subject head noun appears with the plural marker, and the verb
within the postnominal oblique RCl is not marked plural. However, the accusative
pronoun (ame) and the instrumental suffix (-e ‘singular’ in (21) vs. -mea ‘plural’ in
(23)), combined in (23) as the form ammea, mark plural, showing double number
agreement between the head noun and the RCl.
(23) Wa-me mache’eta-m [nim jupa-ta am-mea
dem-pl machete-pl   1sg.gen tree-acc 3pl.acc-inst.pl
chukta-’u] poposiu-la
cut-rel rust-res
‘Those machetes I am cutting the tree with are rusted.’

Because the comitative suffix is invariable, comitative relativization only exhibits

number agreement between the head noun and the accusative pronoun. Compare
(24) with (22).
(24) Wa-me yoeme-m [nim ame-mak yepsa-ka-’u] naamuk-ia
dem-pl man-pl   1sg.gen 3pl.acc-com come-perf-rel get_drunk-res
‘Those men I came with are drunk.’

The same type of construction may be observed in oblique relativization with a

direct object head noun, in instrumental RCls (25) and (26), and in comitative
RCls (27) and (28). No case agreement takes place between the oblique RCl and
 Relative clauses and nominalizations in Yaqui 

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.

(25) Inepo baabu-ta [itom pu’atom a-e

1sg.nom mud-acc   1pl.gen plate-pl 3sg.acc-inst.sg
yaa-ka-’u] tea-k
make-perf-rel find-perf
‘I found the mud we made the plates with.’

(26) Joan wa-me mache’eta-m [nim jupa-ta am-mea

John dem-pl machete-pl   1sg.gen tree-acc 3pl.acc-inst.pl
chukta-ka-’u] jinu-k
cut-perf-rel buy-perf
‘John bought the machetes I cut the tree with.’

(27) U chu’u yoem-ta [nim a-mak yepsa-ka-’u] ke’e-ka

det dog man-acc   1sg.gen 3sg.acc-com come-perf-rel bite-perf
‘The dog bit the man I came with.’

(28) U chu’u yoeme-m [nim ame-mak yepsa-ka-’u] ke’e-ka

det dog man-pl   1sg.gen 3pl.acc-com come-perf-rel bite-perf
‘The dog bit the men I came with.’

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.

(29) Empo yoem-ta-u [nim a-mak

2sg.nom man-acc-dir   1sg.gen 3sg.acc-com
yepsa-ka-’u] karo-ta nenka-k
come-perf-rel car-acc sell-perf
‘You sold the car to the man I came with.’

(30) Empo yoeme-me-u [nim ame-mak yepsa-ka-’u]

2sg.nom man-pl-dir   1sg.gen 3pl.acc-com come-perf-rel
karo-ta nenka-k
car-acc sell-perf
‘You sold the car to the men I came with.’
 Albert Álvarez González

3.5  Locative oblique relativization

Neither the locative nor the temporal functions are included in the accessibility
hierarchy, because they do not seem to fit into the hierarchy. Comrie (1981: 156),
for instance, points out that “in some languages, they are very easy to relativize,
in other languages very difficult to relativize.” However, the locative r­ elativization
deserves special mention here because it is the most economical relativization
strategy in Yaqui, if compared with the other strategies presented above. In Yaqui,
the locative is indeed very easy to relativize, as (31) and (32) show: locative rela-
tivization is done by means of a gap, the postnominal RCl doesn’t exhibit number
agreement with the plural subject head noun, the RCl subject appears in genitive
form (except if it is a plural noun, because of the already mentioned ­incompatibility
between the suffix -ta and the plural suffix), and the locative RCl verb is suffixed
by the marker -’Vpo.

(31) Wa kari [nim bo’e-pea-’apo] ujyooli

dem house  1sg.gen sleep-des-rel pretty
‘That house that I want to sleep in is pretty.’

(32) Wa-me kari-m [jamuch-im to’e-’epo] nim achai atte’ak

dem-pl house-pl  woman-pl sleep-rel 1sg.gen father own
‘My father owns those houses that the women are sleeping in.’

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).

(33) Inepo u-ka kari-ta [nim tomte-ka-’apo] jinu-k

1sg.nom det-acc house-acc   1sg.gen be_born-perf-rel buy-perf
‘I bought the house where I was born.’

(34) Inepo u-m kari-m [em joa-ka-’apo] jinu-k

1sg.nom det-pl house-pl   2sg.gen live-perf-rel buy-perf
‘I bought the houses where you have lived.’

The locative relativization of an indirect object head noun is disallowed in Yaqui

because there is an animacy restriction associated with the indirect object. The
examples below show the locative relativization of a head noun functioning as a
benefactive complement (introduced by the postposition betchi’ibo). Notice that
this construction exhibits, once again, the familiar lack of case and number agree-
ment between the head noun and the locative RCl (35)–(36).

(35) Inepo kari-ta [nim tomte-ka-’apo] betchi’ibo

1sg.nom house-acc   1sg.gen’ be_born-perf-rel for
 Relative clauses and nominalizations in Yaqui 

chu’u-ta jinu-k
dog-acc buy-perf
‘I bought a dog for the house where I was born.’

(36) Inepo kari-m [em joa-ka-’apo] betchi’ibo

1sg.nom house-pl   2sg.gen live-perf-rel for
chu’u-m jinu-k
dog-pl buy-perf
‘I bought dogs for the houses where you have lived.’

Next, I summarize the main features of the Yaqui RCls as follows:

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

The following table synthesizes this outcome:

Table 2.  Main features of the Yaqui RCls

Relativized Marker Relativization Case agreement Number agreement
syntactic function strategy ­between the head between the head
noun and the noun and the RCl
RCl verb verb

Subject suffix -me Gap Yes No

Direct-Object suffix -’u Gap No Yes
Indirect-Object suffix -’u Pronoun No Yes
retention
Oblique suffix -’u Pronoun No No
retention
Locative suffix -’Vpo Gap No No

The relativization scenario presented so far corresponds very accurately to

­ ehmann’s proposal (1986) on the typology of relative clauses, more precisely,
L
with the two correlation rules posited by this scholar: the rule regarding the
­pronominal representation of the head in the RCl, and the rule involving the pres-
ence of agreement affixes within the RCl.
Regarding the first rule, Lehmann (1986) mentions that the use of pronominal
representatives of the head within the RCl depends on the hierarchy of s­ yntactic
functions. He proposes that “if a given strategy requires the pronominal represen-
tation of the head in the RCl at a given position of the hierarchy, it will require such
representation for all LOWER positions” (Lehmann 1986: 674). This rule works in
Yaqui as we have illustrated above (see the previous table). Without considering
locative relativization,6 which is not part of the accessibility hierarchy, I have shown
that the pronoun retention strategy corresponds to indirect-object and oblique
relativization, whereas both subject and direct object may be relativized by means
of a gap. This is in accordance with the observation made by Comrie (1981: 163)
when he observes that “wherever a language has both a more explicit and a less
explicit way of forming relative clauses, then the more explicit type will be used
lower down the hierarchy and the less explicit type higher up the ­hierarchy.” This
generalization has an obvious functional explanation, as ­Comrie (1981: 163) him-
self points out: “the more difficult a position is to relativize, the more explicit indi-
cation is given of what position is being relativized, to facilitate recovery of this
information.”

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)

4.  Nominalization properties of Yaqui RCls

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)

4.1  Nominal characteristics

The clues of nominalization coincide, obviously, with the non-finite characteristics
we can observe in the Yaqui RCls.

4.1.1  Subject-RCl marker: suffix -me

The first easily observable nominal property of the suffix -me corresponds to the
fact that it functions, lexically, as a derivational morpheme. This suffix is used,
indeed, to create agentive nouns from verbal bases. There are numerous words
within the lexicon of the Yaqui language that have been created by the suffixation
of -me to a verbal base, showing that this derivational process is very productive.
In (38), I show more examples of this nominalization type. Notice that this lexical-
creation mechanism is sometimes accompanied by a process in which the first
syllable of the verbal stem undergoes reduplication (see also, (37)). Reduplication
normally indicates habitual aspect in Yaqui, which, in the context of nominaliza-
tion would render the resulting meaning ‘the one who usually X’, where X stands
for the verbal base.
(38) a. ye-ye’e-me
red-dance-nmlz
‘dancer’
b. tekipanoa-me
work-nmlz
‘worker’
c. jita e’-etbwa-me
something red-steal-nmlz
‘thief ’
 Relative clauses and nominalizations in Yaqui 

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).

4.1.2  Non-subject-RCl marker: suffix -’u

The origin of this nominalizing/relativizing suffix is the allative case marker -u, i.e.
a directional marker that indicates the movement towards an objective, as we can
see in (41).
 Albert Álvarez González

(41) In ae pajkoa-u sii-ka

1sg.gen mother party-dir go-perf
‘My mother went to the party.’

If we consider that case markers prototypically appear with nominals, it can be

deduced that this suffix functions as a nominalizing marker if combined with
other categories, as in the case of non-subject-RCls where -u is suffixed to verbs.
An interesting formal difference between the use of -u as an allative case
marker and its use as a non-subject-RCl marker is the insertion of a glottal stop
in the latter case. This epenthesis seems to be morphologically determined, as it
is absent if combined with nominal stems while present after nominalized verbal
stems.
In opposition to the suffix -me, the suffix -(’)u is not used in Yaqui to create
lexical nouns. However, the relativized constructions that take this marker con-
form to nominalized structures in two further respects:

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).

4.1.3  Locative oblique-RCl marker-: suffix -’Vpo

This Yaqui suffix is more frequently used in the formation of locative noun phrases.
It can be seen as a locative case marker, i.e. a marker that is combined with nomi-
nals to express a spatial relation (sometimes also a temporal relation). See example
in (42).
(42) U ili uusi kari-po emo esso-k
det little boy house-loc refx hide-perf
‘The little boy was hiding in the house.’

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

(46) a. Junu’u weyeka-me nakapit

dem be_standing-nmlz deaf
‘This one who is standing is deaf.’
b. Junu’u in ta’a-ka-’u nee alle-tua
dem 1sg.gen learn-perf-nmlz 1sg.acc be_happy-caus
‘What I learned makes me happy.’
(47) a. U-me bwa’am-ta joa-me pu’ato-m tapejti-po joa-k
det-pl lunch-acc do-nmlz plate-pl roof-top-loc put-perf
‘The ones who are cooking, they put the plate on the tapanco.’9
b. U itom nu’upa-ka-’u kaa jaleki
det 1pl.gen bring-perf-nmlz neg enough
‘What we brought is not enough.’
c. U in tekipanoa-’apo mekka taawa
det 1sg.gen work-nmlz far be
‘(The place) where I work is far.’

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 

4.2  Verbal characteristics

I just showed the different footprints of nominalization in Yaqui RCls. Below, I
survey some characteristics associated with finiteness in Yaqui RCls. Since finite-
ness features are practically the same for all the different constructions under con-
sideration, I will treat the Yaqui RCls all at once.
The first property worth mentioning that suggests the finiteness of RCls is
that they may contain verbs exhibiting all tense/aspect/modality marks, just as
any other tensed verb in Yaqui. In fact, this kind of construction does not involve a
loss of TAM morphology in Yaqui, if compared with any prototypical finite verbal
clause.
Another feature that suggests the finiteness of these constructions is the fact
that the element functioning as the object within the RCl is marked with accusa-
tive case. Because of the syncretism existing between accusative and genitive case
in Yaqui, this feature is more obvious when the object is pronominal as in (12)
above, or in the following example:
(51) a. Joan uka chu’u-ta [enchi ke’e-ka-m]-ta me’a-k
John det.acc dog-acc   2sg.acc bite-perf-nmlz-acc kill-perf
‘John killed the dog that bit you.’

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.

(52) a. U yoeme [tu’i-si bwika-me] in sai

det person  good-advz sing-nmlz 1sg.gen brother
‘The person who sings well is my brother.’
b. *U yoeme [tu’i bwika-me] in sai
c. U tu’i yoeme  ‘a good person’
 Albert Álvarez González

(53) a. Inepo uka karta-ta [em kaa tu’i-si

1sg.nom det.acc letter-acc   2sg.gen neg good-advz
ji’ote-ka-’u] taya-k
write-perf-nmlz burn-perf
‘I burned the letter that you wrote wrong.’
b. *Inepo uka karta-ta [em kaa tu’i ji’ote-ka-’u] taya-k
c. U kaa tu’i karta  ‘a wrong letter’

In sum, so far we can argue that relativization in Yaqui is better understood as

a nominalization process by means of which a finite verbal clause is adjusted to
a noun phrase in order to be used as a dependent clause in the combination of
clauses. As mentioned by Lehmann (1986), this scalar phenomenon can be placed
within the continuum of desententialization from clause to noun, which deter-
mines the degree of interrelatedness between two clauses. Within the adjustments
from a prototypical finite verbal clause to a prototypical noun phrase, we have seen
that Yaqui RCls exhibit several nominalization features, showing that RCls in this
language are grammaticalized to the extreme, and treated as noun phrases. On
the other hand, they also retain some finite features of their source verbal clause,
showing that nominalization and finiteness are basically a matter of degree.

5.  Relativization vs. nominalization

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 

In recent work based on a number of wide-spread relativization patterns

around the globe (including Yaqui data), Shibatani (2009) defends this same idea,
i.e. that the structures conventionally identified as relative clauses/sentences in
a large number of languages are in fact nominalized entities, lacking some cru-
cial properties of both full clauses and sentences. One of his main arguments is
grounded in the fact that these head nominalized expressions can function as
arguments or nominal predicates in the same manner as simple nouns can. In
(54), I show examples that I provided to Shibatani (2009: 175) to illustrate these
possibilities in Yaqui. The nominalized form functions as an argument (subject) in
(54a), and as a nominal predicate in (54b).
(54) a. [U nim pu’akta-’u] bette
 det 1sg.gen bear-nmlz be.heavy
‘What I bear is heavy.’ ‘My burden is heavy.’
b. Jabesa [wa jiosam noktua-me]
who  dem book read-nmlz
‘Who is that one that is reading the book?’

Following Shibatani’s (2009) analysis, Yaqui RCls need to be considered as apposi-

tive constructions consisting of a nominal (corresponding to the head nominal)
and a nominalized expression (corresponding to the relative clause) that are juxta-
posed. In this appositive construction, the function of the head noun is to identify
the entity referred to by the RCl nominalized expression, which, in turn, functions
as a modifier of the head noun. Rather than relativization, the phenomenon only
exhibits the juxtaposition of a head noun and an argument nominalized expression
in an appositive phrase. So what has been mentioned as an RCl is in fact a nomi-
nalized modifier linked to the head noun in a modification structure correspond-
ing to the [Noun + Noun] type.11 Consequently, labelling the suffixes -me, -’u,
-’Vpo as relativizers or RCl markers is misleading; a more appropriate label would
be “nominalizer.” Yaqui in fact presents a syncretism between nominalization and
relativization, i.e. the same construction is used for argument nominalization and
relativization purposes. In (55), I show an agent nominalization functioning as a

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.

(55) a. U-me [yeewe-me] Agent nominalization

det-pl  play-nmlz
‘the ones playing’
b. U yoeme u-m usi-m [yeewe-me]
det man det-pl child-pl  play-nmlz
kaka-m maka-k Subject-RCl
candy-pl give-perf
‘The man gave candies to the children who are playing.’

(56) a. Uka [in tea-ka-’u] Patient nominalization

det.acc   1sg.gen find-perf-nmlz
‘what I found’
b. U chu’u [in tea-ka-’u] chukuli Object-RCl
det dog  1sg.gen find-perf-nmlz black
‘The dog that I found is black.’

(57) a. U [nim bo’o-pea-’apo] Locative nominalization

det   1sg.gen sleep-des-nmlz
‘(the place) that I want to sleep in’
b. Wa kari [nim bo’o-pea-’apo]
dem house  1sg.gen sleep-des-nmlz
ujyooli Locative oblique-RCl
pretty
‘That house where I want to sleep is pretty.’

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 

nominalization and clausal nominalization (Koptjevskaja-Tamm 2006; Genetti

et al. 2008; Shibatani 2009).
Strictly speaking, nominalization is a process by means of which something
turns into a noun (Comrie & Thompson 1985: 349). Therefore, this phenom-
enon implies a switch from an initial category to the resulting lexical category
of noun.12 Although this re-categorization process is commonly realized by
means of formal markers, nominalization doesn’t have to obligatorily show a
morphological change, just like in the case of conversion (e.g. the conversion
of the English verb walk into the noun walk). As long as the resulting forms
behave functionally and syntactically like nouns (as arguments or nominal
predicates, for instance), we have to consider that these forms have been nomi-
nalized. Therefore, nominalization is a matter of functional change rather than
formal change, and the result of this operation is an expression that fulfills the
most essential nominal function,13 i.e. the creation of a referring expression
­(Shibatani 2009:  191).
However, these nominalized forms can be located within two different types
of referring expressions, depending on their lexical status: l­exical nominaliza-
tions create nominalized forms with lexical status, while clausal ­nominalizations
don’t. In other words, the forms created by lexical ­nominalization become
“new lexical items belonging to the noun class of the language” (­Shibatani
2009: 187) and the forms generated by clausal nominalization are only refer-
ring ­expressions “created for the nonce” (Shibatani 2009: 192), hence they don’t
become new lexical items. Contrary to what happens in clausal nominaliza-
tions, the result of a lexical nominalization has to be learned and stored in the
memory by the speakers, in order to become a ‘listed object’14 and be part of
their mental lexicon.

12.  Malchukov (2004, 2006) refers to nominalization as a “transcategorial operation”, and

Koptjevskaja-Tamm (2006) as a “transpositional phenomenon”, based on the notion of trans-
position defined by Haspelmath (2002) as a word-class-changing operation.
13.  In his typology of noun classification, Croft characterizes noun classes by reference:
“Noun classes are used in determination, which fixes reference” (Croft 1994: 161). Reference
has been defined, for instance, by Searle as “an act of identifying some entity that the speaker
intends to talk about” (Searle 1969: 85).
14.  In their book On the Definition of Word, Di Sciullo and Williams (1987) coined the term
listeme for this conception of word. Listemes are linguistic expressions memorized and stored
by speakers. The use of this term is meant to highlight the fact that words in this sense must
be listed in the lexicon because they have idiosyncratic properties (not governed by general
principles) that speakers must simply memorize.
 Albert Álvarez González

Lexical nominalizations are in fact a derivational process involving the

­creation of a referring expression that results in a lexical noun, while clausal
nominalization is a grammatical process involving the creation of a referring
expression that results in a noun phrase. As Genetti et al. (2008: 98–99) men-
tioned, these two types of nominalization differ not only in terms of the syntactic
category of the resultant structures (lexical noun vs. noun phrase) but also in
terms of the domains to which they apply (lexical root vs. finite verbal clause).
Lexical nominalizations are then confined to the lexicon, while clausal nominal-
izations convey a syntactic process of interclausal combination, allowing a clause
to function as a noun phrase within a broader syntactic context. Consequently,
lexical nominalizations are a non-gradual process, as opposed to clausal nomi-
nalizations that appear to be an adjustment process from a prototypical finite
verbal clause to a prototypical noun phrase, and hence several degrees may arise
between both types (Givón 2001: 25).
If both types of nominalization allow the creation of a referring expres-
sion, Shibatani (2009: 192–193) points out an important difference regarding
this same general function: lexical nominalizations name the entities referred
to, whereas grammatical nominalizations lack identifications, specifications
or names. It is precisely because of this lack of identification that this kind of
nominalization tends to occur in contexts in which the entity identification is
provided, as is the case of RCls, where the head supplies the identification. This
difference in specification properties between the head noun and the nominal-
ized expression explains why the latter loses its referential function in favor of a
modifying function with regard to the former. The relationship of co-reference
between the head noun and the appositive nominalization is then interpreted as
a modifying relationship between the specified head noun and the non-specified
nominalization.15
Finally, it may be observed that, as a derivational process, lexical nominal-
ization has limited applicability, and is semantically unpredictable (it exhibits
idiosyncratic meanings). In contrast, clausal nominalization exhibits, usually,
unlimited applicability and full predictability.
These differing characteristics between lexical and clausal nominalizations are
summarized in Table 3.

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 

Table 3.  A list of properties of lexical nominalization and clausal nominalization

Lexical nominalization Clausal nominalization

1. Derivational process 1. Grammatical process

2. Creates referring expressions with lexical
status: lexical nouns (listemes)
3. Applies to different lexical roots (verbal but 3. Applies to finite verbal clauses
also adjectival, adverbial roots)
4. Non-gradual process confined to lexicon
5. Identifies the entities referred to
6. Limited applicability and semantic
unpredictability (idiosyncratic meaning)
2. Creates referring expressions with
no lexical status : noun phrases (non
listemes)
4. Gradual process of interclausal
connectivity (can retain some finite
features of the source verbal clause as
TAM markers, etc.)
5. Doesn’t identify the entities referred to
6. Unlimited applicability and full
predictability

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.

6.  Final remarks

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

a function-changing operation. However, in contrast to Shibatani (2009: 186), who

considers nominalization as a functional notion, I prefer to restrict the notion of
nominalization to a transcategorial process and I propose to use the notion of
referentialization in order to distinguish between the category-changing process
(nominalization) and the function associated with this process (referentializa-
tion). This function of referentialization (the act of referring to an entity) is, of
course, the raison d’être of nominalization and the prototypical function of nomi-
nal expressions.
Going ahead with this distinction between category and function, we may
observe that in the same way as clausal nominalization implies desententializa-
tion (Lehmann 1986), the referentialization associated with clausal nominaliza-
tion implies “depredicativization”, i.e. the loss of the sentential predicative function
of the source verbal clause. Moreover, this referentialization function is unable to
specify the identity of what has been referred to by the clausal nominalization. It
is precisely the creation of a referring expression along with its implications (the
lack of predicative function and the lack of identification) that makes grammati-
cally nominalized expressions (i.e. clausal nominalizations) incorporable into a
sentence, allowing them to function as modifiers of a co-referenced head noun
in RCls. The co-reference between the specified head noun and the non-specified
appositive clausal nominalization is then interpreted as a modifying/restrictive
relationship, i.e. the head noun provides the identification of the entity, and the
nominalized expression provides the head noun with additional information.
As the process of nominalization operates with other types of subordinate
modifiers, such as complement clauses or adverbial clauses in a large number of
languages, Shibatani (2009: 166) points out that nominalization is the essence of
clause integration in these languages, i.e. it is “the cognitive ability to convert a
sentence into a non-sentential nominal object,” which allows this now nominal/
referring expression to be legitimately integrated within a matrix sentence. It
seems to be the case for Yaqui, at least for the RC-type clause integration.

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.

Keywords:  relative clauses; nominalization; complementation; Yaqui;

Uto‑Aztecan

1.  Introduction1

A relative clause (henceforth Rel-clause) generally involves a head noun and a

subordinate clause interpreted as attributively modifying that nominal (Lehmann
1984: 276). Notionally, there are two states of affairs involved, one of which (the
dependent one) provides some kind of specification about a participant of the
other (the main one), such as a participant of the main event is identified within
a set of possible referents by mentioning some other event in which she takes part
(Cristofaro 2003: 193).

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

This paper is concerned with true Rel-clauses, as in (1), and non-modifying

­relative-like structures, as in (2). The analysis examines (i) the individual relativ-
izing strategies based on the Accessibility Hierarchy; (ii) the pragmatic uses of
the two major types of Rel-clauses, and (iii) the characterization of the essential
features of true Rel-clauses and structurally similar constructions. Structurally, it
will be suggested that Yaqui relatives show a mixture of nominal and verbal char-
acteristics, so that they establish different points inside a continuum (Comrie &
Thompson 2007; Koptjevskaja-Tamm 1993). Functionally, true Rel-clauses intro-
duce or further establish new information into discourse (Comrie & Thompson
2007; Fox & Thompson 1990), whereas non-modifying nominal clauses serve as
either appositive units – making a comment about a noun without delimiting its
reference (Keenan 1985; Carlson 1977), or as a core argument (van der Auwera
1985). The corpus under analysis consists of almost 400 examples including data
from fieldwork and oral texts, as well as examples from the Diccionario Yaqui-
Español (Estrada et al. 2004) all based on the Sonoran dialect.
The information is organized as follows. Section 2 introduces some basic
morpho-syntactic properties of Yaqui, and characterizes Rel-clauses based on
their internal syntax. Section 3 highlights the nature of restrictive Rel-clauses in
discourse, and outlines the different degrees of nominalization; in addition, the
non-restrictive Rel-clauses (e.g. (2a)) are then introduced. Section 4 deals with the
formal and functional distinction between restrictive Rel-clauses and the other
two non-modifying structures functioning as core arguments (e.g. (2b) and (2c)).
Section 5 discusses the patterns observed in the Yaqui language, and Section 6
concludes this paper.

2.  Characterizing Yaqui Rel-clauses

2.1  Basic morpho-syntactic properties of Yaqui

Yaqui is a language with both synthetic and agglutinative characteristics. In
contrast to the other languages from the Taracahita branch, Yaqui maintains a
morphological case system on nouns. Lexical nominatives are unmarked, accu-
satives are marked by the suffix -ta (3a), and obliques are marked by postposi-
tions (3b). Some postpositions take accusative complements, especially when
referring to animate participants; contrast the directional -u indicating the
hearer of a speech act verb in (3b), and the goal of a motion verb in (3c). The
clause in (3d) shows that accusative and plural marking on nouns are mutually
exclusive. Additionally, double accusative constructions, as in (3e), are also very
common in the language.
 On relative clauses and related constructions in Yaqui 

(3) a. U jamut-Ø Peo-ta bicha-k.

det woman-nom Pedro-acc see-pfv
‘The woman saw Pedro.’
b. U o’ou-Ø jamut-ta-u nooka-k.
det man-nom woman-acc-dir talk-pfv
‘The man talked to the woman.’
c. U goi-Ø u-me chu’u-im ke’e-kan.
det coyote-nom the-pl dog-pl bite-pastc
‘The coyote was biting the dogs.’
d. Aurelia-Ø Ivan-ta mo’obei-ta jinu-ria-k.
Aurelia-nom Ivan-acc hat-acc buy-appl-pfv
‘Aurelia bought Ivan a hat.’

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.

Table 1.  Pronominal system in Yaqui

Nominative Accusative Object of Genitive
postpositions

1Sg inepo = ne nee, ne ne- in, nim

2Sg empo = ’e enchi e- em
3Sg aapo apo’ik/a a- apo’ik/a
1Pl itepo = te itom ito- itom
2Pl eme’e = ’em enchim emo- em, enchim
3Pl bempo apo’im/am ame- bem, bempo’im

Yaqui is characterized as a nucleus-final language, i.e. it employs p

­ ostpositions,
nominal and verbal suffixes. Although nominal arguments tend to precede the verb
(3), it is possible for a core argument to appear extraposed to the right (­Belloro &
Guerrero 2010). When the nominative NP follows the verb (4a), nothing happens.
When the accusative (4b) or postpositional (4c) NPs follow the verb, a clitic pro-
noun co-indexed to the NP must be attached to the verb and, presumably, there
is a pause between the verb and the extraposed NP. Notice that the clitic pronoun
keeps the relevant case and number coded in the extraposed NP.
 Lilián Guerrero

(4) a. Peo-ta bicha-k u jamut-Ø.

Pedro-acc see-pfv det woman-nom
‘As for the woman, (she) saw Pedro.’
b. U jamut-Ø a = bicha-k Peo-ta.
det woman-nom 3sg.acc = see-pfv Pedro-acc
‘As for Pedro, the woman saw him.’
c. U o’ou-Ø a-u = nooka-k jamu-ta-u.
det man-nom 3sg-dir = talk-pfv woman-acc-dir
‘As for the woman, the man talked to her.’

2.2  The internal syntax of Rel-clauses

Languages vary with respect to which syntactic argument can be relativized, e.g.
only subject, only core arguments, core arguments as well as certain peripheral
units. Languages may also employ different morphosyntactic means (i.e. different
relativizing strategies) to indicate the different syntactic-semantic roles of the head
noun inside the Rel-clause, i.e. no-reduction, relative pronoun, pronoun retention,
gapping (Keenan & Comrie 1977, 1979; Givón 2001; Comrie & Kuteva 2005). As for
Yaqui, Rel-clauses consistently mark the functional distinction between subjects
(nominative) vs. non-subjects (object, oblique) (Lindenfeld 1973; Escalante 1990;
Martínez & Langendoen 1996; Dedrick & Casad 1999; Álvarez 2006; ­Guerrero
2005a,b, 2006a,b).4 The Rel-clause always follows the head noun, though the posi-
tion of the relative unit may vary. When the head noun functions as the subject
inside the Rel-clause (e.g. dependent subject), the dependent unit is marked by the
clause linkage marker -m(e), as in (5) below.
(5) a. U o’ou-Øi [__i mesa-ta kokta-ka-me] siika.
det man-nom table-acc break-pfv-clm go.sg.pfv
‘The man who broke the table, left.’
b. Jipi’ikim misi-tai miika-Ø [__i pa’aku weama-m-ta].
milk.pl cat-acc give-prs outside be around-clm-acc
‘(He) gives milk to the cat that is outside.’

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 

Furthermore, some authors also distinguished between correlative and adjunct

Rel-clauses (Dryer 2005: 366). In the former, the head noun occurs inside the rela-
tive (i.e. internally-headed), the linked unit is placed outside the main clause and
it is connected anaphorically to a NP inside the main unit. In the latter, the relative
unit is also outside the main clause, but the head noun functions as an argument
of the main unit (i.e. externally-headed). Yaqui relatives are usually embedded,
meaning they can split the main clause down the middle, as seen in (5a) and (6),
but they may be outside the main unit too. In the construction in (5b), the relative
is extra-posed to the right, but the head noun is kept as an argument of the main
unit, e.g. adjoined relatives. In contrast, in the constructions in (10), both the head
noun and the Rel-clause are extra-posed to the right, but only in (10b) there is
­co-referential pronoun inside the main unit, e.g. correlative clause types.
(10) a. Maria-Ø tomi-ta mabeta-k u-me
María-nom money-acc receive-pfv det-pl
kobanao-m [ waria-po kate-ka-m-ta-betana].
governor-pl  guardia-loc sit.pl-pfv-clm-acc-from
‘María received money from the governors who were sitting in
the Guardia hall.’
b. Ju’unakiachi-si ne ai bicha-Ø
clear-ints 1sg.nom 3sg.acc see-prs
[ju-ka jamut-tai ne-u-bichaa weye-m-ta].
 det-acc woman-acc 1sg-dir-toward come-clm-acc
‘I clearly see a woman walking towards me.’ (Guaymas: 7)

2.3  Other functions of the suffix -me

In addition to mark subject relative clauses, the suffix -me is also the preferred
strategy for lexical nominalization (Shibatani 2009). In Yaqui, deverbal nouns
marked by -me are usually agentive, e.g. the ‘dancer, the one who dance’ in (11a),
or theme/patient oriented, e.g. the ‘food, the one what is eaten’ in (11b). The
­examples in (11c) are slightly more complicated. On the one hand, they incor-
porate ­unspecified objects like yee for human and ji(ta) for inanimates, e.g.
­‘thing-stealing’ for ‘thief ’; some of them even retain other verbal properties like
reduplication and the ­passive suffix -wa (but never tense or mood).5

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

(11) a. yeye’eme ‘dancer’

majtawame ‘student’
b. ko’okoeme ‘sick person’
jo’ome ‘native’
robbojo’okame ‘hunchback’
c. yee-susua-me ‘murderer’ (lit. people-RED.kill-me)
yee-sisibo-me ‘witch’ (lit. people-RED.bewitching-me)
kia-weama-me ‘cheater’ (lit. nice-be walking-me)
ji-e’etb-wa-me ‘thief ’ (lit. thing-steal-PASS-me)
na-susua-wa-me ‘fighter’ (lit.together-become.crazy-PASS-me)
jinko’ola-wa-me ‘competence’ (lit. competing-PASS-me)
allee-wa-me ‘happiness’ (lit. happy-PASS-me)

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).

(12) a. Maria-Ø u-ka bwa’a-m-ta kia bwasa-k.

María-nom det-acc eat-clm-acc nice cook-pfv
‘María cooked a very tasty food.’
b. U jamu-t-Ø tajo’o-ta sewa-ka-m-ta. jipu’e.
det woman-nom cloth-acc flower-pfv-clm-acc have-prs
‘The woman wears a skirt with flowers (lit. a flowered skirt).’

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.

3.  The nature and function of Yaqui Rel-clauses

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 

Yaqui Rel-clauses marked by -me can be considered a kind of reduced clause

due to the following features: (i) the co-referential subject is generally omitted;
(ii) the dependent verb is usually unmarked or marked by the past perfective -ka;
and (iii) the dependent unit take the case and postpositional markers of the head
noun. More examples are presented in (13).
(13) a. Yooko bahim-po kupteo kaba’imi
tomorrow three-loc noon horse.pl
[__i mamahae-me] mansote-ne.
RED.afraid-clm tame-pot
‘Tomorrow at three o’clock, you will be taming the horses that are
afraid.’ (Johnson 218:20)
b. ¡aman = em nee au chachai-ria-bo! = ti
there = 2sg.nom 1sg.acc 3sg.dir RED.talk-appl-mov.purp.pl= clm
te’eka u jamut-Øi [abe __ i muke-me].
say.pfv det woman-nom  almost die-clm
‘“You go there to talk to him about me” said the woman who was dying.’
 (Priest: 25)
In comparison, Rel-clauses marked by -’u can be seen as unreduced clauses since:
(i) they usually keep track of the head noun function by the use of a ­resumptive
pronoun (14b) or indefinite noun (14c); (ii) case marking is disallowed; and
(iii)  the dependent verb shows more independency with respect to temporal,
aspectual and modal operators as well as adverbs. However, the dependent unit
may agree in number with their head (14b).
(14) a. Ne u-ka wikia-tai woita-taite-k
1sg.nom det-acc rope-acc unwinding-start-pfv
[in __ i wiko-saka-’u].
  1sg.gen waist-go-clm
‘I started to unwind the rope that I was having as a belt.’ (Priest: 93)
b. Ne tebotua-ne waate-mi
1sg.nom greet-pot someone-pl
[ketun in ame-ui waate-’u-m].
 yet 1sg.gen 3pl-dir remember-clm-pl
‘Sometimes, I greet some of them who I still remember.’
 (C’s life story, 68)
c. In ji’aniraa ne lijtaroa-k
1sg.gen belongings 1sg.nom prepare-pfv
[si’ime-ta in waiya-bae-’u].
 everything-acc 1sg.gen bring-desid-clm
‘I prepared all my belongings, everything I want to bring.’
 (Little coyotes: 22)
 Lilián Guerrero

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).

Table 2.  Distribution of the relativized syntactic function

S-relatives A-relatives Obj-relatives Obl-relatives Total

Oral texts6 85 16 26 9 136

Dictionary 52 19 28 13 112
137 (55%) 35 (14%) 54 (22%) 22 (9%) 248

This Yaqui preference follows the predictions of the Accessibility Hierarchy

as well as Fox’s (1987) findings on English relatives, whereby subjects are more
relativizable than objects. Even more interesting is the overwhelming correlation
between S- and O- relatives. Along the lines of Du Bois’s (2003: 40) pragmatically-
based theory, S and O roles, in contrast to A, are amenable to the introduction of
new information into discourse, especially as regards human protagonists. As for
Yaqui, we already said that -me derives action nominals, above all those related to
character participants in tales and legends such as the yeebwa’eeme ‘the raven man
(lit. the one eating people)’ and ju’u boobok bo’ojoame ‘the wandering toad (lit. the
toad making the road)’.7

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 

According to the cognitive principle of subject primacy (Keenan 1985),

­S-relatives serve to (i) introduce a new referent by describing it and thereby
­making it relevant to that discourse, and/or (ii) characterize such a noun by
naming “habitual attributes or properties” or describing features of their sub-
jects (cf. Lehmann 1984; Fox 1987: 861–2; Fox & Thompson 1990: 306–7). As said
before, the overwhelming occurrence of S-relatives in Yaqui indicates that their
discourse function is to introduce new referents. This pragmatic function also
explains why they predominantly express stative descriptions of some aspect of
the referent. As illustrated in Table 3, out of 137 intransitive S-relatives, almost
70% involve non-active predicates, the most common positional verbs being joa
‘live’, ane/ayu ‘exist’, and tea ‘be named’ where the head is a proper noun. Stative
transitive predicates like jippue ‘have’, bicha ‘see’, ta’a ‘know’ can relativize the A
and the O roles. In the Table, Su refers to patient or undergoer subjects, while Sa
to actor subjects.

Table 3.  Strong preference for non-active predicates

Su- Sa- A-stative O-stative
relatives relatives relatives relatives
(out of 137) (out of 35) (out of 54)

Oral texts 63 22 5 9
Dictionary 38 14 4 6
101 (67%) 26 (17%) 9 (6%) 15 (10%)

Moreover, it has been said that Rel-clauses introduce or further estab-

lish ­people, objects, time and locations, by linking them to known referents
and ­situations (Fox  & Thompson 1990; Cristofaro 2003: 193). The function of
such restrictions is seen as helping the addressee to identify the referent of a
term, through a ­specification of some state of affairs in which that referent is a
­participant. It makes sense, therefore, that S and O should be the preferred syn-
tactic position for introducing new information by means of nominalization. Take
for example the following passage. The first clause presents the topic of the tale and
makes use of a subject Rel-clause marked by -me in (15a), e.g. the true thing [that
happened here]. In (15b) the first mention of the main character is then introduced
by another -me clause, e.g. a harp person [who was named Loreto]. Once the new
information is set up, the tale keeps using a definite noun phrase for the main
character ini’i achai ‘that man’.
(15) a. Achai-m emo-u ne etejo-bae i’i tua lutu’uria
father-pl 2pl-dir 1sg.nom chat-desid dem ints true
[inim Campani kau-po yeu sika-me]
 here bell hill-loc out go.sg.pfv-clm
 Lilián Guerrero

b. Aman taewai 27-po inım yoeme apa-reo

there year 27-loc here person harp-agt
[Loreto tea-me] jo’a-kan
 Loreto name-clm live-pastc
c. Ini’i achai into yeu siika Costa-u bichaa.
dem father-nom and out go.sg.pfv Costa-dir toward
‘Fathers, I want to chat with you; there is a true thing that happened
here in the Bell’s Hill. Around 1927, a harp person named Loreto lived
here. And that man went towards the Costa…’
 (Fox became human: 1–3)

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 

d. yuku mecha-m sim-su-ko.8

rain moon-clm go.sg-finish-clm
e. U baa’a-Ø kaa sisime-ka buite-n wassukte-n.
det water-nom neg red.go.sg-clm run.sg-pastc have.year-pastc
f. Bea te jita jumak te bicha-k
dm 1pl.nom thing.acc thus 1pl.nom see-pfv
g. [ba’a-po emo yoyoa-ka-m-ta].
 water-loc refl RED.move-pfv-clm-acc
‘Once, we were walking together in the river, we were sat in a small
hill, very close where the water stays (e.g. lake) when the rainy season
was over. Even thought the water goes running all year around. Then,
we saw a thing moving under the water.’ (Grandfather: 1–8)

Nevertheless, in Yaqui oral texts there are a dozen of unexpected examples of

S-relatives introduced by -’u (instead of -me). This construction is the first type of
a non-modifying relative unit, and was first noticed by Buelna (1891) and vaguely
mentioned in Johnson (1962). From the examples below, we may notice that all
are copulative (possessive?) clauses involving noun and adjectives, all are marked
by the past perfective suffix -ka, and in all cases the head noun is either a definite
noun or a proper name and, crucially, refers to human beings.
(18) a. si kubaji-ta poona-Ø [kompae Timo-tu-ka-’u].
inten drum-acc play-prs  compadre Timo-vblz-pfv-clm
‘The one being the Compadre Timo, (he) play well the drum.’
 (H’s life story: 145)
b. [in sai-tu-ka-’u] aman tawa-ne.
  1sg.gen brother-vblz-pfv-clm there stay-pot
‘The one being my brother, (he) will remain there.’ (H’s life story: 127)
c. in maala betana into in
1sg.gen mother from and 1sg.gen
maala [yo’o-tu-ka-’u] Wiibisim betana jo’o-me.
mother  old-vblz-pfv-clm Wiribis from live-clm
‘And from my mother, my grandmother, who is old, was from Wiribis.’
 (C’s life story: 13)
d. [ili’i uusi-tu-ka-’u] yo’o-tu-im-tu-k.
 little child-vblz-pfv-clm old-vblz-pl-vblz-pfv
‘And the ones who were young, became older.’ (Johnson: 218:203)

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.

4.  Distinguishing relatives from complements

In his seminal work on the aspectual classification of predicates, Vendler (1967,

1970) showed that, at least in English, nominalized clauses (e.g. clauses a­ cting as
nominal arguments) can express a few and clearly defined meaning c­ ategories,
predetermined by the semantic type of complement-taking predicates. Thus,
verbs like believe in and recognize take a propositional complement; verbs like
know and regret take a factive complement, whereas verbs like hear or ­continue
take complements referring to events (see also Dixon 2006). ­Nominalized
­sentences can be further divided into two categories, “imperfect nominals”,
where the verb “is still alive as verb”, and “perfect nominals” in which “the verb
 On relative clauses and related constructions in Yaqui 

is dead as a verb, h ­ aving become a noun” (Vendler 1967: 130–1). Thus, tense

markers, auxiliaries and adverbs can occur in the former but not in the latter;
as a result, propositional and factive complements can be encoded by the first
type but not events.
As many other languages, Yaqui makes use of very similar constructions for
relatives and other subordinate relations. With respect to complementation, there
are, at least, four types of complement strategies: the co-lexicalized structure, the
syntactic complement marked by -’u, the nominalized complement marked by
-me, and the ‘participial’ complement marked by -kai; some predicates only take
one type, some can take two, and some can take three (Guerrero 2004, and further
works). Then, the first type consists of a co-lexicalized or morphological complex
structure, where the matrix predicate and the dependent verb are joined together
usually without a clause linkage marker (but see (19b)). The linked verb may be
unmarked or be marked by aspectual and modal suffixes, but not for tense, and
each predicate may take its own set of core arguments, although they used to share
a participant.
(19) a. Peo-Ø Goyo-ta toto’i-m sua-tua-k.
Pedro-nom Goyo-acc hen-pl kill-cause-pfv
‘Pedro made Goyo to kill the hens.’
b. Goyo-Ø Tibu-ta wakas-ta etbwa-ka-t-’ea-n.
Goyo-nom Tibu-acc cow-acc steal-pfv-clm-think-pastc
‘Goyo thought Tibu to have stoles the cow.’

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

c. Nepoi si majae-n [kaa nimi enchi bit-ne-po].

1sg.nom ints be afraid-pstc  neg 1sg.gen 2sg.acc see-pot-clm
‘I was afraid I couldn’t see you.’

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).

(22) 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)

Before starting the discussion of modifying vs. non-modifying clauses types in

Yaqui, a few words on bena are needed. Bena is a very unusual predicate. Outside
clauses like those in (22), it barely appears as a main predicate except in compara-
tive clauses, e.g. X resembles Y in (23a). More commonly, it introduces oblique argu-
ments in the same way as ‘like’ in English (e.g. like your sister) in (23b); it serves as
a clause linkage marker of some adverbial clauses encoding either manner of action
or simultaneous events (e.g. they live under the water, like we live here); it may also
 On relative clauses and related constructions in Yaqui 

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)

Some formal and functional properties defining nominalized units modifying a

head noun vs. nominalized sentences serving as core arguments are outlined next.
(i) Nominal categories. The most obvious difference between modifying
clauses and argument clauses is, of course, the fact that the former offer some
information about a particular entity, while complements do not refer to any
particular individual but rather express a state of affairs in which that indi-
vidual is involved. Thus, the dependent units in (20) or (21) do not p ­ rovide

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

any i­nformation regarding empo ‘2sg.nom’, Aurelia or Ivan but denote an

event in which they act. When the participant is a common noun, we would
be in a borderline area; in this sense, complements are closer in function to
­‘non-restrictive’ clauses. Additionally, a Rel-clause tends to agree in case and
number with the modified noun, but a complement unit cannot. In (24a), the
head noun misi ‘cat’ serves as the object of ‘give’ and hence is marked by the
accusative suffix -ta. The complement unit in (24b) is also marked as accusative
but not because of nominal agreement, but because it is the direct core argu-
ment of the matrix predicate te’a ‘find’. The same is true for -’u constructions; in
(24c), the whole Rel-clause agrees with its plural head noun kaba’im ‘horses’;
regardless of the number of the participants, number agreement is completely
disallowed in nominal complements (24d).
(24) a. Jipi’ikim misi-ta miiika-Ø [pa’aku weama-m-ta].
milk.pl cat-acc give-prs  outside be around-clm-acc
‘Give milk (to) the cat that is outside.’
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 discovered that you already left.’
c. Min-Ø kaba’i-m bicha-k [Anselmo-ta jinu-ka-’u-m].
Fermín-nom horse-pl see-pfv  Anselmo-acc buy-pfv-clm-pl
‘Fermín saw the horses that Anselmo bought.’
d. Min-Ø [Anselmo-ta kaba’i-m jinu-ka-’u] bicha-k.
Fermín-nom  Anselmo-acc horse-pl buy-pfv-clm see-pfv
‘Fermín saw that Anselmo bought the horses.’

(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 

c. Aurelia-Ø ne bicha-k, [tajo’o-ta

Aurelia-nom 1sg.acc see-pfv  cloth-acc
ne/*nim baksia-ka-’u.
1sg.acc/gen wash-pfv-clm
‘Aurelia saw me washing the clothes.’

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.

Table 4.  Distribution of main and dependent subjects

Clause-type Correferential subjets

Sentential nominalization Dependent and main subject must be different

Subject & Agent relatives 83 clauses: the dependent and the main subjects are the same
54 clauses: the dependent and the main subjects are different

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 

(28) a. Mu’u-tai empo bicha-k [___i ito-t wam ne’e-ka-m-ta].

owl-acc 2sg.nom see-pfv     1pl-loc over fly-pfv-clm-acc
‘Did you see the owl that flew over us?’
b. Min-Ø bicha-k [ne kaba’i-ta jinu-ka-m-ta].
Fermín-nom see-pfv   1sg.acc horse-acc buy-pfv-clm-acc
‘Fermín saw me buying the horse.’
c. *Min-Ø ne bicha-k [ ___ i
 Fermín-nom 1sg.acc see-pfv
kaba’i-ta jinu-ka-m-ta].
horse-acc buy-pfv-clm-acc
‘Fermín saw me buying the horse.’
d. Aurelia-Ø ne bicha-k, [tajo’o-ta ne baksia-ka-’u].
Aurelia-nom 1sg.acc see-pfv  cloth-acc 1sg.acc wash-pfv-clm
‘Aurelia saw me washing the clothes.’

Furthermore, the strategy of pronoun retention recalls the syntactic/semantic

role of the head noun inside the Rel-clause (Keenan 1985: 148). Whereas sub-
ject and object relatives involve a ‘gap’, resumptive pronouns are more likely to
appear towards the right end of the accessibility hierarchy, e.g. indirect objects and
oblique relatives. In (29), the head noun wikiata ‘the lasso’ serves as an oblique
argument inside the Rel-clause, such as there is a resumptive pronoun a-e ‘with it’
co-indexed to it.

(29) inepo u-ka wikia-tai tamachia-Ø

1sg.nom the-acc lasso-acc measure-pres
[in a-ei kaba’i-ta jicho’ola-bae-’u].
  1sg.gen 3sg-inst horse-acc rope-desid-clm
‘I am measuring the lasso with which I will rope the horse.’

Complementation involves another type of pronoun retention. In a simple

clause, when a non-actor core argument is extraposed to the right in (4) above
and repeated below as (30a), there is a resumptive pronoun and the two units
are separated by an intonation break (Rude 1996). In the same way, when the
complement unit marked by -’u is extraposed to the right, then the main clause
takes a resumptive pronoun co-indexed to the complement unit as a whole, but
never to a participant. In (30b), the main predicate takes an accusative singular
pronoun a, but there is nothing inside the dependent unit with which a may
agree.
(30) a. U jamut-Ø ai = bicha-k Peo-tai.
det woman-nom 3sg.acc = see-pfv Pedro-acc
‘As for Pedro, the woman saw him.’
 Lilián Guerrero

b. Aurelia-Ø ai bicha-k, [enchii toto’i-m mea-ka-’u].

Aurelia-nom 3sg.acc see-pfv   2sg.acc hen-pl kill-pfv-clm
‘Aurelia saw (it), that you killed the hens.’

(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)

In contrast, the TAM information in a complement crucially depends on the mean-

ing of the matrix predicate. As for perception, for instance, there are important
restrictions. Since a direct perception situation (marked by the -me complement)
expresses a simultaneous perceived event, the dependent verb must be unmarked
or be equally marked in comparison with the matrix clause, it cannot be negated
(32a) and it disallows independent temporal adverbs; but all these features are fine
with the -’u construction (32b) describing a non-immediate perception.
(32) a. Ne kaa [Goyo-ta maska-ta jo’a-m-ta ] bicha-k.
1sg.nom neg  Goyo-acc mask-acc make-clm-acc see-pfv
‘I did not see Goyo making the mask/*I saw Goyo not
making the mask.’
b. Maria-Ø ai bicha-k [chubala enchi
María-nom 3sg.acc see-pfv  time.ago 2sg.acc
serbesa-ta je’e-ka-’u]i.
beer-acc drink-pfv-clm
‘María saw it, that you drank beer sometime ago (saw the
­empty ­bottles).’

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.’

Additionally, it is possible that certain kinds of events simply cannot be perceived

in particular ways. Nominalized complements marked by -me avoid stative predi-
cates (34a), while sentential nominalization marked by -’u does not (34b). Recall
that subject relatives prefer to occur with stative predicates in Table 3, above.
(34) a. *Armando-Ø [kafe-ta ama auka-m-ta] bicha-k.
 Armando-nom  coffee-acc there exist.pfv-clm-acc see-pfv
‘Armando saw there is coffee over there.’
b. Armando-Ø [kafe-ta ama auka-’u] bicha-k.
Armando-nom  coffee-acc there exist.pfv-clm see-pfv
‘Armando saw that there is coffee over there.’

Finally, in some languages, wh-embedded clauses make use of a nominalized

complement also, e.g. Huallaga Quechua (Weber 1989). Wh-embedded clauses
in Yaqui usually make use of the same structure as independent sentences, but
Dedrick and Casad (1999: 376–8) list a pair of examples where the wh-embedded
unit is marked by -’u; in all the examples, the matrix predicate is ju’unea ‘know’.
(34) a. Kan ne ju’unea jakun-bicha=sa bem saja-ka-’u.
neg 1sg.nom know where-site=q 3pl.acc go.pl-pfv-clm
‘I don’t know which way they went.’

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

of this phenomenon, especially those modifying subject and agent participants.

Yet, we have seen that constructions serving as core arguments for complement-
taking predicates have reacquired finite properties, albeit some structures more
than others. Thus, used in complement clauses, the sentential nominalization
marked by -’u shows less signs of nominalization compared to -me clauses, but still
it demands a non-nominative subject; in the same way, the nominalized comple-
ment marked by -me shows a higher degree of deverbalization compared to -’u
complements (i.e. in terms of the state of affairs encoded, the TAM operators and
negation). Finally, the nominal event marked by -me inside a matrix-construction
with bena ‘seem’ shows the highest degree of nominalization, since it avoids sub-
ject coding and any TAM operators.
A major question arises as to what is the association between relative rela-
tions and certain types of matrix predicates but not others? We may also won-
der whether the nominalized unit of complement-perception verb is just relevant
for Yaqui or whether it is found elsewhere. Cross-linguistically, it is not rare that
languages use the same constructions for relatives and other subordinate clauses,
especially when the language in question has few non-independent structures, e.g.
some Australian languages. The crucial point here is that for languages having
more than three subordinate structures, certain complement-taking predicates
systematically take a more nominalized sentence (at least as one of their alterna-
tives), where other predicates take a complement syntactically similar to indepen-
dent clauses, but never a nominalized clause. The use of these ‘pseudo-relatives’ as
syntactic complements occurs, firstly, with direct perception predicates; for Yaqui,
the -me clause also combines with two other image-mental verbs; some languages
include want-type predicates (e.g. Krongo, Reh 1985), and others certain knowl-
edge predicates like French (Lambrecht 1981; Koenig & Lambrecht 1999; van der
Auwera 1985); Huaraz Quechua (Miller 1989); Huallaga Quechua (Weber 1989);
Akatek & Jakaltek (Schüle 2000), Gulf Arabic (cited in Cristofaro 2003: 196), even
for Spanish (Guasti 1989; Borgonovo 1996). Following this cross-linguistic pat-
tern, several of the non-English examples of Malchukov (2006)’s study of the form
and function of nominalization are rather ambiguous between modifying and core
argument units.
A possible explanation may be the need for a participant sharing (Guerrero
2004, 2006a). In Rel-clauses the dependent unit provides a specification attribute
about a single participant. This property is used to uniquely identify this entity
within a set of possible referents (Cristofaro 2003: 197), or to restrict the referent
based on its attributive properties (Lehmann 1984; Bickel 2005). What’s more, an
act of direct perception involves a state of affairs as a whole, that is, the perceived
event and the entity bringing it about. In other words, we simultaneously see, hear,
or otherwise perceive not only the event going on but also the entities involved in
 On relative clauses and related constructions in Yaqui 

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.

6.  Final remarks

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

Álvarez, A. 2006. Cláusulas relativas y nominalización en yaqui. Paper presented at Seminario

de Complejidad Sintáctica. Hermosillo: Universidad de Sonora.
Andrews, A.D. 2007. Relative clauses. In Language Typology and Syntactic Description, Vol.2:
Complex Constructions, T. Shopen (ed.), 206–236. Cambridge: CUP.
Akmajian, A. 1977. The complement structure of perception verbs in an autonomous syntax
framework. In Formal Syntax, P.W. Culicover, T. Wason & A. Akmajian (eds), 427–460.
New York NY: Academic press.
Belloro, V. & L. Guerrero. 2010. Sobre la función discursiva de los argumentos ­pospuestos en
yaqui. In Análisis lingüístico: enfoques sincrónico, diacrónico e interdisciplinario, ­(Estudios
Lingüísticos 2), R.M. Ortiz Ciscomani (ed.), 334–381. Hermosillo: Universidad de
Sonora.
Bickel, B. 2005. On the typological variables of relativization. Paper presented at ‘On the
­Typology, Acquisition & Processing of Relative Clauses’. Leipzig: MPI.
Buelna, E. 1891. Arte de la lengua cahita por un Padre de la Compañía de Jesús. México:
Buelna.
Borgonovo, C. 1996. Gerunds and perception verbs. Langues et Linguistique 22: 1–19.
Carlson, G. 1977. Amount relatives. Language 53: 520–542.
 Lilián Guerrero

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

Pima Bajo a Uto-Aztecan language spoken in Northwestern Mexico, constructs

relative clauses by means of a verbal suffix which has diachronically emerged
from an stative/perfective suffix and a demonstrative; such a nominalizing
strategy is quite distinct from those observed in others Uto-Aztecan
languages from the same area. Furthermore relative clauses are only possible
for subject and object; the latter shows a dependent subject encoded via an
object/possessive pronoun, a characteristic feature observed also in verbal
complements of the verb ‘to want’ as well as in some adverbial clauses.
Functional equivalent constructions of oblique relative clauses show the relative
suffix functioning as a clausal connective, which announces a distinct pathway
of grammaticalization.

Keywords:  Grammaticalization; relative clauses; clause linker; Pima Bajo

1.  Introduction

A relative clause (RC) is usually taken to be express additional ­information about a

referent, a head noun of a noun phrase (NP), and is restricted to appear at the pro-
sodic boundaries of a clause (Wichmann 2000). In functional terms, a RC clause,
as well as other noun modifiers, is part of the grammar of r­ eferential ­coherence
(Givón 2001: 175). Formally, i.e. structurally, RCs have been ­characterized as
one  – together with complement and adverbial clauses –  of three subordinate
clauses which are (a) usually introduced by a particular set of e­lements, and
(b) show a gap, i.e. absence or omission, within the RC, of the element that is
relativized (Comrie & Horie 1995: 65). Semantically, the primary  function of
an RC is to provide “an assertion about the relative NP” (Downing 1978: 378);
 Zarina Estrada-Fernández

that is, it describes or provides a characterization of a referent which is the head

noun or pronoun of a noun phrase, as for example man in the example The man
who answered back immediately, or she, in She who is studying for a Ph.D. at the
­University of York.
Linguistic studies have discussed RCs from several different perspectives;
e.g. Keenan and Comrie (1977) and Comrie and Keenan (1979) approach the
study of RCs from the perspective of the Accessibility Hierarchy, where the
limitations – across languages as well as within a single language – of having a
RC relativizing the different syntactic positions or arguments are explained by
means of a h ­ ierarchy. This hierarchy claims that languages show some prefer-
ences concerning the argument to be relativized by a RC. So languages prefer
in the first place the subject to be relativized by a RC, after that the object,
and so on. Greenberg (1963) addressed the topic of word order between the
basic elements of a sentence – ­subject, object and verb – and its correlations
with the order of other constituents, among them the order of the RC in rela-
tion to the head noun. Keenan and Comrie (1977) and Comrie and Keenan
(1979), and later Comrie (1981, 1998) discuss the different strategies or mor-
phosyntactic means available c­ ross-linguistically for encoding the head noun
within the RC: the relative pronoun strategy, the non-reduction or retention of
the head within the RC, the resumptive pronoun, and finally the gap or omis-
sion, i.e. zero anaphora. Other properties such as the non-finite morphology,
the ­relative marker on the verb, and the similarities that some languages show
between the relative and complement clauses are also considered by ­Comrie
(1996) and Comrie and Horie (1995). Among other discussions, the topic of the
­different degrees of finiteness or nominalization of the RC has been addressed
by ­Lehmann (1984, 1992), while the grammaticalization, or diachronic devel-
opment, or origin of RCs, has been addressed by Hopper and Traugott (2003)
and more recently by (Givón 2009: ­97–120, and this volume).
For Uto-Aztecan languages the linguistic literature specifically concerning
RCs is scarce, although such constructions have been widely described in almost
all the grammars written for the different languages of this linguistic family.
For Yaqui, for example, RCs have been described by Guerrero (2005), as well as
­Lindenfeld (1973), and Dedrick and Casad (1999); for Cora by Casad (1984), and
recently by Vázquez (2002); for O’odham (previously Papago) by Saxton (1982);
for Tarahumara in Burgess (1984). For Pima Bajo, a collateral discussion concern-
ing the diachronic origin of the relative marker -kig is provided by Hale (2002),
and a brief structural description of these constructions by Estrada (1991, 1996).
Most of the discussions addressed in such studies, however, deal with only the
structural properties of such constructions and almost no diachronic analysis has
 From demonstrative to relative marker to clause linker 

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.

2.  Basic facts of 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 

Chart 1.  Personal pronouns

Subject pronouns Non-Subject pronouns

Independent pronouns Reduced pronouns Prefixed pronouns

1SG aani aan in-

2SG aapi aap am-
3SG hig hig a-
1PL aatim aat tit-
2PL aapim aap mim-
3PL higam higam a-

(3) a. ig okosi a’asi’-im.

det.sbj woman laugh-cont
‘The woman is laughing.’
b. kuidag vaak ig tas.
west enter.inside.pfv det.sbj sun
‘The sun goes down in the west.’
c. Marii ik lii oob vak-va.
Maria dem.nsbj dim persona bath-cmpl
‘Maria already bathed the child.’
d. takav Huaan am Mais-tam oimor-tad.
yesterday John loc Maycoba-loc wander-rem
‘Yesterday John was wandering in Maycoba.’
e. aan am dasa-hag ik hod am divar-tam.
1sg.sbj loc leave-fut det.obj stone loc floor-loc
‘I will leave the stone on the floor.’
f. kavlio si mil-ia.
Horse int run-prob
‘The horse will run faster.’
g. ig ki-kil oidig mihi.
det.sbj pl-men land burn.prs
‘The men burn the field.’

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.

(4) a. kareet niar pork kua’ag ab-ivar buagi-an.

wagon buy.pfv sub firewood dir-imp.pl carry-irr
‘I bought a wagon in order to carry firewood!’
 Zarina Estrada-Fernández

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.

3.  Subject and object relative clauses in Pima Bajo

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.

In the majority of languages, the nominalization of a clause is addressed as a scalar

phenomenon or a syntactic process that shows different degrees of adjustment
according to the occurrence of certain morphosyntactic features in the clauses.
Givón (2009: 67) provides the following list of major components as those that are
observed in the process of adjusting from a finite clause into a nominalized one,
i.e. a non-finite clause:

(6) a. The verb becomes a head noun.

b. The verb acquires nominal morphology.
c. The verb loses tense aspect-modal marking.
d. The verb loses pronominal agreement marking.
e. The subject and/or object assume genitive case-marking.
f. Determiners may be added.
g. Adverbs are converted into adjectives.
 From demonstrative to relative marker to clause linker 

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

Furthermore, according to Lehmann (1984: 49), agglutination of the subordinator

to the verb may be optionally accompanied by the presence of a relative pronoun,
and the limited choice of verbal categories may vary from zero to a limited range
of tense-aspect-mode (TAM) markers within the RC.
Based on Givón’s and Lehmann’s list of components, Pima Bajo RCs are
­characterized as not fully nominalized, since RCs, as I will show below, have some,
but not all, of the features which are relevant for the nominalization of a clause.
There are three relevant morphosyntactic aspects that I shall consider for the
analysis of RCs in Pima Bajo: (a) all RCs are post-nominal, so concerning their word
order they do not pattern along with adjectives, but they do qualify as embedded
clauses since they are always placed within the limits of the noun phrase where the
head noun appears;4 (b) RCs are clearly identified by a marker -kig modifying the
verb inside the RC; and (c) the relative marker -kig is now half way to being gram-
maticalized as a general clause linker or connective; as a result of this last stage,
non-conventional cases of an oblique kind of relatives emerge in the grammar of
Pima Bajo. The striking fact in this language is that the relative marker -kig is the
result of an original participial suffix -ka (Hale 2002) that used to mark non-finite
clauses; this suffix evolved, losing its functionality and after being fused into an
emphatic definite independent demonstrative higai (Estrada 2008) becomes rean-
alyzed as a relative marker.5 As a result of this change, the relative marker in Pima
Bajo is now a quite distinct strategy for relativization in comparison with other
nominalizing suffixes observed in other Uto-Aztecan languages from the same
geographical area, e.g. Opata, Yaqui, Tarahumara; as well as the relative marker
described in O’odham by Saxton (1982) (cf. Section 4).6

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.’

Externally-headed subject RC with a relative pronoun introducing the RC:

c. huaan [hig am dah-kig.]
John 3sg.sbj loc be-sit.sg.pfv-rel
‘John is the one that is sitting there.’

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.’

The possessive, genitive, or accusative subject – the NSBJ ‘non-subject’ in this

paper – and the omission (zero anaphora or gap) of the nominal which is coref-
erent with the head noun are two well known relativizing strategies that make a
clear-cut distinction between subject vs. object RC in Pima Bajo.
Furthermore, examples in (8) and (9), but most of all, those provided in
(10) show that RCs in Pima Bajo are not fully nominalized or non-finite, like for
­example the non-finite construction shown in (4). Examples in (8), (9) and (10)
show that the verb within the RCs can be modified by a restricted set of finite tense-
aspect-mode (TAM) suffixes: truncation of the root for ‘perfect’, an unmarked root
(UNMRK) for ‘present’, ‘habitual’, or imperfective, -im for ‘continuous’ or ‘progres-
sive’, and the suffix -tad for the ‘remote’ or ‘past imperfective’:
(10) a. ig okosi [am no’ok-im-kig ] gi-gi-d.
det.sbj woman  loc talk-cont-rel pl-tall-cop
‘The woman who is talking there is tall.’
b. ti-trav [ am=kompig-kig ] sartem-tam.
pl-potatoes  2sg.nsbj=peel.unmrk-rel frying_pan-loc
‘The potatoes that you are peeling are in the frying pan.’
c. lii oob [ suak-im-tad-kig ] si todag niin.
dim person  cry-cont-rem-rel nt afraid wake_up.pfv
‘The boy who was crying woke up afraid.’

Non-finite tempo-aspectual markers may also modify the relativized verb, in

­particular, the future or prospective suffix -(d)ag, in (11):

(11) Marii kav maa ik kilia [ duvi-ag-kig ]

Maria horse give.pfv det.sbj man   come-prosp-rel
‘Maria gave the horse to the man that will come.’

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

(12) a. aan si’ tikpaan ko ap iskueel-tam dah-ka-(d)am.

1sg.sbj int work.pfv sub 2sg.sbj school-loc be-st-nmlz
‘I worked a lot in order for you to enter school.’
b. gogosi in=kii in=mat-ka-ad.
dog 1sg.nsbj=bite.pfv 1sg.nsbj=know-st-cond
‘The dog bit me although he knew me.’
c. aan su-spon gagar huun aan nial-di-an.
1sg.sbj pl-chicken sell.pfv corn 1sg.sbj buy-apl-irr
‘I sold chickens to buy some corn.’

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

According to Andrews (2007: 232) nominal morphology attached to the verb is a

typical indicator of the nominalization of a syntactic element. Determiners are one
of the nominalizing strategies. In Pima Bajo, there are no other traces, vestiges or
relics of other nominalizing markers in RCs except the suffix -kig. In other words,
the relative marker -kig structure is the only available construction for RCs in Pima
Bajo. In this respect, Pima Bajo shows no correlation with other U ­ to-Aztecan lan-
guages from the same geographical area.
Uto-Aztecan languages from Northwestern Mexico, e.g. Opata, Yaqui and
Tarahumara, build RCs by means of nominalizing suffixes, i.e. suffixes that func-
tion to derive nouns as well as to “desentialize” sentences (cf. Lehmann 1988).
Other languages, i.e. Tohono O’odham, have subordinated RCs. In this section I
shall describe the historical development of the relative marker -kig and provide
data from other Uto-Aztecan languages to show that this particular relative con-
struction is only characteristic of Pima Bajo.
At the end of a manuscript dealing with RCs in Dagur (Altaic family), Hale
(2002), briefly sketches what he considers to be the origin of the relative marker
-kig in Pima Bajo. Hale’s analysis proposes a determiner hig to adjoin or attach to
the end of a perfective participial suffix -k(a). Hypothesized examples given by
Hale in support of his proposal are the following:
(15) a. kil [ tikpaan-im-k ] hig in=’aam.
man   work-impf-prtc det 1sg.gen=father
‘The man who is working is my father.’
b. kav [ in=niar-k ] hig gi’id.
horse  1sg.gen=buy-prtc det big
‘The horse I bought is big.’

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.’

Such nominalized structures are completely absent in the grammar of Tohono

O’odham (Papago), a language also from the Tepiman branch, as well as Pima

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 

Bajo. In Tohono O’odham, RCs are, according to Saxton (1982), introduced by

means of a subordinator m(a), and ordered after the head noun,10 e.g. (18a–b).
This is quite a distinct strategy of relativization in comparison to the one that was
described for Pima Bajo.
(18) a. higam a-’al m-o am čičwi ]
those pl-child sub-m loc play
‘Those children that are playing…’ (Saxton: 1982: 250)
b. aapi [ ma-m ia gi-gok]
you  sub-you loc pl-stand
‘…you who are standing here…’ (Saxton: 1982: 252)

Analysis of RCs in other Uto-Aztecan languages from the Sonoran region –

Opata, Yaqui, Tarahumara – also support the evidence that these languages use
­nominalizing suffixes as the most common relativization strategy. In Opata, for
example (cf. Lombardo 1702), another extinct Uto-Aztecan language, RCs show a
stative suffix -ca attached at the end of the verb within the RC:
(19) a. Cauadu [ no eme maquea-ca ] müu.
horse  1sg.nom 2sg.acc give.fut-st die.pfv
‘The horse which I was going to give you died.’
b. Masö-ta [ amo ne mac-ca ] chitta meâue.
deer-acc  2sg.nom 1sg.acc give-st dog.nom kill.pfv
‘The dog killed the deer that you gave me.’

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.’

For numerous languages, however, the development of a demonstrative into a

RC marker, as it is the case of Pima Bajo, has been fully accepted (cf. Heine &
Kuteva 2002). Moreover, the development of the demonstrative higai into a rela-
tive suffix -kig follows the pattern observed for Koyoga-Bambara by Kuteva and
­Comrie (2006). In these languages, the development of the demonstrative starts

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 

in a paratactic construction, i.e. two independent clauses. As a result of this gram-

maticalization pathway, the structures are no longer paratactic, but embedded,
a situation which seems to be identical to the one observed in Pima Bajo if we
­consider examples like (16a).13

5.  Oblique constructions: -kig as a path to a 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.’

Paratactic constructions are also observed when an attribution of a possessor or

a genitive argument is expressed within the RC; in (27) the attribute follows its
­lexical head kaps ‘captain’:

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

(27) higi kaps kaps [ higi vipig no-nov ] ia’i hukan.14

det captain captain  3sg.sbj red pl-sleeve loc up_there
‘The captain, the captain, he whose sleeves are red, was up there.’

Constructions headed with a WH-pronoun as subordinator element are another

possibility for describing a referent with a locative function; (28a–b) are examples
of this:
(28) a. ig kil vuus ki-tav [ ibiga gogos mua.]
det.sbj man go_out.pfv house-dir  where dog die.pfv
‘The man left the house where the dog died.’
b. in-maak-in kaha [ ibiga api serii ab dadas. ]
1sg.nsbj-give-imp box  where 2sg.sbj matches dir keep.rem
‘Give me the box where you used to keep the matches!’

However, in this language a quite different type of construction is also possible to

express an attribute over an oblique argument. In such alternative constructions
what was previously known as a relative marker, the suffix -kig, appears within the
boundaries of the main clause. In examples (29–31), the suffix -kig appears not
directly attached to the verb but to the oblique noun or postpositional phrase. For
this reason, examples in (29–31) show the -kig glossed as lin ‘linker’, and a suf-
fix -ta which I will refer to later. These kinds of constructions are not taken to be
RCs, but as functional equivalent structures for oblique arguments. Example (29)
shows an equivalent structure for an RC of an instrument, (30) for a goal (alla-
tive case), and (31) for a benefactive argument. I will discuss such ­constructions
below:15
(29) aan hivga-di tiipar-ta-kig aap ko’ag in=taata-di.
1sg.sbj lend-appl ax-ta-lin 2sg.sbj firewood 1sg.nsbj=lend-appl
‘I lent you the ax the one that you cut me the firewood with.’
(30) aap timitim maa ik
2sg.sbj tortillas give.pfv det.obj
okosi-vui-ta-kig in=tikpan.
woman-dir-ta-lin 1sg.nsbj=work.pfv
‘You gave tortillas to the woman whom I worked for.’

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 

(31) aan kafee niar ik

1sg.sbj coffee buy.pfv det.obj
kil-vuika-kig am=gigsa-di.
hombre-ben-lin 2sg.nsbj=shake_hands-appl
‘I bought coffee from the man whom you shook hands with.’

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.’

6.  Final remarks: Typological properties and linguistic change

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:

aita’a am-taan higai?

inter 2sg.nsbj-ask.pfv, 3sg.sbj
‘What did she asked you for?’
 Zarina Estrada-Fernández

allow RCs of oblique arguments demonstrates that in terms of the Accessibility

­Hierarchy, Pima Bajo only allows the RCs formation for the first two positions of
the hierarchy provided by Keenan and Comrie (1977).
(33) subject > direct object > indirect object > oblique > genitive > object
of comparison

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 

inherently patterns. Typological affinities among Uto-Aztecan l­anguages of North-

west Mexico, e.g. Pima Bajo, O’odham, Tarahumara, and Yaqui, are well-known, but
the typological properties that draw our attention are those that show that the lan-
guages are following different paths of grammaticalization and typological change.

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

Givón, T. 2009. The Genesis of Syntactic Complexity. Amsterdam: John Benjamins.

Greenberg, J.H. 1963. Some universals of grammar with particular reference to the order of
meaningful elements. In Universals of Language, J. Greenberg (ed.), 73–113. Cambridge
MA: The MIT Press.
Guerrero, L. 2005. Yaqui relative clauses. Proceedings from the Eighth Workshop on American
Indigenous Languages (2005) 16: 15–27. Santa Barbara CA. The University of California.
Hale, K.L. 2002. On the Dagur object relative: Some comparative notes. Journal of East Asian
Linguistics 11: 109–122.
Heine, B. & Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
Hopper, P. & Traugott, E.C. 1993[2003]. Grammaticalization. Cambridge: CUP.
Keenan, E.L. & Comrie, B. 1977. Noun phrase accessibility and universal grammar. Linguistic
Inquiry 8(1): 63–99.
Kuteva, T. & Comrie, B. 2006. The typology of relative clause formation in African languages. In
Studies in African Linguistic Typology [Typological Studies in Language 64], F.K.E. Voeltz
(ed.), 209–228. Amsterdam: John Benjamins.
Langacker, R.W. 1977. Studies in Uto-Aztecan Grammar, Vol. 1: An Overview of Uto-Aztecan
Grammar. Dallas TX: Summer Institute of Linguistics – University of Texas at Arlington.
Lehmann, C. 1984. Der Relativsatz. Typologie seiner Strukturen. Theorie seiner Funktionen. Kom-
pendium seiner Grammatik. Tübingen: Gunter Narr.
Lehmann, C. 1986. On the typology of relative clauses. Linguistics. 24: 663–80.
Lehmann, C. 1988. Towards a typology of clause linkage. In Clause Combining in Grammar
and Discourse [Typological Studies in Language 18], J. Haiman & S.A. Thompson (eds),
181–225. Amsterdam: John Benjamins.
Lehmann, C. 1992. Relativization. In International Encyclopedia of Linguistics, Vol. 3, W. Bright
(ed.), 333–335. Oxford: OUP.
Lindenfeld, J. 1973. Yaqui Syntax [University of California Publications in Linguistics 76].
Berkeley CA: University of California Press.
Lombardo, N. 1702. Arte de la lengua tegüima vulgarmente llamada ópata. México: Miguel de
Ribera.
Saxton, D. 1982. Papago. In Studies in Uto-Aztecan Grammar, Vol 3: Uto-Aztecan Grammatical
Sketches, R.W. Langacker (ed.), 93–266. Arlington TX: Summer Institute of Linguistics.
Smith, B. (ed). 1862. Grammar of the Pima or Névome, A Language of Sonora, from a Manuscript
of the XVIII Century. New York NY: Cramoisy Press. (AMS Press, 1970).
Troike, R.C. 2010. Relative clauses in Coahuilteco, an Indian language of Texas. Southwestern
Journal of Linguistics 29(1): 111–32.
Vázquez Soto, V. 2002. Cláusulas relativas en cora meseño. In Del cora al maya yucateco. Estudios
lingüísticos sobre lenguas indígenas mexicanas, P. Levy (ed.), 269–348. México: Universidad
Nacional Autónoma de México.
Wichmann, A. 2000. Intonation in Text and Discourse. Beginnings, Middles and Ends. London:
Longman.
 Functional underpinnings of diachrony
in relative clause formation
The nominalization-relativization connection
in Northern Paiute

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.

Keywords:  Relative clauses; nominalization; subordination; diachronic syntax

1.  Introduction1

Northern Paiute (Western Numic; Uto-Aztecan) forms relative, or attributive,

clauses through nominalization. Givón (1990) describes nominalization as one
of seven strategies, found cross-linguistically, employed by languages in the for-
mation of relative clauses. One theoretical question is whether to consider nomi-
nalization as a strategy per se available to languages in the formation of relative
clauses or, rather, to treat nominalization and relative clause formation as part of
the same complex functional and grammatical domain. The latter approach has

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

1.1  Relevant basic clause properties/Typological profile

The Northern Paiute language represents the furthest extent to the north and
west of the Uto-Aztecan language family. It consists of two major dialects and
numerous subdialects, although the dialect situation prior to the establishment
of the reservation system is not well understood (Nichols 1974 provides the most
detailed explication of their distribution). Around 400 speakers of the language
currently reside on or are associated with several small reservation communities,
from south-central Idaho across eastern Oregon and north-western Nevada to
parts of eastern California. Along with Mono, it makes up the Western Numic
subbranch of the Numic branch of Uto-Aztecan. The field and supporting data
for this work come mainly from dialects spoken to the north of a major isogloss,
what Nichols (1974: 4) refers to as Oregon Northern Paiute. Most of the field data,
both from text and text-based elicitation, come from the dialect presently spoken
by elder members of the Burns Paiute Tribe in the Harney Valley region of eastern
Oregon. Syntactic differences across the dialects are minimal, and so I assume the
data here to be reasonably representative of the language. Indeed, given the gen-
eral morpho-syntactic profile of Numic more broadly, it is reasonable to assume
that the hypothesized historical developments proposed here pre-date Numic, to
a large extent.
Northern Paiute carries many of the typological features of an SOV language,
including a moderately flexible, verb-final word order. The verb complex consists
of an array of operators on inherent transitivity, verbal aspect, associated motion
and direction, and phasal notions (e.g. of ability or attempt), among other pro-
cesses. The morphological structure of the verb could be considered only moder-
ately polysynthetic, since core arguments are not indexed in the verb. The language
has postpositions, possessors precede possessums, modifiers are prenominal, and
relative clauses are postnominal, with a certain degree of flexibility.
Some of the morphosyntactic properties of Northern Paiute relevant to this
study include the overt coding of two core cases, what I term nominative and
oblique. Core argument alignment follows a nominative-accusative pattern,
although the somewhat threadbare nominal case forms found elsewhere in Numic
are gone in Northern Paiute, leaving case to be overtly specified only on noun phrase
dependents – modifiers and determiners – and distinct pronominal forms. Distinct
suffixes appear on modifiers: -ʔyu denoting nominative case and -(k)u denoting
non-nominative, or oblique case. Definite noun phrases denoting singular entities
are marked with determiner proclitics that follow the same pattern: su = for nomi-
native and ka = for oblique (mi = applies to all plurals, regardless of case).
Pronominal forms are presented in Table 1, with the (always) independent
subject pronouns first, followed in the next column by the direct object/possessor
proclitic forms. The independent object and possessor pronouns appear to serve
 Functional underpinnings of diachrony in relative clause formation 

a more emphatic or contrastive function. Phonological reduction of the indepen-

dent third person subject and object pronouns likely gave rise to the case sensitive
determiner proclitics of Northern Paiute, another symptom, one could argue, of
the paratactic-to-syntactic route of development in complex clause formation.

Table 1.  Northern Paiute subject and object pronouns

Pronouns Subject Object/ Object Possessor
pronouns possessor pronouns pronouns
(Independent) proclitics (Independent) (Independent)

1.Singular nɨ i= nɨka nɨga

1.Plural Exclusive nɨmmi mi=, ni= nɨmmika nɨmmiga
1.Dual ta ta= taka taga
1.Plural. Inclusive tammi ti= tammika tammiga
2.Singular ɨ ɨ= ɨmi ɨmi
3.Singular isu, usu, masu u=, ø= ika, uka, maka iga, uga, maga
(demonstratives)
2/3.Plural ɨmɨ mɨ= umɨ umɨ
Interrogative haga haka hagatɨ
4 (Indefinite) a=
Reflexive Possessor tɨ=
(Logophoric)

Examples (1)–(2) illustrate the core case-sensitivities of noun phrase depen-

dents and pronouns.3
(1) su= udɨ-ʔyu naatsi kima-u-gina
nom= tall-nom boy come-pnc-cisl
‘The tall boy is coming this way.’

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

(2) ni u= punni, ka= udɨ-u naatsi

I 3= see obl= tall-obl boy
‘I see him, the tall boy.’

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

1.2  Possession in Northern Paiute

Possession is accomplished by a simple juxtaposition of full noun phrases in the
order possessor followed by possessum with no special marking on either noun
phrase, as in Tim kaazi (Tim car) ‘Tim’s car’ and ohaʔa pia (baby mother) ‘baby’s
mother.’ When the possessor is pronominal, it most often appears as a partly
bound proclitic, as in i=kaazi ‘my car’ and u=pia ‘its mother’ and less commonly
as an independent pronoun, as in uga kaazi ‘her/his car’ and nɨga pia ‘my mother.’
A special feature of possession in Northern Paiute is the existence of a logo-
phoric (reflexive) possessor proclitic tɨ=, used to indicate coreferentiality with the
grammatical subject (“crane” in (3) and “they” in (4)).5
(3) su= wassa pɨnʔo nainapa mɨda-u tɨ = kaupa
nom= sandhill.crane as.for across stretch-pnc logo= leg
‘... the crane stretched his legs across…’ (NK: ‘Bear and Deer’)
(4) oʔo yaisi umɨ ka= tɨ= nobi naʔuna-wai
dem then they obl= logo= house around-loc
‘… so then they went round and round their house …’
 (NK: ‘Bear and Deer’)

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’)

The role of possession as a property of nominalized clauses is important for the

understanding of relative clause formation in Northern Paiute in that it also serves
to distinguish these from formally similar subordinate clause types, in particu-
lar, adverbial clauses (discussed briefly in Section 4). The development of finite
properties from older nominalizations is well-attested cross-linguistically, and can
account for both formal and functional associations between various subordinate
clause types in languages like Northern Paiute. After a look at the nominalizers
that play a role in relative clause formation in the language in the next section, we
turn specifically to the formal properties of relative clauses themselves.

1.3  Nominalizers in Northern Paiute

There are several ways to derive nouns from verbs in Northern Paiute. Such
nominalizations commonly derive a participant in some particular relation to the
underived verb, such as an instrument or a place, or they may serve to name the
event or state denoted by the verb. For our purposes it is necessary to focus our
attention on the two nominalizers that play a significant role in the formation of
relative clauses.6 These are the subject (S/A) nominalizer -dɨ (NMR) and what I
have elsewhere (Thornes 2003: 127–129) termed the participle suffix -na (PTCP).7
The subject nominalizer provides a means for describing occupations or habit-
ual actions. Like the -er of English (cf. hoawai-dɨ (hunt-NMR) ‘hunter’; tiničui-dɨ

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)”)

The participle suffix performs the function of an action nominalizer in examples

like the following:
(8) i= nossi-na waha-na
1= dream-ptcp tell.of-ptcp
‘…telling my dreams/dreaming…’
(9) yaa i= haba-na
Here 1= shade-ptcp
‘Here’s my shadow/shading.’

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.

2.  Relative clauses in Northern Paiute: A typological profile

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.

2.1  Subject relative clauses: Basic properties

The verb of a subject relative clause in Northern Paiute is marked by the sub-
ject nominalizing suffix -dɨ (NMR). The notional object of the relative clause is
­case-marked oblique, based on its relation to the dependent verb. There is no core-
ferring expression of the subject within the relative clause; that is, the “gap strategy”
is used to recover the case role of the head noun within the relative clause (as well
as the use of the subject nominalizer). The head nominal and the n ­ ominalizing
 Tim Thornes

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’)

Extraposition of an object requires two features, a coreferring pronominal proclitic

and comma intonation (pause) after the main clause verb. Separate intonation
contours, according to Givón (2007, 2009), are a primary diagnostic of paratac-
tic, non-embedded structures – the historical precursor to syntactic, embedded
structures. The Northern Paiute data fit well within his proposed ­non-restrictive

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 

­arenthetical “mega-pathway” in the evolution of embedded (­

p nominalized)
­relative clause types. That the features of extraposed nominal objects should also
hold true for nominalized clauses that are formally and functionally identical to
relative clauses strengthens our supposition that nominalization is not a relative
clause strategy, per se, but that relativization is in fact one function of nominaliza-
tion in the language – that is, one of the purposes toward which nominalization is
put. Nominalizations can still be construed as “restricting the reference to those
referents of which a particular proposition is true.”

2.2  Object relative clauses: Basic properties

The verb of an object relative clause is marked by the participle/action nominal-
izing suffix -na (PTCP). The notional object is “gapped” inside the relative clause.
The notional (A) subject can be interpreted as the possessor (oblique case-marked
object) of the dependent verb and, when pronominal, appears as a verbal proclitic.
In the following example, the head nominal and participle suffix are in boldface, as
well as the pronominal indicating the notional subject. The relative clause appears
in square brackets.
(14) su= miidɨ [i= kuhani-na] kai toki kamma
nom= meat  1= cook-ptcp neg correct taste
‘The meat I cooked doesn’t taste right.’

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…’

The appearance of the logophoric possessor pronoun in this construction helps to

support the analysis of relative clause formation in Northern Paiute as one func-
tion of nominalization. Unlike other Numic languages (e.g. Ute as described by
Givón 2009), Northern Paiute does not mark case directly on nouns and has all
but lost the distinction between accusative and genitive case marking elsewhere.
Due to the object/possessor syncretism in pronominal proclitics described above,
the proclitic i= referring to first person in (14) could be interpretable syntactically
either as object or possessor. The use of the logophoric pronominal proclitic in
(15), however, renders the latter interpretation most appropriate. Further, I would
argue that the comparative evidence from other Numic languages with slightly
 Tim Thornes

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.

2.3  Oblique relative clauses: Basic properties

As with object relative clauses, in an oblique relative clause the verb is marked
with the participle suffix -na (PTCP). Also, the notional subject has the gram-
matical properties of the possessor (object) of the dependent verb – either oblique
(­non-subject) case-marking or verbal proclitic status.
The main difference, however, lies in the fact that the non-core argument is
not gapped inside the relative clause, but expressed by the restrictive (relative) pro-
nominal base pɨ- (RESTR) which serves as the host to a postposition indicating
its case role in the relative clause. The head referring expression, the verbal suffix,
the notional subject of the relative clause, and the restrictive pronominal base all
appear in bold. Again, the nominalized (relative) clause appears in square brackets.
(19) oo [pɨ-kwai ni= maʔnaʔwi-na]
dem  restr-loc 1.pl.excl= do/act-ptcp
‘There where we played…’ (NK: ‘Boarding School’)

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 

(22) [paa pɨ-tu huu-na]

 water restr-loc flow-ptcp
‘(the place where) water flows through it.’ (MS: ‘Autobiography’)

In (21), the speaker is referring to an established character from the narrative

using the headless relative clause, whereas in (22), the speaker employs the same
construction to invoke the idea of an irrigation ditch – something the language has
no clear term for. Such communicative strategies are common in natural speech,
and the presence versus absence of a head, or domain, noun over which reference
is narrowed appears to be a secondary feature of the construction. Example (20)
could, in some analyses, also be headless, since the demonstrative form mean-
ing ‘that kind’ does not carry the necessary semantic features of some referential
domain. Indeed, as I point out in the summary, headless relative clauses are very
common in natural speech, serving themselves a referential function, rather than
restricting the reference of a separate nominal. This, in combination with a ten-
dency toward “refinitization” of nominalized clauses described in Givón (1994),
could represent a process similar to that described by Epps (2007) for Hup – the
development of adverbial (converb) clauses from headless relatives. I return to
these theoretical issues later in the paper. First, I would like to briefly explore the
pragmatic role of (detransitive) voice marking in the formation of relative clauses,
since these also play an important part by altering the referential foci of the result-
ing constructions.

3.  Detransitive (voice) marking and 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

some further support to the postulation of a non-embedded syntactic pathway of

relative clause formation. Further diachronic support needs to be sought through
a comparison of frequency patterns across Uto-Aztecan.

3.1  The middle and antipassive prefixes

Northern Paiute has two prefixes that serve general detransitivizing functions in
the language. As such, they can be considered part of voice-type constructions in
that their appearance typically correlates with alternations in the relative topicality
of a verb’s core arguments. Their interaction with relative clauses is taken up in the
next section.
Elsewhere (Thornes 2008) I have explored the wide range of functions and
proposed a historical account for the verbal prefix na- in Northern Paiute – what
I have termed the middle marker (MM). Two of the most important of these, the
reflexive and passive functions, are exemplified below:

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.’

(26) a. nɨ miidɨ kuhani-kɨ-u-kwɨ

I meat cook-apl-pnc-fut
‘I’ll cook meat for someone.’
b. u-su i=tɨ-kuhani-kɨ
3-nom 1=aps-cook-apl
‘S/He’s cooking for me.’

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.

3.2  Interaction between voice and nominalization/relativization

In an earlier description of relative clause formation in Northern Paiute (Thornes
2003: 428–438), I discuss the phenomena of headless relative clause verbs marked
with either of the two detransitivizing prefixes just described. That discussion con-
sidered just two of the four logical possibilities for combining voice morphology
with one of the two distinct nominalizers discussed above (Section 1.3). These
four possibilities include: (1) the passive in concert with the subject nominal-
izer (the na-Verb-dɨ combination), (2) the antipassive in concert with the subject
nominalizer (the tɨ-Verb-dɨ combination), (3) the passive in concert with the par-
ticiple (the na-Verb-na combination), and (4) the antipassive in concert with the
participle (the tɨ-Verb-na combination), One can readily find examples of each of
these combinations and interpret them as nominalized clauses or headless rela-
tive clauses with different communicative foci. Let us consider examples of each
combination in turn.
The na- Verb -dɨ combination. In this combination, the head, or domain
noun is most typically the promoted subject-of-passive (O > S), but may also be
 Tim Thornes

subject of reflexive, reciprocal, or other construction related to the middle marker.

The relative clause is frequently, if not typically, headless as in (27):
(27) oʔo iwa-ʔyu su= uuni-ʔyu naʔa, su= [na-tɨhona-dɨ]
dem many-nom nom=that.kind-nom grow nom= mm-dig.roots-nmr
‘There’s a lot of that kind growing out there, for the digging.’
(NK: ‘Bear and Deer’)

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)):

(29) mɨ = himma [wɨtsimoʔo pɨ-mma na-kwiba-na]

pl= what  ball restr-instr mm-strike-ptcp
‘those things with which balls are hit (i.e. baseball bats)’
 (NK: ‘Boarding School Days’)

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 

The tɨ- V -na combination. In this combination, it is the would-be head, or

domain noun (notional object) that is unspecified, generating another kind of
headless relative clause, as in the following:
(30) tɨ= tɨ-da-kwɨhɨ-na, pɨnau owi-tu hani-ʔyakwi.
logo= aps-ip/FEET-get-ptcp back there-to do-hab
‘... what it grabs (with its claws), (it) brings back there.’ (NK: N
­ emechozinna)

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’)

The prevalence of all of these combinations make for important consideration

regarding the functional load borne by nominalized clauses in languages like
Northern Paiute. They are themselves referring expressions, with clear traces of
their beginnings as lexical nominalizations that proceeded partway through the
stages of expansion as described by Heine and Kuteva (2007) and Heine (2008).
These stages essentially include the development of more and more clause-like
properties, generally at the expense of, or gradually replacing, nominal prop-
erties – a process reflecting, to a great extent, the reverse of what Givón (1990,
­Chapter 12) describes as “adjustments to the finite-clause prototype due to nomi-
nalization (499)”, or, later, as a process of “finitization” (Givón 1994).
We turn briefly to subordinate clause types that share formal properties
of relative clauses by way of situating relative clauses within a continuum of
 Tim Thornes

­ evelopment. These types include verb complements and adverbial (converb)

d
clauses. The postulation of the notion of “expansion” would appear to entail a
reversal of the u
­ niversal diachronic trend from parataxis to syntaxis.

4.  A (brief) look at other functions of the participle suffix

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].

(33) nɨ [ɨ= čadua-na] naka -supidakwatu.

I  2= talk.sg-ptcp hear -understand
‘I understood what you said.’

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.

(34) u-su [ka= nɨmɨdzoho u= nagi-kya-na] punni.

3-nom  obl= People.Masher 3= chase-trnsl-ptcp see
‘S/He saw the Nemedzoho chasing her/him.’

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:

(35) ɨ ka= ɨ= tuamɨ tammakwɨ -na umɨ oiʔyu-na

you(nom) obl= 2= children abandon-ptcp hey be.there-SIM.CONV
nobi-kwai-ʔyu-na tui-haʔu-tui maʔnaʔwi-na
house-loc-pred-ptcp any-how-any behave-ptcp

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 

ooʔnogasi sɨda ʔmaʔnai-ʔyakwi uu.

at.that.time.then badly do/act-hab thusly
‘If you leave your children behind, and they stay at home, (they) get into
mischief, and do bad things like that.’ (NK: ‘Youth, then and now’)

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

The nominalizing character of relative clauses in Northern Paiute appears to fit

well within Givón’s (2007) diachronic typology, exhibiting features that corrobo-
rate a paratactic-to-syntactic historical trajectory. Clearly headed relative clauses
appear to be somewhat rare in natural speech, however, while headless relative
clause structures are common, either with or without voice alternations.
Northern Paiute has two clearly distinct dependent verb forms used in
relative clause formation, associated here with different nominalizers. There
remain key differences in the formal realization of the subject, particularly their
­genitive-possessive nature in non-subject relative clauses. Properties of nomi-
nalization and the frequency in the use of these constructions as either referring
expressions themselves (that is, without a head noun) or for nominal appo-
sition have been described in detail here as lending possible support for the
­paratactic-to-syntactic path of historical development.
Data that combine relative clause structures with detransitive voice mor-
phology demonstrate a means for narrowing reference to a particular participant
associated with the nominalized clause. Investigation into whether or not data
of this sort represent functional correlates with Shibatani’s (2007) discussion of
­focus-type constructions in Austronesian languages is needed. What is clear, how-
ever, is that detransitive (voice) constructions operate independently of relative
clause type – that is, there is no syntactic “promotion” necessary to make a par-
ticular grammatical relation accessible to relativization (Keenan & Comrie 1977).
As DeLancey (1986) suggests for similar phenomena in Tibeto-Burman
­languages that share relativization-as-nominalization properties, relative clauses
 Tim Thornes

in such languages can readily be interpreted as dependent, appositional, noun

phrases. It is possible to interpret these findings as in line with Givón’s (2009)
suggestion of a non-restrictive parenthetical pathway that gives rise to embedded
relative clauses, in spite of the fact that he represents the Tibeto-Burman data as
more in line with a separate, clause-chaining scenario.
More rigorous data on intonation may contribute to further research into
diachronic syntax in Northern Paiute and other Numic languages as well –
something that is almost completely lacking in published studies on these lan-
guages. As is already well known, the loss of distinct intonational contours may
be considered symptomatic of the first steps in the historical transition from
­paratactic-to-syntactic structures. Morpho-syntactic adjustments can serve to rep-
licate this transition. Dayley (1989: 364ff), for example, presents some intriguing
synchronic variation in relative clause formation, including the presence versus
absence of relative pronouns or even of verbal subordinators (nominalizers), as
well as variation in the position of the relative clause with respect to the assumed
head noun. Such variation would appear to support the paratactic-to-syntactic
trajectory of complex syntax generally, and, more narrowly, the relative indepen-
dence (parenthetical nature) of nominalized clause types in Numic. We could then
view these structures as having entered into relative clause functions, first in appo-
sition to a would-be head noun, then through the development of more classic
restrictive and attributive functions.
The general paucity of embedded complements or headed relative clauses
in natural discourse could either, as Estrada (2007) suggests for Pima Bajo, be
a consequence of language attrition, or, as pointed out in Thornes (2003: 437),
represent one aspect of information flow in less formal discourse situations
than that of direct elicitation. Pragmatic parallels between the backgrounding
function of (adverbial) clause subordination in narrative and the presupposi-
tion of information or referential identity functions of relative clauses need to
be explored further. In terms of their discourse functions, both cases represent
a sensitivity to what the speaker assumes is known to the listener. A compari-
son of the modern corpus with older legacy materials – recordings from the
1950s and 1960s and archival data from the early 1900s – as well as a more
intensive discourse-based study across speakers, may shed light on any appar-
ent trends in the use of embedded nominalized clauses for functions most
associated with relative clauses. The distinct functions and possible historical
developments  explored here have led to a clearer understanding of the com-
plex f­unctional and grammatical domain shared by nominalization and rela-
tivization – one that connects referential identity, apposition, attribution, and
restriction.
 Functional underpinnings of diachrony in relative clause formation 

References

Bunte, P. 1986. Subordinate clauses in Southern Paiute. International Journal of American

­Linguistics 52: 275–300.
Comrie, B. & Kuteva, T. 2005. Relativization strategies. In The World Atlas of Language Struc-
tures, M. Haspelmath, M. Dryer, D. Gil & B. Comrie (eds), 494–497. Oxford: OUP.
Dayley, J.P. 1989. Tümpisa (Panamint) Shoshone Grammar [University of California Publications
in Linguistics 115]. Berkeley CA: University of California Press.
DeLancey, S. 1986. Relativization as nominalization in Tibetan and Newari. Paper presented at
the 19th International Conference on Sino-Tibetan Languages and Linguistics.
Epps, P. 2007. Escape from the noun phrase: The adventures of a relative clause. Paper presented
at the Seminario de Complejidad Sintáctica, Hermosillo, Sonora.
Estrada Fernandez, Z. 2007. Relative clauses in Pima Bajo: The grammaticalization pathways.
Paper presented at the Seminario de Complejidad Sintáctica, Hermosillo, Sonora.
Evans, N. & Dench, A. 2006. Introduction. In Catching Language: The Standing Challenge of
Grammar Writing, F.K. Ameka, A. Dench & N. Evans (eds), 1–39. Berlin: Mouton de
Gruyter.
Givón, T. 1990. Syntax: A Functional-Typological Introduction, Vol. II. Amsterdam: John
Benjamins.
Givón, T. 1994. Nominalized clauses in Ute: The diachronic seesaw of finite and non-finite struc-
ture. In II Encuentro de Lingüística en el Noroeste: Memorias, 269–310. Hermosillo, Sonora:
Universidad de Sonora, Mexico.
Givón, T. 2007. Toward a diachronic typology of relative clauses. Paper presented at the Semi-
nario de Complejidad Sintáctica, Hermosillo, Sonora.
Givón, T. 2009. The Genesis of Syntactic Complexity: Diachrony, Ontogeny, Neuro-cognition,
­Evolution. Amsterdam: John Benjamins.
Haspelmath, M. 1995. The converb as a cross-linguistically valid category. In Converbs in
­Cross-Linguistic Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial
Participles, Gerunds, M. Haspelmath & E. König (eds), 1–55. Berlin: Mouton de Gruyter.
Heine, B. & Kuteva, T. 2007. The Genesis of Grammar: A Reconstruction. Oxford: OUP.
Heine, B. 2008. From nominal to clausal morphosyntax: Complexity via expansion. Paper pre-
sented at International Symposium on the Rise of Syntactic Complexity, Rice University,
Houston TX.
Keenan, E.L. & Comrie, B. 1977. Noun phrase accessibility and Universal Grammar. Linguistic
Inquiry 8: 63–99.
Langacker, R.W. 1977. Studies in Uto-Aztecan Grammar, Vol. 1: An Overview of Uto-Aztecan
Grammar. Arlington TX: SIL Publications in Linguistics.
Marsden, W .L. 1923. “The Northern Paiute language of Oregon.” University of California
­Publications in American Archaeology and Ethnology: The Phoebe Apperson Hearst Memo-
rial Volume 20: 173–191.
Nedjalkov, V .P. 1995. Some typological parameters of converbs. In Converbs in ­Cross-Linguistic
Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial Participles, G
­ erunds,
M. Haspelmath & E. König (eds), 97–136. Berlin: Mouton de Gruyter.
Nichols, Michael J.P. 1974. Northern Paiute Historical Grammar. Ph.D. dissertation, University
of California at Berkeley.
 Tim Thornes

Noonan, M. 1997. Versatile nominalizations. In Essays on Language Function and Language

Type: Dedicated to T. Givón, J. Bybee, J. Haiman & S. Thompson (eds). Amsterdam: John
Benjamins.
Shibatani, M. 2007. Relativization in Sasak and Sumbawa, Eastern Indonesia. Paper presented at
the Seminario de Complejidad Sintáctica, Hermosillo, Sonora.
Snapp, A. & Anderson, J. 1982. Northern Paiute. In Studies in Uto-Aztecan Grammar, Vol. 3:
­Uto-Aztecan Grammatical Sketches, R.W. Langacker (ed.), 1–92. Dallas TX: SIL Publications.
Thornes, T. 2003. A Northern Paiute Grammar with Texts. Ph.D. dissertation, University of
Oregon at Eugene.
Thornes, T. 2008. Polyfunctionality and the na- middle marker in Northern Paiute. Paper pre-
sented at the annual meeting of the Society for the Study of the Indigenous Languages of
the Americas, Chicago, Illinois.
Toosarvandani, M. 2010. Patterns of nominalization in Numic. International Journal of ­American
Linguistics 76: 71–100.
part iii

Elsewhere in the Americas

 Clauses as noun modifiers in Toba
(Guaycuruan)

María Belén Carpio & Marisa Censabella

CONICET-Universidad Nacional del Nordeste*

In this paper we analyze relative clauses and noun complements in a corpus of

narrative texts in Toba. Three different coding devices are identified, described
and explained. One is specific to relative clauses and the other two introduce
noun complements as well. The semantic condition on relative clauses, i.e. the
co-reference between the head noun and the modifying clause, has been central
to distinguish them from noun complements. Regarding relative clauses, we
show that the syntactic function of the head noun within the relative clause is
inferred by subtraction of the missing argument in the relative clause and does
not condition the use of the different relativizers. It is the type of information
encoded in the relative clauses that has proved to be relevant in the selection of
the subordinators. Besides, we describe the synchronic polyfunctionality of the
subordinators.

Keywords:  Relative clauses; noun complements; demonstratives; Toba

1.  Introduction

Toba (Qom lʔaqtaqa) is a Guaycuruan language mainly spoken in the Argentinean

Chaco region (provinces of Chaco, Formosa, and Salta) and also in Rosario (Santa
Fe), and the outskirts of Buenos Aires. There are approximately 50,000 speakers
with different levels of linguistic and communicative competence. These differences
basically depend on the type of demographic concentration: in urban ­settlements,
young Tobas frequently have less communicative competence; in rural areas,
higher communicative competence is found, even among young ­people. Despite
the relatively high number of speakers and the existence of ­‘bilingual i­ ntercultural

* Núcleo de Estudios de Lenguas Minoritarias Americanas (NELMA), Instituto de

­Investigaciones Geohistóricas.
 María Belén Carpio & Marisa Censabella

education’ programs, it is an endangered language ­(Censabella 2009b).1 Only in

the Chaco province, Toba speakers, as well as linguists (Klein 1981), identify four
main dialects: lañaɢashek, noʔolɢranaq, rapigemleʔk, and takshek, all mutually
intelligible, especially the first three.
Complex clauses have recently begun to be studied in the Toba language.
­Messineo and Porta (2009) exclusively analyze restrictive relative clauses,
and ­Carpio (2007b, 2009) describes the synchronic polyfunctionality of the
morphosyntactic morphemes that introduce relative clauses (henceforth
­
­RelCls) and verbal complements. This paper aims to identify, describe and
explain the different coding devices of noun modifying clauses in a corpus of
narrative texts in Toba. We consider RelCls and noun complements as noun
modifying clauses characterized by the presence or absence of co-referentiality
between the noun modified and the modifying clause. We show that the syn-
tactic function of the head noun does not condition the use of the different
subordinators, rather it is inferred by subtraction of the missing argument in
the RelCl. The semantic condition on RelCls has been central to distinguish
them from noun complements. The pragmatic conditions on RelCls, i.e. the
type of information encoded in it, has proved to be relevant in the selection
of the subordinators. Besides, we describe the synchronic polyfunctionality of
the subordinators.
The morphemes that introduce noun modifying clauses found in the corpus
are: (i) maʒe [maʒi], related to a component morpheme of the third person free
pronouns, which is specific to relative clauses, (ii) ra [da], so, na, ʒe, ñi, ka, and
ram, som, nam, ʒem, kam, related to the demonstrative class, which may introduce
relative clauses, noun and verbal complements. Despite the fact that the latter also
introduce verbal complements, we will not treat them in that morphosyntactic
context. In Table 1, we present the different morphemes and their ­morphosyntactic
contexts of use analyzed in this paper.

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) 

Table 1.  Morphosyntactic contexts of the subordinators

Morpheme Noun Third person Relative Noun
determiner pronoun clause complement

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

2.  Overview of the language

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).

Table 2.  Active and middle voice personal markers

Person Active voice Middle voice

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.

labial alveolar pre-palatal palatal velar uvular glottal

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.

4.  Theoretical assumptions

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

complexity. Noun modifiers may be bound morphemes, lexical words, phrases

or clauses. In this paper, we focus the analysis on two types of complex noun
­modifiers in Toba: RelCls and noun complements.
RelCls and noun complements belong to the most complex type of noun
modifiers because they involve the importation of clause-level syntactic organiza-
tion into the noun phrase.
A functional definition of RelCls, as proposed by Givón (2001: 176–177),
requires a semantic condition, valid for all kind of them, and a pragmatic con-
dition whose applicability depends on the definiteness and referential status of
the head noun modified and on the restrictive or non-restrictive character of the
noun-modifying clause.
The semantic condition on RelCls refers to the co-reference between the head
noun modified and a participant within the event or state depicted in the modify-
ing clause. The pragmatic condition on restrictive relative clauses, which narrow
the domain of reference of the head noun and differentiate it from competing ref-
erents, varies if the head noun is definite, referential-indefinite or non-referring.
If the head noun is definite, the proposition in the RelCl is presupposed to be
known or familiar to the hearer. If it is referential-indefinite, the RelCl expresses
new information (salient cataphoric grounding information) to the hearer. If the
head noun is non-referring, the information encoded in RelCl is new to the hearer
depicting a hypothetical state/event.
Non-restrictive RelCls enrich the description of the referent without ­narrowing
the domain of reference and the proposition in the modifying clause is asserted –
not presupposed as in the case of restrictive relative clauses – as new information.
Although the semantic co-reference of the head noun and the missing argu-
ment inside the RelCl permit the recovery of its referential identity, its syntactic
function cannot be likewise recovered. Hence, as Givón (2001: 182) argues, the
various strategies employed by languages to solve the recoverability problem of
the syntactic function of the head noun within the RelCl describe the syntactic
typology of RelCls.
In a crosslinguistic study of the syntactic form of restrictive RelCls, Keenan
and Comrie (1977: 64) consider two criteria to distinguish restrictive RelCl
­forming strategies: the relative position of the head noun and the restricting clause
(postnominal, prenominal or internal RelCl strategy), and the way the syntactic
function of the head noun relativized is indicated in the RelCl, i.e. if there is or not
an element that unequivocally expresses the syntactic function of the head noun
within the restricting clause ([+case RelCl strategy] or [−case RelCl s­trategy]).
Applying these two criteria to the analysis of different languages, Keenan and
­Comrie (1977: 66) propose that languages vary according to the syntactic ­functions
of the head noun that can be relativized and propound a “noun phrase accessibility
 Clauses as noun modifiers in Toba (Guaycuruan) 

hierarchy” to relativization which is implicational from left to right: SU > OD > OI

> OBL > GEN > OCOMP.
Givón (2001: 182–197) describes the following syntactic relativization strate-
gies: (i) the non-embedding or paratactic strategy, (ii) the gap (zero) strategy, (iii)
the anaphoric pronoun strategy, (iv) the relative pronoun strategy, (v) a combina-
tion of verb agreement and the gap strategy, (vi) the word-order strategy, (vii) the
nominalization strategy, (viii) the equi-case strategy, and (ix) verb coding strate-
gies in which relativization interacts with promotion rules.
Concerning complements, Creissels (2006: 191–193) proposes a broad
definition that includes not only verbal complements, but also other types of
complements as well. Complement clauses function as an argument of a verb,
a noun, an adjective or a preposition. Also, he argues that a purely syntactic
definition of relative clauses is not enough to distinguish them from noun
complement clauses. For example, in French L’idée [que tu défends] “the idea
that/which you defend’ (RelCl) and L’idée [que tu partes] ‘the idea that you are
leaving’ (noun complement) cannot be distinguished only on a syntactic basis,
rather the semantic status of the subordinate clause must be taken into account.
Thus, both RelCls and noun complements involve the embedding of a subordi-
nate clause as a noun modifier in the noun phrase, but in noun complements
the semantic condition that characterizes RelCls, i.e. the co-reference between
the missing argument in the subordinate clause and the modified head noun,
is not fulfilled.

5.  Data analysis

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

Sharp distinctions between restrictive and non-restrictive relative clauses did

not prove to be adequate to explain our data.
Furthermore, when the dem-type or dem-top type subordinators introduce
clauses in which there is no co-reference between the modified head noun and
one of the participants in the modifying clause, we describe them as noun comple-
ments. We also show the synchronic polyfunctionality of the subordinator mor-
phemes analyzed.

5.1  The relativizer maʒe ~ maʒi

This is the only subordinator exclusively used to introduce RelCls. The syntactic
function of the head noun within the RelCl is not recoverable by the use of this
relativizer. It introduces postnominal relative clauses, adjacent to the head noun.
The relativizer maʒe is chosen when the information encoded in the dependent
clause is a property or cultural inherent characteristic of that referent, or is retriev-
able from the previous discourse.
The following examples show the use of maʒe, whether the head noun
­functions as S (1), A (2), applied P (3), or is a constituent in a possessive
­construction, ­possessor (4) or possessed (5) noun within the dependent clause.
The argument missing as a noun phrase in the RelCl is the clue to identify the syn-
tactic f­ unction of the head noun within it. This is specially relevant when the head
noun ­fulfills a non-subject syntactic function because, in main clauses, it must be
overtly expressed as a noun phrase.4
(1) S-rel
nachi y-aqonoygi-oʔ na alap, nachi Ø-anol-ek
coord 3t-take-appl:loc dem 3poss.mouth coord 3-vomit-obj
na ntagoʔq [maʒi Ø-chigaqa-ygi na alap]
dem blood  rel 3-come-appl:loc dem 3poss.mouth
‘So he puts (his hand) into his mouth, and vomits the blood that comes
from his mouth.’ O. Sánchez; Nedec 6 - Petoxoi: 74 

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

(3) applied P-rel

nache ʔayim eʔetega-aʔ so-m in-tesqoʔ
coord 1sg 3.say-appl:all dem-top 1poss-uncle
[maʒe s-oʔo-ta-sop]: ʔonaɢay ra
 rel 1-be-cont-appl:loc nice rel
a-w-ʔa ñi la escuela.
2-dir-appl:loc dem the school
‘So, my uncle, who I am around, tells me: how nice if you go to school!’
 AP#41– 42

(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

5.1.1  Free third person pronouns

The relativizer maʒe relates to a component morpheme of free third person
­pronouns. These pronouns are composed of the maʒe morpheme plus one of the
demonstratives with gender markers if the referent is singular or collective, and
without them if it is plural, as is shown in Table 3.
 María Belén Carpio & Marisa Censabella

Table 3.  Third person free pronouns

Person Gender
number
Masculine Feminine

3sg ra-maʒe ‘standing up’ a-ra-maʒe ‘standing up’

ñi-maʒe ‘sitting down’ a-ñi-maʒe ‘sitting down’
so-maʒe ‘moving away’ a-so-maʒe ‘moving away’
na-maʒe ‘coming closer’ a-na-maʒe ‘coming closer’
ʒe-maʒe ‘laying down’ a-ʒe-maʒe ‘laying down’
ka-maʒe ‘invisible or unknown’ a-ka-maʒe ‘invisible or unknown’
3pl  ra-ʔa-maʒe/ra-wa-maʒe ‘standing up’
 ñi-ʔi-maʒe/ñi-wa-maʒe ‘sitting down’
 so-ʔo-maʒe/so-wa-maʒe ‘moving away’
 na-ʔa-maʒe/na-wa-maʒe ‘coming closer’
 ʒe-ʔe-maʒe/ʒe-wa-maʒe ‘laying down’
 ka-ʔa-maʒe/ka-wa-maʒe ‘invisible or unknown’
3coll ra-maʒe-pi ‘standing up’ a-ra-maʒe-pi ‘standing up’
ñi-maʒe-pi ‘sitting down’ a-ñi-maʒe-pi‘sitting down’
so-maʒe-pi ‘moving away’ a-so-maʒe-pi ‘moving away’
na-maʒe-pi ‘coming closer’ a-na-maʒe-pi ‘coming closer’
ʒe-maʒe-pi ‘laying down’ a-ʒe-maʒe-pi ‘laying down’
ka-maʒe-pi ‘invisible or unknown’ a-ka-maʒe-pi ‘invisible or
unknown’

5.2  dem type subordinators

The subordinators ra [da], na, so, ʒe, ñi and ka may introduce relative and
­complement clauses. The distribution of the DEM subordinators in our corpus is
as ­follows: ra [da] is the most frequently used as a relativizer and it also introduces
noun and verbal complements. The na and ʒe subordinators introduce relative
clauses and verbal complements; so only introduces relative clauses, and ka verbal
complements; finally ñi only introduces noun and verbal complements. We just
mention the use of these subordinators to introduce verbal complements in order
to show their non-exclusivity as relativizers or noun complementizers but we do
not analyze verbal complements in this paper.

5.2.1  dem type subordinators introducing RelCls

When these subordinators introduce a RelCl, the head noun is co-referent with
an argument in the dependent clause. This is not so when they introduce noun
complements (§5.2.2).
 Clauses as noun modifiers in Toba (Guaycuruan) 

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

5.2.2  dem type subordinators introducing noun complements

When there is no co-reference between the head noun and a participant in
the event or state depicted in a dependent clause introduced by the DEM type
­subordinator, i.e. when the semantic condition on RelCls is not fulfilled, we
­consider that clause as a noun complement (10)–(12).
 María Belén Carpio & Marisa Censabella

(10) qamaq woʔo da l-chik [da ʔam

coord ex.pres compl 3poss-danger  compl 2sg
ʔaw-asowaɢat na nogotolek qoy a-na ʔalo]
2-be.guilty dem boy coord f-dem woman
‘… but there is the danger that you are guilty of (the death) of the children
and the women.’ O. Sánchez, Nedec 5 - Cabito Leiva: 70

(11) nache n-wiʔ so naʔaq [ra Ø-heʔk

coord 3m-arrive dem day  compl 3-go.pl
seʔeso-wa l-taʔa-l
so Pinʒoʔolek]

dem.pron-dl 3poss-parent-pl dem Pinʒoʔolek
‘So, the day that Pinʒoʔolek parents leave, arrived.’  Pinyo’olec:10

(12) nachi lamaaɢa da l-ʒaɢ-a [da

coord even.worse dem 3poss-desire-nom  compl
de-wotaɢanek a-saʔaso waaka-pi qataq so qay-pi]
3-take.off f-pron.dem cow-coll coord dem horse-coll
‘And it is even worse his desire to steal those cows and horses.’
 O. Sánchez; Nedec 6 - Petoxoi: 74

5.2.3  Demonstratives as noun modifiers

The dem type subordinators are isomorphic to the demonstratives in their
­masculine singular/collective form. In Toba, demonstratives are obligatory noun
determiners used to encode the position of the noun referent:

(13) ra ‘standing up’

ñi ‘sitting down’
ʒe ‘laying down’

and the spatial relation or visibility between the deictic center and the entity
referred to by the noun:

(14) so ‘moving away’

na ‘coming closer’
ka ‘invisible or unknown.’

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):

(15) a. a-na i-wal r-keʔe

f-dem 1poss-grandchild 3-eat
‘My grandchild (fem) eats.’
 Clauses as noun modifiers in Toba (Guaycuruan) 

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.’

 ragmatic conditioning on the selection of maʒe or dem-type

5.3  P
subordinators
In examples (1)–(9) above, we have demonstrated that the use of maʒe or the
dem type relativizers does not depend on the syntactic function of the head noun
within the dependent clause. Instead, it is a pragmatic conditioning that plays a
relevant role. When the information encoded in the dependent clause is familiar
to the hearer (accessible in his/her episodic memory), the speakers will choose
the maʒe relativizer. On the contrary, when new information, cataphorically rel-
evant is encoded in the dependent clause, the speakers will choose the dem type
relativizers.
This distinction in terms of the kind of information encoded in the RelCl is
well illustrated in example (16b) where both types of relativizers are used in coor-
dinated clauses. In na ñaqpyolek [da Ø-paʔayget da l-lewɢa] ‘the children who go
to their death’ the dependent clause adds relevant cataphoric information of the
head noun referent, while in na ʔalpi [maʒi qayka ka yasowaɢat] ‘these women
who are not guilty’ the topic of the dependent clause has been already introduced
in the previous discourse, as it is shown in (16a).
(16) a. chaʔaʒi a-na ʔalo qataq na nogotolek qayka
coord fem-dac woman coord dac child neg.ex.pres
ka l-asowaɢa-shit
compl 3poss-be.guilty-nom
‘…because the woman and the child have no guilt…’
 O. Sánchez, Nedec 5 - Cabito Leiva: 44
b. chaʔaʒi ñimaʒi i-choɢodee-tak na ñaqpyolek [da
coord 3sg 3t-have.pity-prog dem children  rel
Ø-paʔay-get da l-lewɢa] qataq na
3-be.on.ground-appl:dir dem 3poss-death coord dem
ʔal-pi [maʒi qayka ka y-asowaɢat].
woman-coll  rel neg.ex.pres compl 3t-be.guilty
‘… because he is feeling pity for the children that go to their
death and the women who are not guilty.’
 O. Sánchez, Nedec 5 - Cabito Leiva: 69
 María Belén Carpio & Marisa Censabella

5.4  dem-topicalizer (dem-top) subordinators

The dem-top type subordinators may introduce relative and complement clauses.
In our corpus, they are distributed as follows: ram, nam, ʒem and kam introduce
RelCls, and som introduce RelCls and noun complements.5

5.4.1  dem-top type subordinators introducing RelCls

Concerning RelCls, we have observed that the dem-top subordinators are used
when the head noun is topicalized. The head noun can be topicalized through
the following devices: the co-occurrence of the head noun with a pronominal
demonstrative (17), the promotion of the head noun by a valence-changing
operation such as non-promotional passives in the main clause verb (18) or in
the subordinate clause verb (19), and the recurrence of the head noun referent
through the text (20) being one of the main topics of it. As can be seen, the syn-
tactic functions of the head noun within the RelCl are recoverable by the omitted
noun phrase, required by the subordinate verb. The syntactic functions of the
head nouns modified by RelCls introduced by the dem-top type ­subordinators
are: A (17), applied P (18), P (19), and S (20).
(17) A-rel
“ʔam ya-qaya qole a-naʔañi nsoɢoy [ra-m ʔam
  2sg 1poss-brother t.adv f-dem.pron Nsogoi  dem-top 2sg
y-amaɢa-lek ʒe leʔ ] a-kitchikchi-gi”
3t-push-appl:loc dem embers 2-blow-appl:loc
‘You, my brother, when that is Nsogoi the one who pushes you over the
­embers, you blow.’ PR-N#24
(18) applied P-rel
somaʒi y-iʔigena-ʔa-get na doqshi, yoqoʔoʒi
3sg 3t-challenge-go-appl:dir dem White People coord
w-añoʔ ka qo-y-ii-teg-a Charata
3.go.dir.appl:loc compl np.pass-3t-say-prog-appl:all Charata
[ʒi-m w-etaʔa qoʔoʎoɢochiʒi somaʒi
 rel-top 3-go.cont.appl:all t.adv 3sg
na l-awoʔ]


dem 3poss-people
‘He went and challenged the White People, because he was (in the place)
that had been called Charata, where were, in those days, his people.’
 O. Sánchez; Nedec 6 - Petoxoi: 73

5.  All these subordinators introduce verbal complements as well.

 Clauses as noun modifiers in Toba (Guaycuruan) 

(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

5.4.2  dem-topicalizer type subordinators introducing noun complements

Only the dem-top subordinator som was observed introducing noun comple-
ments when the modified noun is not co-referent with an argument within the
dependent clause, as in (21).
(21) nache dos años [so-m s-er-aɢaa-tak] nache
coord dos años  compl-top 1-write-antipass-prog coord
alcancé Segundo grado
I have reached second grade
‘So, there have been two years that I am writing, so I have reached second
grade (at school).’  AP#55

5.4.3  dem-top as noun modifiers

The dem-top morphemes are also used before nouns showing “high topicality,”
i.e. recurrent or prominent referents in the narrative text (22); and with P argu-
ments already topicalized by a non-promotional passive (23).
(22) nache so-m7 Ferocho8 e-nak-o chaaʒe so-m
coord dem-top Ferocho 3t-say-appl:loc coord dem-top
Ferocho woʔo na no-ʔonataɢak : aha…
Ferocho ex.pres dem 3u.poss-power yes
‘And Ferocho says, because Ferocho has power,: -Yes, …’ PR-N#9

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:

so-m yi l-eʔenaGat Ferocho,

dal-top elder.brother 3poss-name Ferocho
‘The name of the older brother (is) Ferocho.’ PR-N#4

Besides, the BENEFACTIVE/MALEFACTIVE (BEN/MAL) applicative is encoded by -m.

8.  Ferocho is the hero of the story.
 María Belén Carpio & Marisa Censabella

(23) nache qa-y-wire-w-ʔa a-so-m Nsoɢoy nache

coord np.pass-3t-arrive-dir-appl:all f-dem-top Nsogoi coord
e-nak-o: aha nyachik i-wal…
3t-decir-appl:loc yes thank.you 1poss-grandchild
‘And someone arrives near to the Nsogoi and (the Nsogoi) says: -Yes, thank
you my grandchild…’ PR-N#12

6.  Conclusions

In this paper, we have analyzed three different coding devices of noun-modifying

clauses: the relativizer maʒe, and two other types of subordinators which intro-
duce not only relative clauses, but also noun and verbal complements: dem type
(ra [da], so, na, ʒe, ñi, ka), and dem-top type (ram, som, nam, ʒem, kam). Besides,
we have described the synchronic polyfunctionality of the subordinators.
The three types of subordinators mark the boundary of the RelCl but they do
not help distinguish the syntactic function of the head noun within the dependent
clause. The syntactic function of the head noun within the RelCl is inferred by
subtraction of the missing argument in that clause.
Pragmatic factors condition the use of the different subordinators to intro-
duce RelCls. The relativizer maʒe introduces RelCls whose propositional content
refers to an inherent property of the head noun referent or to information ana-
phorically retrievable. The dem type subordinators introduces RelCls that pro-
vides new information relevant to the continuity or further development of the
discourse, and the dem-top type subordinators are used when the head noun
is topicalized by a pronominal demonstrative, a non-promotional passive or is
recurrent throughout the text.
dem type and dem-top subordinators also introduce noun complements, i.e.
clauses that modify a head noun non co-referent with an argument within the
dependent clause. dem type subordinators occur, in our corpus, in the following
morphosyntactic contexts: ra introduce RelCls and noun and verbal complements,
na and ʒe introduce RelCls and verbal complements; so only introduces RelCls, ka
introduces verbal complements, and ñi only introduces noun and verbal comple-
ments. The dem-top type subordinators are distributed as follows: ram, nam, ʒem
and kam introduce RelCls, and som introduce RelCls and noun complements. All
these dem-top subordinators also introduce verbal complements, which are not
described in this paper.
Moreover, the synchronic polyfunctionality of the subordinators is presented.
Maʒe plus the demonstratives as prefixes encode third person free pronouns. The
dem type subordinators are isomorphic to the demonstratives in their m ­ asculine
 Clauses as noun modifiers in Toba (Guaycuruan) 

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

Creissels, D. 2006. Syntaxe Générale, une Introduction Typologique 1: Catégories et Constructions

[Collection langues et syntaxe 2]. París: Hermes-Lavoisier.
DGEEC. Dirección General de Estadística, Encuestas y Censos. 2003. II Censo Nacional Indí-
gena de Población y Viviendas 2002. Pueblos indígenas del Paraguay. Resultados finales.
Asunción.
Givón, T. 2001. Syntax: An Introduction, Vol. 2. Amsterdam: John Benjamins.
González, R.E. 2008. Características sintáctico-semánticas de aplicativos -gi e -igi en toba.
V Congreso de Lenguas del MERCOSUR y Primer Encuentro de Regiones Italianas “Paz,
democracia y solidaridad”, 1–11. Resistencia (Chaco, Argentina): Universidad Nacional del
Nordeste, Facultad de Humanidades.
González, R.E. 2009. El aplicativo asociativo en toba (familia guaycurú). IV Coloquio Argentino
de la IADA (International Association for Dialogue Analysis), 1–3 July ,2009, 1–12. Univer-
sidad Nacional de La Plata, Argentina.
Gordon, R. & Grimes, B. (eds). 2005. Ethnologue: Languages of the World, 15th edn. Dallas TX:
SIL International. http://www.ethnologue.com/.
INDEC. 2006. Resultados de la Encuesta Complementaria de Pueblos Indígenas (ECPI) ­2004–2005.
〈www.indec.gov.ar〉, Section ‘Población’, subsection ‘Pueblos Indígenas’.
Instituto Nacional de Estadística. 2002. Censo Nacional de Población y Vivienda 2001. La Paz,
Bolivia: INE.
Keenan, E.L. & Comrie, B. 1977. Noun phrase accessibility and Universal Grammar. Linguistic
Inquiry 8(1): 63–99.
Klein, H.E.M. 1981. Una Gramática de la Lengua Toba: Morfología Verbal y Nominal. Montevi-
deo: Departamento de Lingüística, Universidad de la República.
López, L.E. 2006. Pueblos indígenas, lenguas, política y ecología del lenguaje. In Diversidad y
ecología del lenguaje en Bolivia, L.E. López (ed.), 17–46. La Paz: Plural Editores.
Messineo, C. & Porta, A. 2009. Cláusulas relativas en toba (guaycurú). International Journal of
American Linguistics 75(1): 49–68.
Nate’elpi Nsoquiaxanaxanapi (Keeping Mothers). 2006. So Pinyo’olec. Versión en Qom l’aqtaqa y
Castellano. Córdoba: Editorial Copiar.
Sánchez, O. 2008. Da na’aqtaguec nam qompi (tobas) mayi lma’ na lta’adaic Chaco nam ỹaxat
som lquedoxonecpi na mayipi/Historia de los aborígenes qom (tobas) del Gran Chaco con-
tada por sus ancianos. Resistencia: Author’s Edition.
 Between headed and headless relative clauses*

Patience Epps
University of Texas at Austin

Most syntactic typologies of relative clauses recognize two distinct categories:

‘headed’ and ‘headless’ (or ‘free’) relatives, according to whether or not the relative
clause is associated with a nominal element that refers to a category delimited
by the relative (the ‘domain nominal’, e.g. Andrews 2007). To this, Citko (2004)
adds a third, intermediate category of ‘light-headed’ relatives. However, this paper
considers evidence from Hup, a Nadahup (Makú) language of the northwest
Amazon, to argue that such a strictly categorical approach – even one that makes
room for three categories – is descriptively and typologically inadequate. In
particular, for languages like Hup in which relative clauses are nominalizations
(in an appositional relationship to the domain nominal), elements occurring
as domain ‘nominals’ may be shaped by processes of grammaticalization that
give them a partly lexical, partly grammatical identity, and they may occupy
different points along a lexical–grammatical continuum. In Hup, while relative
clauses may be associated with a full noun phrase (i.e. headed) and with no
domain nominal whatsoever (i.e. headless), they may also appear with a range
of intermediate elements, including ‘bound’ nouns, which differ from full noun
phrases in their syntactic status; classifiers or ‘class terms’; and the enclitic =d’ǝh,
a semi-nominal element that also marks plural number. The Hup data suggest
that the property of ‘headedness’ in relative clauses may be best represented as a
gradient phenomenon, and that this approach is arguably descriptively richer and
typologically more accurate than the alternative.

Keywords:  Headless (free) relatives; nominal classifiers; nominalization; Hup

* 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

Syntactic typologies of relative clauses typically recognize two discrete categories,

‘headed’ and ‘headless’ (or ‘free’) relatives. In the case of headed relatives, the noun
phrase whose reference is delimited by the subordinate clause (i.e. the ‘domain
nominal’) is explicitly stated, whether within the main clause, as is more common,
or within the relative clause itself (for ‘internally headed’ relatives). In the case
of headless or free relatives, the noun phrase whose reference is delimited is left
unstated.1
This paper considers the question of whether this categorical approach to
relative clauses is empirically justified. Just as relative clauses in some languages
require an associated domain nominal (i.e. they lack headless relatives altogether;
see Andrews 2007: 214), it appears to be the case that in other languages relative
clauses may associate not only with full noun phrases (or with no noun phrase
at all), but also with nominals that are in some way less than full lexical nouns in
their own right, and which possess properties different from those of full noun
phrases with respect to relative clause constructions. Note that, in spite of the terms
‘headed’ and ‘headless’, the question of whether the noun phrase that results from
the combination of relative clause and domain nominal – or the lack thereof – can
be understood as headed by something, and what that element is, is to some extent
a separate consideration (see Zwicky 1985; Corbett et al. 1993; Dryer 2004).
The proposal that there might be something intermediate between a headed
and a headless relative clause is not a new one. Citko (2004) argues that a number
of European languages may be understood as having a third, in-between category
of ‘light-headed’ relatives, which share features of both headed and headless rela-
tive clauses but should be classified as neither.2 Based on data from Polish, Citko
argues that this intermediate category has a syntactic reality and differs from its
headed and headless counterparts in interpretation, relative pronoun selection,

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):

(i) Jean lit ce qu’-il aime

Jean reads this that-he likes
‘Jean reads what he likes.’

(ii) Jan czyta to, co Maria czyta

Jan reads this what Maria reads
‘Jan reads whatever Maria reads.’
 Between headed and headless relative clauses 

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.

2.  Relative clauses in Hup

Hup, a member of the small and under-studied Nadahup family,3 is spoken by

approximately 1500 people in the Vaupés region of northwest Brazil and ­eastern
Colombia. Typologically relevant characteristics of the language include its

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

­ rosodic tone and morpheme-level nasalization,4 verb-final basic constituent order,

p
nominative-accusative alignment, dependent-marking (e.g. with c­ ase-marking of
object and oblique arguments), extensive verb compounding, and zero-anaphora,
by which subjects, objects, and other clausal constituents are frequently dropped
(see Epps 2008).
Relative clauses in Hup involve a non-finite verb, which in most cases takes
subordinating morphology and lacks primary aspectual inflection and markers of
main-clause status. Hup relative clauses are best understood as nominalizations
(as is the case for many languages of South America; e.g. Awa Pit, Curnow 1997;
Tariana, Aikhenvald 2003; and languages of the Carib family, Gildea 1998); their
nominal status is evidenced by their occurrence as arguments of verbs and their
ability to take nominal morphology, including case-marking, number marking,
and ‘identity’ negation (see Epps 2008: 828ff). They are formed according to a gap-
ping strategy, such that the relativized (head) nominal occurs outside the relative
clause. In this and the following section, I outline the range of Hup’s relative clause
types with respect to the identity of the domain nominal, or its absence.

2.1  Headed relatives: Domain nominal is a full noun phrase

Examples (1–3)5 illustrate Hup’s fully headed relative clause. These relatives are
followed by a clearly defined domain nominal, a full noun phrase that can func-
tion as an independent clausal constituent. The relative clause itself is identified
by the ‘Dependent’ suffix -Vp on the verb. Where the resulting noun phrase is a
non-subject argument of the main clause (and is not a left-dislocated topic, as in
Example 1), the domain nominal is case-marked accordingly (Example 3).6

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 

(1) [tɨh wæd-ě-p] parátu, wáb-an píd=mah

  3sg eat-perf-dep plate shelf-dir dist=rep
cak-w’ob-yɨʔ-píd-íh
climb-set-tel-dist-decl
‘The plate [he had eaten from], she set (it) back up on the shelf.’
(2) tɨnĭh yǔd wob-d’ǝh-hí-íh, [wág tɨh
3sg.poss clothes set-send-descend-decl  day 3sg
cúd-up] yǔd...
wear-dep clothes
‘His clothes were all laid out, the clothes [he wore in the daytime]...’
́
(3) ʔãh=yɨʔ [ʔám=tæ̃hʔín tih ní-ĩp] hayám-ǎn kéy-éh
1sg=foc   2sg=wife 3sg be-dep village-obj see-decl
‘I have seen the village [that your wife is living in]!’

2.2  Headless relatives: Domain nominal is lacking

Truly headless relative clauses in Hup appear with no domain nominal of any kind.
This absence may be attributed to ellipsis (i.e. the entity described by the relative
clause is understood from the discourse context), to the speaker’s uncertainty of
the identity of the referent, or to the referent’s status as culturally unfamiliar and
thus lacking any conventionalized label in Hup (the default classifier noun teg
‘thing’ is an alternative option in this case; compare 18 below).
Headless relatives in Hup may be indicated simply by the presence of the
Dependent suffix -Vp on the verb, as in (4). As such, the relative clause is in the
unmarked nominative case, and must therefore be interpreted as either the subject
of the main clause or as a predicate nominal. Headless relative clauses of this kind
are relatively rare in Hup discourse, for reasons addressed in §3.3 below.
(4) tɨh tɔhɔ-yíʔ-ay=mah, [tɨnĭh hṹ ni-ʔě-p],
3sg finish-tel-inch=rep   3sg.poss animal be-perf-dep
́
[húp nĭh hũ ni-ʔě-p]
 person poss animal be-perf-dep
‘It was all gone, [that which had been his game animal], [that which had
been the person’s game animal].’ (i.e. it had been fully consumed)

When a relative clause that lacks a domain nominal occurs as an object or

oblique argument in the main clause, the verb inflects with the relevant case

that relative clause + domain nominal form a single syntactic constituent (whereas headless
­relative clauses are directly case-marked; see below).
 Patience Epps

marker – ­principally the Object (5–6) or Oblique (7) suffixes, depending on the

role of the relative in the main clause. In relatives of this kind, the Dependent suf-
fix -Vp does not appear, but the ‘Filler’ syllable -Vw- (or -w- when the preceding
root has undergone the loss of a final consonant) intervenes between the stem
and the case marker. It is possible that this ‘Filler’ syllable derives historically
from the -Vp morpheme (an intervocalic p > w change at morpheme boundaries
is attested elsewhere in Hup), so it may have an equivalent status as a marker of
subordination.7 Case-inflected relative clauses of this kind, which are effectively
headless, are very common in Hup, in contrast to their uninflected (nominative)
counterparts.
(5) hǎt-ǎn=mah yúp, yít [baʔtĭb’ tɨh
alligator-obj=rep that.itg then  spirit 3sg
kɨt-huhuj-yæt-yɨʔ-pó-w-ǎn]=mah yúp,
cut-chop.up-lay-tel-emph1-flr-obj=rep that.itg
húp=b’ay kut-cɔp-hám-b’ay-áh
person=AGAIN bundle-go.up.from.river-go-AGAIN-decl
‘The alligator, [the one the spirit had cut into bits], that person bundled [it]
up and took it up the river, it’s said.’
́
(6) [pɔnǽn nɔ́ -ɔw-ǎn] ʔãh key-níh-ɨʔ yǎ, tɨh nɔ́-ɔ́h
 ponæn say-flr-obj   1sg see-neg-int tag1 3sg say-decl
‘I’ve never seen [the one that says pɔnæn!], he said.’
(7) [máki-and’ǝh nĭh=yɨʔ=ʔĩh ni-ě-ew-ĩt]=yɨʔ-íʔ ?
 Marc-assoc.pl poss =foc=msc be-perf-flr-obl=foc-int
‘It was with [that one who used to be associated with Mark’s group]
(that you went)?’

3.  Intermediately headed relative clauses

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 

3.1  Domain nominal is a bound noun

Relative clauses in Hup commonly take a ‘bound’ noun as a domain nominal. As a
category of noun, bound nouns differ from full noun phrases in their syntactic sta-
tus: they cannot occur as independent constituents of a clause, but are obligatorily
preceded by another nominal element (a noun, demonstrative, pronoun, numeral,
or relative clause; Epps 2008: 232–255). Accordingly, the relative clause and the
bound noun together constitute a full noun phrase, but the domain nominal is
itself less than a syntactically independent element.
Semantic categories of bound nouns include kin, humans, body parts, plant
parts, and others (Epps 2007, 2008). Almost by definition, bound nouns refer to
a category that is in some way generic. For example, =teg refers to the generic
category of ‘tree’, whereas the full noun phrase pæ̌ j=teg ‘umari tree’ specifies a
particular type of tree; similarly, =tip refers generically to eggs, but speakers
typically specify the type of egg, e.g. cadakǎʔ=tip ‘chicken egg’. On first men-
tion, the referent is often explicitly specified by the preceding noun to which it
attaches (such as ‘umari’ or ‘chicken’); later in the discourse, speakers are more
likely to use demonstratives, numerals, or relative clauses in combination with
bound nouns, which refer anaphorically to the previously established referent
(e.g. Example 11 below).
In relative clauses referring to animate entities, the most commonly encoun-
tered bound nouns are terms for ‘man’ (=ʔĩh), ‘woman’ (ʔãý ), ‘old/respected man’
(wǝd), and ‘old/respected woman’ (wa), as in the following examples:
(8) yúp=mah j’ám ̃
ʔayǔp=ʔĩh hɔ̌p kək-hám-ã ́
h...
thus=rep dst.cntr  one=msc fish pull-go-decl
̃
[hɔ̌ p kə́k-əp]=ʔĩh wɨʔ-pǽm-ǽy=mah
 fish pull-dep=msc listen-sit-dynm=rep
‘So, long ago, it’s said, a man went fishing... The man [who was fishing] sat
listening.’
(9) ʔǎn hɨd y’ǽt-ǽh, ʔayǔp=ʔãy-ã ́ t=yiʔ,
́ ʔǎn
  1sg.obj 3pl leave-decl  one=fem-obl=foc   1sg
[hi-kéy-ep]=ʔãy=yiʔ
 fact-see-dep=fem=foc
‘They left me there, with a woman, the woman [who would look after me]’.
(10) [mɔ̌y hǽp-æp]=wa
 house sweep-dep=old/resp.woman
‘old/respected woman [who is sweeping the house]’
(11) yít=mah yúp húp=wǝd wɨʔ-g’ét-éy,
thus=rep that.itg person=old/resp.man listen-stand-dynm
 Patience Epps

[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):

(12) [ʔɨn wǽd-æp]=teg ʔám b’ɔt-yíʔ-íy!

  1pl eat-dep=tree 2sg chop.down-tel-dynm
‘You’ve chopped down the tree [we eat from]!’

(13) [n’ikán wób-op]=tod’

 there rest-dep=jar
‘the jar [sitting over there]’

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’

Furthermore, bound nouns occurring with relative clauses often do not

­correspond to a full noun phrase involving the same bound noun, which also
argues against the possibility that an associated domain nominal is there but
is simply not pronounced. For example, if the type of tree is not known, the
bound noun =teg ‘tree, stick’ (as in 12 above) can only be co-referential with the
generic term tegd’ǔh ‘tree’ (as opposed to, e.g. tĭw=teg ‘Brazil nut tree’); simi-
larly, the bound noun =ʔãý ‘female, woman’ is normally co-referential with the
full noun tãʔãý ‘woman’, which is in fact derived historically from the default
third person pronoun tɨh plus the bound noun =ʔãý (see Epps 2008: 232ff) – in
other words, there is really no full nominal element that could be dropped in
this context.
Finally, the inability of bound nouns to appear without an associated
­nominal is a feature that is shared by various other elements of the noun phrase,
primarily adjectives and numerals (other than ‘one’), likewise suggesting that the
existence of a null nominal head is not a general feature of Hup noun phrases.
For example, an adjective can occur as a noun phrase in its own right only if
preceded by the default form tɨh (just like a bound noun; Example 15), and a
numeral must also occur with an associated nominal element, if only a classify-
ing noun (Example 16):
(15) páh=yɨʔ y’æt-pog-ʔé-y páh yúw-úh
prx.cntr=foc leave-emph1-perf-dynm prx.cntr that-decl
cǝ́c,
tɨh=j’á-aw-áh!

interj 3sg=black-flr-decl
‘Just recently (he) left it, darn it; (it was) a black one!’
́
(16) bodáca, ʔóytu=b’ah yít píd j’ãh ʔín-ǎn
cookie  eight=split thus again dst.cntr   1pl-obj
tɨh nɔ́ʔ-ɔ́y
3sg give-dynm
‘Cookies, eight of them, that’s what she gave to each of us.’
 Patience Epps

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.

3.2  Domain ‘nominal’ is a classifying noun

While a bound noun lacks the fully independent lexical identity of other nouns
(which can function alone as constituents of a main clause), its status as a type of
domain nominal in a relative clause construction is nevertheless in little doubt.
However, the set of bound nouns in Hup is itself closely associated with the set
of classifying nouns, and while like bound nouns these may occur with relative
clauses, their status as true lexical nouns is even more marginal.
Hup’s classifying nouns (classifiers or ‘class terms’; see Epps 2007, 2008; also
Ospina 2002: 217 for a description of the phenomenon in the closely related
­language Yuhup), are formally identical to and semantically and historically
derivative of a subset of the bound nouns. Most derive from plant part terms and
classify inanimates on the basis of shape; for animates, extensions of the bound
nouns ‘man’ and ‘woman’ are used to indicate male and female gender, while the
‘male’ form can also refer generically to animates. The semantic contribution of a
classifying noun in Hup is thus necessarily more abstract than is that of the cor-
responding bound noun. For example, the bound noun ‘split wood’ has developed
a classifying function for flat things; ‘seed’ has come to be used for small, round
things; and ‘tree, stick’ classifies long, thin things, and has developed the further
meaning of ‘thing in general’ (see Epps 2007, 2008: 269). Functionally, therefore,
classifiers may refer to entire sets of entities unified by a relatively abstract ­feature
(principally shape), whereas the referents of bound nouns are more precisely
defined.
The classifying nouns are best understood as an incipient grammatical cate-
gory of noun classifiers in Hup. They appear regularly in only a small set of nouns,
where they have a primarily derivational function. As such, they may attach to
verb roots, where they typically produce an agent/instrument n ­ ominalization
 Between headed and headless relative clauses 

‘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 

3.3  Domain ‘nominal’ is a plural marker

While Hup’s classifying nouns are effectively intermediate between lexical and
grammatical elements, they are arguably more like lexical items (given the incipi-
ent nature of the classifier system in Hup), and thus the relative clauses with which
they associate may be considered somewhat closer to ‘headed’ than to ‘headless’.
However, Hup has a further type of construction that is arguably a form of relative
clause, and as such is even closer to the ‘headless’ end of the continuum.
Clauses of this kind, which I consider to be relatives for reasons given
below, appear with the enclitic =d’ǝh, the primary function of which elsewhere
in Hup is to mark plural number or collectivity on nominal constituents; e.g.
tiyǐʔ=d’ǝh ‘men’, tǎh=d’ǝh ‘tapirs’. As such, the enclitic =d’ǝh may be analyzed as a
­grammatical item, an inflectional element indicating plural number; however, it
also has ­characteristics of a nominal lexical element in its own right, and therefore
has a marginal status as a domain nominal with relative clauses. Examples (19–20)
illustrate clauses of this type:

(19) [nút pǝ́ʔ]=d’ǝh næn-pǽm-ǽh, j’ámyɨʔ

 here give.dabacuri=pl come-sit-decl long.ago
hɨd næn-pæm-ʔě-h
3pl come-sit-perf- decl
‘Here [those who were giving the dabacuri] came in to sit, as long ago they
used to do’.

̃
(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.

3.4  Intermediately headed relative clauses: Diachronic evidence

The preceding discussion offers synchronic evidence that the headedness of rela-
tive clauses – their capacity to appear with or without a domain nominal – is best
understood as a gradient phenomenon in Hup. From the perspective of language
change, then, we would predict that such intermediately headed relatives would
be susceptible to changes that affect headless relatives, but would not undergo
these changes in identical fashion – in other words, that they would likely be
more resistant, but not immune, to processes of reanalysis that affect headless
relatives.
In fact, we do see precisely this phenomenon taking place in Hup. As argued
in Epps (2009), the nominative singular form of the headless relative clause has
undergone a near-complete reanalysis as a converbal clause (i.e. a subordinate
clause having an adverbial function and indicated via a dedicated, nonfinite verb
form; see Haspelmath 1995: 3). The plural variant has undergone a comparable
reanalysis, but one that is much less complete.
Examples of Hup’s standard converbal clause are given in (27–28). As these
examples illustrate, the structure of this clause type is identical to that of the
 Patience Epps

­ 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).

(27) [děh-an hóp-op] tɨh hi-yíʔ-ay-áh

 water-dir immerse-dep 3sg descend-tel-inch-decl
‘[Slipping into the water], he went down (under).’

(28) [děh hɔ́n-ɔ̃p]=mah, tɨh kéy-éy tíh-ǎn-ãh́

 water vomit-dep=rep 3sg see-dynm 3sg-obj-decl
‘While (ritually) vomiting water, he saw her (reflected in the water), it’s said.’

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:

(29) yúp ʔayǔp=ʔĩh, [yɔh-ní-íp]=mah yúp, tɨh

that.itg one=msc  affine-be-dep=rep that.itg 3sg
́
hám-ãh, [tih=yɔ̌
h=n’ǎn wát-ap]
go-decl   3sg(poss)=affine= obj.pl visit-dep
‘So a man, [one who had in-laws], he went [to visit his in-laws].’
RELATIVE
‘So a man, [being affinally related], he went [to visit his in-laws].’
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).

(30) [ciwǐb cóg]=d’ǝh, hɨd ham-píd-ay-áh

 bacaba gather=pl 3pl go-dist-inch-decl
‘They went [gathering bacaba fruit].’
 Between headed and headless relative clauses 

(31) [hup=d’ǝh ni-ě]=d’ǝh b’íyɨʔ=mah hɨd

 person=pl be-perf=pl all=rep 3pl
hidohó-óh, j’ámyɨ-íh
transform-decl long.ago-decl
‘Those (creatures) [who had all been people once], it’s said, they were trans-
formed long ago.’ RELATIVE
‘[Having all been people once], it’s said, they were transformed long ago.’
CONVERB

As these examples illustrate, singular headless relatives and Plural-marked relative

clauses are at different stages in the transition from relative to converb. As argued
in Epps (2009), it was likely the headless nature of the singular relative clause – the
lack of overt indicators of its nominal identity – that facilitated its reanalysis as
a converb. On the other hand, the marginally headed status of the plural variant
(in which =d’ǝh may be interpreted either as a nominal element or as inflectional
morphology) would have made it more resistant – though ultimately still suscep-
tible – to the transition.
In summary, the diachronic changes affecting the Plural-marked relative
clause are evidence of its intermediate status between headed and headless. The
fact that it has lagged behind the singular variant may be attributed to the quali-
ties it shares with headed relatives – which presumably would result in speakers’
greater difficulty in reanalyzing the plural form as a non-nominal construction.
While the Plural marker’s identity as a morpheme that otherwise associates only
with nouns may have contributed to this difficulty, it is also very likely that the
resemblance of Plural =d’ǝh to a domain nominal (a classifier or bound noun)
played an important role. Conversely, that the Plural-marked relative clause could
be reanalyzed at all suggests that speakers were also able to conceptualize it as
a headless relative clause, presumably by attending to the grammatical nature of
=d’ǝh as a marker of number. The fate of Hup’s Plural-marked relative clause thus
indicates that its intermediate identity between headed and headless has had a
certain reality for speakers.

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.

Keywords:  Nominalization; deverbal noun; head-internal; stacking;

relative clauses

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

of this type offered in Andrews (2007: 206) is that an RC “is a subordinate clause

which delimits the reference of an NP [noun phrase] by specifying the role of the
referent of that NP in the situation described by the RC.” Huddleston, P ­ ullum
and Peterson (2002: 1034) refer to this kind of RC as the “integrated relative
clause” and point out that “the prototypical integrated relative serves to restrict
the denotation of the head nominal it modifies.” Comrie (1989: 143) likewise
works with a narrower definition than the traditional concept of RC since the
function of the RC must be to “restrict the range of reference” that the head
might potentially have. All clauses in Seri that fit the cited definitions and char-
acterizations are nominalized clauses (using deverbal noun forms), as shown in
­Section 2, where more details about the forms are given. Therefore RCs always
look quite different from the common finite clauses in the language. But as is
shown in ­Section 4, some formally identical constructions in the language are
not RCs by these definitions.
Before beginning a detailed discussion of the RCs in Seri, a few general facts
about the language are given to help orient the reader. Seri (autoglossonym,
­Cmiique Iitom; ISO 639–3 sei) is typologically a polysynthetic SOV language
that is pro-drop, head-marking, head-final, and without morphological case. It
is spoken by about 900 people in northwestern Mexico and currently consid-
ered an isolate by some (see Campbell 1997; Marlett 2007, 2008b), while at the
same time still viewed by some as a likely member of the putative Hokan family
(Kaufman 1989). The verb has rich cross-referencing morphology and something
akin to relational preverbs (Hale & Craig 1988) rather than postpositional phrases
(­Marlett, in preparation). Clause chaining is common, and different subject mark-
ing appears after the verbs of dependent clauses when the proper conditions are
met. Verb m ­ orphology is agglutinative and mildly fusional. Deverbal nouns are
very common. The determiner phrase is determiner-final. See Marlett (2005) for
more information of a general nature about the language.

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.

(i) Ziix xiica quih c-aai tahac i-ti qu-iij quij

thing things the.fl sn+tr-make md.lc 3.pos-on sn-sit the.cm
hapxa ha-p-áh quij i-t-ataamalca, yo-qu-e.
cottontail sn-pv-say the.cm 3.subj:3.obj-rl-caus+have.antlers dt-us-say
‘The Creator (more literally, ‘the thing that made things’), who was there
(more ­literally, ‘who was sitting on there’), put antlers on the cottontail, it is said.’
 (Cuernos 1)

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

Table 1.  Nominalizers used in forming realis deverbal nouns

Subject-Oriented i- {i-} when the clause is negative
ha- {ʔa-} when the clause is passive (and not negative)
c-/qu- {k-} elsewhere (viz., when the clause is not negative and not
passive)
Direct Object-Oriented Ø when the clause is negative
h- {ʔ-} before {i-} (HAVE)
Ø before/i/(that is not a prefix)
y- {j-} before a long low vowel or {a-} (CAUS, etc.)
oco- {oko-} before a very small class of roots
o- {o-} elsewhere
Oblique-Oriented Ø before stressed short low vowels
Ø before consonants
h- {ʔ-} before {aʔ-} (one suppletive allomorph of the passive prefix)
y- {j-} before a long low vowel or a back vowel of an intransitive
verb (other than the passive prefix mentioned above)†
h- {ʔ-} elsewhere
†This allomorph also causes the back vowel to ablaut to a low vowel. These morphological operation
details are presented in detail in Marlett (1981, in preparation).

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

2.1  Subject-oriented deverbal nouns

The realis DN.S form is fairly straightforward, with three suppletive subject-­
nominalizing (SN) prefixes used: c-/qu- /k/, ha- /ʔa/, and i- /i/. See Table 1. The
prefix used is i- if the clause is negative, ha- if it is passive (and not negative), and
c-/qu- otherwise (i.e., not negative and not passive). These forms are illustrated in
(8) using the transitive verb for ‘drink’.
(8) ‘drink’ DN.S (realis)
sf uf
a. imasi /imasi/ {i-i-m-asi}
‘who does/did not drink it’ sn-tr-neg-drink
b. impasi /impasi/ {i-m-p-asi}
‘that/what is/was not drunk’ sn-neg-pv-drink
c. hapasi /ʔapasi/ {ʔa-p-asi}
‘that/what is/was drunk’ sn-pv-drink
d. quisi /kisi/ {k-i-asi}
‘who drinks/drank it’ sn-tr-drink

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.)

(10) [[cmaam iiquet quih hacx c-miih] quih]

woman 3.pos+on+pregnant.with the.fl apart sn-not.exist the.fl
‘the woman whose child died’
 Relative clauses in Seri 

2.2  Direct object-oriented deverbal nouns

The realis DN.O form is a bit more complicated, with several suppletive object-
nominalizing (ON) prefixes. The prefix used is (a) null if the clause is negative,
(b) h- /ʔ/ before prefixal i, (c) null before other instances of /i/, (d) y- /j/ before
long low vowels and prefixal a, and (e) o- otherwise (Marlett 1981, in prepara-
tion). See Table 1. A few verbs, including the verb for ‘see’ in (4), take the unusual
allomorph oco- /oko/ of the ON prefix. The DN.O form is illustrated in (11)–(12)
using a few forms of the transitive verb for ‘drink’. The realis forms are all mor-
phologically possessed forms, with the possessor prefix indicating the person of
the ­subject. The verb stem (the root and the suffixes that follow it) indicates the
number of the subject.8

(11) DN.O (realis, negative)

a. himasi /ʔimasi/ {ʔi-Ø-m-asi}
‘what I do/did not drink’ 1.pos-on-neg-drink
b. mimasyoj /mimasjox/ {mi-Ø-m-asi-tox}
‘what you (pl.) do/did not drink’ 2.pos-on-neg-drink-pl

(12) DN.O (realis, non-negative)

a. hoosi /ʔoosi/ {ʔi-o-asi}
‘what I drink/drank’ 1.pos-on-drink
b. oosyoj /oosjox/ {i-o-asi-tox}
‘what they drink/drank’ 3.pos-on-drink-pl

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.

(13) DN.O (irrealis, negative)

a. (i)hscmasi ha /(i)ʔskmasi ʔa/ {ʔ-si-m-asi ʔa}
‘that/what I will not drink’ 1s.subj.tr-ir.id-neg-drink aux.o
b. mascmasyoj ha /maskmasjox ʔa/ {ma-si-m-asi-tox ʔa}
‘that/what you will not drink’ 2s.subj-ir.id-neg-drink-pl aux.o

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

(14) DN.O (irrealis, non-negative)

a. ihsiisi ha /(i)ʔsiisi ʔa/ {ʔ-si-asi ʔa}
‘that/what I will drink’ 1s.subj.tr-ir.id-drink aux.o
b. isiisyoj ha /isiisjox ʔa/ {i-si-asi-tox ʔa}
‘that/what s/he/it will drink’ 3.subj:3.obj-ir.id-drink-pl aux.o

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.

(15) [[Cmaam iiquet quih h-oco-h-t] quih]

woman 3.pos+on+pregnant.with the.fl 1.pos-on-see-pl the.fl
ii-sax hant y-aait.
3.pos-spirit land dt-descend
‘The woman whose child we found is happy.’

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.

(16) [[Canoaa ma-hiiy-e] com] qu-iha ha.

boat 2s.do-1.pos+on-give the.hz sn-fast dcl
‘The boat that I gave you is fast.’

2.3  Oblique-oriented deverbal nouns

The realis DN.I form is also somewhat complicated due to several suppletive pre-
fixes that are used.9 The proposition/oblique nominalizing (PON) prefix is (a) null
before stressed short low vowels and before consonants, (b) h- /ʔ/ before the vowel-
initial passive prefix, (c) y- /j/ before a long low vowel or a back vowel of an intran-
sitive verb (by implication, not a vowel of the passive prefix and not stressed), and
(d) h- /ʔ/ elsewhere (Marlett 1981, in preparation). See Table 1. This deverbal form
is illustrated in (17) using a few forms of the transitive verb for ‘drink’. They are all
morphologically possessed forms, with the possessor prefix indicating the person
of the subject.

(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 

a. hiisi /ʔiisi/ {ʔi-Ø-asi} 1.pos-pon-drink

b. himasi /ʔimasi/ {ʔi-Ø-m-asi} 1.pos-pon-neg-drink
c. ipasi /ipasi/ {i-Ø-p-asi} 3.pos-pon-pv-drink
d. impasi /impasi/ {i-Ø-m-p-asi} 3.pos-pon-neg-pv-drink

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.

(19) [[Oot i-naail ih-y-acötim] quih]...

coyote 3.pos-skin 1.pos-pon-cover.oneself the.fl
‘The coyote hide that I used as a blanket (more literally, ‘that I covered
­myself up with’) …’ (Topete 95)

(20) [[comcaac quih Ø ii-cp y-aii] hac]

Seri.people the.fl 3.pos-near 3.pos+pon-be.located/pl the.lc
‘(the place) where the Seri people were staying’ (Topete 63)

3.  Syntax

In this section several syntactic facts about the RC constructions are presented.

3.1  Types of heads

The head of an RC is commonly a noun; a simple example is given in (21).

(21) [[cmaam c-oos-tim] quih]

woman sn-sing-impf the.fl
‘the singing woman, the woman who sings, the woman who sang’
 Stephen A. Marlett

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

The head may also be a personal pronoun, as shown in (24).

(24) [[Me satoj quih qu-i-iit-oj ]] ma-s-moqueep-t a-ha.
2.pro mussel(s) the.fl sn-tr-eat-pl 2p.subj-ir.id-sick-pl aux-dcl
‘You who ate mussels are going to be sick.’

The head cannot be a demonstrative pronoun, however, as illustrated by example

(25), but it may be a name in some contexts, as in (26).
(25) *Taax hap qu-i-cö-toj (coi) ox m-imjöc.
dem.dt.pl mule.deer sn-tr-kill-pl the.pl thus px-think/pl
Intended reading: ‘That’s how those who are expert mule deer hunters
think.’
(26) [[María qu-isil] quih]
sn-little the.fl
‘the younger María’ (Marlett 2008a: 50)

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 

And the head may also be null, as in (27).

(27) [[Ø Hapxa ha-p-áh] quij]...
cottontail sn-pv-say the.cm
‘The one that is called cottontail …’ (Conejo_Cuernos 007; also in 001)

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.

3.2  Position of the head

All of the RCs in the language appear to be analyzable as being head-internal, and
some of them must be so analyzed. Therefore, to avoid positing unnecessary struc-
tures, I claim that they are all in fact head-internal. As a result, Seri is one more
North American language to add to the small sample listed in Dryer (2008) with
internally headed relative clauses. Jacobsen (1998) presents a much more exten-
sive list of North American languages of this type, building on work by Langdon
(1977); Jacobsen uses the term headless relative clauses for these structures. Like
some of those languages, Seri presents no examples of RCs preceding the head,
and no examples of extraposed RCs.12 Comrie and Kuteva’s (2005) typology lists
head-internal RCs under the non-reduction strategy.
Since the basic word order is SOV and arguments are commonly not present
as DPs, this means that the vast majority of RCs in Seri actually look like they are

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

Figure 1.  Example (32)

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.’

(35) a. [[ctam canoaa quih hiiye] quij]

man boat the.fl 1.pos+on-give the.cm
b. *[[canoaa quih ctam hiiye] quij]
‘the man whom I gave the boat’

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

Variability in position of the head is also illustrated in examples (36)–(37) for

two obliques (instrumental, and location as object of relational preverb iti,
respectively).
(36) a. [[eenim ziix i-pxasi quih mii-zix] cop]
knife thing 3.pos-flesh the.fl 2.pos+Pon-saw the.vt
b. [[ziix ipxasi quih eenim miizix ] cop]
‘the knife with which you cut the meat’

(37) a. [[ctam-cö coi hant iti haxoj ii-zcam] hac]

man-pl the.pl land 3.pos.on shore 3.pos+pon-arrive/pl the.lc
b. [[hant ctamcö coi iti haxoj iizcam] hac]
‘the place where the men beached’

3.3  Inheritance of properties of the head noun

The properties of the head noun are inherited by the complete DP. Therefore, if
the head noun is plural, a subject-oriented RC requires a denominal verb with
a plural stem, and the DP determines plural agreement on any verb that agrees
with it, as shown in example (38). The head noun caaytaj ‘horses’ is explicitly
plural and the denominal verb coopl in the RC is explicitly plural (indicated by
deletion of the vowel of the second syllable since the singular is coopol). The verb
of the main clause, caacöl, is also explicitly plural (compare the singular form
caacoj).

(38) [[Caay-taj quih c-oopl] coi] c-aacöl iha.

horse-pl the.fl sn-black+pl the.pl sn-big+pl dcl
‘The black horses are big.’

3.4  Definiteness of the head and the presence of definite articles

Williamson (1987: 169), based on facts from Lakhota, proposes that languages
with head-internal RCs universally have an indefinite restriction – the head must
be indefinite although the phrase is definite. Basilico (1996: 526) and Andrews
(2007: 271) repeat this claim. However, this proposal is incorrect for Seri as well as
for Haida (see Enrico 2003: 577–578) and several other languages (Cinque 2009).
The heads in Seri can be marked as definite by an accompanying definite article,
or they can be presented without the definite article, but the head of a definite RC
is never presented as indefinite.
The definite article quih (the.fl) has an important use within DPs that have
RCs. In these situations it is often very lightly pronounced, sometimes being
 Relative clauses in Seri 

­ 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

canoaa quih quisil com

‘boat’ ‘the’ ‘small’ ‘the’

Figure 2.  Example (39)

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

(41) [[[[[ [Haxz]NP quih]DP [qu-isil]VP xah [c-aacoj]VP]S quih]DP

dog the.fl  sn-little and/or  sn-big the.fl
[hant quih i-ti c-aap]VP]S cop]DP ziix z iij
 land the.fl 3.pos-on sn-stand the.fl thing a differently
cö-po-m-aai-tim ta x, ...
3.io-ir.dp-neg-do-impf ds ut
‘If there is a small or big dog somewhere that isn’t doing anything wrong, …’
 (Consejos_Perro 01)

(42) [[[[[ [Hap ]NP quih]DP [c-aacöl]VP]S quih]DP [qu-iipe ]VP ]S

mule.deer the.fl  sn-big+pl the.fl  sn-good
tacoi ]DP i-t-amjc
xo...

md.pl 3.subj:3.obj-rl-bring although
‘Although he brought those big good mule deer, …’ (Gigante_Comelon 187)

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 

(46) a. ctam haxz quih oco-ho cop

man dog the.fl 3.pos+on-see the.vt
b. haxz quih ctam ocoho cop
c. *haxz quih ctam quih ocoho cop
d. *ctam quih haxz quih ocoho cop
‘the man that the dog saw’
(47) a. Ø ctam ocoho cop
b. *Ø ctam quih ocoho cop
‘the man that s/he/it saw’
The situation is a bit different when the head noun is indefinite. Under very
unusual conditions relating to negation, but not uncommonly, the DP with an RC
and also the head noun occur with the singular indefinite article.
(48) [[cmiique zo ziix z imoz.cöiihca17 ]S z]DP
Seri.person a thing a who.hoards a
i-m-á ha.


sn+tr-neg-know dcl
‘… he had never known anyone who hoarded.’ (Gigante_Comelon 477)
(49) … [[Ø ziix zo cö-i-y-asíyal-am ]S zo ]DP haquix
thing a 3.io-3.pos-on-use/pl-impf a somewhere
i-m-iih iha.
sn-neg-be.located.fl dcl
‘…there was nothing else [no other shell] that they used as much.’ (Xtiip 06)

3.5  Recursive RCs

Various examples of recursive RCs have already been presented. See Section 3.1,
and particularly examples (22), (23), and (30). In those particular examples, the
embedding is quite simple, with the RCs ultimately modifying a single nominal.
This is the kind of recursion that has been called stacking (see Note 10). Some
recursive RCs are structurally different in that the RCs modify different nomi-
nals. Such RCs have not been found commonly in texts, but certain ones may be
elicited without too much difficulty (albeit with patience since they are hard to
process). Example (50) shows one such example where each subject RC is modify-
ing a different nominal; it was judged as being completely grammatical although I
doubt that anyone has said anything quite like it. In (50) I label the brackets with
numbers to indicate the pairings and include only bracketing for the DPs (which
contain the RCs, of course).

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

(50) 1[Haxz 2[ziic 3[zixcám siimet quih qu-i-hit quih]3 qu-i-cö

dog  bird  fish bread the.fl sn-tr-eat the.fl sn-tr-kill
quih 2 cö-c-aaitim cop]1 y-aacö.
the.fl] 3.io-sn-chase the.vt dt-uo.bark
‘The dog that chased the bird that killed the fish that ate the bread barked.’

Attempts to elicit recursive relatives involving an object RC have been completely

unsuccessful, in sharp contrast to the situation with all recursive subject relatives.
See unacceptable examples (51)–(52), for which no grammatical examples could
be produced.

(51) a. *1[Ziic quij 2[zixcám com siimet oo-hit]2

 bird the.cm fish the.hz bread 3.pos+on-eat
ooi-cö quih]1 m-ooxp.
3.pos+on-kill the.fl px-white
Intended reading: ‘The bread that the fish ate that the bird killed was
white.’
*Ziic quih zixcám quih siimet oohit ooicö quih mooxp.
b. 
c. *Siimet zixcám quih oohit ziic quih ooicö quih mooxp.

(52) *1[Ziic quij 2[zixcám siimet quih qu-i-hit (quih)]

 bird the.cm   fish bread the.fl sn-tr-eat the.fl
ooi-cö com] c-aacoj iha.
3.pos+on-kill the.hz sn-big dcl
Intended reading: ‘The fish that ate the bread that the bird killed was big.’

The problems with sentences containing multiple center-embeddings have been

of interest for a long time; see Miller and Chomsky (1963), Kuno (1974), de Roeck
et al. (1982), and Hudson (1996) for a part of the literature. The sharp contrast
in judgments on these examples compared to those like (50) casts doubt on any
proposal that center-embedding is itself the problem, however, since (50) as well as
(51)–(52) would be characterized as involving center-embedding.18
Whatever disallows examples such as (51)–(52) in Seri must not be allowed to
block grammatical examples such as (53), which has a pair of stacked relatives and
then a subject relative modifying the noun “possessed” by that complex nominal.
I do not have any formal proposal to make.

18.  De Roeck et al. (1982: 328) clarify that A is center-embedded in B if some material of B

both precedes and follows A.
 Relative clauses in Seri 

(53) [[[[hehe cö-c-ootij] hoox qu-isoj] án qu-inej] zo]

plant 3io-sn-dry intns sn-have.body 3.pos/area sn-empty a
‘a very large dry bush with an empty area under it’ (More literally,
‘a very large dry bush whose area is empty’) (Conejo_Puma 17.2)

3.6  Verbless RCs

It is very common to express the relationship of identity between two nominals
using a construction that consists of the subject nominal (typically a DP) followed
by a predicate nominal (an NP, never a DP) and a declarative or interrogative
enclitic modal. A simple pair of examples is given in (54).
(54) Subject DP Predicate NP Mood
a. [Ma-ta quih] [cmaam qu-iipe] ya?
 2.pos-mother the.fl  woman sn-good dcl
‘Is your mother a good woman?’
b. [Hipiix] [hast] iha.
 dem.px  stone dcl
‘This is a stone.’

Similarly, an RC may be a predicate nominal that has no accompanying verb, as

shown in (55)–(57) and illustrated in Figure 3 for example (55).

DP

S D

DP NP

NP N

haxz ctam ticop

‘dog’ ‘male’ ‘that’

Figure 3.  Predicate nominal as RC (example (55))

(55) [[[Haxz]DP [ctam]NP]S ticop]DP hin-yo-catxla.

dog man/male  md.vt 1s.do-dt-bite
‘That male dog bit me.’
(56) [[[He]DP [cmajiic]NP]S ]DP ha-s-cm-alx a-ha.
1.pro woman/pl 1p.subj-ir.id-neg-go/pl aux-dcl
‘We women will not go.’
 Stephen A. Marlett

(57) … cmaam zo toc cö-t-iij, [[[[[[[[[ziix]DP [cmiique]NP ]S]DP

woman a there 3.io-rl-sit  thing  Seri.person
[caii]AP]S]DP [ctam]np ]S]DP [aal]NP]S quij,] DP tiix
 mature  man/male  3.pos+spouse the.cm dem.dt
ah Bariil im-p-áh.


foc B. px-pv-say
‘… there was a woman, the adult male [and now deceased] who was her
spouse was called Barrel.’ (Hombre_Barril 26–27)

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.

3.7  Embedded finite clauses

An RC may have an embedded dependent clause with a finite verb. Examples
­(58–59) both have overt head nouns in initial position. In both of these examples
the head noun is also subject of the embedded finite clause.
(58) … comcaac [Socaaix ano mo-t-a-t
Seri.people  S. 3.pos/from twd-rl-move-pl
he-c-aazcam]S coi …
1.io-sn-arrive/pl the.pl
‘… the people who, coming from Socaaix, arrived to me …’

(59) … cmiique [i-t-al tahac

 person 3.subj:3.obj-rl-accompany md.lc
co-nti-c-a]S tintica...
3.io-aw-sn-move md.aw
‘… the person going there who had accompanied him …’

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.

4.  The frequency of RCs in Seri discourse

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.

4.1  Four types of constructions to be considered separately

Most quantifiers, including numbers, are predicates in Seri, and so when one men-
tions the quantity of some item, an RC-looking structure is used. The quantifier in
most cases is not delimiting the reference of the nominal, but rather simply telling
more about it by indicating how much or how many. For that reason they are not
considered RCs here. Two examples are given in (62)–(63).
(62) Hapxa c-oocj ih-yoo-ho.
cottontail19 sn-two 1s.subj.tr-dt-see
‘I saw two cottontail rabbits.’

(63) zaah c-cooo cop

sun/day sn-entire the.vt
‘the whole day’ (Topete 055)

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)

(65) ziix qu-iisax

thing sn-have.breath
‘person’

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

(67) [Ø yaazi] < {i-j-aaʃi}

‘his child’ (specifically child of male)    3.pos-on-carry

(68) [Ø moop] < {mi-o-ap}

‘basket that you are making’    2.pos-on-sew.basket
‘basket that you have made’

(69) …, [Ø c-oos] zo toc cö-p-iij ta, [Ø c-ooila]

 sn-sing a there 3.io-ir.dp-sit DS   sn-dance/impf
quih toc cö-s-oii a-ha.
the.fl there 3.io-ir.id-stand/pl aux-dcl
‘…, there will be a singer there, (and) there will be dancers.’ (Topete 127)

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)

(71) Xaasj21 C-acöla quih ano t-oii, …

cardon.cactus sn-tall/pl the.fl 3.pos/in rl-be.located.fl/pl
‘They are at “Tall Cardones”, …’ (Topete 098)

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 2.  Basic information about texts included in the analysis

Short title Genre # of sentences # of clauses

Conejo_Cuernos folktale  8  24

Conejo_Puma folktale  20  60
Topete personal history 132 379

4.2  A small statistical analysis

For the purposes of this article an inventory was made of RCs in three Seri texts.
These texts varied in length and genre, as shown Table 2. The clause count in that
table does not include RCs or any of the types of clause mentioned in Section 4.1.
The clause count also excludes the short clause yoque ‘one said’ used as an evi-
dential in the two folktales (5 instances in the Conejo_Cuernos text and 11 in the
Conejo_Puma text). In addition, the clause count excludes the clausal expression
ox oo mpacta xo ‘however’ (literally, ‘it was like that but’), and the very common
sentence-initial transitional clause ox tpacta ma ‘thus it was’, which is usually best
translated as sequential ‘then’ (19 instances in the Topete text).
The number of RCs in these texts is shown in Table 3, as well as the ratio of
RCs to the clauses. Of course, this is a very small sampling of text, but the data are
quite suggestive for the language generally.22

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

Conejo_Cuernos 3 1:8 (3:24) 1:3.3 (3:10)

Conejo_Puma 8 1:7.5 (8:60) 1:5.5 (4:22)
Topete 29 1:13.1 (29:379) 1:5.5 (25:138)

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.

Table 4.  Forms and transitivity of RCs

Short title DN.S DN.I Ratio of null head Ratio of transitive
to overt head RC to intransitive RC

Conejo_Cuernos 3 0 1:2 0:3

Conejo_Puma 8 0 0:8 0:8
Topete 14 7 1:4.8 1:13.5

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

The language of the Gavião of Rondônia, a Tupian language spoken in western

Brazil, constructs relative clauses by syntactic nominalization, using either of
two particles, mát ‘concrete nominalization’ or méne ‘abstract nominalization’,
which are derived diachronically from discourse pronouns. The resulting
nominalized clause may or may not have an internal head, which, if it occurs, is
not marked, leading to a certain degree of ambiguity. The nominalized clause can
modify a following noun stem, which serves as an external head. Relative clauses
and complement clauses are not distinct constructions in this language. The
typological significance of these constructions is discussed.

Keywords:  relative clauses; Gavião of Rondônia; nominalization; complement

clauses

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.

2.  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.’)

(2) mát ká bó tá-máà ci-ko-e

that in focus 3p-aux.past 3s-mouth-abstract (=words)
ábemáá teé-á
follow continuing-end
‘They were there [in that] repeating his words.’ (Previous sentence: ‘“Where
do I enter to drink manioc beer with you?” he asked.’)
(3) méne ajálá bó máà a-ti
that leave focus (3s)-aux.past 3c-mother
kay-á kí-ip
involve-end evidence-past
‘He left that with his mother.’ (Previous sentences: ‘He roasted many birds
before leaving on his journey to bring darkness. The birds will be signs of
his return.’)
(4) “méne mi paní pa-vít aka kí̦-á”
that use 1pi.aux.let 1pi-food kill again-end
bó vit aká-ày máà kí̦-á
focus food kill-agent.pl aux.past again-end
‘“Let’s kill our food in that way again”, the hunters said.’ (Previous sentence:
‘There he performed magic for their hunting to fail.’)

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).

3.  Nominalization of verb phrases

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

(6) natáó ká mát sep ma-’-óló

Christmas in nmnlz.concrete leaf.obj trans-come
e̦èt tó-koy-á
2s-(aux.imperat.def) 1pe-to-end
‘Give the Christmas photos to us.’

(7) sep tóló ká bó, méep, ñá natáó ká

leaf.obj long.pl in focus hesitation that Christmas in
méne sé-e-na
kí-nap

nmnlz.abstract aux.subjective-that-manner evidence-indef
‘It’s in the long photos that there are pictures of Christmas.’

(8) ó̦o té me-sá a̦á sérék

neg non.affirm 2p-aux.non.affirm this (3s)-clothing
tá méne íkinì
with nmnlz.abstract see
‘You don’t see how she is with her clothing?’ (i.e. pregnant)

A nominalization is defined by Givón (1990: 498) as “the process by which a

prototypical verbal clause – either a complete sentence (including the sub-
ject) or a verb phrase (excluding the subject) is converted into a noun phrase”.
However, the syntactic nominalizations formed by mát and méne are not full
noun phrases. Evidence for this is that these syntactic nominalizations (see (9)
below), paralleling morphological nominalizations (see (10) below), can mod-
ify a noun stem which follows them, which is impossible for a noun phrase in
Gavião, as it is in English (*these red bricks house vs. brick house). The syntac-
tic nominalizations are considered here as syntactic nouns or bar-one nouns –
nominal constructions larger than morphological nouns but smaller than
noun phrases. In Gavião each major lexical category has a bar-one projection
(Moore 1984). Within the phrase, these syntactic nouns have the same distri-
bution as any other noun, for example being optionally modified by following
adjectives. Their modification of following noun stems is exactly parallel to the
­modification of following noun stems by nonderived nouns, for example in (11)
below.

(9) ñá natáó ká mát sep

that Christmas in nmnlz.concrete leaf.obj
‘photos of that Christmas’

(10) vit ma’á̦-p sep

food get-nmnlz.concrete leaf.obj
‘money to buy food’
 Relative clauses in Gavião of Rondônia 

(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’

4.  Nominalization of clauses

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’.

(14) pa-bábe ákinì mága ó̦òp saká-néè

1pi-hand see (3s)-make neg 3s.aux-nominal
mát
sé-e-na

nmnlz.concrete aux.subjective-that-manner
a-ma-kóbáá kí-nap
3c-trans-learn evidence-indef
‘Who doesn’t make [pottery] learns watching our hands.’

In (14), the internal head is the third person subject and the nominalized clause is
the subject of the matrix sentence.

(15) “náapó tígi paá-néè náapó

that time.of 1pi-(aux.past.def)-nominal that
kávo ká méne mi teé pa-máge-é-na
year in nmnlz.abstract use continuing 1pl-aux-that-manner
ibal-é-na kí̦-á” mâà
dance-that-manner again-end 1s.aux.past
‘I say that we are going to dance again as we did at that time in the past year.’

In (15) there is no internal head in the nominalized clause, which is the object of
the verb ‘use’

(16) “eé-na tá-máà a-ma-ágá̦á,

that-manner 3p-aux.past 3c-trans-begin.day
a-kapá̦p ma-‘-í̦í aá-néè
3c-darkness trans-enter 3c-(aux.past.def)-nominal
méne
ká-á” kí-ip

nmnlz.abstract in-end evidence-past
‘In that manner they made it become day when they put darkness back
again.’ (i.e. into the container)

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 

máà tá-kay-á” kí-ip

(3s)-aux.past 3p-involve-end evidence-past
‘“I got them behind where you were calling them”, he said to them.’ (A man
explains to the others where he found the monkeys.)

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:

(18) natáó ká eé-néè sep

Christmas in 2s-(aux.past.def)-nominal leaf.obj
ígí mát sep íkinì maté̦é
take nmnlz.concrete leaf.obj see causative
e̦èt kí̦-gáre-ále-á
2s-(aux.imperat.def) again-yet-future-end
‘Show us the photos that you took at Christmas.’

In (18), the internal and external head is ‘leaflike object’, which designates photos.

(19) até a̦á sérék tá zâ-néè mát

affirm this clothing with 1s.aux-nominal nmnlz.concrete
sérék sá káre kí-nap
clothing aux.subjective still evidence-indef
‘This clothing which I am using still exists.’ (viewing photo)

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:

(20) ñá natáó ká eé-néè paá-co ígí

that Christmas in 2s-(aux.past.def)-nominal 1pi-photo take
méne sep
nmnlz.abstract leaf.obj
‘The photos that you took of us on that Christmas’
 Denny Moore

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.’

As an example of the ambiguity of the nominalizations, which have no marking of

the internal head, consider the example below which, according to Gavião consul-
tants, has three interpretations:
(22) bay ánéè paderè va
snake (aux.past.def)-nominal person bite
mát sáp


nmnlz.concrete house
a. ‘the house of the person that the snake bit’ (‘person’ being the internal
head)
b. ‘the house of the snake which bit the person’ (‘snake’ being the internal
head)
c. ‘the house where the snake bit the person’ (with no internal head)
In the example below there is no internal head and the external head has no coun-
terpart inside the clause. This external head is a nonderived noun stem of a type
which, like a morphological nominalization, forms a syntactic noun with a pre-
ceding NP. Note that the translation is given as a complement clause instead of a
relative clause.
(23) bolíp ígí ó̦-ga-néè méne kázóp
fish get 1s-(aux.past.def)-go-nominal nmnlz.abstract sign
‘sign that I went fishing’

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

This paper deals with a number of relative constructions in Yucatec Maya

(the Mayan language of the Yucatan Peninsula) where no overt nominal head
is observed. In these constructions however, several elements are observed
in the domain of the relative clause, such as determiners and quantifiers. The
paper addresses the question of whether these elements should be analyzed
as light heads in the sense of Citko (2004). Evidence is presented to argue that
this is not the case. Instead, the evidence indicates that the relatives under
consideration do have a nominal head, albeit one that is phonetically null. This
null head, however, can still be modified by the ordinary nominal modifiers
of the language, including determiners, quantifiers and relative clauses.

Keywords:  Yucatec Maya; relative clauses; null heads; noun phrase

1.  Introduction

Descriptive works on relative clauses make a standard distinction between headed

and headless relatives (Comrie (1989), Kroeger (2005), Andrews (2007)). In a
recent study, Citko (2004) proposes that Polish displays a third kind of relative,
light-headed relative clauses. An example of this kind of relative is presented in (1),
where apparently there is no nominal head that could function as the head of the
relative. However, Citko (2004) crucially argues that in these cases the determiner
that introduces the relative is formally and structurally the head of the relative
clause.

*  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 

(4) Leti’ le ts’a-ik-ø to’on le janal-o’.

3sg dm give-ind-abs.3sg 1pl dm food-cl
‘He is the one that gives us food.’ (MDG-B: 280)

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.  Preliminary descriptive considerations

2.1  Basic clause structure

Yucatec Maya is a language that has an ergative-absolutive cross-referencing
­system that is split on the basis of aspect, but the language as a whole displays
mostly nominative-accusative syntax. It is a strictly head-marking language. The
verb agrees with the object through a set of pronominal suffixes (glossed ABS
in what follows), whereas the transitive subject is cross-referenced by a series of
­proclitics (glossed ERG in what follows) that may attach prosodically to either the
verb or to a number of different preverbal elements, most often auxiliary particles.
The ergative proclitics are also used to cross-reference the possessor of the head of
a noun phrase. Yucatec is a systematically head-initial language, except for the fact
that the basic word order of its transitive constructions is SVO (see Briceño Chel
(2002), Gutiérrez-Bravo & Monforte (2008), Gutiérrez-Bravo & Monforte (2010))
which makes it different from most other Mayan languages.2 The basic structure of
the clause in Yucatec (in brackets in the example below) consists of the main verb,
the ergative proclitic cross-referencing the subject, and an a­ uxiliary particle that

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)

2.2  Relative clauses and relativization strategies

Previous works dealing with the description of relatives in Yucatec are scarce.
Brief descriptions can be found in Bricker (1978) and Tonhauser (2003). A more
elaborate description of the structure of relatives in this language can be found in
Gutiérrez-Bravo and Monforte (2009). However, these three works focus mostly
on the agent focus form of the verb found in some relative clauses, rather than
in their general typological properties.3 Hence I first provide a basic typological
description of relative clauses in this language. The overall description of rela-
tives that follows relies heavily on the descriptive terminology of Comrie (1989)
and Andrews (2007). In all of the examples that follow, the head of the relative is
underlined for ease of exposition.
Relatives in Yucatec are clausal in nature. With the exception of subject inver-
sion when the subject of the relative is expressed overtly, they show no significant
syntactic or morphological asymmetries in comparison with matrix clauses. They
display finite morphology identical to that of matrix clauses and they are not intro-
duced by a complementizer or any other element signaling their subordinated sta-
tus. This is illustrated with the example in (6), where the basic clause structure
illustrated in (5) is equally observed in the relative clause.
(6) Le tunich [RC tu’ux k-u pak’-a’a-l le graasia]-o’.
dm stone where aux-erg.3 sow-pass-ind dm stuff-cl
‘The stones where (in between which) the stuff (corn) is sowed.’
 (MDG-B: 15)

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)

Relatives in Yucatec display two different relativization strategies, the relative

­pronoun strategy (as in (6)) and the gap strategy. As noted in Gutiérrez-Bravo
and Monforte (2009) and Gutiérrez-Bravo (2010), each kind in turn displays
a different behavior with respect to the properties of the head of the relative.
­Specifically, all kinds of gap relatives may or may not show an overt head, but
subject and object pronominal relatives never display an overt head: pronominal
relatives can occur with an overt head, however, when an oblique argument or
adjunct is ­relativized, such as the locative expression relativized in (6). In the
following subsections I briefly describe the properties of both pronominal and
gap relatives.

2.2.1  Pronominal relatives

Pronominal relatives in Yucatec show a set of relative pronouns that are a s­ ubset
of the corresponding interrogative pronouns (ba’ax ‘what/which’, máax ‘who/
whom’ and tu’ux ‘where’). Both relative and interrogative pronouns are obliga-
torily fronted to the left edge of the clause in Yucatec.5 This is illustrated for
the pronoun máax ‘who’ in the following examples, where (9) corresponds to a
subject relative.
(8) Pues máax ts’-u paak’áal way-e’?
well who term-erg.3 sow here-cl
‘So who from here has already sowed?’ (MDG-B: 271)

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

(9) Lekan taa-k-ø le [RC máax bi-s-ik-ø le

when come-irr-abs.3sg dm who go-caus-ind-abs.3sg dm
paca ti’ Enlace]-o’…
bale prep Enlace-cl
‘When the one who takes the (henequen) bales to Enlace comes…’
 (MDG-B: 105)

Pronominal relatives in Yucatec are observed for subjects, objects, indirect

objects, prepositional complements and locations. Further examples are pre-
sented below.6
(10) Le [RC ba’ax k-in tsikbal-t-ik-ø te’ex]-a’.
dm what hab-erg.1sg chat-trns-ind-abs.3sg 2.pl-cl
‘This (thing) which I’m telling you.’ (MDG-B: 108)

(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)

(12) Le lu’um [RC tu’ux ken a pak’-ø xan]-o’.7

dm soil where aux erg.2 sow-abs.3sg also-cl
‘The soil where you’re going to sow it too.’ (MDG-B: 224)

Lastly, as mentioned briefly in the preceding section, pronominal relatives in

Yucatec may only display an overt nominal head when an oblique is relativized.
Hence an overt head can be observed in location relatives like (6) and in indirect
object relatives like (11), but not in subject relatives like (9) or object relatives like
(10). This restriction is illustrated in the examples below:

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 

(13) T-in kax-t-ik-ø le [ba’ax

dur-erg.1sg look.for-trns-ind-abs.3 dm  what
k-u y-ok-ol t-in
kool]-o’.

hab-erg.3 ep-enter-ind prep-erg.1sg cornfield-cl
‘I’m looking for the (thing) which goes into my cornfield.’

(14) *T-in kax-t-ik-ø le kitam [ba’ax k-u

dur-erg.1sg look.for-trns-ind-abs.3 dm boar  what hab-erg.3
y-ok-ol t-in
kool]-o’.

ep-enter-ind prep-erg.1sg cornfield-cl
(‘I’m looking for the boar which goes into my cornfield.’)

2.2.2  Gap relatives

As mentioned in 2.2, relative clauses in Yucatec are not introduced by a comple-
mentizer or any other element signaling their subordinate status. As a result,
gap relatives in Yucatec are akin to English contact relatives like (15), with the
addition that in Yucatec this kind of relative can also be observed for subject
relativization. This is illustrated by the object relative in (16) and the subject
relative in (17).
(15) The book [I read ___].

(16) Mina’an-ø u chan p’óok [RC k-u p’at-ik-ø

neg.ex-abs.3sg erg.3 little hat hab-erg.3 leave-ind-abs.3sg
t-u
___ yáanal u k’áan]-o’.
prep-erg.3 under erg.3 hammock-cl
‘His little hat he used to leave under his hammock was not there.’
 (MDG-B: 48)

(17) Jmeen, jaaj, jun túul jmeen [RC ____ ku

priest true one numc priest hab-erg.3
meen-t-ik-ø waajil kool].
make-trns-ind-abs.3sg cornfield.ceremony
‘He was a priest, it’s true, a priest that performed the cornfield ceremony.’
 (MDG-B: 61)

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.

3.  Light heads vs. null nominal domain

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 

(22) Leti’ le [ts’a-ik-ø to’on le janal]-o’.

3sg dm  give-ind-abs.3sg 1pl dm food-cl
‘He is the one that gives us food.’ (MDG-B: 280)

Before addressing this issue, a terminological note is necessary here. An anony-

mous reviewer asks whether the relative constructions in (21) and (22) should
be taken to be free relatives, given their absence of an overt nominal head. It’s
worth noting here that whereas typological works often use the terms head-
less relative and free relative interchangeably, theoretical works often provide
a more specific characterization of free relatives. Specifically, free relatives are
taken to be pronominal relatives that lack the referential head noun that is typi-
cally ­modified by restrictive (i.e. headed) relative clauses, as in English I brought
[what you need] (see van Riemsdijk 2006 for a recent survey of free relative con-
structions). I further assume that free relatives are not modified by any of the
modifiers that regularly appear with the missing head noun. This distinction
between headless relatives and free relatives is not purely terminological. The
relative pronoun of English free relatives, for instance, is analyzed by Bresnan
& Grimshaw (1978) as being in fact the external head of the relative construc-
tion. Hence, in some ­languages at least, free relatives are not strictly speaking
headless. ­Furthermore,  in contrast with other kinds of relatives, free relatives
commonly show some properties that are akin with the properties of embedded
interrogatives; again, see van Riemsdijk (2006) for a summary of these proper-
ties. For the purposes of what follows, though, the relevant point is that (21) and
(22) are (apparently) headless relatives, but not free relatives under the definition
I adopt. This is because (21) is a pronominal relative without a nominal head,
but it is introduced by the determiner le, and (22) is a gap and not a pronominal
relative. Yucatec does have free relatives as defined above, but their analysis goes
beyond the scope of this paper. Instead I refer the reader to Gutiérrez-Bravo
(2010) for detailed discussion.
Now, returning to our central discussion, in an analysis of cases like (21) and
(22) as light-headed relatives, it is crucial that the light element in the domain of
the relative (the demonstrative le in the examples above) is nonetheless the struc-
tural head of the noun phrase in which the relative is embedded (Citko 2004: 110).
I now present evidence that this is not what is observed in Yucatec relatives.
Instead, I argue that the demonstrative le in (21), (22), and similar structures is
a modifier of a null nominal head, which is represented as Ø in what follows. As
such, (21) and (22) are neither light-headed relatives nor headless relatives, but a
relative structure in which the head of the noun phrase happens to be phonetically
null, a typological possibility already considered in Lehmann (1984). Following
Gutiérrez-Bravo and Monforte (2009), I thus analyze the noun phrase in (22) and
its relative clause as in (23).
 Rodrigo Gutiérrez-Bravo

(23) [NP Le Ø [RC ts’a-ik-ø to’on le janal]]-o’.

dm give-ind-abs.3sg 1pl dm food-cl
‘The (one) that gives us food.’

It is a crucial part of the argumentation in favor of a structure such as the one

in (23) to show that null nominal heads are not dependent on the presence of a
relative clause. In other words, the possibility of having a null nominal heading
the noun phrase appears to be a typological property available in some languages
(but not in others, i.e. English) for any given noun phrase (see Gutiérrez-Bravo &
Monforte (2009)). This is indeed what is observed in Yucatec. As illustrated in (24)
the possibility of having a null nominal is a regular property of noun phrases in
Yucatec (see Gutiérrez-Bravo (2002)).
(24) [NP Le boox Ø] -a’
dm black cl
‘The/this black one.’

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.

3.1  No distinct set of elements that function as light heads

A first argument against analyzing the Yucatec relatives under consideration as
light-headed relatives has to do with the range of elements that can appear in their
domain. In Polish, the set of elements that can appear in the domain of a light
headed relative includes, but is not limited to, NP determiners. Other possible
light heads include indefinites, negative indefinites and universal pronominals
(i.e.  everything, everywhere, etc.). In contrast, in Yucatec the elements that can
appear in the domain of relatives like (21) and (22) is exactly the same as the set
of elements that can precede the head noun in any kind of noun phrase. Thus the
determiner le of (21) and (22) can equally be observed in the full headed relative in
(6), or in noun phrases without relative clauses such as those in (7) and (24). This
is further illustrated below with examples that show the quantifier tuláakal, ‘all’.
(25) Tuláakal le meyaj-o’ob-o’.
all dm worker-plur-cl
‘All the workers.’ (MDG-B: 103)
(26) Tuláakal le gente [RC k-u taal]-o’, k-u
all dm people hab-erg.3 come-cl hab-erg.3
ts’a-ik-ø u janal.
give-ind-abs.3sg erg.3 food
‘All the people that came, he gave them their food.’ (MDG-B: 107)
 Relative clauses in Yucatec Maya 

(27) Ka ka’ans-a’a-k-ø … tuláakal le Ø

comp teach-pass-irr-abs.3sg all dm
[RC ts’o’ok in tsikbal-t-ik-ø te’ex]-a’.
term erg.1sg chat-trns-ind-abs.3sg 2pl-cl
‘So that all this I have already told you about… be taught.’ (MDG-B: 109)

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.

3.2  Same set of relative pronouns as headed relatives

Citko observes that headed and light-headed relatives in Polish make use of dif-
ferent sets of relative pronouns. This is in fact entirely coherent with Citko’s pro-
posal. In her analysis, full-content nominals and the light nominals that function
as the head of light headed relatives are two distinct subclasses of nouns. Hence it
is not surprising that they make use of different pronominal elements inside the
relative clause when they head one (see also Andrews 2007: 218). This distinc-
tion, ­however, is not observed in Yucatec. The relative pronouns of the relatives
under consideration are exactly the same as those of fully headed relatives. This is
­illustrated for the relative pronoun tu’ux ‘where’, in the examples below.
(28) Le lu’um [RC tu’ux ken a pak’-ø xan]-o’.
dm soil where aux erg.2 sow-abs.3sg also-cl
‘The soil where you’re going to sow it too.’ (MDG-B: 224)
(29) Le Ø [RC tu’ux ts’-u yáax máan le meyaj]-o’.
dm where term-erg.3 first pass dm work-cl
‘Where they first passed (i.e. went) to work.’ (MDG-B: 105)

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.

3.3  Distribution in existential contexts

The third argument that supports the null head analysis in (23) comes from the dis-
tribution of bare nouns and null bare nouns modified by a relative clause. Yucatec
readily allows for bare nouns (i.e. nouns that are not introduced by a determiner),
 Rodrigo Gutiérrez-Bravo

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)

It is important to note that the existential yaan in Yucatec is an adjective. As such,

the relevant parts of (30) and (31) are non-verbal predications where the adjective
yaan selects for a single nominal argument.8 Agreement between yaan and this
nominal is expressed with the series of absolutive suffixes of the language. Now,
in principle we can expect these bare nouns to be modified by a relative clause.
Given that Yucatec nouns have the possibility to be null we thus expect to find
these n­ ull-head relatives in adjacency to the existential yaan. This is indeed what
is observed in Yucatec, as illustrated in (32) and (33).
(32) Yaan-ø [k-u púut-ik-ø-o’ob le fibra]-o’.
ex-abs.3sg hab-erg.3 carry-ind-abs.3sg-plur dm fiber-cl
‘There were those that carried the fiber.’ (MDG-B: 101)
(33) Yaan-ø [k-u wéej taal bejla]-e’.
ex-abs.3sg hab-erg.3 still come today-cl
‘There are those that still come today.’ (MDG-B: 270)

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

In this paper I have analyzed a number of relative constructions in Yucatec Maya

where the nominal head of the relative clauses is absent. I have presented ­evidence
that these relatives are not light-headed relatives, but rather relatives with a null
nominal head, a possibility first considered in Lehmann (1984), and indeed
­considered by Citko (2004) herself in her analysis of light headed relatives in

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

In a number of languages, interrogative and relative pronouns show the same

forms. The pattern is not distributed evenly around the globe, however: it is
concentrated in Europe. It does appear elsewhere, for example in South America
in Tariana (in contact with Portuguese), and in Mesoamerica in Nahuatl (in
contact with Spanish). It also appears in North America, in Tuscarora, a Northern
Iroquoian language.
On the basis of centuries of documentation of European languages,
Heine and Kuteva (2006) propose a recurring sequence of extensions which can
result in such patterns. A marker begins in lexical gap questions (Who came?).
It is extended to indefinite subordinate clauses (I don’t know who came.). It is
then generalized to definite subordinate clauses (You also know who came.),
sometimes interpreted as headless relative clauses (You know the one who came.).
Finally, it may be extended to headed relative clauses (Do you know the woman
who came?). Each of these developments could happen spontaneously, but
contact could stimulate progress along the path.
Comparisons of 19th century documentation of connected Tuscarora
speech with that through the next century reveal the step-by-step
development of all of the major interrogative pronouns along this trajectory.
The pronouns did not all develop at the same rate, or in the same order as in
German or English, but all have now progressed to use in headless relatives.
The perfect coincidence of these Tuscarora developments with bilingualism
in English adds evidence of the potential effect of contact in stimulating such
evolution.

Keywords:  Interrogative pronouns; relative pronouns; contact; extension;

headless relatives

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?’

(2) Tuscarora káhne’ ‘who’ relative: Elton Greene, speaker

Thwé:’n wa’kayę’na’nit’úthahs ha’ káhne’ kayę’nę’né̜ nhyahr.
all he put them to sleep the who they are guarding him
‘He put to sleep all those who were guarding him.’

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?’

(4) Mohawk nahò:ten’ ‘what’ relative: Harry Miller, speaker

Takhné:kanont kati’ kí:ken, ohné:ka’ nahò:ten’ sathrorià:ton.
liquid-serve me then this water what you are talking about
‘Then give me this water that you’re talking about.’

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 

2.  The Iroquoian languages

The genetic relationships among the modern Iroquoian languages for which we
have documentation of connected speech are sketched in Figure 1.

Iroquoian

Southern Iroquoian Northern Iroquoian

Five Nations Iroquois

Cherokee Tuscarora Seneca Cayuga Onondaga Oneida Mohawk

Figure 1.  Genetic relationships among the modern Iroquoian languages

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

(7) Cherokee kato ‘what’ question: Montgomery-Anderson 2008: 481

Kato uùwáakhthi?
what it means
‘What does it mean?’
(8) Cherokee ji- relative: Montgomery-Anderson 2008: 132
Ahnawo ji-skihnéehn v̋. jitheeska.
Shirt rel-you.gave.me.flexible.subor I.am.ironing
‘I am ironing the shirt that you gave me.’

Without an interrogative/relative match in Southern Iroquoian there is no evi-

dence that the pattern was present in Proto-Iroquoian. The fact that there are
matches between Tuscarora and other Northern languages might at first suggest
that the pattern was established by the time of Proto-Northern-Iroquoian. But in
most of the languages, only some interrogative and relative pronoun categories
match (of the possible human ‘who’, non-human ‘what’, place ‘where’, time ‘when’,
etc.), and the matching categories are not the same from one language to the next.
Furthermore, the individual pronouns are not cognate across all of the languages.
It is of course possible that all indefinite and relative pronoun categories showed
perfect matches in Proto-Northern-Iroquoian, but that various pronouns have
been replaced in individual languages. It is more likely that the matches developed
more recently in the individual languages.

3.  The transfer of grammar

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 Tariana speakers, bilingual in Brazilian Portuguese, observed that

­Portuguese relative clauses contain pronouns that match those in questions. The
relative pronoun quem ‘who’, for example, has the same shape as the interrogative
pronoun ‘who?’.
 Questionable relatives 

(10) Portuguese: Aikhenvald 2002: 183

Quem sabia, falava assim.
who new spoke thus
‘Those who knew spoke 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?

Each of these developments could happen spontaneously within a single language.

Contact, however, could stimulate progress along this path. Speakers might extend
the use of a marker in one of their languages to contexts of the next stage, on the
model of its counterpart in their other language. As shown by Haspelmath (2001),
Heine and Kuteva (2006), and others, interrogative/relative pronoun matches in
Europe are in fact not restricted to Indo-European languages. Matches also appear
in genetically unrelated languages in the area, among them Basque, Hungarian,
Balkan Turkish, and Georgian. Heine and Kuteva discuss a number of individual
shifts in specific European languages that were apparently triggered by contact.
Few if any languages indigenous to the Americas are represented by ­philological
records comparable to those for many Romance, Slavic, and G ­ ermanic languages.
 Marianne Mithun

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).

(13) Identified Tuscarora speakers

19th century texts
Joseph Williams born 1817
Joseph Henry born 1840
Lucinda Thompson born 1850
20th century texts
Nellie Gansworth born 1884
Edith Jonathan born 1886
Elton Greene born 1889
David Hewitt born 1894
Howard Hill born 1923

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?’).

4.  Tuscarora té’ ‘what’

The Tuscarora interrogative pronoun ‘what?’ was used in questions in the

19th ­century essentially as it is today. (Material from earlier sources is presented
here in the modern orthography to facilitate comparison. The abbreviation RC
stands for the 19th century Hewitt and Thompson material published in Rudes &
Crouse 1987.)
(14) 19th century té’ ‘what?’ question: 1888 ms 432, RC 1987: 590
Te’ sè:ri?
what you want
‘What do you want?’
(15) 20th century té’ ‘what?’ question: Elton Greene, speaker, p.c. 1971
Té’ rá:’nę?
what he is saying
‘What is he saying?’

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.’

(20) 19th century indefinite comp of cognition: Joseph Williams, speaker

1897 ms 411: RC 1987: 357
Thyá:ryęhst ta’awé̜:te yu’nehá’nę węheyé̜:či hęh.
it is not known what it causes it it died q
‘They would be unable to determine what had brought about the
­person’s death.’

(21) 19th century indefinite complement of perception: 1888 ms 432:

RC 1987: 189
É̜:kkę’ te’ awé̜:te θkáhews.
I will see what thing it brings back
‘I’ll see what she brought back.’

These uses have persisted through the 20th century into the modern language.
 Questionable relatives 

(22) 20th century indefinite complement of speech: Nellie Gansworth 1948,

to Wallace
Wa’kayę’nahrù:yę’ té’ tihra’nyérhahk.
they asked what he was doing
‘They asked him what he had been doing.’

(23) 20th century indefinite complement of cognition: Elton Greene,

­speaker p.c. 1972
Šyę’nè̜:ri hé̜h tawé̜:te wa’neha’tha’?
you know q what it causes it
‘Do you know what causes it?’

(24) 20th century indefinite complement of perception: Elton Greene,

­speaker p.c. 1972
Wa’nyę’tkę’θéhre’ tawé̜:te hení:kę:.
they went to investigate what that
‘They went to see what it was.’

Prosodically these constructions constitute a single sentence, with a coherent,

descending pitch contour, visible in Figure 2. (The first pitch peak is the stressed
syllable θéh of the first verb ‘they went to see’.).

Wa’nyę’ké’0éhre’ tawę:te hení:kę:.

They went to see what that
0 2.426
Time (s)

Figure 2.  ‘They went to see what it was.’

These examples correspond to Heine and Kuteva’s Stage 2 of development.

Various mechanisms can be imagined for the shifts from Stage 1 to 2, most involv-
ing sequences of independent sentences as a starting point: ‘They asked him.
“What was wrong?” ’; ‘Do you know? What causes it?’. With frequent use, such
sequences could become integrated into single constructions.
 Marianne Mithun

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.’

(26) 20th century definite complement of cognition: Elton Greene

­speaker p.c. 1972
Kyę’né:ri: te’ rayá:θę.
I know what he is named
‘I know what his name is.’

(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.’

These constructions are examples of Heine and Kuteva’s Stage 3 of d ­ evelopment.

Some sentences with definite complements are potentially ambiguous in a sub-
tle way. The sentence ‘I saw what she brought back’ could mean either ‘I ­discovered
the answer to the question “What did she bring back?” ’, or ‘I saw the object she
brought back’. The ‘what’ clause could be interpreted either as a kind of embedded
question or as a headless relative designating a specific referent. Such ambiguity
provides a bridge for the extension of the construction to contexts in which it is
interpretable only as a headless relative. There is no evidence of the use of té’ ‘what’
in headless relatives in the 19th century, but we do see it in the modern language.
(28) 20th century headless relative ‘what’: Elton Greene, speaker p.c. 1972: 256
Wa’nętíhęhs tawé̜:te ęwęnhéhkę.
it stores what it will live on
‘It is storing what it will live on.’

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’.

Ha’ ękayętęhninéhek hé’thu ha’ …ę tawé:te kayakyeti:yahs u’tíkste tawé:te hé’thu.

They’ll be selling there the … uh things they make beadwork things there.
0 5.538
Time (s)

Figure 3.  ‘They’ll be selling the beadwork they make.’

5.  Tuscarora káhne’ ‘who’

The interrogative use of the pronoun káhne’ ‘who?’ has remained unchanged over
the past century.
 Marianne Mithun

(33) 19th century ‘who?’ question: Thompson 1888 ms 422, RC 1987: 233

Káhne’ ęwásθhara’ kyé:nę: í:kę: raká:θ’a?
who one will care for this it is boy
‘Who will care for this boy?’

(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.’

In the modern language, the pronoun appears robustly in indefinite complements.

(37) 20th century ‘who’ indefinite complement: Elton Greene, speaker
p.c. 1972: 202
The’ a:kyę’nè:ri:k kahne’ wa’na’rì:yu’.
not I know who one killed one
‘I don’t know who killed him.’
(38) 20th century ‘who’ indefinite complement: Elton Greene, speaker
p.c. 1972: 182
Ìskah wa’kayé̜:kkę’ káhne’ hé’thuh íhre’θ.
not they saw one who there he is walking around
‘They didn’t see who was walking there in the crowd.’
(39) 20th century ‘who’ indefinite complement: Elton Greene, speaker
p.c., 1972: 219
E̜ kayerá:ku’ káhne’ ha’
they will choose who the
yęyętkyérhę:ht nyuhtà:wé̜’ę.
she will throw her body down there Falls
‘They will choose who will go over the falls.’
 Questionable relatives 

These constructions correspond to Heine and Kuteva’s Stage 2 of development.

The 19th century material shows no uses of káhne’ ‘who’ in definite
c­omplement constructions. Definite complements with káhne’ do appear in the
20th century.

(40) 20th century ‘who’ definite complement: Elton Greene, speaker

p.c. 1972: 412.
ę’né:ri: káhne’ wahranęhsá:tya’t.
I know who he house bought
I know who bought the house.’

This construction corresponds to Heine and Kuteva’s Stage 3.

The 19th century material also shows no headless relatives containing káhne’.
They do, however, appear in the modern language.

(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.’

There is no evidence of extension of káhne’ ‘who’ to Heine and Kuteva’s Stage 4, in

headed relative clause constructions.

6.  Tuscarora hè̜:we ‘where’

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.).

(42) 19th century hè̜:we ‘where?’ question: 1888 ms 422, RC 1987: 437

Hè̜:we ha’ è̜:kwe tyahwá’ę:t?
where the person there one went
‘Where did the people go?’

(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 same marker appears in indefinite complement clauses throughout the

record.
 Marianne Mithun

(44) 19th century indefinite ‘where’ complement: Thompson 1889 ms 411,

RC 1987: 494
E̜hsahrù:yę’ hè̜:we tiké̜’rę’ uké̜ ’ tì:wa’θ thuh
you will ask where it sits or it is so much there
‘Ask it to locate for you the place where can be found the greatest numbers of
ste’awé̜:te ha’ θa’neθwé:ki.
something the you want
whatever kind of game you want.’

(45) 20th century indefinite ‘where’ complement: David Hewitt, speaker,

1951: 47 to Wallace
Ù:nę hésnę: wa’kayenawa’tiyé̜:tha’ hè̜:we tikakuhyá’kę.
now then they went to find where so they have crossed
‘They went back to find where they had come across.’

(46) 20th century indefinite ‘where’ complement: Elton Greene, speaker

p.c. 1973: 479
Thé’ akyę’né:ri:k è̜:weh tyahwáhse:t.
not would I know where you went to
‘I don’t know where you went.’

These constructions correspond to Heine and Kuteva’s Stage 2 of d

­ evelopment.
This particle does not introduce definite complement clauses in the 19th cen-
tury material, but it does in the 20th century.

(47) 20th century definite ‘where’ complement: Elton Greene, speaker

p.c. 1972: 183
Wahrá:kę’ è:we tihrá:ta’č.
he saw where there he is lying
‘He saw where he was lying.’

(48) 20th century definite ‘where’ complement: Elton Greene, speaker

p.c. 1972: 435
Kyę’né:ri: è:we nękheya’čirá:’nihr.
I know where I will sting him
‘I know where I’ll sting him.’

(49) 20th century definite ‘where’ complement: Howard Hill, speaker to

­Francene Patterson 2000
Thwé:’n ha’ é̜:kwe, kayeyę’né:ri hè̜:we,
all the person they know where
‘All the people know where
ękayekúhe’ ané̜hsnači’.
they will get dark seed
to get sassafras.’
 Questionable relatives 

These constructions correspond to Heine and Kuteva’s Stage 3 of development.

The modern language shows further extension of this marker to headless rela-
tives designating places: (h)ę̀:we ‘the place where’. These constructions can func-
tion either as arguments or, more commonly, as locative adverbial clauses.
(50) 20th century headless ‘where’ relative: Elton Greene, speaker p.c. 1972: 412
È̜:we tihru’nè:nę’ weyúhre: á:thu’.
where there he lives it is amazing cold
‘Where he lives it’s cold.’
(51) 20th century headless ‘where’ relative: Edith Jonathan, speaker to
­Lounsbury, 1950:  528
Ù:nę wá’ę’w hé’i’ è̜:we tíkta’č.
then it came I where I was lying
‘Then it came over to where I was lying.’
(52) 20th century headless ‘where’ relative: Elton Greene, speaker p.c. 1972: 335
Hé’thu yękęhá:wi:t è̜:we kyení:kę:, unęhráhse.
there I will take you there where this milk
‘I’ll take you there where there’s milk [that I drink].’

7.  Tuscarora kahné̜’kye ‘when’

The time interrogative is káhnę or kahné̜’kye ‘when?’. (-ę’kye is a place nominalizer.)

(53) 19th century ‘when’ question’: Hewitt ms 2895: 125
Káhnę tičiθarhékwę?
when you have gone back
‘When did you go home?’
(54) 20th century ‘when’ question: Elton Greene, speaker p.c. 1972: 269
Kahné̜’kye nęčhárku’?
when so you will go back
‘When are you going home?’

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.’

This construction corresponds to Heine and Kuteva’s Stage 3 of development.

The same marker also appears in the modern language in headless relatives,
‘the time when’. These usually function syntactically as temporal adverbial clauses.
 Marianne Mithun

(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.’

8.  Progression through the stages

This single century of Tuscarora documentation shows development of each of

the indefinite pronouns along the path predicted by Heine and Kuteva. The coin-
cidence of this progress with increasing bilingualism in English suggests stimula-
tion by contact. The pronouns have not progressed in unison, however. Figure 4
provides a summary of the uses of each in the 19th and 20th centuries. (Heine and
Kuteva’s Stage 3 has been split into two stages here, iii and iv.)

‘what’ ‘who’ ‘where’ ‘when’

19th 20th 19th 20th 19th 20th 19th 20th
i Simple questions x x x x x x x x
ii Indefinite complements x x x x x x -- x
iii Definite complements x x -- x -- x -- x
iv Headless relatives -- x -- x -- x -- x
v Headed relatives -- -- -- -- -- -- -- --

Figure 4.  Development of Tuscarora Indefinite Pronouns

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 Stage of development

what, who, when (was, wer, wann) 3
where (wo) 3–4
English
What 3
who, when, where, which 4
Tuscarora 19th century
when (kahné’kye) 1
where, who (hé:we, káhnhe’) 2
what (te’) 3

Figure 5.  Comparative stages of development

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

can of course be evaluated only in light of substantial information about many

more languages. One inviting direction of inquiry concerns the system in which
the innovation takes place, whether the expansion of a marker consists in the sim-
ple substitution of one marker for another in an existing structure, or creates a
brand new grammatical construction where there was none.

9.  Innovations in light of existing structure

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.

9.1  Té’ ‘what’

As shown earlier, the 20th century saw an extension of the indefinite pronoun té’
‘what’ into headless relative clause constructions.

19th century 20th century

Simple questions x x
Indefinite complements x x
Definite complements x x
Headless relatives -- x
Headed relatives -- --

Figure 6.  Development of te’ ‘what’

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 

(59) 19th century clausal nominal: 1888 ms 432, RC 160

Wahrakyéhrę’ ha’ θhrahà:wi’.
he put up the he carries back
‘He set out what he had brought back.’
This construction continues in the modern language, though it is less frequent.
Many of the verbs now used in this way are lexicalized as referring expressions.
(60) 20th century lexicalized clause: Elton Greene, speaker p.c. 1972: 329
Wahrá:kę’ hésnę: ká’θre’.
he saw then it drags
‘Then he saw a wagon.’

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

In German and English, interrogative pronouns have replaced demonstrative

pronouns in complement and relative constructions. A few Tuscarora examples
might suggest that the similar subordinate constructions existed in Tuscarora,
such as that below.
(63) Demonstrative construction: Edith Jonathan 1950 to Lounsbury MM 522
È̜:ruh yękhiya’tkahríθe ha’ ké̜ ’ nyękwa’nè:nę’
she she was telling us the where so we reside
hení:kę: tyakuyéhsawę.
that so it happened to her
‘She was telling us at home what happened to her.’

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.’

Wahękhęhsyę’ hení:kę:. Nahrakwa’nętí:ye:t hésnę: ékáhne’t hení:kę:.

I heard that one He sent me thus to destroy them, those.
0 5.584
Time (s)

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.’

Wa’tkaháhihθ hení:kę:, ruya’kwáher.

it met it that one he body carries
0 3.727
Time (s)

Figure 8.  ‘It met it, that one, a dinosaur’

The headless relative constructions such as ‘what they promised’ and

‘what happened to her’ did not develop as in German and English, through
the substitution of an interrogative pronoun for a demonstrative in an existing
relative construction. There was no relative construction, rather only a gen-
eral dependent clause structure. The extension of té’ produced a more specific
construction.

9.2  Káhne’ ‘who’

The 19th century material shows no uses of káhne’ in definite complements or
headless relatives.
Ideas translated with English definite complements were conveyed with
a ­simple Tuscarora finite clause, without overt nominalization, an indefinite
­pronoun, or complementizer. The clause was usually preceded by the article ha’,
which also served to mark its dependent status.
 Marianne Mithun

19th century 20th century

Simple questions x x
Indefinite complements x x
Definite complements -- x
Headless relatives -- x
Headed relatives -- --

Figure 9.  Development of káhne’ ‘who’

(66) 19th century absence of ‘who’ in definite complement: 1888 ms 432,

RC 1987: 193
Ruyę’nè:rih ha’ rané̜:’nę tíhsnę’ runé̜ sne’.
it knows the he feeds it and it loves him
‘It knew [who he was and loved him for having fed it].’
Ideas translated with headless relatives were also conveyed by simple clauses with-
out an indefinite pronoun, also usually with the article ha’. This was a major con-
struction in the 19th century, occurring pervasively in the texts.
(67) 19th century headless relative: Hewitt ms 433, RC 1987: 40–41
Ha’ hé’thu kayetá:kre’ kwé̜ hs akayaiheyé̜hθek.
the there they dwell not would they die
‘They who dwelt there did not die.’
(68) 19th century headless relative: Joseph Williams to Hewitt 1888 ms
438, RC 412
Ru’ríhę:t rahęwúhahs ranęwęthúhtha’
he has as his business he sinks boats he causes to drown
ha’ yera’nawé̜:’nye’.
the one is going along in water
‘His business was to sink boats and to drown those who were in them.’
This construction persists today, somewhat more robustly than its ‘what’
counterpart.
(69) 20th century headless relative: Elton Greene, speaker p.c. 1972: 326
Nęθwa’nęnuré̜hkwhek ha’ neθwa’nętyákhę’.
you all will continue to love each other the you all are married
‘Love one another, you who are married couples.’
(70) Tuscarora headless relative: Howard Hill, speaker, to Francene
­Patterson 2000
Há:ne:’ ha’ kakuneha’ké̜:ha’nę’ kayé̜:’nę.
that the they are old variously they say
‘That is what the old people say.’
 Questionable relatives 

In German and English, as noted in the previous section, interrogative pronouns

have been replacing demonstratives in subordinate constructions. In the 19th
century Tuscarora material, headless relatives designating animates also occur in
apposition to demonstratives.

(71) 19th century apposition to demonstrative: 1888 ms 432, RC 157

Kwé̜hs akakwè:ni’ ęθę’né’ku’
not was it able it will escape
‘It was not fast enough to escape
ha’ áhkwir ha’ há:ne:’ θę’né’ku’.
the doe the that it ran back
the fawn’s mother [who had returned in the interim].’

This pattern continues in the modern language.

(72) 20th century apposition to demonstrative: Edith Jonathan, speaker, to

Lounsbury 1950: 528
We’eteyú’knę’ hení:kę: kayeta’θá:rye’ kayekętí:θ’a.
she went to them those they are lying down they are small
‘She went over to those lying down, the children.’

(73) 20th century apposition to demonstrative: Elton Greene, speaker

p.c. 1972: 180
E̜kwehè̜:we, kyení:kę: kayetá:re’ skarù:rę’.
real persons these they live here Tuscarora
‘The Indians living here were Tuscarora.’

A typical use of demonstratives is below. Without consideration of prosody, this

could be interpreted as a relative clause structure: ‘the one that was named Oba-
diah went there.’

(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.

(75) 20th century passage with demonstrative: Elton Greene, speaker

p.c. 1972: 183
Yahwahrárku’ kyení:kę:.. . Čá:ks rayá:θę. Yahwahrárku’
he went there this one Obadiah he is name he went there
kyení:kę:.. . , ęčę’na’tahskúhči
this one he will release him
 Marianne Mithun

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.).

Yahwahrárku’ kyení:kę, Čáks rayá:sę. Yahwahrárku’ kyení:kę:, ęčęnaktahskúhči’.

He went away this one he is named Obadiah. He went away this one he will release him.

0 8.499
Time (s)

Figure 10.  Managing the Flow of Information

The development of the Tuscarora animate definite complements and headless

relatives in the 20th century thus does not parallel those in German or ­English. The
pronoun káhne’ ‘who’ was added to clauses rather than substituted for a demon-
strative in existing complement or relative constructions.
Constructions like that in (74) could of course evolve into relative clauses.
Even today, not all sequences of clauses containing demonstratives are separated
prosodically by such marked pauses and pitch resets.
Although té’/tawé̜:te ‘what’ and káhne’ ‘who’ show the same profile in the
modern language, both appearing in questions, indefinite complements, definite
complements, and headless relatives, they do not show precisely the same degree
of development. The té’/tawé̜:te’ ‘what’ complements and headless relative con-
structions have essentially replaced simple juxtaposed clauses serving comparable
functions, but the káhne’ ‘who’ complements and headless relatives are relatively
rare in the 19th century, and in the modern language, they still coexist with the
earlier, less grammaticalized alternatives.

9.3  Hè̜:we ‘where’

In the 19th century, hè̜:we ‘where’ had just begun to be extended to indefinite
complements.
The marker was in competition with a much more robust particle kę’. Kę’
appears in both indefinite and definite complements.
(76) 19th century kę’ indefinite complement: Thompson 1888 ms 432, RC
1987: 579
Ù:nę ha’ kę’ weθkata’ríheh wa’kayęnęnhyà:rę’.
now the where it puts head back on they examined
‘They sought to find out where the head had been taken to.’
 Questionable relatives 

19th century 20th century

Simple questions x x
Indefinite complements x x
Definite complements -- x
Headless relatives -- x
Headed relatives -- --

Figure 11.  Development of hè̜:we ‘when’

(77) 19th century kę’ definite complement: 1888 ms 422, RC 1987: 438–9

Ù:nę wahrá:kę’ kę’ newé̜ hryę’.
now he saw where they two enter dirt
‘He was able to see where the children had escaped through the ground.’

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.’

It occurs in numerous lexical items in the manuscript dictionary compiled by

Erminnie Smith and Hewitt (BAE ms 2850), such as kę’ yetakwthráhkhwa’ ‘where
one-bed-puts.up-with’ = ‘bed chamber’ and kę’ yehyatęhstayę’náhkhwa’ ‘where
one-paper-lays-with’ = ‘bookcase, desk’ (Rudes 1999: 298–9).
The particle kę’ continues in modern Tuscarora, but it is much less frequent
and has a narrower sense: ‘right where’. It does appear in headless relatives meaning
‘exactly the place where’, usually functioning as a locative adverbial clause marker.
(79) 20th century kę’ headless relative: Elton Greene, speaker p.c. 1972: 340
É̜:či hésnę: ha’ učísneh yahwahé̜:’ni’ u’téhsnakwt
one then the live coal she threw behind
‘She threw one coal back behind her,
ha’ kę’ thru’na’níhręh.
the right where so he was standing
right where he was standing.’

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

The particle persists in certain lexicalized constructions, such as terms for

‘home’.
(81) 20th century lexicalized kę’: Edith Jonathan, speaker to F.G. Lounsbury
1950: MM487
Ù:nę kyení:kę: wakwa’ruhčré̜ ’nę’ ké̜ ’ nyękwa’nené̜:tyę’.
now this we gathered ourselves where so we reside here and there
‘Then we held meetings right in our houses.’

(82) 20th century lexicalized kę’ construction: Elton Greene, speaker

p.c. 1971: 17
θá:ku’ ha’ kę’ wak’nè:nę’.
I went back the where I live
‘I went back home.’

Kę’ is an old demonstrative adverb, ‘right here’. An entry in Hewitt’s notebook

reads: Kę’ sá’ ‘Look here!’ (Rudes 1999: 298). Cognates persist in other North-
ern Iroquoian languages with this function: Oneida kʌ˛ ‘here (close to speaker)’,
Mohawk kèn:’en or kèn: ‘right here’, as in Kèn: sátien ‘Sit right here’.
The interrogative hè̜:we thus developed into a subordinator by replacing a
demonstrative adverb in existing overt subordinate constructions. It has not com-
pletely displaced the demonstrative, but it has narrowed its meaning.

9.4  Kahné̜:’kye ‘when’

In the 19th century, the pronoun kahné̜’kye ‘when’ had not expanded into contexts
beyond direct questions. Within just a century, it has been extended to indefinite
complements, definite complements, and headless relatives, which function as
temporal adverbial clauses.

19th century 20th century

Simple questions x x
Indefinite complements -- x
Definite complements -- x
Headless relatives -- x
Headed relatives --

Figure 12.  Development of kahné̜ ’kye ‘when’

In the 19th century, these subordinating functions were filled by several

c­ onstructions, all of which persist in the modern language. One was a simple
clause preceded by the article ha’.
 Questionable relatives 

(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.’

A number of overt temporal subordinators were also already established in the

19th century, particularly ù:nę ‘at the time, now, then, when’, áθę’ ‘when, finally’,
and kanyú’ ‘as soon as’.
(84) 19th century ù:nę ‘when’ clause: Thompson 1889 ms 441, RC 1987: 486
Ù:nę ú’nę’ θáhra’w wa’ętkáhrye’ ha’ wé’θhahk.
now as well he came back she told the it was walking around
‘When he returned, she told him it had been there.’
(85) 19th century áθę’ ‘when’ clause: Joseph Williams, speaker 1888 ms 438,
RC 1987: 412
Áθę’ ranyatarì:re’ wa’ná:tkę’.
when he lake crosses one met another
‘While he was wading in the lake he met him.’
(86) 19th century kanyú’ ‘as soon as’ clause: Thompson 1890 ms 445,
RC 1987: 122
Kanyu’ hésnę: θayuhθá:thu’ na’ é̜:čę’w
as soon as then it got dark again much she will return
‘When she returned in the evening
ù:nę hé’thu yahwá’nyę:t kę’ nyuta’čuhkwa’níhrę.
then there they two went there where there it heap stood
they two went over to the heap.’
All persist robustly in the modern language.
(87) 20th century ù: nę, kanyú’ ‘when’ clauses: Elton Greene, speaker
p.c. 1972: 184
Ha’ ù:nę ę’nwa’thè̜:wúha’
the when it will sail
‘When it sails,
kanyú’ ęθka’rát’a:’,
as soon as I will ride again
as soon as I go back,
nęθkayętkáhnę’w ha’ θú:tar.
they will wake back up the soldiers
the soliders will wake up again.’
The new subordinator kahné̜’kye tends to be used in irrealis contexts: ‘when in the
future, whenever’. Speakers report that it is usually possible to substitute the more
general ù:nę or ha’ ù:nę ‘when’ for kahné̜ ’kye.
 Marianne Mithun

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.

9.5  The variation

The interrogative pronouns ‘what’, ‘who’, ‘where’, and ‘when’ have now all reached
the second step of Heine and Kuteva’s Stage 3 of development, but apparently
at different rates. At the end of the 19th century, ‘what’ was at the initial step
of Stage 3, ‘who’ and ‘where’ at Stage 2, and ‘when’ just at Stage 1. All have now
been extended to the second step of Stage 3, but with varying degrees of robust-
ness. Tawé̜:te ‘what’ is now the usual basis for headless relatives. Káhne’ ‘who’
does appear in headless relatives, but these constructions are still in competi-
tion with basic finite clauses containing no indefinite pronoun. Hę̀:we ‘where’
appears in headless relatives functioning as locative adverbials, but the marker it
is replacing, kę’, persists with a specialized meaning ‘right where’. Finally, kahné̜
’kye ‘when’ can appear in headless relatives, but it is still a minority pattern with
a specialized irrealis meaning.

‘what’ ‘who’ ‘where’ ‘when’

19th 20th 19th 20th 19th 20th 19th 20th
Simple questions x x x x x x x x
Indefinite complements x x x x x x -- x
Definite complements x x -- x -- x -- x
Headless relatives -- basic -- robust, -- major -- minor
competing
Headed relatives -- -- -- -- -- -- -- --

Figure 13.  Varied Rates of Development

It is of course not possible to determine with certainty why the pronouns

developed in the order they did. One variable might be worthy of attention. The
two interrogative pronouns that were the most advanced in the 19th century and
 Questionable relatives 

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.’

These same demonstrative adverbs now also function as part of conventionalized

adverbial clauses, which form a prosodic unit with the preceding, now main clause.
These adverbial clauses can now occur either before or after the main clauses they
modify.
(91) 20th century ù:nę ‘when’ clause: Elton Greene, speaker p.c. 1972: 223
Yahwakuwé:rhu’ ha’ ù:nę yahé̜ čyęht.
it covered her the when she went down,
‘It [the mist] covered her when she went down.’

100
Pitch (Hz)

70

50 Yahwakuwé:rhu’ ha’ ù:nę yahé:čyęht.

It covered her when she went down.
0 2.49
Time (s)

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

Aikhenvald, A. 2002. Language Contact in Amazonia. Oxford: OUP.

Campbell, L. 1987. Syntactic change in Pipil. International Journal of American Linguistics 53(3):
253–80.
Haspelmath, M. 2001. The European linguistic area: Standard Average European. In Language
Typology and Language Universals: An International Handbook, Vol. 2, M. Haspelmath,
E. Konig, W. Osterreicher & W. Raible (eds), 1043–1068. Berlin: Walter de Gruyter.
Heine, B. & Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
Heine, B. & Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: CUP.
Heine, B. & Kuteva, T. 2006. The Changing Languages of Europe. Oxford: OUP.
Hill, J.H. & Hill, K.C. 1986. Speaking Mexicano: Dynamics of Syncretic Language in Central
­Mexico. Tucson AZ: University of Arizona.
Karttunen, F. 1976. Uto-Aztecan and Spanish-type dependent clauses in Nahuatl. In Chicago
Linguistic Society Papers from the Parasession on Diachronic Syntax, 150–158.
Kortmann, B. 1998. Adverbial subordinators in the languages of Europe. In Adverbial Construc-
tions in the Languages of Europe [Empirical Approaches to Language Typology, EURO-
TYP], J. van der Auwera (ed.), 457–561. Berlin: Mouton de Gruyter.
Le Goffic, P. & Wang, X.L. 2002. Les pronoms interrogative – indéfinis du chinois: l’Example de
shéi ‘qui/quelqu’un/quiconque’. Verbum 24(4): 451–71.
Lehmann, C. 1984. Der Relativsatz. Tübingen: Gunter Narr.
Lounsbury, F.G. 1961. Iroquois-Cherokee linguistic relations. In Symposium on Cherokee and
Iroquois Culture [Smithsonian Institution Bureau of American Ethnology Bulletin 180],
W.N. Fenton & J. Gulick (eds), 11–17. Washington DC: U.S. Government Printing Office.
Matras, Y. 1996. Prozedurale Fusion: Grammatische Interferenzschichten im Romanes.
Sprachtypologie und Universalienforschung 49(1): 60–78.
 Marianne Mithun

Montgomery-Anderson, B. 2008. A Reference Grammar of Oklahoma Cherokee. PhD disserta-

tion, University of Kansas.
Romaine, S. 1984. Towards a typology of relative-clause formation strategies in Germanic. In
Historical Syntax [Trends in Linguistics 23], J. Fišiak (ed.), 437–70. Berlin: Mouton de
Gruyter.
Rudes, B. 1999. Tuscarora-English/English-Tuscarora Dictionary. Toronto: Toronto University.
Rudes, B. & Crouse, D. 1987. The Tuscarora Legacy of J.N.B. Hewitt: J.N.B. Hewitt Wa’ękhiríhwayęʔθ
Skarù: ręʔ [Canadian Ethnology Service Paper No. 108]. Canadian Museum of Civilization
Mercury Series. Ottawa: National Museums of Canada.
Schafroth, E. 1993. Zur Entstehung und vergleichenden Typologie der Relativpronomina in den
romanischen Sprachen. Tübingen: Max Niemeyer.
Thomason, S. & Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics.
Berkeley CA: University of California.
Wallace, A.F.C. 1952. The Modal Personality Structure of the Tuscarora Indians ­[Smithsonian
Institution Bureau of American Ethnology Bulletin 150]. Washington DC: U.S. G ­ overnment
Printing Office.
 Language and language family index

A 274–275, 284–285, 288–289, Iroquoian languages  270–271,

African languages  40 291–292, 296–297, 299 274, 294, 297
Akatek  122 European languages  192–193, Israeli Hebrew see Hebrew 
Altaic  137 238, 254, 269–270
Ancash Quechua  48 J
Arawakan  270 F Jakaltek  122
Atlantic Creoles  40 French  31, 42, 122, 179, 192 Japanese  3, 7–8, 14, 19, 21, 243
Australian languages  122, 208 Jupde see Hup, Hupda
Austronesian languages  167 G
Awa Pit  194 Gavião  243–247, 249–252 K
Georgian  273 Kambera  102
B German  3, 9, 11, 13–14, 17, Kannada  38
Balkan Turkish  273 21–22, 35, 43, 50–51, 54, 269, Khmer  243
see Turkish 285, 288–289, 291–292 Kihungan  20–21
Bambara  5–9, 11, 14, 21, Germanic languages  13, 273 Konkani see Saraswat
59–62, 140 Gothic  17 Brahmins Konkani  38, 43
Basque  43, 273 Greek  22 Koyoga-Bambara  140
Berber  102 Guarijío see Warihío  Krio  22, 42
Bislama  32–33, 42 Guaycuruan  173–174 Krongo  122
Bodic  15, 18 Gulf Arabic  122 Kwa languages  40
Gullah  31, 40, 42
C Guyanese Creole English  L
Capeverdean  32 32, 42 Lakhota  51, 62, 226, 228
Carib  194 Lhasa Tibetan  17, 102
Caribbean  15 H Louisiana Creole  31, 42
Cayuga  271 Haida  228 Luiseño  102
Cherokee  271–272 Hawaiian Creole  31
Chuave  10 Hebrew  13–14, 17, 22 M
Cinta Larga  243 Hittite  9–11, 14, 21 Makú family  191
Cmiique Iitom see Seri Hokan  214 Malagasy  54–55
Cora  105, 128 Hopi  102 Maale  30
Cupeño  102 Huichol  105 Mandarin Chinese  19
Huallaga Quechua see Mayan languages  253–256, 258
D Quechua  Middle English  296
Dagur  137 Huaraz Quechua Miraña  202
Dâw  193 see Quechua  Mohawk  8, 270–271, 294
Dravidian languages  38 Hungarian  273 Mondé  243
Hup see Hupda, Jupde  149, Mono  148, 150, 255
E 161, 191–203, 205–210
English  11, 13, 17–18, 31–32, Hupda see Hup, Jupde N
42–43, 47–48, 50–54, 56–58, Nadahup family  149, 193
91, 108, 112, 114, 118, 122, I Nadëb  193
152–153, 155, 237, 239, Indo-European languages  Nahuatl  269–270
245–246, 259, 261–262, 269, 54, 273 Negerhollands  34–36, 42
 Language and language family index

Névome  138 R Tok Pisin  31–33, 41, 43

Ngemba  30 Ralámuli see Tarahumara Tukang Besi  102
North American Romance languages  273 Tukanoan languages  198
Languages  225 Tupian languages  243
Northern Iroquoian  269–271, S Tümpisa Shoshone  158
294, 297 Salamãy  243 Turkic  15, 148
Northern Paiute  147–150, Saraswat Brahmins Konkani Turkish  102, 273
152–159, 161–163, 165, see Konkani Tuscarora  269–272, 274–275,
167–168 Seneca  271 279, 281, 283–286, 288–293,
Numic languages  15, 147, Seri see Cmiique Iitom  296, 298–299
149–150, 153, 157–159, 168 213–215, 217–218, 223–225,
227–229, 231–232, 234–236, U
O 238–239 Ute  15–17, 20–22, 102, 116, 157
Old English  13 Shoshoni  102 Uto-Aztecan family  15, 67, 70,
Old Norse  13 Sioux  226 97–98, 102, 121, 127–129, 137,
Oneida  294 Slavic languages  273 139, 142–145
Onondaga  271 Spanish  43, 122, 177, 215, 254, 269
O’odham see Papago  Southern German  22 W
128, 138 Southern Iroquoian  272 Warihío see Guarijío  105
Opata  129, 133, 137, 139 Southern Numic  158 Western Numic see Numic
Southern Paiute  158 languages
P South Seas Pidgin  41
Palenquero  36, 43 Sumerian  15 Y 
Papago see O’odham  Supyire  7, 21 Yaqui  67–68, 70–73, 75,
Pima Bajo  127–130, 132–133, Surui of Rondônia  243 78–82, 84–89, 93–94,
135–145, 168 97–109, 111–114, 117–118,
Pipil  270 T 121–123, 128–129, 133,
Pirahã  8–9 Tarahumara see Ralámuli  105, 137, 139–140, 142,
Philippine  21, 43 128–129, 133, 137, 139–140, 145  145, 218
Polish  192–193, 253–254, Tariana  38–39, 43, 194, 198, Yucatec  253–267
262–263, 267 269–270, 272–273 Yuhup  193, 200, 203
Portuguese  31, 38–39, 42–43, Tibetan  17–21, 102 Yuman languages  218 
201, 269, 272–273 Tibeto–Burman  15, 148,
Proto-Iroquoian  272 167–168 Z
Proto-Northern- Toba  173–178, 181, 183–184, Zapotec  265
Iroquoian  272 187, 189 Zoró  243
 Name index
A 154–155, 167, 178, 193, 214,
Abney, S. P.  226
Aikhenvald, A. Y.  27, 29,
37–39, 43, 194, 198, 270,
272–273
Aissen, J.  258
Alsagoff, L.  43
Anderson, J.  148, 162, 271–272
Andrews, A. D.  69, 98, 106,
136–137, 191–193, 214, 217,
228, 253, 255–256, 263
B De Vries, M.  112
Basilico, D.  48, 226, 228
Baxter, A.  27, 42
Bell, A.  43
Belloro, V.  101
Bickel, B.  122
Bird, C.  5–6, 59, 232
Borgonovo, C.  122
Bresnan, J.  261
Briceño Chel, F.  255
Bricker, V.  256
Bruyn, A.  31, 36, 43
Buelna, E.  105, 111
Bunte, P.  158
C Escalante, F.  98, 102
Campbell, L.  214, 218, 270
Carlson, G.  7, 97, 99–100, 112
Casad, E. H.  85, 98, 102,
121, 128
Chapuis, D.  42
Chomsky, N.  232
Cinque, G.  226, 228
Citko, B.  191–192, 198, 253–255,
260–263, 266
Clements, C.  27, 42
Cole, P.  48
Comrie, B.  27–31, 42, 54,
69–70, 78–80, 89, 91, 97–98,
100, 102–103, 106, 127–128,
134, 136–137, 140–141, 144,
165, 167–168, 177–179, 237,
225, 243, 250, 253, 256 246–247, 251
Corbett, G. G.  192 Gómez, P.  105
Craig, C.  214 Green, L.  36, 43
Cristofaro, S.  97, 109, 117, 122 Greenberg, J. H.  128
Crouse, D.  274–275 Grimshaw, J.  261
Crowley, T.  43 Grinevald, C.  202
Crumrine, L. S.  98 Guasti, M. T.  122
Guerrero, L.  97–99, 101–102,
D 105, 113–114, 118, 122, 128, 140
Dayley, J. P.  158, 168 Gumperz, J. J.  29, 40
De Roeck, A.  232
H
Dedrick, J. M.  85, 98, 102, 121, 128 Hagège, C.  28
DeLancey, S.  17–18, 148, 167 Hale, K. L.  128–129, 133, 137,
Dench, A.  148 208, 214
Dick, F.  239 Hansen, K.  43
Dixon, R. M. W.  113  Haspelmath, M.  91, 166, 207,
Dolan, T. P.  43 270, 273
Dryer, M. S.  105, 192, 218, 225 Heath, J.  29, 37
Heine, B.  13, 27–29, 37–38, 43,
E 47, 129, 133, 140, 149, 165,
Elman, J. L.  239 269–270, 272–279, 281–285,
Enrico, J.  228 296, 298–299
Epps, P.  149, 161, 191, 194, Hernández Doode, G.  98
196–197, 199–200, 202, Hill, J. H.  109–111, 270, 274,
206–209 281–282, 284, 290
Hill, K. C.  109–111, 270, 274,
Estrada Fernández, Z.  98, 214 281–282, 284, 290
Evans, N.  148 Hinskens, F.  34–36, 42
Hoekstra, J.  43
F Holm, J.  42–43
Félix Armendáriz, R. G.  98 Hopper, P.  128
Fought, C.  43 Horie, K.  127–128, 134, 136, 243
Fox, B.  100, 108–110 Huddleston, R.  214, 236–237
Franklin, G. L.  203 Hudson, R.  232
Fraser, N. M.  202 Hutchinson, J.  32, 42–43
Huttar, G.  42
G Huttar, M. L.  42
Givón, T.  3–5, 9, 11, 14–16, 18,
20–23, 69, 82, 92, 102, 116, I
121, 127–128, 132–133, 136, 141, Islas, B.  105
147–149, 155–157, 159, 161, Iturrioz, J. L.  105
 Name index
J Marlett, S. A.  213–214, 219,
Jackendoff, R.  49
Jacobsen, W. H., Jr.  225
Jelinek, E.  98
Johnson, J. B.  98, 103, 107, 111
K McWhorter, J.  27–29, 36
Karttunen, F.  270
Kaufman, T.  214, 284
Keenan, E. L.  54, 69–70, 79,
97–100, 102, 109, 112, 119, 128,
137, 141, 144, 155, 167, 178,
225, 250–251
King, M.  9–10
Klinger, T.  42
Koenig, J. P.  122
Koptjevskaja-Tamm, M.  68,
91, 97, 100, 106, 117
Kouwenberg, S.  42
Kroeger, P.  253
Kuno, S.  232
Kuteva, T.  3, 13, 27–31, 37–38,
42–43, 69–70, 98, 102–103,
129, 133–134, 140, 144, 149, 154,
165, 225, 269–270, 272–279,
281–285, 296, 298–299
L Nichols, M. J. P.  150
Lambrecht, K.  99, 122, 265
Langacker, R.  143, 152,
162, 218
Langdon, M.  218, 225
Langendoen, T.  102, 108
LaPolla, R.  48
Le Goffic, P.  284
Lehmann, Ch.  70, 80–81, 88,
94, 97, 106, 109, 112, 122, 128,
133–134, 137, 261, 266, 284
Lick, H. C.  43
Lindenfeld, J.  98, 102, 128
Look, R.  32, 115, 140, 149, 153,
160–161, 166, 197, 214, 225,
235, 259, 270, 294
Lorenzino, G. A.  42
Lounsbury, E. G.  271, 274, 283,
288, 291, 294
M Romaine, S.  43, 296
Malchukov, A. L.  91, 106,
122, 216
S
221–222, 224, 227, 230, Schachter, P.  20, 224
235–237 Schafroth, E.  284
Martínez, C.  81, 98, 102, 108 Schneider, E.  27, 42
Matras, Y.  284 Schüle, S.  122
McCawley, J. D.  237 Schwegler, A.  36, 43
Sebba, M.  33, 41
Mesthrie, R.  43 Shibatani, M.  3, 19, 22, 68,
Migge, B.  27, 40, 42 89–94, 105, 167, 266
Miller, G.  122, 148, 232, 270 Silva Encinas, M. C.  98
Monforte, J.  255–257, 261–262 Skopeteas, S.  255
Montaño Herrera, R.  213–214 Smith-Stark, T.  258
Montgomery-Anderson, B.  Snapp, A.  162
271–272 Stiebels, B.  256
Moore, B.  203 Stockwell, R. P.  224
Moore, D.  243, 245–246
Moreno Herrera, F. X.  T
213–214 Thomason, S. G.  40, 284
Moser, M. B.  213–214, 230, Thompson, S. A.  19, 89,
235–236 91, 97–98, 100, 106, 109,
Mufwene, S.  31–32, 40, 42 193, 274–276, 280, 282,
Muysken, P.  42 292, 295
Myers-Scotton, C.  40 Thornes, T.  147, 152–153, 158,
162–163, 168
N Toosarvandani, M.  153
Nedjalkov, V. P.  153 Tryon, D. T.  32–33, 42
Nichols, J.  37 Trudgill, P.  29, 37
Noonan, M.  148, 245, 247 V
Van der Auwera, J.  99–100, 122
O Van der Riemsdijk, H.  261
Ospina Bozzi, A. M.  200, 203 Van Rossem, C.  34–36, 42
Van Valin, R. D., Jr.  48–52, 55,
P 62, 98
Partee, B. H.  224 Vázquez, V.  105, 128
Peet, W., Jr.  31, 42 Veenstra, T.  42
Perlmutter, D. M.  222 Vendler, Z.  112–113
Peterson, P.  214, 236–237 Verhoeven, E.  255
Plag, I.  28
Prescod, P.  27, 43 W
Pullum, G. K.  214, 236–237 Wallace, A. F. C.  274–75,
277–278, 282
R Wang, X. L.  284
Radford, A.  237 Weber, D. J.  121–122
Reh, M.  122 Wilson, R.  29, 40
Rojas, R. M.  177, 265 Williamson, J.  62, 226, 228
Roland, D.  239 Wurm, S.  33
Rude, N. E.  119 Z
Rudes, B.  274–275, 293–294 Zwicky, A.  192
 Subject index
A see also complementation  194–196, 201–205,
Accessibility Hierarchy  70,
78–80, 100, 108, 119,
128, 137, 141, 143–144,
178, 250
Adverbial clauses  69, 94, 98,
114–115, 127, 149, 153, 161,
166–168, 296–298
see also locative adverbial
clause  293
Agent  89–90, 108, 117, 122,
164–165, 200–202,
205–206, 244
Agreement  71–85, 201–202
Ambiguity  208, 245,
249–256, 278  203, 210
Anaphoric pronoun  8, 22, 179
Attributive  55–63, 137, 141,
147, 275
B Converbal clause  207–208
Bilingual  173, 272, 297
see also bilingualism  37,
269, 275, 284, 297, 299
C 160, 162
Cleft  20–21, 86, 237, 254
Complement clauses  69, 94,
122, 128, 136, 139, 166, 179,
182, 186, 222, 243, 251, 264,
276, 281–282, 286
see also complements  67–68,
71, 77, 100, 112–113, 115–117,
121–122, 127, 144, 149, 158,
166, 168, 173–174, 178–180,
182–183, 186–189, 243, 245,
247, 258–259, 276, 278,
280–281, 283–284, 289,
292, 294, 298
see also complement
strategies  113
see also complement-taking
predicates  112, 114, 122
97–98, 113, 117–119
Complementizer  30, 53–56,
59, 140, 180, 245, 256,
259–260, 289
Conjunction  5, 9–11, 23,
30, 130
see also clause linkage
marker  99, 102,
112–114, 117
Contact  27, 29–30, 37–40, 43,
218, 259–260, 269, 273,
284, 299
Continuum  30, 88, 97, 100,
117, 123, 149, 165, 191, 193,
see also gradient
phenomenon  97, 191, 193,
207, 210
Constraints  98
Co-reference  92, 94, 173,
178–180, 183
Core argument  97, 100–103,
115, 117–120, 122, 150,
D G
Definite article  33, 213,
228–229, 239
Demonstrative  3, 5–9, 11–14,
21, 23, 33, 41, 85, 127, 129,
133, 137, 140–141, 151, 159,
161, 165, 173–175, 180, 181,
184, 186–188, 193, 197,
200–201, 215, 224, 230,
245, 254, 261, 286–289,
291–292, 294, 296–298
Dependent  5, 69, 88, 97–99,
101–104, 107, 110, 112–123,
127, 138, 143, 155, 157–158,
160, 167–168, 175, 179–180,
182–183, 185, 187–188,
207–208, 214–215, 234,
262, 286, 289
Deverbal noun  213–218, 220,
226, 235
Different subject  214–215, 235
E
Embedding  5–9, 14, 21, 59, 69,
179, 217, 231–232, 239
see also embedded
clause  48, 54, 113, 118, 121,
132–133, 141, 234, 248, 251
see also embedded
structures  156–157
see also embedded relative
clauses  159, 168
F
Finite  15, 22–23, 53, 82, 88,
91–93, 122, 135–136,
165–167, 214–217, 219, 221,
234–235, 237, 247, 256
see also finiteness  67–68,
82, 87–88, 90, 120, 128,
132, 135
Gap  70, 72–73, 78–80, 104,
119, 127–128, 134–135, 155,
179, 254, 257, 259–261,
265–266, 269, 273
see also gapping
strategy  194
Genitive  15, 17, 23, 71, 73–74,
76–78, 84–85, 87, 99,
101, 103, 113, 116–118, 130,
132–136, 141, 144, 148, 157,
167, 247, 250
Grammaticalization  9, 28,
127–129, 133, 141, 144–145,
189, 191, 193, 202, 210,
273, 275
 Subject index
H Missing argument  173–174,
Head noun  4, 12, 18–19, 30,
33, 48–49, 53–63, 67–80,
84–85, 87, 89, 92, 94,
97–98, 102–105, 107–108,
110–112, 115–116, 118–119,
127–128, 132–136, 139, 144,
154–156, 158–160, 164,
167–168, 173–174, 178–180,
182–186, 188, 218, 220, 222,
224, 227–228, 230–231,
234, 236, 239, 251, 261–262
see also bound noun  193,
197–200, 202–203, 207, 209
see also domain
nominal  191–207,
209–210
see also head nominal  89,
155, 157, 214
Headless relatives  161, 192,
195, 204, 207–210, 237,
253–254, 261, 269, 278, 281,
283–285, 287, 289–294,
296, 298
I 158, 160, 165–167, 180,
Indirect object  17, 54, 67–68,
70, 72–79, 81, 84, 87, 98,
144, 215–216, 258
Internal relative clauses  251
Interrogative pronouns  257,
266, 269, 273, 275, 288,
291, 296–299
Intonation contour  5–6, 8,
11–12, 19–20, 156, 160
see also intonational
contour  159–160, 168   140–141, 143–144, 150–151,
Irrealis  130–131, 135–136,
215–217, 219–223, 235, 254,
295–296
L 247–250, 255, 257–259, 278
Language contact  27, 29,
37–40, 299
Light head  251, 254, 260,
262–263, 266
Locative relativization  78–80
M 155–157, 159–160, 164–166,
Matrix clause  48, 54, 69,
120, 264
178–179, 188
Modality  50, 82, 87
N Perception  99, 114, 120, 122,
New information  97, 100,
108–109, 178, 183, 185, 188
Nominal domain  255, 260
Nominalization  3, 16, 18–21,
30, 67–69, 81–83, 85,
87–94, 97–98, 100,
105–106, 108–109, 117,
121–123, 128–129, 132–133,
137, 140, 144, 147–148,
152–154, 157–159, 161, 163,
165–168, 179, 191, 200,
202, 205–206, 210, 213,
216–217, 243–252,
266, 289
Non-finite  18, 67–69, 82,
128, 131–133, 135–136,
166, 194
Non-subject  15, 79, 83–85, 97,
99, 102, 104, 123, 129–131,
134–136, 143–144, 149, 156,
194, 215
Null head  201, 223, 225, 239,
253, 262–265
O 74, 77, 91–92, 94, 98–100,
Object  8, 15–17, 19, 23, 33–34,
41, 48, 54, 67–68, 70,
72–84, 86–87, 90–91, 94,
98, 101–102, 108, 110, 116,
119, 127–130, 132, 134–137,
155–161, 163–166, 180,
194–196, 200, 205–206,
208, 215–219, 221–222, 224,
226–228, 232, 238, 244,
Object relative clause  155, 157
Oblique  15, 34, 42, 67–68,
70–71, 75–81, 84–85, 87,
90, 99, 102–104, 108, 110,
114, 119, 127, 129–130, 133,
137, 141–142, 144, 150–151,
194–196, 208, 215, 217, 219,
222, 257–258, 264
P
Participle  23, 140, 149, 151,
153–154, 157–158, 160, 163,
166–167
276–277
Possessive  34, 71, 111, 117, 127,
130, 135, 148, 167, 175,
180, 194
Possessor  70, 79, 84, 141,
150–152, 157–160,
165–166, 180–181, 206,
215–216, 218, 221–223,
255, 264
Pragmatic conditioning 185
see also function  109
Pragmatically-based
theory  108
Predicate nominal  195, 217, 233
Privileged syntactic
argument  48, 51
Pronoun retention  74, 79–80,
102, 119
Prosody  279, 287, 289, 291
R
Realis  215–216, 219–223
Recursion  213, 217, 224,
231, 239
Reference  47–50, 62, 69, 72,
108, 112, 148–149, 154–155,
157–158, 161, 167, 173,
177–180, 183, 192–193, 198,
200, 213–214, 236, 255,
258, 265
see also referent  4–5, 7, 11,
14, 69, 98, 106, 109, 112,
122, 127–128, 142, 164,
178–182, 184–188, 195, 197,
201, 205, 214, 236, 245, 265,
278, 280
see also referring
expressions  52, 91–93,
164–165, 167, 278, 287, 289
Relative pronouns  34, 50, 54,
59–60, 138, 168, 217, 257,
263, 266, 269, 271, 273, 298
Relative constructions  253,
261, 265–267, 271,
287–289, 292–293
 Subject index 

Relativization strategies  21, 112–114, 116–119, 121–123, T

68, 70–71, 129, 144, 179, 127–137, 140–141, 143–144, Topic  23, 98, 109–110, 128, 149,
256–257 149–158, 160–167, 180, 153, 165, 185, 187, 194, 213,
Restrictive relative clause  99, 194–195, 206, 208, 235, 254
160, 164, 251 214–224, 226, 228, 231–235,
Resumptive pronoun  7, 76, 238, 244, 246–249, U
107, 110, 119, 128 254–259, 265 Undergoer  51–52, 55, 59, 61, 109
Subordination  40, 69, 98, 147,
S 149, 166, 168, 196, 204, W
Subject  8, 15–17, 19, 23, 29, 271, 285 Wh-pronoun  142, 266
31–34, 38–40, 42, 48, see also subordinate Wh-questions  20
51–52, 54–55, 61, 67–68, clauses  15, 58, 122, 127,
70, 72–85, 87, 89–90, 158, 167, 235, 258, 269, Z
97–99, 102–105, 107–110, 273, 297 Zero relativization  31–32